COST PERFORMANCE OF MULTI-JUNCTION, GALLIUM ARSENIDE, AND

SILICON SOLAR CELLS ON SPACECRAFT

Edward M. Gaddy
Goddard Space Flight Center, Greenbett, MD 20771

silicon and .20 mn thick silicon solar cells on this

spacecraft is compared below. The attributes of these
cells and their associated blanket arrays is shown in
Table I for the spacecraft's nominal 5,600 W end of life
(EOL) array. The thin silicon cells have metallization
that adds significant weight to them and explains why
the silicon cells' weight does not ratio with thickness
and efficiency. The blanket weight in Table I includes
the cell weight. If extremely lightweight blanket
components are used, the thin silicon cells will have a
weight advantage over the GaAs/Ge cells. However, at
present this type of blanket is not yet practicable.

ABSTRACT
Spacecraft solar array engineers now have cell
choices running from inexpensive and less efficient
silicon (Si) cells, to gallium arsenide on germanium
(GaAdGe) cells, to more expensive and efficient multijunction (MJ) cells. This paper finds that the more array
weight can be reduced by using more efficient cells,
even when they are a very expensive option in terms of
the array alone, and put into the spacecraft payload
i. e. the scientific instruments or in the case of
commercial spacecraft, the communications equipment,
the more cost effective the spacecraft array system.
This is true for a wide variety of spacecraft. This is
because of the very high price of launching a spacecraft
payload and supporting it with a spacecraft.

Table I
Blanket Array Attributes Populated With Different Cells

Attribute

INTRODUCTION

Nominal Efficiency
This paper reports the cost effectiveness of using
Si, G W G e and MJ solar cells on varying types of
spacecraft and solar arrays. The spacecraft selected
are mainly differentiated by their weight. The largest
weighs approximately 4,700 Kg, the next 3,500 Kg and
the lightest 240 Kg. All of the spacecraft are tow earth
orbiters.
In all of the cases studied, the more efficient solar
cells offer substantial performance advantages to the
spacecraft. Namely, more spacecraft payload can be
launched and used. If any reasonable cost estimate is
assigned to launching the payload and supporting it in
orbit, the more efficient cells offer an enormous price
advantage to less efficient cells.

Si

Thin
Si

GaAs

14.8

14.0

18.5

7.8

11.8

16

52.9

35.0

25.8

EOL Efficiency
7
59.0
Array Area (d)
Cell Wt. (Kg)
36.7
Blanket Wt. (Kg)
102.1
Blkt. Hsing Wt. (Kg) 72.8
Mast Assy. Wt.(Kg) 33.6
Other Weight (Kg)
24.6
Total Weight (Kg)
233.2
Cost (M$)

28.0

MJ
24

74.5

32.3

23.8

73.8

72

53.6

70.4

61.5

56.2

32.1

26.8

23.6

23.5
199.9

19.9
180.2

18.1
151.5

29.5

38.0

33.3

The table uses cell costs per square centimeter of

$1.25 for Si, $2.08 for thin Si, $20.83 for GaAdGe, and
$26.04for MJ cells. The price for the GaAs/Ge and MJ
cells is on the high end. The price for the MJ cells is
particularly uncertain because they are not yet in
production. Nonetheless, the above should be a
reasonable estimate in that the production cost of MJ
cells is readily compared to GaAs/Ge cells. The major
difference is that the MJ cells must have more active
layers grown. As this can be done under automatic
control without moving the cell from the reactor where
the layers are grown, the cost increase is minimal. The

LARGE SPACECRAFT
The large spacecraft studied is a typical low earth
orbiting spacecraft. It will have a life of 5 years, a weight
of 4,690 Kg, an altitude of 700 km, an inclination of 98
degrees and a single side whose normal is parallelto the
spacecraft's velocity and another side whose normal
always points to the nadir. The spacecraft will be
equipped with a flexible blanket array that is sun
tracking.
The cost effectiveness of .14 m thick multijunction, .14 mm thick gallium arsenide, .062 mn thick

293

U.S. Government work not protected by U S . copyright

25th PVSC;May 13-17, 1996; Washington, D.C.

P P PWR (28.7-ADATA-A.OOMM-dELEC-AFVR&A)/INSTR = APWR

(10)
S P P P SA (28.7-ADATA-~COMM-&LEC-APPWR(11)
&A)/INSTR = &A

in price is primarily due to a necessarily lower

the addition of reverse bias protection.
cells for the first
are used will
gnif~antlyhigher than that given in
of course, for contingency to cover
d costs for the first few uses of the
the GaAdGe cell
igher than used in
uHs of this paper
higher efficiency
cells offer substantial cost benefits for a typical blanket

CF = . 9

(12)

PF

.451

(13)

SF

.863

(14)

where CF is the fraction of the comma~d

and ~ a ~ a
handling system that is used l o s
the fraction of power that is used
is the fraction of wei~htthat

xt to last row of the table shows a

asing array weight with more

difference in weight in
The solution to these equations is:

increases some of the spacecraft

pport the payload must also increase.

ADATA

4.61

(15)

2.55

(16)

dELEC =

1.87

(17)

APWR =

2.71

(18)

0.83

(19)

Total = 12.57

(20)

"A

m capabil~jesthat support the payload are

propo~ionaito the payload's needs, and the

&SA =

load
Not all of the 28.7 Kg can
because as the payload
the
spacecraft subsystem
to support it. From

to increase ten percent and

and data handling system is

weight constant, the net instrumen

16.1 Kg.

spacecraft payload weight. For the case of the array

ing from GaAs to MJ cells the following equations
are applicable:
INSTR

1163.3

(1)

DATA =

369.7

(2)

COMM =

204.1

(3)

ELEC =

294.8

(4)

PWR =

427.6

(5)

SA

151.5

(6)

1 1 63 Kg. On this basis it takes

MJ array is priced $4.7 million dollars less than

account for the increase in array
requiredby the increased ins~rument
obtained from equations (6) and (19) that show that the
MJ array increases in

In the above INSTR, DATA, COMM, ELEC, PWR,

espectively the weight in kilograms of the
command and data handling system, the
commun~ationssystem, the electrical system, the
stem less the solar array and the solar array.
simultaneous equations that compu?e the
hangs of the these subsystems are:

3.5 million. Thus t

complete picture in
other s u b s y s ~ eof~ the
spacecraft must also
ncreased in capability and
price to accommod
larger i n s ~ r u ~ e nIf ~ this
.
increase in capability and price is p ~ o p o ~ ~ to
o nIah~
weight of the subsystem the s acecraft increases in
price $1.3 million, using data not in this paper. Thus the
cost to launch and service the additional 16.1 Kg is

CPDATA (28.7-ADATA-&OMM-&LEC-APWR&A) /INSTR = h A T A

(7)
C P COMM (28.7-ADATA-AWMM-&LEC-AFWRBSA)/INSTR = A.COMM
(8)
P P ELEC (28.7-hDATA-AWMM-hELEC-APWR&A) /INSTR = B L E C
(9)

294

inclination of 98 degrees and a single side whose normal

points at the sun. The spacecraft will be equipped with a
fixed array with an aluminumface sheet and honeycomb
core substrate.
The computations for the cost effectiveness of the
solar cells for these spacecraft were updated to the cell
priies directly under Table 1. Results are provided in
Tables Ill and IV below. The updated GaAs/Ge and MJ
cells are much more expensive than in the references.
Even so, these cells retain a very large price advantage
over silicon cells.

$3.2 million. This data, as well as data for the silicon,

and thin silicon arrays is summarized in Table II.
Table I1
Cost Effectiveness of Solar Cells by Type
For a Large Spi

/rem

Cost Per Unit W.

to launch &
Support Science
(K@W
Instrument Webht

ells by Type

Additional
Instrument Wf.
Over Si Array (Kg)
Cost Per Unit
Wejght to "BUY
Science with High
Perf. Cells ( M g )

Si

Item

Cost Per Unit wr. to

Launch & Support
Science ( K W g )

552

531

519

Instrument Weight

633

664

678

31

45

152

127

Additional Instrument
Wf. Over Si Array (Kg)
Cost Per Unit Weight to
"Buy"Science with High
Perf. Cells (UKg)

The above computations are not exact as there are

many uncertainties and inefficiencies that factor into
the fabrication of spacecraft, particularly a one of a kind
spacecraft. Many subsystems are not the optimum
weight because they are based on earlier subsystems
to obtain heritage. The weight uncertainty for certain
new equipment or instruments may be high and this will
cause the spacecraft to carry a large weight
contingency for a while. When this contingency is
resolved, the spacecraft may then have the ability to
carry additional weight that can be traded for dollars by
using a less expensive array. But, on the average, the
computations should be valid.
To use Table I to obtain the $3.2 million dollar
advantage computed another way above, the
"Additional Instrument" weight row in GaAs is multiplied
by the "Cost Per Unit Weighr for GaAs and the
corresponding product is computed for the MJ array.
The two products are subtracted. The Table value is
$3.1 million, the difference from the $3.2 million being
due to rounding.

NIA

Table IV
Cost Effectiveness of Solar Cells by Type
For a Small Space aft with i qigid Ari r

Item
Cost Per Unit Wt.
to Launch & Support
Science (K$/Kg)
Instrument Weight
Additional Instrument
Wt Over Si Array (Kg)
Cost Per Unit Weight to
'Buy"Science with High
Perf. Cells (YKg)

MEDIUM SPACECRAFT

Computations similar to the above have been

performed and published for medium and small size
space~raft.~'~
The medium size spacecraft will have a
l i e of 3.5 years, a weight of 3,512 Kg, an altitude of
350 km, an inclination of 35 degrees and a single side
whose normal is parallel to the spacecraft's velocity and
another side whose normal always points to the nadir.
The spacecraft will be equipped with a sun tracking
array with an aluminumface sheet and honeycomb core
substrate. The small spacecraft will have a life of 2
years, a weight of 241 Kg, an altitude of 700 km, an

MJ

Si

GaAs

680

641

617

58.2

62.4

64.9

4.2

6.7

NIA

189

164

For the small and medium spacecraft, the

computation for the thin silicon cells was not performed.
This is because the weight of these spacecraft's rigid
substrate arrays swamps the weight of the cell stack.
As a result, the use of thin solar cells has only a very
slight effect on the weight of the array.
CONCLUSIONS

In all cases studied, the MJ solar cells are effective

in increasing the payload weight of the spacecraft. This

295

It is particularly so in the
size spacecraft. In these

o $164,000 per kilogram.

support costs that range

price is the most

Ths multi-junction cells are also very effective for

blanket array, but not nearly so
raft with rigid arrays. For blanket
purchase additional spacecraft
payload at a cost of about $250,000 per kilogram. The
reduced effectivene~of the cells is primarily because
of blanket arrays is less sensitive to size and
hive to the heavier cell weight of the GaAdGe
11s. For the same reason, the GaMGe cells
show almost no advantage relative to the silicon cells
and a disadvantage relative to thin silicon cells.
This paper uses a methodology that is different
than that used by Ralph. His estimates are based
primar~lyon the cost to launch the array into orbit and
this cost is repotted by him to be about $1 1KKg for a
bw earth orbit. He does not include the cost of the
spacecraft support to the payload. This paper does
include that cost. As a result, the price advantage of the
more efficient cells is significantly greater as computed
in this paper. Nonetheless the qualitative ranking of the
cost eff~tivenessof the cells is the same.

Gene Ralph, High Efficiency Solar Cell Arrays System

Tradeoffs, IEEE First World Conference on
Photovoltaic Energy Conversion, Hawaii, Figure 1,
p. 1998.
Edward M. Gaddy, Cost Trade Between MultiJunction, Gallium Arsenide, and Silicon Solar Cells,
Proceedings of the 14th Space Photovoltaic Research
and Thno/ogy Conference,Cleveland, OH, Oct.,
Edward M. Gaddy and Hsiao L. Smith, Relative Cost
Effectiveness of Multi-Junction, Gallium Arsenide, and
Silicon Solar Cells, Proceedings of the 34th Aerospace
Sciences Meeting and Exhibit, Reno, NV, Jan., 1996.

296

DESIGN and MANUFACTURING EXPERIENCE

with GaAs / Ge SOLAR CELLS
as GAINED with the TEMPO SOLAR GENERATOR PROGRAM
Author: Wolf D. SCHULTZE, Solar Array Technologies at DASA (Daimler Benz
Aerospace, former MBB), Dornier Satellite Systems GmbH,

81663 Muenchen, Germany

Besides few relativelysmall applications up to now
practically all spacecraft have been alimented by solar
arrays based on silicon solar cells. Different types of
cells have been developed resulting in different characteristics and cost. The relation is, however relatively proportional.
The TEMPO digital communication satellite is the first to
use GaAs solar cells in large quantities. For three identical arrays about 80 000 cells have been worked so far;
further rebuilts have been started.
In the following the main application relevant differences
of these Ga-As cells are compared with standard and
advanced (hi q) Silicon cells as preferably used, in the
field of
Cell Data and Properties and consequently
Array Design and Analysis
* Manufacture (Repair) and
Testing.
The critical and driving elements relevant to the implementation of the cells into a program are:

be the optimum size and utilisation, if fitting. In practically

all applications external limitations and requirements
(e.g. panel dimensions, voltage level and current sectioning, degradation...) do not allow the full size use; all minor dimensions represent accordingly loss of power and
Consequently economy.
The smaller cell area requires automatically about 50 %
more cells and covers to be worked/assembled for the
same panel area, compared to usual Si sizes.
The lattice orientation and consequently the preferred
breakage direction of the very brittle material is such that
defects propagate practically always parallel to the cell
edges. As detailed in para. 3.1 testing showed the evidence of cracks crossing the whole cell. Silicon cells are
specified to angles of k 45 deg, usually reducing the risk
of total circuit interruption.

1)

1.3) Cell Thickness/Mass

In order to counteract the Ge fragility for cell manufacturing and array assembly a minimum thickness of the
base material has to be provided (here: > 145 pm = 57
mil). The higher specific weight follows a total mass coverage, including 100 pm cover glasses of at least l ,2
kg/m2 compared to 0,9kg/mZ for 200 pm Si and twice as
much as achievable with thin Si cells. Obviously the panel substrates must take into account the higher loads by
added mass.

Cell Data and Properties

While the silicon cells can be provided by all cell

manufactureres in different types and configurations (Jimited to some degree the high efficient cells > 15+16 %
'concerning) Ga-As cells are manufactured by two companies only, with limited availability.
In Table I the main facts and data, as relevant to the topic fields are listed for comparison.

1.4)

Cell Reverse Characteristics

1.2) Dimensions and preferred Breakage Direction

As shown in Figure I the reverse characteristics of

Ga-As cells differ significantly from Si: the breakdown
voltage is between 1,5 and 6 V, while Si stands around
3 0 6 0 V, at typically more uniform behaviour for each type. The reverse sensitivity differs again basically. While
Si cells are damaged by thermal breakdown, often recovering, the Ga-As cells are preferably damaged electrically, in the p/n junction, the more the better the cell.
Consequently all cells have to undergo a reverse load
test resulting in damages, lost for further use. In case some shadowing is expected on certain parts of the array,
dedicated cells will see a significantly more stringent test
load, resulting again in damaged and lost cells.

Another relevant design criterion is the limitation

in cell size and preferred breakage direction of the cells
available on the market. In fact the cells are fabricated
on square germanium wafers (43 x 43 mm), which would

In any application there exists the probability, that certain

numbers of cells will see reverse operation points, preferably at BOM. For Ga-As cells the probability is significantly higher, the damage potential also. In TEMPO e.g.

1.1)

Cell Efficiency

The prime comparison figure, not very meaningful,

however for the applications, is the conversion efficiency
at AM0 conditions. The Ga-As advantage of about 45 %
will be reduced due to the higher operating femperatures
and the high specific particle degradation towards the
end of the mission. Relative advantages exist in proton
rich orbits due to the lower conversion ratio of about
1000 e- per p+ compared to 3000 as applicable for Si.

357

0-7803-3166-4/96/$5.000 1996 IEEE

25th PVSC; May 13-17, 1996; Washington,

D.C.

follow the series of weld spots and extend few mm along

the preferred breakage directions. The propagation as
result of thermal cycling is minor and considered tolerable by the program.
The formation of these HL's and, if more severe, of
cracks occurred during all manufactur~ngsteps, practically homogenously distributed over processes, time
and cell lots. Especially after thermal cycling new cracks
appear and existing cracks propagate (see below). Mechanical loading (sinus, acoustic, depl~ymen~shocks~
practically has no detremental effect.

139 cells are wired in series.

Based on the damage sensitivity and especially the fact,
that no applicable mission experience is available on
ground operational tests can practically not be performed
it shall be pointed out that the standard use of shortening
shunts or AC power systems are a potential source of problems. Consequently open circuited systems can ovoid
the problematic, resulting for most users in significant reconfigurations of the entire power systems. The operation
points would have to stay in the Uoc part of the array
electrtical characteristics, so far forbidden from conservitive reasons.

3)
2)

Experienc@sgained during Manufacture

Coupon T h ~ r m a~ ~y ~ ~ i R

A standard coupon test at the beginn~n~

of the
TEMPO program did not show any damage or unexpected result. In the progress of the manu~actu~~ng
the HL's
have been detected. In order to evaluate the behaviour
of cracks and HL's under mechanical and thermal loads
(GEO mission time + 50 %, + 85 to - 175 "C) a series of
tests have been performed. One of the major activities
was the manufacture and (TV + thermal) cycling testing
of special coupons: a vide variety of p~e-damagedHLand cracked cells was selected ~no~mal~y
red). The main findings of these tests were

The implementation of the Ga-As cells for mass production is definitely different compared to the development
and pre-production activities. The significantly higher brittleness has resulted in "softer"manufacturing approaches,
especially in the welding and bonding processes. Another
improvement was introduced by better matching of the
thermal expansion coefficients and consequently the inplane forces between Ge-substrate and cover glass. For
the low temperatures (down to - 185 "C) the differences
below about -100 "C,i.e. where the glass adhesive is frozen, are lowest when using CMZ type glass.

3.1) Crack and Hairline Propagation

2.1) Cell Edge Damage Sensitivity

At certain intervals the propagation of the defects

has been registered, as well as the influence on power
performance. The data have been analised wrt. the direction of propagation (parallel to the gridlines or crossing), correlation to previous lengths, nr. off etc.
The specific total growth rates are shown in Fig. 3.
The normalised growth rate of hairlines as well as
cracks decreases significantly, by almost 4 orders of
magnitude. A s shown the trend is still down in the "bathtub-couwe", allowing the statement, that there is obviousty no real danger visible.
Hairlines:
" The initial total length did increase by only 6,4 %.
" Very seldom the HL's became real cracks.
No new hairlines have been formed.
" With higher cycling life further growth tends to become
parallel to the gridlines.

During the cell -> p contact welding -> calm glassing process it was detected, that repeated measurements
in order to gain a feeling about measurement reproducibility - exhibited permanent and consecutive degradation
(see Fig. 2). The degradation was caused by mechanically
contacting the cell edges in the magazines and with jigs
and fixtures. Application of a "friendly" coating cured the
effect.

2.2) Cell Performance Bandwidth

Generally the Ga-AdGe solar cells as manufactured

result in a wider Gaussian performance distribution as Si
populations ( = 3 times, > -+ 10 %, depending on limit conditions). P-welding and glassing generally resulted in widening of the incoming cell class, to a higher degree at lower classes. The average glassing loss is about 0,3 %.
Due to the matching bandwith at least 1,5 % of power was
lost compared to 5 1 YOwith Si.

Cracks:
* The majority of the defects appear after few thermal

cycles. This prooves, that the performance of the panel

2.3) Contact Welding

N cycles is able to "trigger" infant cracks and enable, if

The frontside welding did not cause any problems;

necessary, repair.

the cell degradations I losses were below 1 %.

* New cracks appeared, starting from cell edge as well

a s inside. The latter was never experienced with Si
cells. Two of the new cracks went all the way through
the cell. A power degradation of about 5

The rearside welding exhibited a so far unknown phenomenon: generation of "hairlineeffects" (HL's). The investigations let assume, that in fact the focal and temporal heat
weave generated during the welding (Ag rearside to Ag interconnector, 123 pm thick) causes some softening of the
Ge substrate and by this opening a crystal gap which closes again after cool-down. The defects are very narrow,
hard to detect and, a s shown in the following, obviously
"contained".They generally start at the welding locations,

circuit has been measured.

The development of some cracks is shown in Fig. 4.
" The accumulated increase in total crack length, propagated and new, was above 200 %.This is to be seen in
relation to the HL's, because in the programs cracks are
repaired (at relatively high expenses).
358

sian yield population and potential correlation to the mechanical defects.

Generally the crack generation and propagation should

be watched seriously, especially since the reasons for
the high rates are not understood so far. It is assumed,
that the Ge crystal itself has irregularities and built-in
stresses which cause new cracks and propagate through
the whole cell with statistical relevance. This is worse
than the experience with Si cells.
4)

5)

The application of the Ga-AdGe solar cells to large

arrays is under way. No directly applicable mission results are available, especially not conceming the discussed fields as

Manufacturing Results

O influence of the generation and propagation of mainly

the cracks. According to the authors opinion the potential of a crack problem cannot be excluded.

A total of about 10 % of cells with HL and crack

defects have been repaired during the whole manufacturing and testing process for TEMPO and PanAmSat
#l.
The 10 TV cycles were intended to find more
cracked cells for subsequent repair. The reference is obviously the "natural" distribution of samples compared to
the biased and pre-damaged selection on the coupons.
The defect generation and growth rate of cracks and
HL's consequently differs somewhat from the coupon results:
O

O The knowledge of the reverse cell behaviour, its stability and longtime behaviour is currently insufficient and
should be improved by carefully observing the on orbit
performance.
O
Not directly comparable cycling tests for LEO missions indicate potential degradation problems. Results
will be published, when available.

New HL's have been detected (or first seen?) after the

The currently planned Ga-As cells for further programs

will be improved on the rear sides in two ways: the substrate surface will become polished (was raw) and the
Ag layer will be increased, as on the cell front side. Better reproducibility of welding process and lower defect
starting conditions are expected.

N cycles
O Basically the distribution of the defects looks well randomly distributed over the panel areas, i.e. no biasing
due to built-in substrate discontinuities as inserts, doublers
The number of new and total cracks (they were already repaired before the bonding process) is significantly
higher than of the HL's.

...

While searching for potential reasons of the defects the

following correlation has been found with statistical relevance (basis: 65 000 cells or 10 panels with about 250
circuits):
O
HL's seem to be related to the cell = circuit current
class such that the lowest class (=16 % below average)
has 4 4 times more defects. This appears for rate as
well as total number after N.But
O Cracks show an eventually negligible such correlation
(see Figs. 5 and 6).

mm

V.l,.,.

(VI

"cm

1.1

nor

RMRC CUUCII~

Specific Cost

tv no1

z
' 0

@e
vat,.,.

(VI

V.I.,.

(VI

Figure 1: Ga-As Cell Reverse Characteristics

Si,Hi rl

Si, BSR

17
2 from 4(5)" 0 wafer
200
100

123

20.000

'Ow/

1Array:
EOM Specific Power
EOM Sp. Mass

IVD

Ga-AdGe

Basics:
AM0 (28 "C)
Cell Size
Cell Thickness
Breakage
Typ. cost / m2

bl8a..

f ; ;[-*;J
EKRX

Further analyses may concentrate on the reasons for the

formation of the different cell qualitioes within the Gaus-

Cell Type

Conclusion

'4 L o o

25

143
17,7

1,s

117

18,3

43x43"
145
hi, 0 190
120.000

168

17,8

-2-

Table 1: Main Cell and Array Data Comparison

359

Formation of new Cracks

# of Actions

Magazin
coated
+avg X lop
Magazin

-4%

uncoated
-5 %

o
l

o
R

M
F

Thermal

Cycles

Crack and Hairline Propagation

Figure 2: lop Damage by Edge Contacting

I;

100

I F
\

10

5
a

40

accum.

0.1

--E

-5

30
Y

F 20
2

0.01

Hairlines
accum.

0,001

0.0001

1000

10

100

Thermal

Cycles

10

Figure 3: Specific Growth Rate of Cracks and

Hairlines after Thermal Cycling

0
1

10

100

1000

Thermal Cycles (GEO, Coupon)

Figure 4: Growth History of Thru Cracks

Normalised new Defect Rate

Normalised post N D e f e c t s

.;3

.-a

i3

i3

23

+ 2

0
1

Current

9 1 0 1 1 1 2

Class

Current

9 1 0 1 1 1 2

Class

Figure 6: Total Defects after Panel TV

Figure 5: Defect Generation after Panel TV

360

UV STABILITY OF HIGHEST-QUALITY PLASMA SILICON NITRIDE PASSIVATION

OF SILICON SOLAR CELLS
Thomas Lauinger, Jens Moschner, Armin G. Aberle, and Rudolf Hezel
lnstitut fur Solarenergieforschung HamelnlEmmerthal (ISFH)
D-31860 Emmerthal, Germany

ABSTRACT

thermal oxide without a diffused surface region is not

stable under concentrated illumination [7].
The microwave downstream (remote) PECVD
(RPECVD) of SiN,:H has recently turned out in our lab to
be extremely successful as passivation technique for
p-type Si wafers [8]. In this technique the damage to the
substrate surface during the deposition process is much
lower than it is in a parallel-lplate reactor. Together with a
different plasma chemistry [9], very low SRVs and, as will
be shown below, an excellent UV stability are achieved.
This study investigates in detail the UVstability of
different ISFH silicon solar cells passivated by SiN,:H
monolayers and thin-thermall-oxide/SiN,:H double layers.
If covered by a standard P'd module glass plate (cut-off
wavelength -340 nm), all cells are found to be perfectly
stable under UV exposure. In order to gain improved
insight into the UV behavior of SiN,:H-passivated p-Si
substrates, we investigate test samples that do not
possess any metal contacts. As shown in Ref. I O ,
symmetrical SiN,:H/Si/Sihl,:H
structures allow to
accurately determine extremely low surface recombination velocities by means of the MW-PCD method.

The UV stability of Si solar cells passivated by lowtemperature remote PECVD silicon nitride films is tested.
Perfect stability of the front surface passivation and the
rear surface passivation of both p-n junction as well as
MIS-IL Si solar cells is observed. Using the microwavedetected photoconductance decay (MW-PCD) method, a
very small and slow degradation of the differential
effective surface recombination velocity S e f d is observed
at silicon nitride-passivated p-Si surfaces corresponding
to the non-metallized rear surface regions of bifacial cells.
However, the degradation is too small to have any impact
on the long-term stability of encapsulated 17-18% rear
efficient bifacial cells. Thin-silicon-oxide/silicon-nitride
double layers incorporating Cs as used at the front
surface of MIS-IL solar cells provide perfectly stable and
excellently low differential s & d values of 23 cmls on
1.5-Qcm wafers. Applied to the rear surface of bifacial Si
solar cells, this double-layer scheme gives the potential of
stable rear efficiencies of even 20%.

INTRODUCTION
Many investigations in the past have shown that lowtemperature (< 500C) plasma-enhanced chemical vapor
deposition (PECVD) of silicon nitride (SiN,:H) is well
suited to provide excellent surface passivation of silicon
solar cells [I-51. There are two possible applications of
SiN,:H films: (a) passivation of diffused surfaces, e.g. at
the front surface of diffused p-n junction silicon solar
cells, (b) passivation of the pure silicon bulk material, e.g.
at the front surface of metal-insulator-semiconductor
inversion layer (MIS-IL) solar cells or at the rear surface
of both p-n junction as well as MIS-IL solar cells.
In case (a) UV stability is not a critical problem since
the doping profile reduces the surface recombination
velocity (SRV) at the SiN,:H/Si interface to a much lower
effective SRV at the edge of the space charge region [6].
The UV stability is much more important, however, if
SiN,:H is used in case (b) to passivate the non-diffused
surfaces of high-efficiency Si cells. In these applications
the silicon nitride must directly pravide excellent Seff
values, which have to be stable under UV exposure since
an increase directly affects the surface passivation. As an
example, for 18% rear efficiency of bifacial silicon solar
cells, Seffmust be better than about 200 cm/s [15]. This
problem is already known from the Stanford concentrator
Si solar cells, where a front surface passivation by a

UV TESTING OF SOLAR CELLS

A. Experimental details
An accelerated UV tesl is performed with a 15.1%
efficient standard MIS-IL cell [ I l l , a 16.7% efficient
truncated-pyramid MIS-IL cell [12], a 18.5% efficient p-n
junction cell [13], and a 17.2% rear efficient bifacial p-n
junction cell [13]. All cells are fabricated on 0.5 or
1.5 Qcm boron-doped singile-crystalline float-zone (FZ)
silicon and are using SiN,:H as combined passivation and
antireflection layer. Both MIIS-IL cell types receive a short
thermal oxidation and a C:s incorporation prior to the
SiN,:H deposition, resulting in a double-layer passivation
scheme on the front surface. This is done to achieve a
very high positive insulator charge density for a low sheet
resistance of the induced inversion-layer emitter.
RPECVD silicon nitride dleposition is performed in a
microwave downstream reactor (Plasma Technology).
Using 'design-of-experiments' methods, we determined
an optimized deposition parameter set [14]: substrate
temperature 375"C, pressure 270 mTorr, 50 sccm NH,
3 sccm SiH,, microwave power 60 W. These parameters
result in silicon nitride films with an refractive index of
about 2.3, indicating a silicon-rich nitride. Since the nitride
film of standard MIS-IL cells is deposited onto the existing
metallization, this cell type is restricted to a maximum

417
0-7803-3166-4/96/$5.00 0 1996 EEE

25th PVSC; May 1-3-17, 19%; Washington, D.C.

nitride deposition temperature of 250C. This is known to

51.
lead to a poor surface passivation quality [I
The cells are illuminated by a 400-W metal halide UV
lamp (Hoehnle 400F). The spectrum of the lamp is shown
in Fig. 1. For comparison, the standard solar spectrum at
the Earths surface (AM1.5G) is included. To perform a
realistic test for solar cells, a UV filter with a cut-off wavelength of 340 nm is used, representing the module
encapsulation. As demonstrated later by the UV tests on
PCD samples, the radiation damage of our silicon nitride
passivation is exclusively due to wavelengths below
400 nm. In the wavelength range 340-400 nm, the
averaged intensity of the UV lamp is 7 times the intensity
of the AM1.5G spectrum. A weighted intensity ratio of 9 is
obtained using a linear weighting function decreasing
from unity at 340 nm to zero at 400 nm. This weighting is
also based on the MW-PCD results, which show faster
degradation for shorter wavelengths (see Section UV
testing of PCD-samples, part C). For the yearly
illumination of the module we use I000 h/year of AMI .5G
radiation, which is rather typical for Germany. The cell
temperature was about 60C during the UV test.
At several stages during the UV test, the cell
parameters efficiency q, fill factor FF, open-circuit voltage
V
,,
and short-circuit current density J are measured
B
under a light intensity of 100 mW/cm and at a cell
temperature of 25C.
1 1 1 1

1 1 1 1

1 1 1 1

1 1 1 1

1 1 1 1

a)

- 2018

Outdoor operation [years]

0 1 2 3 4 5

b)

uv

16

5 700
650

p0 600
30
0

J
v)

200 400 600

UV exposure [h]

Fig. 2 (a) Measured parameters of a SiNx:H passivated p-n

junction Si solar cell during UV exposure. The solid lines are firstorder regressions. (b) Illuminated fron

a)

- 19

Outdoor operation [years]

0 2 4 6 810

b)

E 17

p-SI

15
80
75
70

650

-600

I l l 1

U_

r TuvT t

a
-

p 550

40
35
30

$
v)

200 400 600

UV exposure [h]

Fig. 3 (a) Measured rear-illuminated cell parameters of a SiNxH

passivated bifacial p-n junction silicon solar cell during UV
exposure. The solid lines are first-order regressions
(b) Schematic of the illuminated rear surface design.

300 350 400 450 500 550 600

wavelength [nm]

a)

- 1917

Fig. 1 Spectrum of the 400-W metal halide UV lamp in comparison to the AMI .5Gspectrum. The intensity of the UV lamp was
measured at the test plane (24 cm from the bulb).

Outdoor operation [years]

0 1 2 3 4 5

b)

1 iuvl 1

15

B. Front illumination of p-n junction silicon solar cells

Passivation of n-diffused emitters by silicon nitride is
known to be well suited for Si solar cells [16, 171. Ref. 16
reports that silicon nitride-passivated, phosphorousdiffused cells are perfectly stable under UV exposure.
Fig. 2 shows the cell parameters of a SiN,:H-passivated
diffused p-n junction Si solar cell fabricated at ISFH
during UV exposure, together with a schematic of the
illuminated front surface design of the cell. In agreement
with Ref. 16, we observe no degradation of all cell parameters during 600 hours of UV exposure, corresponding
to 5.5 years of outdoor operation in non-concentrating,
glass-encapsulated modules.

650

- 600

a-

550

-409
35
30
0

200 400 600

UV exposure [h]

2
3

p-si

Fig. 4 (a) Measured cell parameters of standard and truncatedpyramid MIS-IL silicon solar cells during UV exposure. The solid
lines are first-order regressions. (b) Schematic of the illuminated
front surface of the standard MIS-IL cell.

418

C. Rear illumination of bifacial silicon solar cells

Fig. 3 shows the rear-illuminated cell parameters of a

the SiN,:H film thickness on the UV stability. A set of PCD

samples with different film thicknesses was illuminated up
to 120 hours on each surface, using a 320-nm UV cut-off
filter. To eliminate effects from different reflection and
absorption properties, Seffdis plotted versus the total
number of UV photons reaching the Si/SiNx:H interface.
As can be seen, the UV stability increases with increasing
40 nm is
film thickness. A minimumi thickness of
required to ensure long-term Se.d values below 100 cm/s.

SiN,:H passivated bifacial cell fabricated at ISFH during

UV exposure, together with a scheme of the illuminated
rear surface of the cell. Perfect stability of all cell
parameters is observed during the test. In typical bifacial
applications, the rear intensity of the PV module does not
exceed 50% of the front intensity. Thus, our 600-hour UV
test corresponds to an outdoor operation of the cells of
more than 10 years.

film thickness

D. Front illumination of MIS-IL silicon solar cells

Fig. 4 shows the cell parameters of both a standard
MIS-IL cell and a truncated-pyramid MIS-IL cell fabricated
at ISFH during UV exposure. As can be seen, all
parameters of both cells are stable during the 600-hour
UV test, corresponding to an outdoor operation of the
cells of 5.5 years.

80

UV TESTING OF PCD SAMPLES

Solar cell parameters are related to an area-averaged
SRV including the metal contacts. For highest-quality
surface passivation layers this averaged SRV is
dominated by the metal contacts [IO, 131. Thus, small
degradations of the passivation layer properties cannot
be observed by the solar cell parameters. Therefore, in
order to gain improved insight into the UV behavior of
SiN,:H
passivated non-diffused p-Si substrates, we
investigate PCD samples that do not possess any metal
contacts. Hence, a very accurate determination of
extremely low surface recombination velocities is
possible [IO]. The sensitivity of PCD samples with regard
to Sefi variations is about 10 times better than the
sensitivity of our present solar cells.

50

No. of photons with h < 400 nm

at interface [IO'* cm-21

100

Fig. 5 Measured differential SRV Sefd of p-Si/SiNx:H

passivations as a function of the number of the UV photons
reaching the Si/SiNx:H interface and the film thickness (UV
cut-off wavelength 320 nm, total UV exposure 120 hours).

A. Experimental details
All PCD samples are (100)-oriented boron-doped
i wafers with a
single-crystalline float-zone (FZ) S
resistivity of 1-2 Qcm and a thickness of 300 pm. The
wafers exhibit excellent minority-carrier lifetimes above
1.7 ms. Prior to the Si nitride depositions onto both
surfaces, the samples receive a standard RCA clean.
Samples with double-layer passivation schemes are
additionally oxidized at 800C and dipped into a Cs
solution. The MW-PCD measurements are performed
under a bias light intensity of about 50 mW/cm2 at room
temperature. Differences in the measured effective
minority-carrier lifetime zeff are exclusively related to
surface passivation effects, since we observe no variation
of the bulk minority-carrier lifetime during all processing
steps. From zeff we calculate a differenfial SRV Seffd [I81
and use its increase to characterize the passivation
stability. It should be noted that the acfuaf SRV at silicon
nitride-passivated p-Si surfaces is less than a factor of 2
higher than the differential SRV Seffd [19].
B. Film thickness dependence
To obtain clear effects within reasonable times, a
more extreme UV test is performed using harder UV
radiation with h,,=320
nm. Fig. 5 shows the influence of

419

C. Wavelength dependence1
Fig. 6 shows the waveleingth dependence of the UV
degradation. A set of identical PCD samples was
illuminated using different 1JV cut-off filters. To obtain
clear effects within reasonable times, samples with thin
SiN,:H films (35 nm) were iinvestigated. With increasing
UV cut-off wavelength the stability improves enormously.
For wavelengths greater than 400 nm no degradation is
detectable. To verify this result, an additional test was
performed using a 250-W halogen lamp. The films are
perfectly stable under the halogen illumination (3.7 suns,
400 hours).

80

50

100

150 50yt 10000

No. of photons with h c 600 nm at interface [ l o

Fig. 6 Measured differential SRV Seffd of p-Si/SiNx:H

passivations as a function of the number of the UV photons
reaching the Si/SiNx:H interfaa?and the UV cut-off wavelength
(total UV lamp exposure 120 hours, film thickness 35 nm).

D. Passivation schemes a s used on solar cells

In Fig. 7, a SiN,:H monolayer as used for the rear
surface passivation of bifacial cells is compared with a
SiO,/Cs/SiN,:H
double layer as used for the front surface
passivation of our MIS-IL solar cells. The samples are
illuminated for up to 120 hours on both surfaces using a
320-nm cut-off filter. In spite of the more extreme UV
exposure (h,,,=320 nm), the SiO,/Cs/SiN,:H
double-layer
passivations are perfectly stable at excellently low
differential Sefid values of 23 cm/s. This explains the
observed perfect UV stability of the front
ation scheme of our MIS-IL silicon solar
cells. Furthermore, if applied to the rear surface, this
passivation scheme has the potential to achieve stable
rear efficiencies of about 20%.

-3

Using PCD samples as a much more sensitive test

structure for the investigation of the stability of the
passivation quality, a small degrada
detectable at silicon nitride pass
corresponding to the non-metallized rear surface regions
of bifacial cells. A detailed study on the dependence of
the degradation on the silicon nitride film thickness and
on the wavelength of the UV photons is presented. We
observe no measurable degradation for wavelengths
greater than 400 nm. The detected slight i
wavelengths below 400 nm is too small t
impact on the long-term stability of encapsul
rear efficient bifacial cells. Perfectly stable differentia
S e f f d values of 23 cm/s on 1.5-Qcm p-silicon are obtained
with SiO,/Cs/SiN,:H
double-layer pas
presently used on the front surface of ou
This passivation scheme is also well suite
surface of bifacial solar cells, giving the pot
rear efficiencies of bifacial silicon solar cells of even 20%.
AC

30

The authors would like to thank all

photovoltaic department at ISFH fo
this work. The support by the State
the German Bundesministerium
schaft, Forschung und Technologie IS gra
ledged. The ISFH is a member of the German
Forschungsverbund Sonnenenergie.

y$

20

0)

10

50

100

No. of photons with h <: 400 nm at interface [I 0'

150
cm -*I

Fig 7 Measured differential SRV Seffdof p-Si/SiNx:H monolayer

films and Si/SiOx/Cs/SiNx.H double-layer passivations versus
the number of UV photons reaching the siliconlinsulator
interface. UV exposure is performed for up to 120 hours using
320-nm and 340-nm UV cut-off filters, respectively

For the monolayer passivation an additional U V test

using a 340-nm cut-off filter is performed. The 120 hours
of UV exposure on both surfaces correspond to about
2 years of outdoor operation of rear module surfaces. As
can be seen, after this time, the SiN,:H monolayer
passivation still provides excellently low differential Sefid
values of 20 cm/s. A linear extrapolation as a worst-case
scenario (since the measured , ,S
curve shows a
negative curvature) gives a differential Sef, of about
80 cm/s after 10 years of outdoor operation. This implies
that SiN,:H
monolayer passivations provide stable
17-18% rear efficiencies of bifacial silicon solar cells.
CONCLUSIONS
Accelerated UV testing on low-temperature remote
PECVD silicon nitride passivated silicon solar cells is
reported. The front surface passivation of p-n junction and
MIS-IL silicon solar cells and the rear surface passivation
of bifacial silicon solar cells are investigated. We observe
perfect stability during a 600-hour UV test, corresponding
to about 5.5 years of outdoor operation of the cells.
Hence, we expect our remote PECVD silicon nitride
passivation to be stable during the anticipated PV module
lifetime of about 20 years.

420

[I] R. Hezel, K. Jaeger, J. Electrochem. Soc. 136,277

[2] Z. Chen, P. Sana, J. Salami, A. Rohatgi, /E Trans.
Electron Dev. 40, 1161 (1 993)
[3] J. Szlufzik et al., 72th European PVSEC, Amsterdam 1994,
p 1018
[4] A.G. Aberle, T. Lauinger, J. Schmidt, R. Hezel, Appl. Pbys.
Left 66,2828 (1995)
[5] C. Leguijt et al., 72th European PVSE
p. 720
[6] A.G. Aberle, S. Glunz, W. Warta, J. Appl. Phys. 71, 4422
(1992)
[7] P.E. Gruenbaum, R.A.
, R.M. Swanson, 20th lEEE
P VSC, Las Vegas 1988, p. 423
[8] T. Lauinger, J. Schmidt,
berle, R Hezel, Appl. Phys.
Lett. 68,1232 (1996)
[9] G. Lucovsky and D.V. Tsu,J. Vac. Sci. Techno/.. A 5 (4),
2231 (1987)
[ I O ] J. Schmidt, T.Lauinger, A.G. Aberle, R. Hezel, this conf
[Ill R. Hezel, W. Hoffmann, K. Jaeger, 70th European PVSEC,
Lisbon 1991, p. 51 1
[I21 M. Grauvogl, A.G. Aberle, R. Hezel,
[I31 A. Hubner, C Hampe, A.G. Aberle,
[I41 T Lauinger, J Moschner, A G Aberle, R. Hezel, to be
published
[I51 A.G. Aberle and R. Hezel, this c
[I61 Z. Chen, A. Rohatgi, D. Ruby, 7st WC PVEC, Hawaii 1994,
p. 1331
[I71 D.S. Ruby, W.L. Wilbanks, C B. Fleddermann, 1st WC
P VEC, Hawaii, 1994, p. 1335
[I81 R. Brendel, Appl Phys. A 60, 523 (1995)
[I91 A G. Aberle, J. Schmidt, R. Brendel, J. Appl. Pbys. 79, 1491
(1996)

Efficiency evaluation of a-Si and c-Si solar cells for outdoor use
M. Kameda, S. Sakai, M. Isomura, K. Sayama, Y. Hishikawa, S. Matsumi, H. Haku,
K. Wakisaka, M. Tanaka, S. Kiyama, S. Tsuda and S. Nakano
New Materials Research Center, Sanyo Electric Co., Ltd.
1-1, Dainichi Higashimachi, Moriguchi, Osaka 570,Japan

ABSTRACT
We have achieved the world's highest
stabilized efficiency of 8.9% for an a-Si single-junction
solar cell (lcm2) and 10.6% for an a-Si/a-SiGe tandem
solar cell (I cm2).
To apply these results to practical outdoor
power use, the annual output power of a-Si singlejunction (a-Si) and tandem (a-Si/a-Si) solar cells and a
crystalline silicon (c-Si) solar cell was calculated based
on annual meteorological data and the solar spectrum in
Tokyo, Japan, including the effect of the output power
dependence on temperature, incident irradiance and
solar spectrum. As a result, this simulation revealed that
the annual change of efficiency for the c-Si solar cell is
most affected by the solar spectrum among the three
types of solar cells, and the annual fluctuations of the aSi and a-Si/a-Sisolar cells are mostly caused by recovery
by the annealing effect.
INTRODUCTION
We have been investigating improvements in

the conversion efficiency and reliability of a-Si solar cells.

As a result, we have achieved a stabilized efficiency of
8.9% for an a-Si single-junction solar cell (lcm2) and
10.6% for an a-Si/a-SiGe tandem solar cell (lcm2) shown
in Fig. 1[1]. Both values are the highest ever reported. To
apply these results to practical outdoor power use, the
cell design should be optimized not under the standard
conditions (AM15,100mW/cm2i25C) but under practical
outdoor conditions. In this study, the annual output power
of an a-Si, an a-Si/a-Si solar cell and a c-Si solar cell
was calculated based on annual meteorological data and
the solar spectrum, including the effect of the output
power dependence on temperature, incident irradiance
and solar spectrum, with the cells installed in Tokyo,
Japan.
TEMPERATURE DEPENDENCE
Figure 2 shows the measurement results of
the dependence on temperature for each solar cell. The
temperature coefficients obtained from Fig. 2 are shown
in Table 1. The temperature coefficient of the c-Si solar
cell is about 2.4 times and 1.5 times higher than that of

light
v-r7

5sun(AM-l.5), 25"C,
160min;open circuit

360min, open circuit

AM-1.5. 100 mW/cm2

------------

Stabilized,t'
Cell area : 1cm2

0.5

1.o

0.5

1.5

Voltage (V)

Voltage (V)

Fig. 1 Illuminated I-V characteristics of a-Si and a-Si/a-SiGe tandem solar cells.

1049
0-7803-3166-4/96/$5.000 1996 IEEE

25th PVSC; May 13-17, 1996; Washington, D.C.

20

0 Single a-Si
A Tandem a-Si (Si/Si)

10

20

30

40

50

60

Temperature('C )
Fig. 2 Temperature dependence of a-Si single-junction
and c-Si
solar cell (O),a-Si/a-Si tandem solar cell (A)
solar cell (U).
Table 1 Temperature coefficients.
~~

Single a-Si

Tandem a-Si(Si/Si)

c-Si

the a-Si and a-Si/a-Si solar cell, respectively. The

temperature of solar cells outdoors is often over 50C in
summer. So, the efficiency of c-Si solar celk on summer
days is more than 10% lower than that at the temperature
of 25C. The difference between the a-Si cell and a-Si/aSi cell comes from the current mismatching at the
optimum voltage in the a-Si/a-Si solar cell. In this study,
the output power dependence on cell temperature is
simulated by using the coefficients as shown in Table 1,
and then also the effects of irradiance and wind velocity
on the cell temperature were taken into account[2].
I

DEP

CE

It was found that the illumination-intensity

dependence of the current-voltage characteristics I(V) is
approximated by the following newly-developed and
simple equation proposed by the authors[2].

"

-0.2

0.2
0.4
0.6 0.8
1
Voltage (V)
Fig.3 Dependence of I (V) on the illumination intensity.
Symbols indicate experimental data.

characteristics is well approximated by Eq. (1). In this

study, it is assumed for simplicity of calculation that the
output power of each solar cell is approximately in
proportion to the illumination intensity based on
experimental data of dependence on illumination
intensity.

SP

N@

In general, the change

a clear days' solar
spectrum corresponds to the Air
ss value at that time,
which can be calculated by a function of the solar
altitude. The relationship between AM values and the Isc
of each solar cell, which is calculated using the measured
solar spectrum data for several clear days for each
season in Tokyoj31 and the collection efficiency of each
solar cell are shown in Fig. 4. The solid lines in Fig. 4 are
fitting curves. The Isc of each solar cell is well
approximated as a function of the air mass value. So, in
this simulation, as a new method, it is assumed that the
solar spectrum dependence of the output power of each
solar cell is approximately calculated by the solar

where,
lE2(V): current-voltage characteristics at E, mW/cm2,
I&/):
current-voltage characteristics at OmW/cm2
(dark),
lloo(V): current-voltage characteristics at 100 mW/cm2,
V:
applied voltage.

6 25

20

2 15
10

The experimental I(V) for the a-Si single-junction solar

cell is shown by the symbol in Fig. 3. The result of the
calculation using Eq. (1) is also shown by the solid line in
Fig. 3. The good agreement of the experiment and the
calculation in Fig. 3 indicates that the illuminationintensity dependence of the current-voltage

0 0.5 1

1.5 2 2.5 3 3.5 4 4.5

Air Mass value
Fig. 4 Relationship between AM and I, which is
and outdoor measured
calculated by measured 7;,
solar spectrum (9:OO-15:OO : clear days of four seasons).
k

1050

spectrum dependent Is, expressed by each function of

the air mass value as shown in Fig. 4.

Sreal(A ,t) : spectrum measured outdoors

AM(t) : Air mass value
6 : correction factor of irradiance

CALCULATION
RESULTS
In this simulation, the following equations are
used to calculate the output power of each solar cell,
including the output power dependence on temperature,
irradiance and solar spectrum as mentioned above.
Annual total output power Ptotal is calculated as a sum of
instant output power P(t) through a year by Eq. (2),and
instant output power P(t) can be expressed by a product
of Pstd and a coefficient of temperature CTemp(t),a
coefficient of irradiation Clr(t), and a coefficient of
spectrum CSpec(t)by Eq. (3),and here Pstd is the output
power measured under the standard conditions of AM1.5,
100mW/cm2, 25C. CTemp(t),Clr(t), and CSpec(t)are
calculated by normalizing each parameter dependent
output power PTemp(t),Plr(t), and PSpec(t)by the standard
output power, respectively, by Eqs. (4)-(6). The efficiency
of the total output power is expressed by a ratio of each
total output power to each total irradiance by Eq. (7).
Each parameter dependent output power is calculated
by equations (8)-(IO).

where

Ptotal:annual output power

P(t) : instant output power

P(t) = P,,d.XC,e,p(t)XClr(t)XCSpec(t)

(3)
(4)

(7)

The calculated normalized monthly average total

output efficiency of each solar cell is shown in Fig. 5.The
annual behaviors of the efficiency of the a-Si and a-SilaSi solar cells are more stable than that of the c-Si solar
cell. These results indicate that the annual behaviors of
the efficiency of a-Si and a-Si/a-Si cells are less affected
by both temperature and spectrum through the year than
that of the c-Si solar cell.
The results of this simulation are compared with
the annual behaviors of actually outdoor measured data
for the three types of solar cells. The case of c-Si is
shown in Fig. 6. The annual behavior of the efficiency of
c-Si is well represented by this simulation. This indicates
that this simulation technique can reasonably express the
annual behavior of output power of the c-Si solar cells.
The case of the a-Si single-junction solar cell is
shown in Fig. 7. The actually measured efficiency of a-Si
single-junction solar cell increases in summer time as is
well known. On the other hand, the result of the
simulation is almost stable through the year. Therefore,
the increase of outdoor efficiency in summer is supposed
to be mostly caused by the recovery by the annealing
effect, rather than by spectrum dependence nor by
temperature dependence.
The case of the a-Si/a-Si tandem solar cell is
shown in Fig. 8.A slight increase of measured efficiency
in summer time appears here also, similar to the case of
the a-Si single solar cell, although the annual behavior of
the measured efficiency is close to the simulation. So the
a-Si/a-Si tandem solar cell is less affected by both
spectrum and recovery by the annealing effect than the cSi and a-Si solar cells respectively.
The difference between the amorphous silicon

where

Irtotal:total irradiance calculated

using meteorological data
Ptemp(t) = Pstdx {I P/IOO(T(t)-25)}
(8)
(Tcell(t) = Tatm+ AT(lr, Wind)).

1.20

x c-Si

a-Si
A a-Si/a-Si

(9)

0.8%
'1

where
(x : temperature coefficient (measured)
S,,,( A ,t) : spectrum at irradiance 100mW/cm2

'

'

'

'

6 ' 7
Month

'

'

9 '10 '11 '12

Fig. 5 Normalized monthly average efficiency of solar

cells by simulation.

1051

solar cells and the c-Si cell comes mainly from material
properties, such as bandgap, and the difference between
the a-Si and a-Si/a-Si solar cells comes mainly from cell
design. The a-Si/a-Si solar cell is less affected by the
annealing effect than the a-Si solar cell, and is less
affected by spectrum dependence than c-Si, but it is
inversely affected through a year compared with c-Si
because of an imbalance in photocurrent between the top
1.50

c-Si
0

6
C

a,
.0

Simulated
Measured (outdoortest)

U 1.00 -

.m

1.5

a-Si single
0

0
%

a,
.0

Simulated
Measured (outdoor test)

U 1.0

.m

and bottom cells from the optimized value under the

standard conditions[5].

(1)
A technique for simulating the annual output
power of solar cells has been developed using only the
indoor data of solar cells, including output power
dependence on temperature, irradiance and spectrum,
using meteorological data and the solar spectrum in
Tokyo, Japan.
(2)
The annual behavior of the actual output
efficiency of the c-Si solar cell has been well expressed
using this simulation technique, and It is confirmed that
this simulation technique can reasonably express the
annual behavior of the output power of solar cells
including only the dependence on temperature, irradiance
and spectrum.
(3)
As a result, it has turned out that the annual
behaviors of the efficiency of a-Si and a-Si/a-Si solar cells
are more stable against both temperature and spectrum
than that of c-Si by successfully including the spectrum
dependence in this simulation, and it is confirmed that the
annual fluctuation of the efficiency of a-Si and a-Si/a-Si
solar cells is mostly caused by recovery by the annealing
effect.
(4)
This simulation can be used to estimate the
annealing effect, which is a main factor of the annual
fluctuation of a-Si solar cells, and can also be used to
optimize cell design, including the practical temperature,
irradiance, and, especially, solar spectra as mentioned
above.
This work is supported by NEDQ as a part of the
New Sunshine Program under the Ministry of
InternationalTrade and Industry.

0.5

9 1 0 1 1 1 2

Month
Fig. 7 Normalized monthly average efficiency of a-Si
solar cell by simulation and outdoor measurement.
1.5c

a-Si/a-Sitandem
0

ox
C

a,
.0

Simulated
Measured (outdoor test)

p 1.oc
N
._
-

0.50
C

9 1 0 1 1 1 2

Month
Fig. 8 Normalized monthly average efficiency of a-Si/a-Si
solar cell by simulation and outdoor measurement.

REFERENCES
[ l ] Y. Hishikawa et al., "Approaches for stable multijunction a-Si solar cells", First WCPEC, 1994, pp. 386393.
[2] M. Kobayashi and T. Tani, "The characteristics of PV
modules output considered the environmental factors", J.
Japan. Solar Energy Soc., 19, 1993, pp. 39-46.
[3] Y. Hishikawa and S. Qkamoto, "Dependence of the IV characteristics of amorphous silicon solar cells on
illumination intensity and temperature", Solar Energy
Mate. and Solar Cells, 33, 1994, pp. 157-168.
[4] M. Habu et al., "Measurement of solar spectral
irradiance at Tanashi, Tokyo (Ill)", Res. Nectrotech. Lab.,
830, 1983, pp. 35-175.
[5] D. L. Morel, R. D. Wieting and K. W. Mitchell, "Design
and performance of thin film si1icon:hydrogen tandem
modules", Technical digest of Ist PVSEC, 1984, pp. 567570.

1052

IMPROVED STABILITY IN ECR-DEPOSITED A-SI SOLAR CELLS

Vikram Dalal, S.Kaushal, R. Girvan, S . Hariasra and L. Sipahi
Iowa State University, Microelectronics Research Center, Ames, Iowa 50011

ABSTRACT
We report on fabrication, properties and stability of aSi:H and a-(Si,Ge):H solar cells made using remote low
pressure ECR deposition. We have fabricated both
substrate and superstrate type solar cells. We can make
solar cells with high fill factors in both geometries, but
the voltages are higher with substrate-type solar cells than
with superstrate type cells. Special problems related to
diffusion of B have to be solved in superstrate cells
because the deposition is done at higher temperatures(
350-375 C). Several novel player grading schemes and
buffer layers which allow us to fabricate these types of
cells are described. The substrate cells were made with
both H-ECR and He-ECR discharges We fmd that while
the cells prepared with He discharge have lower H
concentration, and lower H content, they are less stable
than cells prepared using H, discharges. The stability of
cells was measured using ELH and xenon lamps, and
compared with the stability of cells made using standard
glow discharge techniques. We find that the cells
prepared using H,-ECR discharges are more stable than
standard glow discharge cells with comparable fill
factors, voltages and thicknesses of i layers. We also
report on a new type of graded gap a-(Si,Ge):H cell,
which appears to show improved stability.

Introduction
Improving the stability of a-Si:H solar cells remains a
problem of interest and fervent research. Several new
deposition techniques, among them remote ECR
growth[ 1,2],hot wire growth[3] and W-annealing[4] have
been used to improve the stability of a-Si:H solar cells.
While reports on film data are plentiful, there is little
work on making and studying a-Si;H solar cells made
using these new techniques. In this papr,we describe the
fabrication of a-Si:H solar cells on both stainless and tin
oxide substrates using remote, low pressure ECR growth
techniques. We show that good cells can be made in both
geometries. We also report on the special problems one
faces when making these cells, and how to solve them.

We also report on the fabrication of a new type of graded

gap a-(Si,Ge):H solar cell, also using the ECR technique,
and discuss its stability.

Growth Technique
The ECR( Electron-Cyclotron-Resonance) plasma-CVD
growth technique we use has been described
previously.[ 11. To recapitulate, it uses microwaves and
magnetic fields to create a highly ionized beam of either
hydrogen or helium atoms, which stream towards the
substrate. The streaming beam of hydrogen reacts with
silane to create primarily SiH, species, which leads to the
growth of the material. During growth, the material is
subjected to controlled ion bombardment and etching by
the H ions and radicals, thus achieving the condition of
etchingduring-growth.Thisetchingduring-growthshould
lead to better localized microstructure and hence, better
stability.
The previous work on films showed that indeed, the film
properties were comparable to the best glow-discharge aSi:H, and that the degradation in the material properties
upon light exposure was significantly less than
comparable glowdischarge films[11. In particular,we
found that the increase in defect density upon light
soaking was only about a factor of 2.5-3, and that the
addition of minute levels of B compensationreduced both
the initial and final defect state densities.

Device Fabrication
We have fabricated two types of p-i-n devices: superstrate
devices on tin oxide substrates, and substrate devices on
bright-polished stainless steel. Note that no special
reflectors were provided in either case. The stainless
substrate, in particular, was not electro-polished, nor
provided with a textures Ag reflector or any other
reflection enhancement means.
Asuperstrate devices. These were made on low haze
Asahi tin oxide, and were of standard p-i-n types, with a

1069
0-7803-3166-4/96/$5.00 0 1996 IEEE

25th PVSC; May 13-17, 1996; Washington, D.C.

buffer layer between p and i. Since the device deposition

takes place at high temperatures, the reduction of tin
oxide by the H plasma, and the diffusion of B from the
e i layer, were special problems that had to be
solved. To solve these problem, we used a special
graded p layer, and a buffer layer with high C content.
The special feature is a simultaneous grading in both B
towards the end of the p layer so as to
e p layer with the subsequent i layer,and
to reduce the diffusion of B[2]. Using such grading
schemes, we can achieve cells with decent voltages and
fill factors. The cell data is shown in Fig. 1. While the
fill factor is good, the voltage is somewhat low, only
about 0.81 V. We are still working on improving this
voltage.

B. For fabricating substrate-type cells, we used the

structure : ss/n/i/p/semi transparent Cr. The Cr was
about 20-30 % transmitting, and had a central Al bussbar
to collect the current while putting a probe on.
s arrangement because it gave very
reproducible contacts, as opposed to no. We used both
Hydrogen and Helium discharges for producing the
devices, and we found systematic differences in device
properties when using Hydrogen vs. Helium.

0.17

On problem we faced was contamination from the n layer

deposition. To reduce such contamination, we used two
n layers, a thick (1000-2
A thick) n layer deposited
in a glowdischarge reactor, and a thinner ( 100-200 A)
n layer in the ECR reactor. Then, we followed with a
very thorough purge and plasm clean procedure,
followed by a H-plasma etch of the substrate. Then, the
i layer was deposited, followed by a buffer layer, and
finally the p layer. The i layer growth rates were
1-2
AIsec rate,dilution ratios [(H,/He)/SiHJ were 20: 1,
tures
350-375 C,and pressures 10 mT.

Fip.2 I(V) curve for

semi-transparent Cr

strate cell wi

1.0
0.8

evice 1:

The I(V) curve for a solar cell produced using Hdischarge is shown in Fig. 2. The corresponding QE
curve is shown in Fig.3, and the QE ratio curve ( Ratio
of QE at OV/QE at + O S V) in Fig. 4. From these
figures, it is apparent that one can produce good cells
using H-ECR. The voltage is 0.786 V, which translates
tion, significantlybetter
metries. The fill factor
d the QE ratio shows no collection
a similar result when
stead of Hydrogen.
efficiencies of the two
and other using He, we

0.

f
1070

found a significant difference in baradgaps. The one

produced using He had a lower gap by about 0.035 eV,
consistent with a lower H content in He-produced
material (about 4% vs. 7%).

1.8

Stability of ceUs

1.6$

A) Superstrate cells:
To measure the stability of superstrate cells, we
compared two cells, one produced using glow discharge,
and one using ECR. Both had the same thickness of i
layer(within lo%),and very similar voltages(0.78V) and
fill factors(0.67) initially. The QE curves also were very
similar. Then, we proceeded to light soak these cells
using 2 x sun xenon illumination from an Oriel simulator.
In Fig.5, we show the results of the degradation of these
cells as a function of time. It is clear from Fig. 5 that the
ECR-produced cell is more stable than its glow
discharge-produced cousin. We measured the QE and
QE ratio for the two cells before and after degradation.
These results are shown in Fig. 6, and we see that the
QE ratio after degradation for the glowdischarge cell is
much higher in the red region of the spectrum than for
the ECR cel1,implying a worse problem with hole
collection, i.e. greater i layer degradation for the glowdischarge cell. Thus, we have shown that the ECR cell is
more stable because the i layer itself is more stable.

.A

1*2f+.t.
1.0
.;0
I
4 0 0 4 5 0 5 0 0 5 5 0 600 650 7 0 0 7 5 0 8 0

Wavelength (nm)

Fig.4 (QE at OV/QE at +0.4 V) for cell of

Fig.2

ECR

B) Substrate cells
We again selected three cells which were as near identical
as we could make them, one deposited using H-ECR, one
with He-ECR, and one, glow discharge. All three had
similar ss substrates, top Cr contacts, and i layer
thicknesses, and initial fill factors in the range of 0.67.
We subjected the three cells to light soaking from an
ELH lamp which was focussed down to give incident
intensities of about 6x sun, which is about 1.9 sun into
the cel!. The cells were maintained at temperatures of 20
C using a cooled mounting block.

Fig, 5 Relative degradation in fill factor of

comparable ECR vs. Glow Discharge cells

1.5

The results for the degradation of the fill factor for the
three cells, which is the primary indicator of material
quality degradation, are shown in Fig.7. From this
figure, we see that the H-ECR cell degrads much less
than either the glow discharge or He-ECR cells, whose
relative degradations are very similar. This is a striking
result. What it says is that H content by itself is not an
indicator of material stability. Even though the
He-ECR film had less bonded H, it degraded just as
badly as the glowdischarge device. We therefore,
conclude that only by changing the localized
microstructure, and therefore H bonding, can we
change stability. H-etching during growth does change

'

1-t

503

803

6co

Wavelength (nm)

Fig. 6 QE ratio after degradation for GD and

ECR cells. The increase in QE ratio for GD
cell compared to ECR clearly indicates more
problems with hole collection,i.e. greater
degradation in i layer properties
1071

microstructure, and presumably makes

robustagainst degradation.

it

more

1.0

%
Graded gap a-(Si,Ge):H Cells

s
II
Y

Previously, we had suggested that the ideal graded gap

structure was to use a grading in the middle of the i
layer, because that is where the field collapse is the
greatest[5]. We had shown that such a grading improved
the stability of a-Si solar cells[5]. We have now
p r d d to make such a graded gap a-(Si,Ge):H cell,
again using H-ECR growth techniques. The substrate-type
device result is in Fig. 8, and the preliminary stabilitydata
in Fig.9.
For this experiment, we kept the Ge
concentration low, in the 1.6-1.62 eV bandgap range.
The device fill factors are excellent, and so is the
stability, indications that our grading scheme works.
More experiments to reduce the gap and study stability
are in progress.

e --.
\ -

H-ECR

.- _- -__
H+ECR

GD

2a

0.5

+!

25

50

75

Time ( Hours )

Fi2.7 Degradation in fill factor for H-ECR,

He-ECR and GD cells. All the cells had
comparrable initial fill factors, arad
comparable i layer

In conclusion, we have been able to make good solar

cells in both superstrate and substrate geometries. We
find that it is easier to get higher voltages in substrate
geometries than in superstrate geometries. We also find
that the stability of both types of cells, when deposited
using the H-ECR technique, is better than the stability of
comparable quality glow discharge cells. A striking result
is that H-ECR growth producs much more stable cells,
even though He-ECR material has a lower H content than
H-ECR material. We speculate that this effect is a
consequence of the beneficial effects on microstructure of
etchingduring-growth produced by the energetic H ions
and radicals.
This work was partially supported by a contract from
NREL.

Fig.8 1 0 curve of a-(Si,Ge):H cell with

grading in the middle of the i layer

1.0

References

Degradation of 1.6eV a-(Si,Ge):H cell

[Initial FF= 0.681
A

0.9

5
1. V. L. Dalal,E. Ping, S. Kaushal, M. K. Bhan and M.
Leonard, Appl. Phys. Lett. 64,1862(1994)
2. V. L. Dalal et al. J.NonCryst.Sol(1996)To be publ.
3. A. H. Mahan et al, J. Appl. Phys. 69,6728(1991)
4. H. Shirai, J. Hanna and I. Shimizu, Proc. AIP,

268,388( 199 1)

0.5

5. V. Dala1,G. Baldwin and P. Garikepati, Proc. of

23rd.IEEE PV Conf. ,p.816(1993)

25

50

75

Fig. 9 Degradation in f
a factor of H-ECR a(Si,Ge):H cell with grading in the middle.

1072

IMPROVED RADIATION HARDNESS OF SILICON SOLAR CELLS

HidetoshiWashio, Yoshifumi Tonomura, Minoru Kaneiwa & Tatsuo Saga
SHARP Corporation, 282-1 Hajikami, Shinjo-cho, Kitakatsuragi-gun,Nara 639-2198,Japan
Osamu Anzawa & Sumio Matsuda
National Space Development Agency of Japan, 2-1-1 Sengen, Tsukuba, lbaraki 305-8505, Japan

ABSTRACTS

N-Contact(front) Si02Layer
NfLayer
\
1

SHARP and NASDA (National Space Development

Agency of Japan) have been engaged in the development
of silicon space solar cells since 1970s. We have started
the project to improve the radiation hardness of silicon
solar cells in 1998[1]. This project gave fruitful results in
BJ (Both-side Junction) and AHES (Advanced High
Efficiency Silicon) structure. The design and
manufacturing process for the BJ cells and the AHES-1
cells were finalized and their characteristicswere qualified.
The experimental results of these solar cells performance
in comparison with conventional HES (High Efficiency
Silicon) cells. The BJ, the AHES-1 and AHES-2 cells
showed 13.5%, 13.1% and 13.4% efficiency, respectively,
at AMO, 28OC after the irradiation of 1MeV electrons of 1 X
10i5e/m2. The EOL efficiency of 13.5% of BJ cell is the
highest ever reported for space silicon solar cells. This
paper presents the main features of the BJ, the AHES-1
and AHES-2 design and qualification status.

ARC
/

P-Substrate

Si02tayer P-Contact P Layer N+Layer ~ - ~ o n t $(rear)

ct
Fig. 1. Schematic Cross-section of BJ cell and its
Equivalent Electrical Circuit

DEVELOPMENT OF SOLAR CELLS

Both-side Junction Solar Cell Development

INTRODUCTION
In 1995, some kinds of high efficiency silicon (HES)
solar cells designated as NRS/LBSF (Non-Reflective
Surface / Local Back Surface Field) cells, NRS/BSF (NonReflective Surface / Back Surface Field) cells and
NRS/BSF cells with IBF (Integrated Bypass Function)
have been developed and commercialized[l][2][3]. These
HES cells were categorized as the first generation of high
efficiency silicon solar cells and had 12.2 % to 12.5%
efficiencies at EOL condition (1MeV electron fluence of
lX105 e/cm2). Totally about 1.5 million HES cells have
been manufactured and supplied not only for NASDA
satellites but also for extensive satellites in the world
(totally over 50 satellites).
In 1998, the development project of the second
generation of high efficiency silicon solar cells designated
as BJ and AHES cells were started to meet the customer
requests for higher EOL efficiencies, This project was
executed under the cooperative development between
NASDA and SHARP in the same manner as the first
generation of HES cells. Preliminary data of BJ and
AHES-1 cells were presented in the previous papers[2][4].
This paper summarizes the main features and
qualification status of BJ, AHES-1 and AHES-2 cells
generated in this project.

0-7803-5772-8/00/$10.00 0 2000 IEEE

Figure 1 shows the schematic cross-section of BJ

structure. As p-n junctions are located at both front and
rear surfaces of cells, photo-generated minority carriers
have more opportunities to reach the p-n junction and to
contribute to the electrical output after electron irradiation.
This is equivalent to increasing the diffusion length of
minority carrier and to reducing the effect of decreased
carrier lifetime caused by radiation damage. As shown in
Figure 1, the front surface of the BJ cell is quite similar to
the AHES-1 cells[l]. The rear surface structure of BJ cell
is rather different.
The design of rear structure is shown in Figure 2. The
BJ cell also has a p (back surface field) layer, but it is
localized to the finger-grid shape. The n+ layer is formed
between p grids. The rear n+ layer occupation rate is
about 75%. The carriers generated by long wavelength
light absorbed at deep positions from the front surface are
collected effectively to the rear p-n junction. So even if the
lifetime is reduced by irradiation, carriers can reach to the
p-n junction and contribute to the electrical output.
Narrower back contact gridlines with tapered collecting
bars give better result at EOL. The rear contact is
prepared by lift-off technique onto the n+and p+layers and
the grid width is kept in minimum to transmit the infrared
light from the front surface to the outside of the rear
surface in order to reduce solar absorptance.

1114

The BJ cells have an additional advantage of lower

solar absorptance, which will lower the operating
temperature of cells in case of being used with flexible
transparent solar array substrate. Figure 4 shows the
reflectance and transmittance data of BJ cells. These
data were taken for BJ cells with kapton back films.
Table 2 shows the performance of the BJ cells
comparing with HES, AHES-1 and AHES-2 cells. The BJ
cell assembled on kapton film showed lower solar
absorptance and the calculated operating temperature
was reduced. Taking the weight of solar cells into account,
the BJ cells aive 159% hinher EOL efficiencv
~-weight ratio
than HES c e h in case of flexible array.

N-Contact
c'

N+ Layer
Contact Win dow

P-Contact

P+ Layer

.&
(Rear View a t Left-Top Corner7

0.7

$
QI

o.6
0.5

E
0.4
U)

(A-A' Cross-sectional View)

-E

[r 0.3

ti, 0.2

Fig. 2. Rear Structure of BJ Cell

*a

We compared two types of BJ cells with the 10S2cm

substrate made by FZ (floating zone) method and the
lOS2cm substrate made by CZ (Czochralski) method,
respectivery. From our experiments, the BJ cells made by
FZ substrate showed higher VOCand Pmax at EOL. But
Voc was decreased by exposing of AM0 light for 10 hours
after irradiation. As a result, we selected the 10Qcm
substrate materials made by CZ method.
Figure 3 shows the spectral response of the BJ cells
with lOQcm, CZ, 70pm thick substrate[4]. Here the
response is separated into the front and the rear p-n
junction contribution. The spectral response of BJ cell is
slightly degraded by electron irradiation and the front and
the rear p-n junction contribute almost equally to this
degradation in relative values.
Table 1 summarizes the characteristics of the BJ cells
comparing with HES, AHES-1 and AHES-2 cells. Although
BJ cells showed lower BOL performance due to thinner
substrates, the EOL performancewas superior than those
of other cells. The BJ cells showed higher radiation
hardness, and its remaining factors more than 98% and
80% for the Isc and Pmax, respectively.The BJ cells have
achieved 13.5% EOL efficiency, which is the highest
efficiency ever fabricated as mass-production space
silicon solar cells.

0.1
0

300

400

500

600

700

800

1000

1100

1200

Fig. 3. Spectral Response of BJ Cell and Separated

Front and Rear PN Junction Contribution before and
after 1MeV Electron irradiation up to 1 X I Oi5 e/cm2
70

-8

E
5

60

a,

Reflectance & Transmittance

50

.E
40

'm

g
d

30

20

5:

10
n
" l

-~

300 500 700 900 1100 1300 1500 1700 1900 2100 2300 2500
Wavelength(nm)

Fig. 4. Reflectance and Transmittance of BJ Cell

with AR coverglass

Table 1. Electrical Output Characteristicsof W, AHES-1 and AHES-2 Cells

Comparingvith HES Solar Cells a
t BOL and EOL(1MeV electron 1xl Oi5e/cm2)

1115

900

Wavelength (nm)

Its front surface has inverted pyramids and a shallow n+

diffused layer (530.1
pm) passivated by thermally grown
SiO, layer. The Ti0,-AI,O,
dual layer anti-reflective coating
is applied to the front surface. The entire rear surface is
diffused with boron (back surface field) and passivated
with SiO, layer in order to reduce minority carrier
recombination loss on the back silicon surface. This SiO,
layer also increases the effect of AI back surface
reflector(BSR).
The thickness of silicon was reduced to 75pm from
1OOpm of HES cell, and the depth of the surface pyramids
was increased from 14pm to 28pm.
As shown in Figure 7, by magnifying the surface
pyramid size, pyramid depth becomes deeper, and the
average thickness reduction is 2/3 of the pyramid depth.
As a result, the average cell thickness is reduced from
93pm to 66pm. By reducing substrate thickness, the
radiation tolerance of cells is improved due to the
increased photo-inducedcarrier density closer to the front
p-n junction. Although the initial short-circuit current (Isc)
and maximum power (Pmax) are lower than those of HES
cells. Those parameters of AHES-1 cells become higher
at EOL condition (1 x loi5 e/cm2 1MeV electrons
irradiated). The AHES-1 cell gives 71 .ImW Pmax at EOL,
which is equivalent to 13.1% efficiency.
Figure 8 shows the spectral response of HES cell and

As the solar cell size for today's satellites are 4cm x

6cm or larger, we selected the maximum size (3.6cm x
6.9cm) as a target of development. As for enlarging, we
tried to optimize the design of front grid patterns by
numerical analysis. And rear contact and diffusion layer
designs of enlarged cells were taken from 2cm x 2cm cells.
These enlarged cells showed the similar electric power
proportional to the cell areas.
The BJ cell has N-contacts at both front and rear
surface. Considering its unique structure, the
interconnector has newly designed to branched contact
pads. And the interconnector with out-of-plane stressrelief was developed to be suitable for BJ cells because
our standard interconnector with inplane stress-relief
needs more P-contact areas and causes decrease of
performance. And the assembly technology was also
established. Figure 5 shows the newly developed
interconnector.
Interconnector welded a t rear surface P-contact pad
/

\
/

7
/
/.

'

,'

,,

/
'

BJ solar cell-

Interconnector welded a t rear surface N-contact pad

lnterconncctor welded a t front surface N-contact pad

P-Substrate

P' Layer
Fig. 5. Newly developed interconnector

-/ I,
Si02 Layer

P-Contact

Fig. 6. Schematic Cross-sectionof AHES-1 Cell

AHES Solar Cell Development

Figure 6 shows the schematic cross-section of AHES-

'
-

1 cells that are based on the HES (High Efficiency Silicon)

cells previously designated as the NRS/BSF cell[l]. The

10Qcm CZ 6-doped silicon wafer is used as a substrate.

Fig. 7. EnlargedTextures

Table 2. Performanceof AHES-1 and ELI Cells

Comparingwith HES Solar Cells at EOL(1MeV electron 1x1Oi5elcm2)

I
Type
HES (bare cell)
AHES-1 (bare cell)
BJ (with opaque substrate)
BJ (with transparent kapton substrate)

v(%) I
@28"C
12.5
13.1
13.5
13.5

I v(%) I EfficiencylClCWeighqEfficiency/CICWeigh4

I
as
0.86
0.84
0.84
0.80

Top("C)
51.8
48.9
48.5
43.8

1116

Ratio
Ratio
@Top Ratio
11.2
100
100
100
11.8
105.4
78.1
135
12.2
108.9
70.1
155
111.6
12.5
70.1
159
Measurement Condition : AM0 135.3mW/cm2,28C
Top:Operating Temperature Solar Cell Size:2x2cm2

AHES-1 cell before and after electron irradiation. The EOL

spectral response of AHES-1 cell shows higher values in
the range of longer wavelength than that of HES cell.
The AHES-2 cell has been designed to be more
radiation resistant than AHES-1 cell. The thickness of
silicon was reduced 60,um for AHES-2 cell from 75,um
for AHES-1 cell. Current performance data of AHES-2
cells are given in Table 1.
0.8

0.7

9
a 0.6
0.5

3 0.4

E2 0.3

'

& 0.2

0. I
0 '

300

400

500

600

700

800

900

1000

1100

I200

Wavelength (nm)

Fig. 8 . Spectral Response of HESUHES-I Cells

CONCLUSION
The BJ structure improvedthe EOL efficiency of silicon
cells moreover. These cells have p-n junctions at both
front and rear surface and showed less radiation
degradation and better remaining factor than previous
silicon cells. Comparing with HES, the BJ cells improved
EOL performanceat operating temperature by about 8.9%
and the cell weight was reduced by about 0.7 times. So
the power weight ratio of the BJ cells is greater than HES
cells by about 1.5 times.
Development of BJ cells were accomplished and
qualified as space solar cells in accordance with the
NASDAs specification. The detailed design and process
of the cells were settled for their mass production.
The radiation resistance of silicon space solar cells has
been improved by the thinning substrate and enlarging
surface texture size. The decreased substrate thickness
and enlarged textures give the effects of reduced
recombination centers and carrier generation closer to the
front p-n junction. The AHES-1 cells have been developed
and commercialized. Their EOL efficiency has reached
13.1%.
The AHES-2 cells also have been developed. Current
performance data of AHES-2 cells were presented.

NASDA QUALIFICATION TEST

REFERENCES

The QT (qualification test) of the BJ cell was

successfully performed according to the NASDA solar cell
specification. The NASDA QT consists of nine test groups
(humidity, electron irradiation , thermal shock and contact
pull test , non-operating life test , high temperature
vacuum , radiometric properties , interconnector fatigue ,
thermal shock and anti-reflection coating , anti-solvent
property). Table 4 shows an excerption of QT results of BJ
cells.
Both the performance evaluation tests to obtain the
applicable data necessary for the design of solar panel
and the marginal assessment tests to evaluate limits of
the environmental characteristics were performed almost
at the same time. The results of all the tests successfully
according to meet the requirements specified in NASDAs
specification,
The AHES-1 cells have been qualified for commercial
applications and commercialized in advance of BJ cells.
The AHES-I and AHES-2 cells will be qualified NASDA
solar cell specification in a short time.

[I] ASuzuki et al., "Progress and Future View of Silicon

Space Solar Cells in Japan", /E Transactions on
Electron Devices, 46, 1999, pp.

Test
Humidity
Thermal shock

[2] Y. Tonomura et al., "Development of High End-of-life

Efficiency Silicon Space Solar Cell", 2nd World
Conferenceon PVSEC, 1998, pp. 351 1-3514
[3] T.Hisamatsu et al., "Reverse bias characteristics of
modules made of solar cells with and without
integrated bypass function", 25th /EPVSC, 1996,
pp.251-254
[4] Y. Tonomura et al., "Development of Both-side
Junction Silicon Space Solar Cells", PVSEC- 11, 1999,
pp.179-180

Table 4. An excerption of QT results of BJ cells

I
Test condition
45"C,90%RH,720hrs
-186- t 140C. 5cycles and
rcycies
~.
~

-\-

1117

Chanae rate(%\

SHUNT-ANALYSIS OF EPITAXIAL SILICON THIN-FILM SOLAR CELLS BY

LOCK-IN THERMOGRAPHY
Sandra Bau, Dominik M. HuljiC. Jtirg Isenberg, Jochen Rentsch
Fraunhofer Institutefor Solar Energy Systems ISE, Heidenhofstr. 2, 79110 Freiburg, Germany

ABSTRACT
Lock-in thermography has been applied for shuntanalysis on epitaxial siiiwn thin-film solar cells. The solar
cell material was made by epitaxial deposition of the base
layer on highly doped monwrystaliine (Czochralski) and
multicrystalline silicon substrates in an APCVD-system.
Solar cells were prepared in a laboratory-type and an
industrial-type process.
Characterization of the solar cells by infrared Lock-in
thermography and microscopy revealed a clear correlation between shunts and epitaxial defects in case of the
lab-type solar cells. Furthermore an increased concenlration of shunts located under the emitter grid lines of the
industrial-type solar cells compared to the lab-type solar
cells was observed. The analysis by thermography thus
gave insight Into the quality of the epitaxial layers and into
problems concerning a transfer of the solar cell process
from laboratory to industrialscale manufacturing.
INTRODUCTION
Within the high-temperature route for crystalline silicon solar cells the epitaxial deposition of silicon layers
plays an important role in most solar cell concepts. Starting from a recrystallized silicon seeding layer or from a
low-grade silicon substrate the base layer Is usually deposited by epitaxy. The performance of the thin-film solar
cell strongly depends on the crystal quality of the deposited base layer. Research activities in the field of silicon
deposition for an application in silicon thin-film solar cells
therefore concentrate on the development of inexpensive
deposition methods which are capable of producing layers
of high crystal quality.
Another important topic conceming crystalline silicon
thin-film solar cells is the transfer of solar cell processing
from laboratory to industrial-scale manufacturing. Epitaxial
thin-film solar cells represent a wafer equivalent which
can be directly introduced into standard solar cell fabrication.
In this paper we report the manufacturingof epitaxial
thin-film solar cells by applying a laboratory-type and an
industrial-type solar cell process. The solar cells were
characterized by I-V measurements, microscopy and lockin thermography. Lock-in thermography has proved to be
a powerful tool for an analysis of leakage currents
(shunts) affecting the solar cell performance. However up
to now this technique has mainly been applied for the

0-7803-7471-1/021$17.0002002 IEEE

characterization of solar cells from silicon bulk material

111, PI.

SOLAR CELL STRUCTURE

The solar cell structure under investigation is the

epitaxial silicon thin-film solar cell. Starting f 5x5 cm
highly doped monocrystalline (Cz) and multicrystalline
(mc) silicon substrates the base layer with a thickness of
25 to 35 pm was deposited. The resistivity is 0.02 Rcm
and 0.3 Rcm for the substrate and the base. respectively.
The silicon depositions were performed at atmospheric pressure in an optically heated APCVD-reactor
(Atmospheric Pressure Chemical Vapor Deposition) constructed at Fraunhofer ISE. A mixture of trichlorosilane
and hydrogen was used as precursor gas. Boron doping
of the deposited silicon layers was achieved by adding
diborane diluted in hydrogen to the precursor gas. For
deposition a process with deposition rates of 8 pmlmin
was used. A detailed description of the deposition principle can be found In 131.
From this material solar cells were manufactured in
two different processes. Half of the material was processed in a laboratory-type solar cell process with POCb
diffused eminer (80Wsq) and photoMhographical eminer
grid. In addition a double layer antireflection coating
(Ti02/MgFZ)as well as hydrogen passivation were a p
plied. The other haif was processed via an industrial-type
solar cell process based on in-line diffused phosphorous
emitter (35Nsq) and screen printed contacts fired
through a PECVD-SIN. antireflection layer. Process details can be found in 141.

I-V CHARACTERISTICS
A comparison between the I-V characteristics of the
iab-type and the industrial-type solar cells reveals that the
latter suffer from extremely low fill factors (<30%) as well
as low open-circuit voltages. The reference cells have
equally low fill factors but high open-circuit voltages of
600 mV. In Table 1 the best solar cell results obtained
with both solar cell processes are listed.
In Fig. 1 two typical dark current density curves of
epitaxial thin-film solar cells are shown: one cell was
manufactured via lab-type the other via industrial-type
solar cell process. An analysis of the dark current characteristics revealed that in case of the industrial-type prccess all solar cells, including references, suffer from high
series resistance in the range of > l o Rcm. An Increased
thickness of the SIN. antireneclion coating combined with

1335

a low firing temperature is responsible for the high series

resistance. in general fill factors >77 % are achieved for
monocrystalline (Cz) solar cells manufactured by the
same industrial-type process but with optimized parameters for the SIN. layer thickness and the firing temperature
151.
In addition, the epitaxial cells show extremely low
parallel resistance in the range of 4 0 0 Qcm'.
Vm

FF

lsc

RP

PA]

[mv] [mn/cm']
FA]
Lab. I 6 1 9
30.0
73.9
79.8
630
33.7
Ind. 521
15.4
28.9
604
20.8
29.2
'References: Cz material
Solar ceN area: 21.16 cm' (lab.).

13.8 2.6~10'
17.0 7.8xld

*I
I
'1

49
65x18

2.3
3.7

RS
ram?

0.54

Lab-type solar cells

For a laboratory-type epitaxial thin-film solar cell
based on monocrystalline substrate the lock-in thermcgrams (0" image) measured in forward and reverse bias
direction are shown in Fig. 2 and Fig. 3 respectively.
Brighter contrasts correspond to higher local current densities. in forward direction many point-like shunts distrib
uted over the whole cell area are present. In reverse direction these shunts are missing which means that they
all show non-ohmic behavior except the one in the lower
right comer.

0.54
32.5

20.3

23.04 cm' (ind.)

Table 1. Best solar cell results obtained for both solar cell
processes. Thin-fiim cells based on Cz substrate.

In contrast the dark current characteristics for the

epitaxial thin-film cell manufactured via lab-type solar cell
process represents features which are typical for such a
cell structure.

. , . , .

10
.

1~

.( .. ,.
I

' :..
.

..

Fig. 2. Lock-in thermogram (0" image) of an epitaxial thinfilm solar cell on monocrystalline substrate at 0.6 V forward bias.

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Voltage M
Fig. 1. Dark current density curves of WO epitaxial thinfilm solar cells on monocrystalline silicon substrate.
LOCK-IN THERMOGRAPHY
Lock-in thermography was carried out for an analysis
of shunts affecting the performance of the manufactured
solar cells. A detailed description of the measurement
principle can be found in [Z].

0-7803-7471-1/02/$17.00 02002 IEEE

Fig. 3. Thermogram (0" image) of the same solar cell as

in Fig. 2 but measured at 0.6 V reverse bias.

An investigation of the cell surface by microscopy revealed that three types of surface irregularities are present: pure silver droplets, epitaxial spikes (mostly coated
with silver) and epitaxial stacking faults. The spikes are
the most frequent defect type followed by the stacking
faults. A microscopical analysis (SEM. Nomarski microscopy) revealed that most spikes are grown on top of a
stacking fault.

1336

A comparison of the location of shunts and surface

topography revealed that all shunts visible in the thermogram can be directly correlated to spikes. Fig. 4 shows a
typical correlation of point-like shunts and spikes.
Neither stacking faults alone nor silver droplets present on the surface of the solar cell seem to introduce severe shunts in the solar cell.
The spikes typically feature a height of up to 30 pm
and a diameter of up to 70 pm. The crystal Structure is
polycrystalline. The SEM image of an as-grown epitaxial
spike as seen on the surface of the epitaxial thin-film
material is displayed in Fig. 5.
Stacking faults usually originate at the interface of the
substrate and the epitaxial layer. On <loo> oriented epitaxial surfaces they appear as squares with side lengths
being directly related to the thickness of the epitaxial
layer.
The silver droplets on the surface and the coating of
the spikes by silver are remainders of the thickening of
the emitter grid by electrc-plating. The distortion of the
crystal structure by defects present on the surface enhances an electrolytical crystallization of silver on these
sites.

The correlation between epikaxial defects and shunts

is not unexpected since it is well known that epitaxial
growth defects in general have a negative effect on the
performance of any semiconductor device. Stacking faults
are assumed to be comparatively harmless whereas
spikes are often referred to be crystal defects having a
detrimental effect on the epitaxial layer
The evocation of shunts by spikes can be explained
as a consequence of the polycrystalline crystal structure
which is interfering with the pn-junction in this region.

m.

Industrial-type solar cells

For comparison a corresponding sample prepared via
the industrial-type solar cell process was investigated by
thermography. The performance of these solar cells is
affected by severe shunts as the analysis of the I-V characteristics has shown. The result of the lock-in thermography measurement can be seen in Fig. 6 with the scaling
beina the same as in Fig. 2.

Fig. 4. Section of lock-in thermogram (left) and plan view

of corresponding topographic image (right).
Fig. 6. Thermogram (0" image) of an epitaxial solar cell
manufactured via industrial-type process (0.6V forward
bias). The scaling is the same as in Fig. 2. The emitter
grid has been supplementary added.

Fig. 5. SEM image of an as-grown epitaxial spike (angle

of view: 45")
For an interpretation of the dark regions visible in the
0' image of the thermogram (e.g. Fig. 2. upper right cor-

ner) the corresponding -90" image was studied. This image indicated that regions which appear black in the
0' image actually are affected by shunts. The crystal
structure in these areas is strongly disturbed due to a
comparatively high density of spikes. A similar relation
between shunts and regions of poor crystal quality or regions with a high density of grain boundaries is reported
in [6].

0-7803-7471-1/02/$17.0002002 IEEE
,

Again point-like shunts dominate the thermogram,

with a large fraction of shunts located directly underneath
the emitter grid lines. The shunts in Fig. 6 do not show a
uniform behavior: some of them are ohmic others are
non-ohmic. Obviously there are different mechanisms
working which give rise to shunts of different characteristics.
Some of the shunts visible in the thermogram correspond to spikes located directly underneath the grid lines.
However an exclusive correlation of shunts and epitaxial
defects or any other surface irregularityvisible by microscopy could not be determined.
A thennogram of an identically processed epitaxial
thin-film cell on mc-substrate showed similar features:
point-like shunts located underneath the emitter metallization grid dominate the thermogram but again no exclusive correlation of shunts and crystal defects could be
determined.
The observed accumulation of shunts undemeath the

1337

emitter grid lines can be explained by the following

mechanism which is herewith proposed: the silver paste
which is used for the screen printing of the emitter w n tacts contains glass frit which etches silicon to a certain
degree [E]. The silicon Is etched more rapidly for defective
crystal structures or grain boundaries [9]. If the Ag-paste
is screen printed over epitaxial spikes the etching effect of
the glass frit strongly affects the space charge region or
might even lead to the formation of holes in the emitter
structure.
In [Z] and [lo] an accumulate occurrence of shunts
underneath the emitter metallization is also reported. Silicon humps located underneath the emitter grid lines were
found to be responsible for these shunts [lo]. As a possible shunting-mechanism it is proposed that a thin emitter
diffusion at the edges of the humps results in a direct
contact between the base of the solar cell and the emitter
metallization.
For reference a monocrystalline solar cell which run
through the same industrial-type solar cell process was
also characterized by lock-in thermography. The thermgram revealed a domination of edge shunts and a negligible effect of shunts within the cell area and under the grid
lines. These are typical features observed for solar cells
manufactured in a standard industrial solar cell process
with the edge shunts being the result of an incomplete
separation of emitter and base [6],121.
In conclusion, the results gained by the analysis of
the lock-in thermograms lead to the assumption that the
observed accumulation of shunts underneath emitter
metallization grid lines hasto be attributed to the interaction between the epitaxial thin-film material and the industrial-type solar cell process. To clarify the influence of
a screen-printed eminer and a screen-printed contact
when applied to thin-film material further investigations
have to be carried out.
SUMMARY
We reported the manufacturing of epitaxial siliwn
thin-film solar cells by twa different solar cell processes
and their characterization by infrared lock-in thermography. The distribution and formation of shunts affecting the
solar cell performance were found to depend on the applied solar cell process as well as the quality of the epitaxial layer. For solar cells manufactured in a laboratorytype process the existence of shunts could be directly
traced back to epitaxial defects. An enhanced occurrence
of shunts located undemeath the emitter grid lines could
be determined in case of the solar cells made by an industrial-type process.
The results achieved in this paper show that lock-in
thermography can give important information about the
origins of shunt-formation. For a successful transfer of the
silicon thin-film solar cell concept from laboratory to industrialscale manufacturing it will be necessary to adapt
the industrial standard solar cell processes to epitaxial
wafer equivalents.

0-7803-7471-1/02/$17.00 02002 lEEE

ACKNOWLEDGEMENTS
This work was supported by the European Commission in
contract no. ERK6-CT-1999-00014 and by the Federal
Ministry of Economics and Technology under contract
no. 0329836 (TEKSI). D.M. Huljie is financially Supported
by the scholarship program of the Deutsche Bundesstiftung Umwelt. Germany and a scholarship of the
Albert-Ludwigs-Universityof Freiburg, Germany.
REFERENCES
[ l ] 0. Breitenstein, M. Langenkamp, 0. Lang,
A. Schirrmacher, Solar Energy Materials and Solar Cells
65, 2001, pp. 55-62.

[2] 0. Breitenstein. M. Langenkamp, J.P. Rakotoniaina,

J. Zettner, Pmc. I f h EC Photovoltaic Solar Energy Conference, 2001, in print.
[3] F.R. Faller, A. Hurrle, /E Transactions on Electron
Devices46 1999, pp. 2048.
[4] D.M. HuljiC, D. Bid, R. Preu, C. Craff Castillo,
R. Ludemann, Pmc. 2ff /E Photovoftaic Specialists
Conference. 2000, pp. 378-382.

[5]J. Rentsch, D.M. Huljit, S. Reber. R. Preu, R. Ludemann, "Progress in screen printed front side metallization
schemes for CSiTF solar cells", this conference.

161 I. Konovalov, 0. Breitenstein. K. Iwig, Solar Energy

Materials and Solar Cells 48.1997. pp. 53-60.

F. Passek, R. Schmolke, U. Lambert, G. Puppe.

P. Wagner. Proceedings of the Symposium on Crystalline
Defects and Contamination, Journal of the Electmchemical Society, 1997, pp. 438-447.

[SI C. Ballif, D.M. HuljiC, A. Hessler-Wyser. G. Willeke,

'Nature of the AgSi interface in screen-printed contacts:
a detailed transmission electron microscopy study of
cross-sectional structures", this conference.
191 R. Mertens, M. Eyckmans, G. Cheek, M. Honore, R.
Van Overstraeten, Pmc. I? lEEE Photovoltaic Specialists Conference, 1984, pp. 1347-1351.
[IO] M. Langenkamp, 0. Breitenstein, M.E. Nell, A. Braun.
H.-G. Wagemann, L. Elstner. Pmc.
EC Photovoltaic
Solar Energy Conference. 2000, pp. 1643-1646.

1338

Thin film silicon materials and solar cells grown by

pulsed PECVD technique
U. Das, S. Morrison, E. Centurioni and A. Madan
Abstract: Pulsed plasma enhanced chemical vapour deposition (PECVD) involves modulation of
standard 13.56 MHz RF plasma in the kilohertz range. This allows an increase in the electron
density during the ON cycle, while in the OFF cycle, neutralising the ions responsible for dust
formation in the plasma. The authors report the development of state-of-the-art nanocrystalline Si
(nc-Si:H) materials using a pulsed PECVD technique with 220 crystallite orientation, grain size of
B200 A, low O concentration and a minority carrier diffusion length Ld of B1.2 mm. The crucial
effects of the p/i interface and the incubation layer have been investigated and an efciency of
B7.5% for a single junction nc-Si:H p-i-n device has been achieved for an i-layer thickness
of 1.4 mm, using non-optimised textured substrates.

Introduction

Amorphous silicon (a-Si:H) materials are used in such

diverse applications as active matrix displays, electrophotography, image sensors and solar cells. Although solar cell
conversion efciency Z (B15.0% initial, small area) has
been obtained with amorphous silicon (a-Si:H) [1], prepared
by the plasma enhanced chemical vapour deposition
(PECVD) technique, the intrinsic light-induced degradation
remains an impediment to large scale usage. Despite
advances in device design, the efciency of commercial
modules is low, B8%. Moreover, the deposition rate (DR)
for device quality a-Si:H lms remains low (B1 A/s) and
the gas utilisation rate (GUR) is generally poor (o8%)
leading to high production costs.
Because instability should not be a major issue, the use of
micro- (or rather nano-) crystalline silicon (nc-Si:H) is
attracting much attention, via deposition techniques, such
as PECVD [2, 3], VHF-PECVD [4], gas jet [5] and
HWCVD (hot wire CVD) [6]. All of these techniques have
resulted in Z of 79% and the so-called micro-morph cell
(comprising a-Si:H and nc-Si:H devices) using the PECVD
technique has resulted in ZB13% at a deposition rate of
B1 A/s [7]. As the lm thickness requirement in the device
(for nc-Si:H) is in the range 13 mm, this is thus an
impractical approach. Using a similar deposition approach
and device conguration, the Kaneka Co. has reported
large area modules with a stabilised Z of B10% [8]. Using
the HWCVD technique, Klein et al. [6] have reported
ZB9.0% for n-i-p solar cells, while it is B8% for a stable
device, but at a DR B1 A/s. Using a VHF-PECVD
technique, the DR has been increased to 5 A/s with ZB7%,
in a single junction conguration. Using a conventional
PECVD (high pressure and low substrate temperature)
technique, Z49% [2, 3] has been achieved, but these process
r IEE, 2003
IEE Proceedings online no. 20030627
doi:10.1049/ip-cds:20030627
Paper rst received 10th February and in revised form 2nd May 2003
The authors are with MVSystems Inc., 17301 W. Colfax Ave. Suite 305,
Golden, CO 80401, USA

282

conditions are not suitable for production due to potential

yield problems. The scale-up of VHF-PECVD is problematical [9] as would also be expected for the jet deposition
technique. Hence all the techniques so far studied suffer
from low DR, dust formation, and/or the compatibility
issue of large area deposition.
The technique considered in this paper uses pulsed
PECVD. This can, in addition to increasing the DR,
suppress dust formation [10]. This is accomplished by
modulating the plasma at frequencies in the range 1 to
100 kHz, and with an ON- to OFF-time ratio of 1050%,
so that negatively charged particles can be extracted before
they grow to sizes that can cause manufacturing yield
problems. The time averaged plasma properties when so
modulated also differ markedly from those generated using
continuous wave (CW) excitation. Because the discharge in
the plasma is not in equilibrium, time modulation permits
tuning of processing conditions and hence the growth and
surface of the resulting lms can be manipulated in a rapid
fashion. Previously we have shown that state-of-the-art
a-Si:H materials and solar cells can be produced using this
technique [11, 12] and we have successfully produced
devices on a larger area (30 cm
40 cm). It should be
emphasised that the technique is consistent with all existing
systems in the eld as the inclusion of this technique
requires only a minor modication. In this paper, we extend
the pulsed PECVD study to grow state-of-the-art nc-Si:H
lms and single junction solar cells based on nc-Si:H.

Experimental methods

The nc-Si:H lms and solar cells were deposited in a

commercially available PECVD cluster tool system specically designed for the thin lm semiconductor market and
manufactured by MVSystems, Inc. The pulsing (between 1
and 100 kHz) was superimposed onto a 13.56 MHz signal
applied to the RF plate. The process conditions for nc-Si:H
depositions were, RF power 1070 W, ow rate of silane
250 sccm, hydrogen dilution 9099%, deposition pressure
2009000 mTorr and heater temperature (TH) during lm
growth 100450 1C.
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

Results and discussion

3.1

Nanocrystalline silicon: material properties

There are several factors which determine the optoelectronic

properties of nc-Si:H such as the orientation and passivation of grains, oxygen concentration, voids, crystalline
fraction etc. From a device point of view the critical factors
are minimisation of the incubation layer, control of
interfaces and the effect of textured substrate.
We have used a proprietary modied pulsed PECVD
technique to deposit nc-Si:H absorber intrinsic layer on a
superstrate type (glass/TCO/ZnO/nc-p/nc-i/a-n/Ag) conguration, while doped layers (nc-p and a-n+) were grown
using the normal CW (xed frequency of 13.56 MHz)
technique. The structural studies were performed on actual
p-i-n devices grown simultaneously on c-Si wafer using
XRD, FTIR and cross-sectional TEM. The minority carrier
diffusion length (Ld) was estimated from a reverse bias
quantum efciency (QE) experiment after determining the
depletion width from capacitancevoltage (CV) measurements.
Figure 1 shows the dark conductivity sd and DR as a
function of deposition pressure, while Fig. 2 shows the
XRD spectra of p-i-n devices grown at different deposition
temperatures, TH on c-Si wafers. nc-Si:H lms consist

growth rate
4

dark conductivity
103

2
105

d, cm1

growth rate, /s

107

0
0

2000

4000

6000

pressure, mtorr

Fig. 1 Deposition rate and dark conductivity against deposition

pressure for nc-Si:H films

mostly of (220) oriented grains with a size B1518 nm and

crystalline fraction B80%. However, the integrated intensity ratio [(220) to (111)] I220/I111 varies with TH. The
high efciency (7.5%) discussed below is (220) oriented with
I220/I111 ratio B2.0.
Figure 3, of the FTIR spectra, shows that no change in
the hydrogen content (B7 at%) in the lm with TH was
observed. The SiO peak (B1090 cm1) intensity, however,
increases with increase of TH, which implies an increase in
the porosity of the lm at higher TH. There is a large body
of evidence that the control of oxygen plays a crucial role in
the transport properties of the material. For instance,
oxygen is known to act as a donor (O3+) in a-Si:H [13],
i.e. O of about 1019 cm3 increases sd to about 109
(ohm cm)1 from its normal value of B1010 (ohm cm)1.
Similar effects in nc-Si:H have also been noted, in which the
mobility of electrons (as determined by the Hall effect)
remained constant at about 1 cm2/(Vs)1 while the electron
carrier density increased from 1013 to 1014 cm3 (i.e an
increase in sd) when the O concentration increased from
1018 cm3 to 1019 cm3 [14]. It was surmised that the O may
segregate at the grain boundaries and hence the O3+ donors
lead to an increase in recombination at the surface, i.e. high
surface recombination velocity. With the use of appropriate
process conditions a passivation reaction, such as
OSi+
3 +H- SiOSi+SiH can take place [15], which
promotes passivation of bonds at the grain boundaries, with
the result that there is minimisation of recombination at the
grain boundaries and an improvement in the properties of
the material. Evidence for this is also provided by the
related work on ne-grained (size B0.25 mm) poly-silicon
thin lm transistors (TFTs) [16]; in this, the ON/OFF
characteristics, Vth (threshold voltage) and eld effect
mobility are improved when the as-deposited TFT is
subjected to H, O and H+O plasmas, with the latter
exhibiting the best characteristics. For a gate voltage,
Vg4Vth, the technique allows the investigation of transport
of carriers above the conduction band and across the grain
boundaries. Hence, the appropriate parameter to measure is
the activation energy of such transport, as any impediment
to its ow will be reected in its changes. It was shown that
the activation energy (for VgVth 10 V) was the lowest
(o30 meV) for the O+H plasma treatment while in the asdeposited TFT, the activation energy 4140 meV. This
study would lend support to the suggestion of the
passivation reaction discussed above, leading to good
passivation of the grain boundaries.

(220)

160
(111)

(311)

2000

Si-O

1500
80

, cm1

Intensity, arb. units

Topt 20C
120

Topt

Topt + 140C

1000

Topt + 20C
500

40
Topt + 20C

0
20

30

40

Topt
Topt 20C

50

500

60

IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

1500

2000

2500

wavenumber, cm1

2, deg

Fig. 2 XRD spectra of p-i-n devices grown at optimum TH (Topt)

and 201C deviation either side

1000

Fig. 3

FTIR spectra of p-i-n devices grown at different TH

The Y-scale is shifted for clarity

283

We have measured Ld in an actual solar cell conguration, using a CV technique at a frequency of 100 Hz. At
such a frequency, all states in the intrinsic layer of the cell up
to B0.5 eV below the conduction band should respond to
the capacitance signal. Figure 4 shows the CV curve and
the slopes of the curves give values of doping density
2.1
1015 cm3 near the p layer, and 7.6
1015 cm3 at a
distance 0.7 mm from the p layer. Knowing the depletion
width from CV measurements, Ld can be estimated from
the reverse bias quantum efciency (QE); this involves
deposition of ITO on the n+ layer, and the measurement of
QE from a 500 nm light beam at different bias. At this
wavelength, as most of the photons are absorbed within
o0.1 mm, then the holes are generated essentially at the
back surface of the intrinsic layer. By varying the bias
voltage, the depletion width is changed, and therefore, the
eld-free distance that the hole diffuses before it is swept
away by the depletion eld, also changes. By matching this
curve with a simple exp(x/Ld) relationship for carrier
collection, where x is the thickness of the eld free region,
we estimated the diffusion length Ld to be B1.2 mm.
10

gas), and that it increases several orders of magnitude in the

thickness range 2030 nm, suggesting the presence of an
amorphous incubation phase in the nc-p layer deposited on
glass.
Figure 6 shows a comparison of illuminated JV
(currentvoltage) characteristics of three types of p-i-n solar
cells with the same nc-p (60 nm) and a-n+ (30 nm) layer but
with three different i-layer types; (i) a-i layer (400 nm), (ii)
nc-i layer (750 nm), (iii) a-i layer (B2 nm) followed by nc-i
layer (750 nm). A good FF (0.68) for a-Si:H i-layer indicates
that the doped layers (nc-p and a-n) are reasonably good,
however, the device with a nc-i layer exhibits poor device
performance. Deliberate insertion of an a-i (even a few
nanometers thick) at the p/i interface results in severe
deterioration of the device, i.e. low FF. To reveal the
structure of the lm at the p/i interface (initial growth of
nc-i layer on fully nucleated nc-p), we studied the reectance
spectra in the UV region. Additional reections (at 365 nm
and 275 nm) appear due to the presence of crystalline phase
in the lm, as shown in Fig. 7. The gure also shows
reectance spectra for (i) nc-p layer, (ii) B7 nm nc-i layer on
nc-p, and (iii) B7 nm nc-i layer on nc-p. Curve (i) clearly
shows shoulders at 365 nm and 275 nm, indicative of good
crystallinity in the nc-p layer. However, the B7 nm nc-i
layer directly grown on nc-p appears amorphous (curve (ii)).

1/C 2 , pF2 (x1016)

8
6

voltage, V
0
0.1

0.1

0.3

0.5

0.7

0.9

0.5

0.3

0.1

0.1

0.3

0.5

0.7

0.9

reverse voltage, V

J, mA/cm2

5
0

Fig. 4 1/C2 (estimated from CV measurement) against applied

reverse voltage for a typical pulsed PECVD grown device

3.2

10
nc-p/a-i/a-n
nc-p/nc-i/a-n
nc-p/a-i(2 nm)/nc-i/a-n

Nanocrystalline silicon: solar cells

The p/i interface, in a superstrate-type (p-i-n) device, plays a

crucial role in determining the device performance, as ncSi:H often starts with an amorphous incubation phase, the
extent of which sensitively varies with the lm growth
conditions.
Figure 5 shows sdB0.3 (ohm cm)1 (lateral conductivity)
of nc-p layer (using tri-methyl boron (TMB) as a dopant

15

Fig. 6 Illuminated JV curve of p-i-n device with three different

i-layer types (see text)

reflectance, %

lateral
d, cm1

101
103

a-Si:H

60

(ii)
(i)
40

105

(iii)
20

107

process B
process A
200

109

300

400

500

600

wavelength, nm
0

500

1000

1500

2000

2500

thickness,

Fig. 5 Variation of lateral dark conductivity using TMB as a

dopant gas for different film thicknesses
284

nc-Si:H

80

10

Fig. 7 Reflectance spectra of (i) nc-p, (ii) nc-p/7 nm nc-Si:H

(pulsed PECVD), (iii) nc-p/7 nm nc-Si:H (modified pulsed
PECVD)
The nc-Si:H and a-Si:H spectra have been shifted in Y-scale for clarity
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

Topt + 140C

1.0

Topt
Topt 20C

0.8
relative QE

The use of our proprietary modied pulsed PECVD results

in a signicant increase in crystallinity in the overlying nc-i
layer (curve (iii)). Either or both of the following would
result in such an initial nucleation of nc-i: (a) the technique
modies the nc-p surface, resulting in a high density of
nucleation sites for subsequent nc-i layer growth; and/or (b)
removal of surface oxygen, which may form during
interchamber (dope-to-intrinsic) transfer, from the nc-p
lm, leading to better nucleation of nc-i. Figure 8 shows the
illuminated JV characteristics of nc-Si:H p-i-n devices, and
a signicant improvement in the device characteristics is
observed when the incubation layer has been eliminiated.

0.6
0.4
0.2
0
400

600

800

1000

wavelength, nm
voltage, V

Fig. 10 Relative quantum efficiency (QE) of nc-Si:H p-i-n devices

grown at three different TH

0
0.1

0.1

0.3

0.5

0.7

3
1.0

25
Topt

12

0.9

Topt + 20C

Topt + 140C

15

0.8

FF

20

Jsc, mA/cm2

J, mA/cm2

pulsed PECVD

15

Topt

10

modified pulsed PECVD

0.7

18
5

Fig. 8 Illuminated JV curve of nc-Si:H p-i-n with and without

modified pulsed PECVD technique

Lines are guide to the eyes

thickness of the i- layer, which is a consequence of improved

i-layer quality in terms of improved Ld (presumably, better
grain boundary defect passivation) at lower TH.
Figure 12 shows the dark JV characteristics of two
devices (0.8 mm and 2.1 mm thickness) grown at Topt.
Interestingly, the dark JV characteristics show better
device performance for thicker i-layers, which is indicative
of better i-layer quality for the thicker lm. There is some
evidence in the literature that nc-Si:H material properties
can in fact improve with thickness. For instance, using the
SPV technique, Ld improved from 0.8 mm to 1.2 mm when

100

200

0.75
0.70
0.65
0.60
0.55
100

100

200

TH- T opt, C

2.0
1.8
1.6
1.4
1.2
1.0
0.5 1.0 1.5 2.0 2.5
i-layer thickness, m

6
efficiency ,%

FF

TH- T opt, C

log (J)

Jsc , mA/cm2

Voc, mV

100

25
20
15
10
5
100

0.6
2.5

2.0

Fig. 11 Variation of Jsc and FF as a function of i-layer thickness

at different TH

2
550
500
450
400
350

1.5

i-layer thickness, m

diode quality factor

Figure 9 shows nc-Si:H p-i-n device performance (Voc,

Jsc, FF, Z), under AM1.5 illumination, as a deviation from
optimum heater temperature (Topt) of the nc-i layer growth.
We have achieved ZB7.5% (FF of 0.69) with an i-layer
thickness of B1.4 mm. Figure 10 shows the relative
quantum efciency (QE) for an nc-i layer fabricated at
three different heater temperatures. All three devices exhibit
similar QE in the long wavelength (7501000 nm) region
indicating that the i-layer properties (crystalline fraction and
size) are identical, as also noted from the XRD spectra in
Fig. 2. However, there is a small difference in the short
wavelength region (400650 nm) for the device grown at
Topt, resulting in a slight increase in Jsc (Fig. 9). Figure 11
shows the variation of Jsc as a function of i-layer thickness
at three different TH. The value of Jsc starts to drop for
i-layers thicker than B1.1 mm for higher TH However at
lower TH (closer to Topt), Jsc increases continuously with

1.0

100

TH- T opt, C

200

8
7
6
5
4
3
100

0.84 m
1.4 m

100

200

TH- T opt, C

Fig. 9 Variation of open-circuit voltage (Voc), short circuit current

(Jsc), fill factor (FF) and efficiency with deviation from optimum
heater temperature (ToptTH) during nc-Si:H i-layer growth
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

0.8 m
2.1 m
8
0.4

0.2

0.0

0.2

0.4

0.6

voltage, V

Fig. 12 Dark J-V characteristics for nc-Si:H p-i-n devices grown

at Topt for two different thicknesses
Inset shows the variation of diode quality factor, n, as a function of
thickness of nc-Si:H in a p-i-n device structure
285

the thickness was increased from 1 mm to 3 mm [17].

Kitagawa et al. reported that the intensity ratio of (220)
to (111), I220/I111 increase with thickness [18]. In contrast,
Klein et al. [6] report that the diode quality factor (estimated
from dark JV characteristics using the diode equation,
Jdark J0[exp(eV/nkT)1], where n is the diode quality
factor) increases from 1.4 to 1.8 when the device thickness is
changed from 1 mm to 43 mm, and attribute this to
increased recombination within the bulk. However, in the
inset of Fig. 12, we show that for devices produced by the
pulsed PECVD technique, n decreases as the device
thickness increases, which lends support to the notion that
the material properties actually improve with thickness.
Lastly, we show in Fig. 13, a cross-sectional TEM for a
high efciency device. It is to be noted that the grain grows
perpendicular to the substrate and tends to collide
(indicated by two arrows in Fig. 13) with the neighbouring
grains, and is due to the texturing of Asahi TCO substrate.
Such a grain collision during lm growth may lead to
poorer (defective) grain boundaries [2]. It should be noted
that the texture of Asahi TCO has been optimised for
a-Si:H based devices, and Fig. 13 clearly indicates that such
texturing is not suited to nc-Si:H based devices. Therefore,
there is signicant room for further improvement in device
efciency by developing TCO texturing suitable for nc-Si:H
lm growth and using ZnO/Ag as back contact.

Fig. 13

Cross-sectional TEM of high efficiency p-i-n device

Conclusions

The pulsed PECVD technique, of necessity, leads to a nonequilibrium plasma condition and appears to suppress the
nucleation centres that subsequently lead to dust, which in
turn could reduce structural defects in the lm. This
reduction in particulate count in the plasma has manufacturing yield implications. In this work, an existing pulsed
PECVD technique was modied for the growth of nc-Si:H
lms and solar cells. The technique has been demonstrated
to reduce/eliminate the amorphous incubation phase at the
p/i interface (which has a deleterious effect on nc-Si:H
devices). An efciency of B7.5% (FF of 0.69) at an i-layer

286

thickness of B1.4 mm has been achieved on a non-optimised

texture of substrates (Ashai U-type TCO) by reducing
amorphous incubation at the p/i interface and by optimising
the i-layer growth temperature. A minority carrier diffusion
length of more 1 mm has been achieved.
5

Acknowledgments

This work was nancially supported in part by National

Renewable Energy Laboratory, Golden, Colorado, under
contract ZDJ-2-30630-31. The authors sincerely thank Dr
Don Williamson and Dr Sutter of Colorado School of
Mines for XRD and cross-sectional TEM measurements,
respectively, and Dr Vic Dalal of Iowa State University for
diffusion length measurements.
6

References

1 Yang, J., and Guha, S.: Amorphous silicon alloy materials and solar
cells near the threshold of microcrystallinity, Mater. Res. Soc. Symp.
Proc., 1999, 557, pp. 239250
2 Nasuno, Y., Kondo, M., and Matsuda, A.: Microcrystalline silicon
thin-lm solar cells prepared at low temperature using RF-PECVD.
Proc. 28th IEEE Photovoltaic Spec. Conf., Anchorage, USA,
September 2000, pp. 142145
3 Roschek, T., Repmann, T., Kluth, O., M.uller, J., Rech, B., and
Wagner, H.: High rate deposition of microcrystalline silicon solar cells
using 13.56 MHz PECVD prerequisites and limiting factors, Mater.
Res. Soc. Symp. Proc., 2002, 664, pp. A4.3.1A4.3.12
4 Shah, A., Meier, J., Torres, P., Kroll, U., Fischer., D., Beck, N.,
Wyrsch, N., and Keppner, H.: Recent progress on microcrystalline
solar cells. Proc. 26th IEEE Photovoltaic Spec. Conf., Anaheim,
USA, October 1997, pp. 569574
5 Jones, S.J., Crucet, R., and Izu, M.: Use of a gas jet deposition
technique to prepare microcrystalline Si solar cells. Proc. 28th IEEE
Photovoltaic Spec. Conf., Anchorage, USA, September 2000,
pp. 134137
6 Klein, S., Finger, F., Carius, R., Rech, B., Houben, L., Luysberg, M.,
and Stutzmann, M.: High efciency thin lm solar cells with intrinsic
microcrystalline silicon prepared by hot wire CVD, Mater. Res. Soc.
Symp. Proc., 2002, 664, pp. A4.3.1A4.3.12
7 Meier, J., Vallat-Sauvain, E., Dubail, S., Kroll, U., Dubail, J., Golay,
S., Feitknecht, L., Torres, P., Fay, S., Fischer, D., and Shah, A.:
Microcrystalline/micromorph silicon thin lm solar cells prepared by
VHF-GD techinque, Sol. Energy Mater. Sol. Cells, 2001, 66,
pp. 7384
8 Yamamoto, K., Yoshimi, M., Suzuki, T., Nakata, T., Sawada, T.,
Nakajima, A., and Hayashi, K.: Large-area and high efciency a-Si/
poly-Si stacked solar cell module. Proc. 28th IEEE Photovoltaic Spec.
Conf., Anchorage, USA, September 2000, pp. 14281432
9 Sansonnens, L., Howling, A.A., and Hollenstein, C.H.: Large area
deposition of amorphous and microcrystalline silicon by very high
frequency plasma, Mater. Res. Soc. Symp. Proc., 1998, 507,
pp. 541546
10 Kirimura, H., Maeda, H., Murakami, H., Nakahigashi, T., Ohtani, S.,
Tabata, T., Hayashi, T., Kobayashi, M., Mitsuda, Y., Nakamura, N.,
Kuwahara, H., and Doi, A.: Study of deposition process in
modulated silane plasma, Jpn. J. Appl. Phys., 1994, 33, pp. 43894393
11 Morrison, S., XI, J., and Madan, A.: High deposition rate amorphous
silicon solar cells and thin lm transistors using the pulsed plasma
PECVD technique, Mater. Res. Soc. Symp. Proc., 1998, 507,
pp. 559564
12 Morrison, S., and Madan, A.: Deposition of amorphous silicon solar
cells via the pulsed PECVD technique. Proc. 28th IEEE Photovoltaic
Spec. Conf., Anchorage, USA, September 2002, pp. 928931
13 Morimoto, A., Matsumoto, M., Yoshita, M., and Kumeda, M.:
Doping effects of oxygen or nitogen impurity in hydrogenated silicon
lms, Appl. Phys. Lett., 1991, 59, pp. 21302132
14 Kamei, T., Wada, T., and Matsuda, A.: Effects of oxygen impurity
on microcrystalline silicon lms. Proc. 28th IEEE Photovoltaic Spec.
Conf., Anchorage, USA, September 2002, pp. 784787
15 Nasuno, Y., Kondo, M., and Matsuda, A.: Passivation of oxygenrelated donors in microcrystalline silicon by low temperature
deposition, Appl. Phys. Lett., 2001, 78, pp. 23302332
16 Chern, H.N., Lee, C.L., and Lei, T.F.: The effects of H2O2 plasma
treatment on the charcteristics of polysilicon thin lm transistors,
IEEE Trans. Electron Devices, 1993, 40, p. 2301
17 Kocka, J., Fejfar, A., Vorlicek, V., Stuchlikova, H., and Stuchlik, J.:
Microcrystalline silicon relation of transport properties and
microstructure, Mater. Res. Soc. Symp. Proc., 1999, 557, pp. 483493
18 Kitagawa, T., Kondo, M., and Matsuda, A.: Control of crystallinity
and orientation of microcrystalline silicon using in situ RHEED
observation. Proc. 28th IEEE Photovoltaic Spec. Conf., Anchorage,
USA, September 2000, pp. 780783

IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

Microcrystalline Si:H solar cells fabricated using ECR

plasma deposition
V.L. Dalal, J.H. Zhu, M. Welsh and M. Noack
Abstract: The properties of microcrystalline Si:H materials and solar cells fabricated using remote,
low pressure ECR (electron cyclotron resonance) plasma deposition are described. p+nn+
junction solar cells were deposited at 2753251C on stainless steel substrates using mixtures of silane
and hydrogen. Microcrystalline layers and solar cells could be produced even for low dilution ratio
of hydrogen/silane of 8:1. It was found that once crystallisation started, one could decrease the
hydrogen/silane ratio and still obtain microcrystalline Si:H solar cells. The voltage of the solar cells
could be improved by tailoring the interface between p+ and n layers. An amorphous interfacial
layer improved the voltage. A thin amorphous Si:H layer at the back, between n+ and n layers
was used to signicantly reduce the shunt resistance. Standard device analyses, including dark I(V)
curves and capacitance measured at several frequencies, revealed that device characteristics could
be understood in terms of a standard Si diode model. The doping densities in the n layer were
found to be in the 1
1015 to 2
1016/cm3 range and could be adjusted by altering the amount of
compensatory B doping of the layer. The inuence of the addition of He dilution to the mixture
was also studied, and it was found that He degraded the crystallinity, though it increased the
growth rate and open-circuit voltage.

Introduction

Microcrystalline Si:H is an important technological

material, which is being used extensively for solar cells
and thin-lm transistors [14]. Basically, it is a small-grained
Si material, the grain size being approximately 10 nm, with
grain boundaries saturated with H or with thin a-Si:H tissue
layers separating grains [57] . The presence of signicant H
bonding at the grain boundaries, and the presence of the
thin amorphous tissue, leads to excellent minority carrier
transport through the grains and minimises grain boundary
recombination. The optical absorption characteristics of the
material are similar to that of crystalline Si, with enhancement due to internal scattering at the grain boundaries [8].
In contrast to a-Si:H, most microcrystalline Si:H materials
and devices are very stable against light-induced degradation, although some material with excessive porosity do
suffer from rapid oxidation and changes in device
characteristics.
Four techniques have been used in the past to deposit
microcrystalline Si:H solar cells, namely standard glow
discharge at 13.56 MHz [9], VHF discharge at higher
frequencies [10, 11], hot wire deposition [12, 13], and low
pressure, remote electron cyclotron resonance (ECR)
plasma deposition. Our group has used the latter technique
to deposit lms and devices in both microcrystalline Si:H
and microcrystalline (Si,Ge):H [1418]. In this paper, we
explore device fabrication techniques for ECR-deposited
microcrystalline Si:H solar cells, and show that appropriate
design of both the back (n+n) and front (p+n) interfaces
r IEE, 2003
IEE Proceedings online no. 20030629
doi:10.1049/ip-cds:20030629
Paper rst received 17th February and in revised form 19th May 2003
The authors are with the Microelectronics Research Center, Iowa State
University, Ames, Iowa 50011, USA

316

can signicantly improve device properties. We also study

the dark IV characteristics of these devices, and
capacitancevoltage spectroscopy at different frequencies,
and show that the ECR-deposited base n layers in solar cells
behave very similarly to standard crystalline Si materials.
The ECR-deposited cells are found to be very stable against
both environment-induced oxidation and light soaking,
with reproducible results being obtained three years after
initial deposition.
2

Growth of materials and solar cells

The materials and solar cells were prepared using the

remote, low pressure ECR plasma technique previously
described [14]. Figure 1 shows a schematic diagram of the
reactor system used. A beam of hydrogen ions and radicals
is produced using an ECR plasma and travels towards the
substrate, which is about 25 cm away from the resonance

ECR zone
H2 in
microwaves

magnets
vacuum

silane in

substrate

Fig. 1 Schematic diagram of ECR plasma reactor used to grow

microcrystalline Si:H devices
It is a single chamber machine, operated in a remote plasma mode at
pressures of B5 mTorr. Hydrogen is introduced upstream near the
microwave inlet, and silane downstream near the substrate
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

zone. Silane is introduced near the substrate. Some silane

diffuses upstream towards the plasma resonance zone;
however, the reaction products of silane produced upstream
do not contribute to growth since their mean free paths are
much smaller than the resonance zonesubstrate distance of
25 cm. The electrons and ions in the plasma interact with
silane near the substrate and give rise to lm-producing
radicals. The ECR plasma deposition technique has some
unique advantages compared to the standard RF glow
discharge technique. Among these advantages are:
(i) The plasma can be maintained at very low pressures
because of the very long mean free path of the electrons
rotating around the magnetic eld lines. The advantage of
low pressures is that there is little reaction in the gas phase
between the different radicals (e.g. SiH2 and SiH3) generated
by the plasma. This means that higher order radicals, such
as Si2H5, do not play a role in growth, which makes the lm
more homogeneous. Indeed, we have previously shown that
only the three silane radicals, SiH, SiH2 and SiH3, play a
role in growth of lms in the ECR reactor [19].
(ii) The ECR plasma, because of the high frequency
involved (2.45 GHz), produces ions of only about 1015 eV
energy [21], considerably less than the energy of ions in a
RF glow discharge reactor, which can approach 50100 eV
under conditions typically used for microcrystalline Si:H
growth.
(iii) The fact that the resonance zone is remote from the
substrate allows for good control over the energy and
density of the species arriving at the substrate, which may
play a role in promoting crystallisation. For example, by
controlling pressure, one can change the ratio of H*/H2
arriving at the substrate [21], with higher pressure leading to
a smaller ratio. We nd, indeed, that when the H*/H2 ratio
at the substrate decreases, one obtains an amorphous lm
as opposed to a microcrystalline lm.
The incident microwave power in the reactor was maintained at B200 W. The substrate temperature was varied
between 2701C and 3701C. No signicant differences in the
properties of lms and devices were observed so long as the
temperature was below about 3401C. However, increases in
temperature beyond B3501C led to reductions in opencircuit voltage. This may be related to a loss of hydrogen
from either the SiH bonds at the grain boundaries or from
the a-Si:H tissue coating the grain boundaries at higher
temperatures.
One of the experimental factors investigated was the ratio
of hydrogen/silane. In contrast to other groups, it was
found that excellent lms and devices could be grown even
when the hydrogen/silane ratio was as low as 8:1, although
more typically, we used a ratio of about 10:1. It is believed
that the benecial effects of low energy ion bombardment
may be responsible for our being able to use a lower
dilution ratio of hydrogen to silane. This lower dilution
allows are to grow the materials and devices at higher
( /s). Even higher growth rates (up to
growth rates (B3 A
( /s) can be obtained if He is used in addition to
10 A
hydrogen. We will briey discuss the properties of solar cells
made using He and hydrogen dilutions. Small amounts of
boron (sub-ppm levels) are added to the mixture so as to
compensate for the inevitable oxygen donors present in
plasma reactors.
3

Device structure

The basic device structure is shown in Fig. 2. It consists of a

polished, non-textured, steel substrate. A 100 nm thick n+
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

ITO
p+ a-(Si,C):H layer

buffer layer

micro Si:H n-base layer

thin a-Si:H layer

n+ a-Si:H
steel

Fig. 2 Schematic diagram of the basic device, deposited on

untextured steel substrates

a-Si:H layer is rst deposited to form the back contact.

Next, a thin (2050 nm) undoped a-Si:H layer is deposited.
It was discovered that the presence of this thin a-Si:H layer
helped seal up any shorts and signicantly improved device
properties such as shunt resistance. Next, the n microcrystalline Si:H base layer is deposited to the given
thickness, typically about 0.81.0 mm thick. This is followed
by a thin, graded gap interfacial buffer layer to provide for
bandgap matching between the n layer and the p+ layer
and also to reduce any electron recombination at this
interface. The previous buffer layer used was an amorphous
layer, whose bandgap was varied between B1.4 eV and
1.8 eV by changing the Ge:Si:C ratios [17]. In this work, we
report a buffer layer which is in two parts. First, the
amorphous/crystalline fraction in the interfacial layer is
increased continuously by increasing the silane ow (from
1:10 silane:hdyrogen ratio to 1:6 ratio) as we proceed
towards the p layer. This crystalline layer with decreasing
crystallinity is approximately 100 nm thick. It is followed by
an amorphous (Si,C) layer whose function is to match the
base layer to the p-a-(Si,C):H layer. The amorphous (Si,C)
buffer layer is very thin, about 1530 nm. If it is too thick,
or if too high a bandgap is used for a-(Si,C):H, inection
points occur in the IV curve. This point will be discussed
later. The nal active layer is a p+ a-(Si,C):H layer
deposited at lower temperatures (1701C). A sputtered ITO
contact completes the cell. The thickness of the base n layer
was generally in the range 0.81.0 mm when only hydrogen
was used, but was higher (B2.0 mm) when He was also used
as a diluent gas.
The energy band diagram of the cell is shown in Fig. 3. It
is seen that, unlike the case for a-Si:H cells, this band
diagram resembles a standard crystalline Si p+nn+ junction
solar cell. Unlike a-Si:H, the electric eld does not extend
over the entire thickness of the n layer. Nor is the base layer
undoped; it is, rather, n-doped. In our cells, and in cells
made by MV Systems (a company in Golden, Colorado)
that were measured [20], capacitance data revealed that the
so-called i layers were really n layers. Therefore, there is no
point in calling the cells p-i-n, as some authors do. The
doping in these base layers is quite high, in the 1015 to
2.5
1016/cm3 range. Note that the presence of amorphous
n+ and p+ layers alters the band diagram from that used
for standard c-Si devices. It will be recognised that the
presence of a n+ a-Si:H back layer is potentially very useful
in allowing for efcient transfer of photons after reection
back into the crystalline base layers, though we do not use
any efcient light trapping or back reection techniques for
cells as yet. The back amorphous layers also serve to reduce
317

buffer layer

1.0

n+

0.8

QE

0.6

0.4

0.2

0
400

thin a-Si

The eld does not extend across the entire base layer

600

700

wavelength, nm

Fig. 5 Quantum efficiency of cell of Fig. 4 plotted against

wavelength

104

shorts and increase shunt resistance, while having minimal

effects on series resistance.

QE

101

QE

4 Device results and influence of p+n interface

layer
The IV curve for one of the better devices produced is
shown in Fig. 4. The devices are quite reproducible
(voltages and ll factors within 5%), except for the few
which suffer from the inevitable pinholes. The efciency,
without any signicant back reection and no deliberate
light trapping, is B4.7%, with a voltage of 0.48 V and ll
factor of 0.7. The corresponding quantum efciency (QE)
curve is shown in Fig. 5, indicating a QE from 400 to
800 nm. A more detailed QE curve, showing that the QE
follows the absorption of c-Si out to 1.1 eV energy is shown
in Fig. 6. For comparison, we also show the absorption
coefcient for a crystalline Si wafer over the same range of
energy. Thus, it is conrmed that the base layer in the device
is primarily microcrystalline Si:H.

800

103

102

102

103

10

104
1.0

1.2

1.4

1.6

1.8

absorption coeff. of Si

Fig. 3 Band diagram of device, showing back thin n+n interfacial

a-Si:H layer, n microcrystalline Si:H base layer and graded gap
buffer layer at p+n interface

500

1
2.0

E, eV

Fig. 6 Quantum efficiency of cell plotted on log scale, and

absorption coefficient of Si, both plotted against energy, showing
crystalline Si type absorption behaviour in microcrystalline cell at
low energies
The deviation in QE at high energies may be a result of absorption in
the back amorphous layer and enhanced scattering at grain boundaries

1.2
FF = 0.7
1.0

I, mA

0.8

0.6

0.4

0.2

0
0

0.1

0.2

0.3

0.4

0.5

V, V

Fig. 4 I(V) curve for microcrystalline Si:H cell with base layer
thickness of B0.8 mm

The devices are extremely stable against both exposure to

room air, and against light-induced degradation. Some of
the devices have now lasted about 3 years in ambient
atmosphere with essentially the same IV curves being
318

reproduced over time, showing that there is no ambientinduced oxidation and degradation in these devices. It is
believed that this result may be a consequence of the fact
that the lms and devices are subjected to B1015 eV ion
bombardment during growth [21], thereby producing dense
materials resistant to oxygen percolation along grain
boundaries.
The inuence of changing the hydrogen/silane ratio was
studied, and Fig. 7 shows that as the silane/hydrogen ratio
increased, the open-circuit voltage increased. Good microcrystalline cell characteristics were obtained, even at silane/
hydrogen ratios of 1:8, in contrast to the results of other
groups where for higher silane content beyond about 1:16
ratio, the material became primarily amorphous [11]. As the
voltage increased, the QE at 800 nm decreased, indicating
the more amorphous-like nature of the basic n material
(Fig. 7).
We also studied the inuence of varying the thickness of
the interfacial buffer layer for a xed composition
microcrystalline n base layer. Figure 8 shows the affect of
increasing the thickness of the a-(Si,C):H buffer layer from
15 nm to 30 nm on the IV curve. Clearly, the voltage has
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

0.20

0.6

0.15

Voc, V

0.4
0.3

0.10

QE at 800 nm

0.5

0.2
0.05
0.1
0
0.05

0
0.15

0.10
silane/hydrogen ratio

Fig. 7 Open-circuit voltage and QE at 800 nm against silane/

hydrogen ratio used to grow base layer

Fig. 9 Proposed band diagram at pn interface showing development of a notch as bandgap of buffer layer is increased
The notch traps holes, and a thick notch prevents tunneling of photogenerated holes into the p layer when electric eld is reduced in the
power quadrant

0.8

0.6
1.0

0.4

0.8

0.2

0.6

I, mA

I, mA

FF = 0.67

0.4

0
0

0.2

0.4

0.6
0.2

V, V

Fig. 8 Influence of increasing thickness of pn buffer layer to

30 nm on I(V) curve

0
0

Results for 15 nm buffer layer were shown in Fig. 4. In extreme cases,

an inection point may occur in the I(V) curve

increased as the buffer layer thickness increased, but the ll

factor decreased drastically. In extreme cases, the decrease
in ll factor shows up most prominently as an inection
point in IV curves, indicating troubles with hole collection
as the electric eld across the interfacial layer decreased.
This result may be a consequence of the fact that as one
adds C to the buffer layer, or makes the buffer layer more
amorphous, the valence band edge moves down with
respect to the microcrystalline Si:H layer, forming a notch
to trap the holes (see Fig. 9 for an energy band diagram
showing the notch). Previously, we had shown that such a
notch results even for a-Si:H cells when an improper graded
gap a-(Si,C):H buffer layer is used [22]. If the thickness of
the buffer layer is too high, the holes cannot tunnel through
this notch, particularly in forward bias (power quadrant).
This effect does not show up in reverse bias because the
elds are high, assisting tunneling across the notch. It also
does not show up in far forward bias, because the holes are
then injected from the p into the n layer, and they do not get
trapped by the notch. We recognise that this is only a
qualitative model; an exact quantitative model will require
measurement of band edge discontinuities between microcrystalline Si and a-(Si,C):H using internal photoemissiontype experiments.
Finally, we studied the addition of He to the mixture to
( /s. Figure 10 shows the
increase the growth rates to B5 A
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

0.2

0.4

0.6

V, V

Fig. 10 I(V) curve of cell made with base layer fabricated with
helium dilution in addition to hydrogen dilution

IV curve for such a device. The thickness of the n layer in

this cell is B1.0 mm; the ll factor is still reasonable (0.67)
indicating reasonable diffusion length. However, QE data
reveal that the QE at 800 nm has actually decreased to 0.04
compared to a device with similar thickness made with
hydrogen dilution alone where the QE at 800 nm was 0.07.
This decrease is a clear indication that the material has
become more amorphous-like and less crystalline, a result
which agrees with Raman measurements on lms that show
more amorphous structure as He is added to the gas
mixture. The voltage generally increases as more He is
added, another indication that the material is becoming
more amorphous-like.
5

Device analysis and material measurements

It is well known that in a-Si:H devices, measurement

techniques, such as CV curves, can lead to misleading
results unless one is careful about using the right
temperaturefrequency regime. This fact arises from the
nature of the a-Si:H material, where there is a continuous
distribution of deep states, extending across the bandgap
range. During a CV measurement, the states which are
B0.80.9 eV below the conduction band do not respond to
319

standard measurement frequencies. For a state B0.85 eV

below the conduction band, one can calculate the attemptto-escape frequency v from:

12

v for the 0.85 eV deep state is 5.6
10 /s, using the

standard value of 1
1011/s for v0. (If we were to use a value
of 1
1012/s for v0, as some authors have used [23], the
escape time would be 5.6
103 s, still requiring sub-1-Hz
frequency in CV measurements.)
Therefore, to see such a state, one would have to use
frequencies of 0.0002 Hz, or use higher temperatures. For
microcrystalline Si:H, since the bandgap is 1.1 eV, the midgap deep state levels are only about 0.5 eV from the
conduction band, and vB413 Hz. Therefore, a measurement at 100 Hz should yield meaningful data about the total
defect and donor density. Of course, if there were few deep
states, then a measurement at higher frequencies would not
be too different from a measurement at 100 Hz. In this case,
a measurement at 501C or 751C should also yield results
similar to those obtained at 251C.
We have measured the CV curves of these devices at
100 Hz, 120 Hz, 1 kHz and 10 kHz using a Stanford
Research System CV meter. The signal voltage was kept
small, 0.1 V. The capacitance as a function of frequency is
shown in Fig. 11, indicating that there is little change in
capacitance at low frequencies up to 1 kHz, but it reduces
by a factor of about 2 at 10 kHz, as expected. In all cases,
care was taken to make sure that true parallel-circuit
capacitance was being measured; i.e. the shunt resistance
was high enough (B100 kO). If the shunt resistance is low
(o1 kO), accuracy of the measurement signicantly
decreases.

10
C 2 (1016), F 2

4

8
6
4
2
0
1.0

0.5

0.5

1.0

reverse voltage, V

Fig. 12 Capacitancevoltage curve of device, showing classical

crystalline Si type behaviour, with a doping density of 2.2
1016/cm3
and built-in voltage of 0.8 V

103
104
105
106
I, A

v v0 expEc  Et =kT 

14

107
108
109

10

1010
0

0.2

0.4

0.6

0.8

1.0

capacitance, nF

V, V

Fig. 13 Dark I(V) curve for device, showing three regions; a

generationrecombination at low voltages, a diffusion at higher
voltages, and a resistive loss at the highest voltages

The curves indicate classical c-Si type behaviour

0
10

102

103

104

frequency, Hz

Fig. 11 Capacitancevoltage curve for a standard device against

frequency

Figure 12 plots the results of a CV curve on one of

these normal, high ll factor (40.65) devices. The total
defect density is B2
1016/cm3, and the shallow donor
density is about half of that, or B1
1016/cm3. The built-in
voltage is approximated to B0.8 V, as expected for this
doping density in Si.
We also measured the dark IV curves in this device.
The IV curve is shown in Fig. 13, and it can be tted
extremely well to a standard Si diode three-regime curve,
one representing current arising from generationrecombination in the depletion region I  qV =2kT , one representing diffusion-dominated current I  qV =kT , and one
320

representing the series resistance of the base layer. This

point is made clearer by replotting the exponential data
as the sum of two exponential curves, one representing
expqV =2kT and one, expqV =2kT (Fig. 14). Thus, unlike
the results of some other groups [10], we do not get strange
IV curves with varying diode factors. Rather, we get the
standard result expected for c-Si, in agreement with the
standard capacitance results.
The combination of standard dark IV curve and
standard CV curve, virtually independent of temperature
and frequency in the range studied, clearly indicates that the
band diagram shown in Fig. 3 is correct, representing a
crystalline type device. In the accompanying paper from
MV Systems, they show that their capacitance and diode
dark current data can also be tted with this standard
model [20].
Note that in [20], we also measured diffusion lengths
using a reverse-biased QE technique for a double-transparent cell. The diffusion length was found to be 1.2 mm in their
material, whose doping density was in the 1015/cm3 range.
Since our cells are deposited on non-transparent substrates,
such measurements are difcult to make; we are now in the
IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

thank Josh Koch, Puneet Sharma, Keqin Han and Durga

Panda for their technical help.

104
exp(qV /kT )

105

I, A

106
107

exp(qV /2kT )

108
109
1010
0

0.1

0.2

0.3

0.4

0.5

0.6

V, V

Fig. 14 Detailed analysis of I(V) curve of Fig. 13, showing the

two diode regions
The exp(qV/kT) curve is generated by taking the actual I(V) data and
subtracting the exp (qV/2kT) data from it

process of making double transparent structures for our

cells.
6

Conclusions

In conclusion, it has been shown that ECR deposition at

low pressures can lead to the growth of high quality
microcrystalline Si:H solar cells. The solar cells were very
stable against both light-induced and environment-induced
degradation. The base layers in the cell behaved very
similarly to the base layers in crystalline Si solar cells, with
dark currents being controlled by the standard diode
generationrecombination and diffusion models. The measurement of capacitance against voltage curves at low
frequencies (100 Hz) yielded doping densities in the
1
10152.5
1016/cm3 range. Capacitance did not change
signicantly as frequency was increased to 1 kHz, and then
changed only by a factor of 2, indicating that the deep lying
defect density was comparable to the donor density. The
performance of the solar cells can be improved by
appropriate design of the pn interfacial layer, with a
graded amorphous interfacial layer leading to higher
voltages. Too thick an interfacial layer possibly leads to a
notch in the valence band, which may result in poor hole
transport and poor ll factor, with perhaps even an
inection point in the IV curve. A thin back amorphous
Si layer next to a-Si n+ layer improved the shunt
characteristics of the cell. It was found that good devices
could be made even when the hydrogen/silane dilution ratio
was only about 8:1, though more typically, a ratio of 10:1
was used. Good cells were made using a mixture of helium
and hydrogen dilution. As the helium/hydrogen ratio
increased, the QE at 800 nm decreased and the open-circuit
voltage increased.
7

Acknowledgments

This work was partially supported by a sub-contract from

National Renewable Energy Laboratory. The authors

IEE Proc.-Circuits Devices Syst., Vol. 150, No. 4, August 2003

References

1 Yamamoto, K., Yoshimi, M., Tawada, Y., Fukuda, S., Sawada, T.,
Meguro, T., Takata, H., Suezaki, T., Koi, Y., and Hayashi, K.: Large
area thin lm Si module, Sol. Energy Mater. Sol. Cells, 2002, 74,
pp. 449455
2 Rech, B., Kluth, O., Repmann, T., Roschek, T., Springer, J.,
M.uller, J., Finger, F., Stiebig, H., and Wagner, H.: New materials
and deposition techniques for highly efcient silicon thin lm solar
cells, Sol. Energy Mater. Sol. Cells, 2002, 74, pp. 439447
3 Mulato, M., Chen, Y., Wagner, S., and Zanatta, A.R.: Microcrystalline Si with high electron eld-effect mobility deposited at 2301C,
J. Non-Cryst. Solids, 2000, 266269, pp. 12601264
4 Chen, Y., Pangal, K., Sturm, J.C., and Wagner, S.: p channel thin
lm transistor and complementary metal-oxide-silicon inverter made
of microcrystalline silicon directly deposited at 3201C J. Non-Cryst.
Solids, 2000, 266269, pp. 12741278
5 Vetterl, O., Gro, A., Jana, T., Ray, S., Lambertz, A., Carius, R., and
Finger, F.: Changes in electric and optical properties of intrinsic
microcrystalline silicon upon variation of the structural composition,
J. Non-Cryst. Solids, 2002, 299302, pp. 772777
6 Unold, T., Br.uggemann, R., Kleider, J.P., and Longeaud, C.:
Anisotropy in the transport of microcrystalline silicon, J. Non-Cryst.
Solids, 2000, 266269, pp. 325330
7 Kocka, J., Stuchl!kov!a, H., Stuchl!k, J., Rezek, B., Mates, T., Vrk, V.,
Fojt!k, P., Pelant, I., and Fejfar, A.: Model of transport in
microcrystalline silicon, J. Non-Cryst. Solids, 2002, 299302, pp.
355360
8 Vanacek, M., Poruba, A., Remes, Z., Rosa, J., Kamba, S., Vorlicek,
V., Meier, J., and Shah, A.: Electron spin resonance and optical
characterization of defects in microcrystalline Si, J. Non-Cryst. Solids,
2000, 266269, pp. 519523
9 Nasuno, Y., Kondo, M., and Matsuda, A.: Microcrystalline silicon
thin-lm solar cells prepared at low temperature using PECVD, Sol.
Energy Mater. Sol. Cells, 2002, 74, pp. 497503
10 Vetterl, O., Finger, F., Carius, R., Hapke, P., Houben, L., Kluth, O.,
Lambertz, A., M.uck, A., Rech, B., and Wagner, H.: Intrinsic
microcrystalline silicon: a new material for photovoltaics, Sol. Energy
Mater. Sol. Cells, 2000, 62, pp. 97108
11 Meier, J., Dubail, S., Golay, S., Kroll, U., Fa.y, S., Vallat-Sauvain, E.,
Feitknecht, L., Dubail, J., and Shah, A.: Microcrystalline silicon and
the impact on micromorph tandem solar cells, Sol. Energy Mater. Sol.
Cells, 2002, 74, pp. 457467
12 Klein, S., Finger, F., Carius, R., Rech, B., Houben, L., Luysberg, M.,
and Stutzmann, M.: High efciency thin lm solar cells with intrinsic
microcrystalline Si prepared by hot wire CVD, Mater. Res. Soc.
Symp. Proc., 2002, 715, p. A26.2
13 van Veenendaal, P.A.T.T., van der Werf, C.M.H., Rath, J.K., and
Schropp, R.E.I.: Inuence of grain environment on open circuit
voltage of hot-wire chemical vapour deposited Si:H solar cells, J. NonCryst. Solids, 2002, 299302, pp. 11841188
14 Dalal, V.L., Kaushal, S., Ping, E.X., Xu, J., Knox, R., and Han, K.:
Microcrystalline and mixed-phase Si:H: Preparation, properties and
potential for devices, Mater. Res. Soc. Symp. Proc., 1995, 377, p. 137
15 Erickson, K., and Dalal, V.L.: Growth of microcrystalline Si:H
and (Si,Ge):H lms on polyimide substrates using ECR deposition
techniques, Mater. Res. Soc. Symp. Proc., 1997, 467, p. 409
16 Erickson, K., and Dalal, V.L.: Preparation and properties of
microcrystalline (Si,Ge) on plastic substrates, Mater. Res. Soc. Symp.
Proc., 1998, 507, p. 987
17 Dalal, V.L., and Erickson, K.: Microcrystalline Si solar cells, Mater.
Res. Soc. Symp. Proc., 2000, 609
18 Dalal, V.L., and Erickson, K.: Microcrystalline Si and (Si,Ge) solar
cells on plastic substrates. Proc. 28th IEEE Photovoltaic Specialists
Conf., 2000, pp. 792795
19 Pontoh, M., Dalal, V., and Gandhi, N.: Characterization of ECR
plasma, Mater. Res. Soc. Symp. Proc., 2002, 715
20 Das, U., Morrison, S., Centurioni, E., and Madan, A.: Thin lm
silicon materials and solar cells grown by pulsed PECVD technique,
IEE Proc., Circuits Devices Syst., 2003, 150, pp. 282286
21 Kaushal, S., Dalal, V.L., and Xu, J.: Growth of high quality a(Si,Ge):H lms using low pressure remote ECR discharge, J. NonCryst. Solids, 1996, 198200, p. 563
22 Dalal, V.L., and Baldwin, G.: Design and fabrication of graded
bandgap solar cells in a-Si and alloys, Mater. Res. Soc. Symp. Proc.,
1993, 297, p. 833
23 Street, R.A.: Hydrogenated amorphous Si (Cambridge University
Press, 1993)

321

Photovoltaic Solar E1ectricity:State of the Art and Future Prospects

T.M.Razykov
Physical-Technical Institute, Scientific Association Physics-Sun
Uzbek Academy of Sciences, G.Mavlyanov Street 2B, Tashkent
700084, Uzbekistan
Phone: +998-7 1-235-4103, Fax: +998-71-235-4291, E-mail: razykov@uzsci.net
Absfrocf-We review here the technical progress made in the
past several years in the area of single- and polycrystalline thin
film P V (photovoltaic) technologies based on Si, Ill-V, II-VI
and I-Ill-VI, semiconductors. PV electricity is one of the most
options identified for sustainable providing of future energy
requirements of the world. At the moment PV market is growing very rapidly by 30-40 % each year like telecommunication
and computers. PV production was 520 MW in 2002. Si and
CaAs single crystalline solar cells performed high efficiency
very closed to theoretically predicted maximum value. Singleand polycrystalline Si solar cells still remain the predominant
PV technology at the present. The cost of these solar cells is
rather high (US%3.5-4.5). Thin film PV was developed as a
means of substantially reducing the cost of solar cells. Remarkable progress has been achieved in this field for recent years.
CdTe and Cu(ln, Ca)Se2 thin film solar cells performed a record
efficiencies of 16.17% and 18-19% respectively. These values
are highest for thin film solar cells. The thermodynamic limit of
the solar energy conversion to electricity is 93-95%. This goal
can be achieved on thin film quantum well solar cells.

model [2] for fabrication of high efficiency solar cells. However, research of these solar cells was interrupted in the beginning of 1980s because of the degradation of Cu2S layer. In
spite of too many efforts amorphous (a) Si based thin film
solar cells performed degradation as well. The physical
mechanism of this is Stabler-Wronsky effect (interruption of
Si=H bond under illumination).
Then, R & D in this field was continurd on CulnSe2(ClS)
and CdTe films based solar cells. From 1980s until present
their efficiency has been increased significantly from 5-6 %
to 18-19 % for CIS and to 16-17% for CdTe cells. While,
single junction c-Si and GaAs solar cells are approaching
their upper limits in terms of the theoretical maximum efficiency.
Current status of the PV market and recent results on different main materials based solar cells: c- and pc-and a-Si,
Ill-V, II-VI and I-IIl-V12 and their alloys are reviewed in this
report.

1. 1NTRODLICTlON

II. PV MARKET

It is well known that effect was discovered in 1839 by E.

Becquerel mostly on selenium and curpous oxide devices. A

long period required to get sufficiently good efficiency. Solar
cells developed rapidly in the 1950s owing to space programs and were used on satellites (crystalline (c)-Si solar
cells with efficiency of 6-10%). Energy crises of the 1970s
stimulated Rand D of PV very well.
Compound semiconductors (Ill-V and 11-VI) based solar
cells were elaborated in the 1960s. This is due to the fact that
significant decreasing of the interface states in AI,Gal-,AsGaAs heterostructures and the tailoring of the energy band
structure became possible. At the same time polycrystalline
(pc) Si and thin film solar cells technologies were developed
to provide high production capacity at reduced material consumption and energy input in the fabrication process and integration in the structure of modules by the deposition process and consequently cost reduction for large scale terrestrial
application.
Among thin film solar cells mostly polycrystalline Cu2SCdS structure had been investigated. Results of these investigations allowed to understand many physical processes in
thin film solar cells and more than 10% efficiency had been
achieved. In particular, the investigation of the doping level
on electrical and photoelectrical properties of thin film Cu2SCdS structures [I] allowed to suggest the first time p-i-n

Rapid growh of PV market was started in the 1980s owing

to application of multimegawatt PV plants for power generation. The present PV market grows with very high rates of
30-40 % like telecommunication and computer sectors.
World PV production in 2002 increased to 520 MW (Fig.1)
[3, 41. This became possible owing to technology cost reduction and market development,, reflecting the increasing
awareness of the versatility, reliability and economy of PV
electric supply systems. Of particular interest is the strong
differential growth rate in the rural application, which now
accounts for nearly half of the total PV market. The second
largest market is industrial applications.
PV applications are progressively finding their market
mainly in the US, Japan and EU (mostly Germany). The annual productions of PV cells and modules in the US, Japan
and EU in 2002 were 113.36 MW, 254 MW and 112.29 MW
respectively. These are more than 80 % of the total world PV
production. Recently, Japan became leading PV country. One
of possible explanations of this might be the unfailing support given by the Japanese government to the PV sector. Target PV productions of Japan, US and EU in 2010 are 5 GW,
1 GW and 3 GW respectively.
The top 10 producers of PV cells and modules are prsented
in the Table 1. These companies produced 336.24 MW in
2001. It is 86 % ofthe world production.

- 291 -

--

TABLE 1 . TOP IOPV CELUMOOULE

PROOUCERS~MW)
~

OWorld
0

AJapan
m
.

Company
Kyocera
Sharp
BP Solar
Siemens Solar
Astropower

OUSA

RWE

Sanyo
lsofoton
Mitsubishi
Photowan

0
A

~~

1 1997 1 1998 I 1999 I2000 I2001

I 15.4 1 24.5 I 30.3 I 42 I 54
10.6
11.3
22.0
4.3
4.0

13.5
20
7
4

10.0

4.7

6.5

13.0
6.1

5.7

9.5

10.0

14.0

30.3
32.5
22.2

12.0

50.4

41.9
28
18.
14.0
17
9.S
12.0
14.0

75.0
54.2
39.0
26.0
23.0
19.0
18.0
14.0
i4~n

TABLE 2. 2001 WORLDCELLJMOOULE

PRODUCTION
BY T E C H S O L ~(MW)
Y

Technology
,c+,

7-3

,.I %
.SI

!,cl

2-3

,me

c-Si

v..

pc-si
a-Si

Fig1 Evolution ofworld celllmodule praduction (MW)

Kyocera was leading company during the past several

years, while at the present Sharp gained the first position with
annual production of 75.02 MW in 2001, It should be pointed
out that about half of the national production of the US was
due to companies with capital of Japan or EU. The best results in the US in terms of the production volume were
achieved by Astropower (26 MW in 2001).
In terms of technology, c-Si was the main material for the
world PV industry for a long period (136.78 MW in 2001,
Table 2). Recently pc -Si based solar cells gained leading
position with 184.85 MW in 2001 mostly owing to Japanese
manufacturers.
At present, over 80 % of the world PV industry is based on
c-and pc-Si technologies. Technology of a-Si grows slowly
and annual production in 2001 was 33.68 MW. 8.0 MW of
these solar cells were used for indoor applications (calculators, watches, etc.)
While thin film CIS and CdTe based PV productions are
still in the beginning stage in spite of the remarkable results
of R & D many years ago. It might be due to the reasonable
difficulties between laboratory and large scale production
technologies. Annual world productions of CIS and CdTe
based cellhodules in 2001 were 0.7 MW and 1.51 MW respectively. 1.2 MW CdTe cells were produced by Matsushita
Battely for indoor applications. Several new multimegawatt
thin film plants ready for production of these type solar cells
and their contribution to the world PV market might be significantly expanded soon. It is expected that thin film PV
technologies will play the main role in the world PV market
in the future.
At present the PV market is dominated (more than 40%)
by grid connected residential systems and module prices
are in the range of US$ 2.5-4/WP and whole systems pricesare in the range of US% 4-8lWp depending on technology
and size. According to the US program the cost effective so
lar cells for large scale terrestrial application ( I 5% efficiency,
US$ 0.33/W, and module manufacturing costs of about
US$ 50/m2) should be achieved to 2010.

14.40
4.0

c-Si

concentrators
Ribbon4
CdTe
CulnSq
Si on low cost subs.
Si on Cz slicc
Tola1

II

5.4
0..33

]I

100.32

I ::;

I .2

]I

18.0

171.22

86.38

2.62
1

111. SILICON
SOLAR
CELLS

3. I Crystalline Si Solar Cells

The evolution of Si laboratory cell efficiency is shown in
F i g 2 This dependence can be divided to four stages. Each
stage corresponds to a new solutions in technology or structure of cell. In the beginning of the semiconductor era
(after discovery of the bipolar transistor in 1948) rapid progress of silicon technologv allowed to produce Si solar cells
with 15 % efficiency. In the second stage (1970s) 17 % efficiency Si solar cells were fabricated owing to achievements
of microelectronics (photolithography, etc). The most significant results have been obtained in the stages 3 (1980s) and 4
(2000) and Si cell eficiency close to 25 % has been achieved.
These were due to improved contact and surface passivation
of the cell, along both front and rear surfaces, as well as improved understanding of the significant role of light trapping
in Si devices [SI. For Si cell of 80pm thickness the maximum
efficiency is 28.8%.
The main features of Si solar cells developments were noted in [6, 71. Optically, the inverted pyramids on the top of
surface significantly reduce reflection loss. The metal rear
contact serves as an efficient reflector. Consequently, light is
very effectively trapped into the cell up to 40 times [7]. For
low quality wafers the buried contact cells becomes less important because the open-circuit voltage becomes dominated
by recombination in the bulk regions of the cells rather than
at the contacts.
Passivated emitter and rear locally d i i h e d (PERL) cell on
a FZ wafer has performed 24.7%, efficiency (V,=706 mV,

- 298 -

3 y O d

3.4 Thin Film Amorphous Si Solar Cells

Fig.2 Evolution ofihe efficiency of different Solar cells

1,=42.2 mA/cm' and FF=0,828) which is the highest ever

reported efficiency for Si solar cells [SI. Commercial c-Si
cells have efficiencies of 15-17 %.

Thin a-Si films usually contain a few percent of hydrogen

to passivate dangling bonds and increase material quality.
The plasma CVD is widely used for fabrication of these films.
The design of the cells optimises the collection of current by
having a very thin n-and p-layers, with an intrinsic intermediate layer, thickness of which enough to absorb almost all the
incident light, to give a p-i-n structure. However, the physical properties of the i- layer are degraded under illumination.
Because, Si-H bonds are destroied under visible light. This
effect can be reduced (but not eliminated) by careful control
of the deposition process, decreasing of the thickness of ilayer and using multiple junctions [16]. The best initial efficiencies of 13.7%. and 9.8% were achieved on triple junction
cells and modules respectively many years ago [17]. However, stabilised efficiencies remain low (6-7% for the best
commercial modules). Nevertheless, at present, about IO% of
world wide PV production uses a-Si technology. The best Si
based solar cells efficiencies are presented in the Table 3.
TARLE 3. THEBEST SI SOLARCELLS
EFFICIENCIES

I Dcvice

3.2 Polycrystalline Si Solar Cells

Pc-Si wafers can be fabricated on large area. Plasma processing of lower cost pc-Si is used to form a highly transmissive surface and to increase the light absorption. This known
as reactive ion ekhing allowed about 40 % relative increase
in the absorption. Importance of texturing pc-Si to exhibit
full potential was considered in [9, IO]. 19.8% efficiency
textured pc-Si solar cells was fabricated [ 1 I].Bulk hydrogenation and nitride passivation of cell surface have given
good results. Despite a number of advantages of pc- Si there
is no significant difference between costs of c- and pc-Si solar cells. Commercial cells have efficiency of 12-15 % for
pc-Si wafers.
3.3 Polycrysfalline Thin Film Si Solar Cells
Thin film Si solar cells have the following important advantages in comparison with crystalline cells: I.Thickness of
Si material can be drastically reduced to more than 50pm; 2.
Thin films can be deposited on low cost substrates; 3. Thin
films can be fabricated on module-sized substrates and in
integrally-interconnected structures. According to calculations of [5] thickness of Si films can be reduced up to 1 pm.
Despite different technologies of Si films were developed
in microelectronics many years ago, a long period (around 30
years), has been required to fabricate thin film pc-Si solar
cells with reasonable efficiency. Thin film Si solar cells were
fabricated on ceramic, carbon and glass substrates. Pc-Si thin
films for solar cells with a few pm thickness were deposited
by plasma CVD, hot-wire CVD, rapid thermal CVD, zone melting recrystallization and liquid phase epitaxy. 16.6%
efficiency for small area cells and 11.8% efficiency for large
area module of 238.7~111~havebeen achieved [12-14]. The
cell with a thickness of 2 . 0 p has performed efficiency of
10.7% [15].

I Eff. Yo

I Kef.

c-Si (wafer)
pc-Si (wafer)

24.1

PI

19.8

[Ill

pc-Si (thin film)

a-Si (thin film)

16..6

[12. 131
Cl71

13.7

IV. THIN FILM CHALCOGENIDE SOLAR CELLS

4. I . 11- VI Solar Cells

Among Il-VI compounds CdTe is used as base material in

polycrystalline thin film solar cells owing to its optimal band
gap of 1.46 eV. Progress in thin film CdTe solar cells has
also been very impressive. Over I O % efficiency was
achieved on CdTe cells fabricated by different technologies.
Most of the progress was made by improving contacts to
CdTe and by reducing the thickness of CdS layer. Fabrication
of ohmic contacts to p-CdTe is quite difficult. Quality of contacts is related to efficient doping of p-CdTe. Fabrication of
high conductivity p-CdTe films is difficult because of selfcompensation. However, despite these problems remarkable
results have been achieved over the last years [18-20]. More
than 16% efficiency has been obtained [21]. 10.6% and
11.2% efficiencies were obtained on very thin 0.55pm and
Ipm thick CdTe respectively [22]. Parameters of CdTe thin
film solar modules fabricated by different companies are presented in the Table 4 1231. CdTe modules performed very
good stability for sufficiently long period (800 days) [ZO].
4.2. 1-111- Vi2Solar Cells

Among I-Ill-VI? compounds CulnSe? has very attractive

properties for thin film solar cells. Very high optical absorption (=IOScm~')and consequently very thin layer (O.I-0.3pm),
possibility to change type o f conductivity (p-or n-) and

- 299 -

TAI E 4. BEST LARGE-AREA

THINFILMMOOULES
r4NOART CONDITION
Company/Date
Device
Power
Size
CdSICdTe
CdS/CdTe
CdS/CdTe
CdS/CdTe
CdS/CdTe
a-Si/a-SiGea-Si/SS
a-Sila-SiGe
a-Si/c-Si/glass

BP Solar, 5/00
Fin1 Solar, 6/96
Matsushita, 6/00
BP Solar, 4/00
Golden Photon
United Solar, 6/97
BP Solar, 9/96
Kaneka. 9/00

38.0

a-Si triple

United Solar. 6/97

35.7

Siemens Solar. 3199

44.3

59.0
53.9
31.0
70.8
56.0

stab.

._/

1.6
10.0
stab.

0.01

5.4
3.7
10.2
8.2
4. I

.COS

a, P . C d T B
,

n-ClnTa*
P-CulnTe,

0.02

junction

CdS/CIGS

Fig.3 Dependence orlhe intergrain energy barrier height

an the valency 0110w valent elcnient of ditkrent
Semiconductorsmaterials

4.0

resistivity, formation of solid solutions (different band gaps)

are suitable for fabrication of high efficiency thin film solar
cells. Its band gap of I eV is rather IOW for solar cell and it is
usual to replace indium with an alloy of indium and gallium
Cu (In, Ga)Sel to increase the band gap and hence the open
circuit voltage. The best results have been obtained on cells
fabricated by coevaporation. World wide record eficiency of
19.2 % (the highest value for thin film cells) has been
achieved at NREL and 18.5% at Matsushita Electric [24).
In terms of large scale terrestrial application this material
has disadvantage because indium and gallium have very limited resources. Recently, in these solar cells aluminum was
used instead of gallium and more than 17% efficiency has
been obtained for Cu (In, AI) Se2cell [25]. It should be noted
that the passivation process (key step for a- and pc-Si technologies) is not necessary for Il-VI and l-lll-V12 thin film
solar cells. This can be due to the effect of selfpassivation
suggested in [26] (Fig.3). Very good stability was observed
for these thin film modules for more than 5 years [20]

v. 1Il-v SINGLE-AND MULTIJUNCTION

SOLAR CELLS
It was mentioned above that the theoretical efficiency of
the single junction cell is around 30 %. This value can be
increased up to more than 50 % for four junction cell (Table
5 ) [27]. Bener efficiencies could be obtained with more efficient use of the solar spectrum. Combining two or more cells
of different band gaps into a tandem arrangement increases
the amount of work done per photon. To increase power, the
cells should be current matched and joined in se ries, separated for example by a tunnel diod.
In terms of space power applications Si cell efficiency is
low and susceptible to radiation damage. At the end of 1980s
GaAs cells have been used to fabricate flat plate arrays owing
to high efficiencies and lower radiation induced degradation
than Si cells. GaAs single junction solar cells with efficiencies of around 25 % have been reported for concentrated
AM0 sunlight many years ago [28, 291. Metalorganic chemical vapor deposition, liquid phase epitaxy and molecular
beam epitaxy are usual methods for fabrication of Ill-V solar
cells.

TABLE 5.THERI~ODYXALIIC
EFFICIE~~CY
UNCOHCEUTRATED
SUNLIGHT
LIMITS,

I Numberof I Efliciency I E,(eV)

I[:

I
2

30%

I.3

42%

I.9-I .O
2.3-1.4-0.8
2.6-1 .8-1.2-0.8

149%
Infinity

53%
68%

The mechanically stacked 3-junction cells and InGaAs

bottom cells have reached hish efficiency of 33.3 % [30].
Enhanced efficiencies may be realised by stacking up GaAs
either monolithically o r mechanically on booster cell of lower
band gap materials, such as S i , Ge, CIS, GaSb or InGaAs. In
order to apply super high efficiency cells widely, it is
necessary to improve convenion efficiency and reduce cost.
One of approaches is fabrication of GaAs/Si cells. However,
in this case GaAs films contain dislocations. Another problem is the fact that impurity diffusion from highly doped
tunnel junction during overgrowth of the top cell increases
the resistivity of the tunnel junction. Despite these problems,
a 40% efficiency goal is achievable in the nearest future [31].
VI. THE QUANTUM
WELL SOLAR CELLS
The quantum well solar cells (QWSC) is novel device with
the potential to achieve high efficiency in an altemative approach to tandem or cascade solar cells. QWSC was reviewed
in [32]. Different losses in a standard p-n-junction limiting
the efficiency around 30 % and ways to avoid these losses
to significant increase the efficiency of QWSC are analysed
in [33]. In the simple QWSC quantum wells of a smaller
band gap material are grown within the space charge region
of p-n or p-i-n- structures. Molecular beam epitaxy or metal
organic chemical vapour deposition are used for fabrication
of QWSC.
The incorporation of CdTe quantum wells into Cd,MnlxTe p-i-n structures has been shown to be successful in extending the photoresponse to longer wavelengths [34]. The

- 300 -

QWSC can enhance the short circuit current as well as the

open circuit voltage (V,=1.07 V for QWSC and 1.022 V for
usual cell) over a comparable convetitional cell formed from
the barrier material and that AIGaAs/GaAs QWSC enhance
efficiency over comparable AlGaAs cells. This is due to the
reduction of the recombination losses [32]. Thermodynamic
limit of 93-95 % efficiency might be achieved on QWSC.
VII. CONCLUSION
The present PV market grows with very high rates of 3040 % and the world PV production in 2002 was 520 MW.
Over SO O h of the world PV industry is based on c- and pc-Si
technologies. Single junction c-Si and GaAs solar cells are
approaching their upper limits in terms of the theoretical
maximum efficiency. Remarkable results have been achieved
in the field of thin film solar cells. 18-19 % efficient ClGS
and 16-17 % efficient CdTe and Si polycrystalline thin film
solar cells were fabricated. For Ill-V multijunction solar cells
33.3% efficiency was obtained. Further RgLD will be
directed on the increasing of the efficiency ofthin film CIGS,
CdTe and Si and multijunction Ill-V cells and QWSC.
Thermodynamic limit of 93-95 % efficiency might be
obtained on QWSC.
REFERENCES
111-

121.
131.
141.
151.
161.
171.

1x1.
191.

1131.

T.M. Razykov e l al. Influence of the doping level an electrical and

photoelectrical properties of CwS-CdS solar cells. Appl. Solar Ene r a , VOI. 3, pp. 20-24,1977.
T.M. Razykov and M.A. Razykova, Thin film Cu2SCdS solar cells
with dual distribution of the impurities, Phys. Serniconducrors, vol.
15,pp. 1402- 1405, 1981.
P V N ~ s s , v o l21,N.l,2002.
.
PlNews, vol. 22, Ne3,2003.
M.A. Green, J. Zhao, A. Wang and S.R. WenhanProgress and outlook for high efficiency crystalline silicon solar cells. Solar Energy
MareriolsandSolor Cells. vol. 65, pp. 9-16,2001.
W. Wettling, High efficiency silicon solar cells: state of the art and
trends. Solar Energy Marerials andSolor Cells. vol. 38, pp. 487-500,
1995.
M.A. Green, Status of crystalline photovoltaic technology. World
Renewable E n e r a Congress V1,pp. 2630-2635.2000.
J. Zhao, A. Wang and M.A. Green High eficicncy PERL and PERT
silicon solar cells on FZ and MCZ sub straten: Solar Energy Marer;.
alsand Solar Cells. vo1.65, pp. 429435,2001.
J.C. Zolper, S . Narayanan, S.R. Wenham S.R. and M.A. Green,
16.7% efficiency, laser textured. buried contact polycrystalline silicon solar cells. Applied Physics Letters. vol. 55, pp. 2363-2365, 1989.
V. lnomota K. Fukui and K. Shirasawa Surface texturing of large
area multicrystalline silicon solar cells using reactive ion etching
method. Technical Digesr. !Yh lnlernorional PhorovolroicScience and
Engineering Conference,Japan, pp. 109-1 10. 1996.
J. Zhao, A. Wang and M:A. Green, 19.8 % efficient honeycomb
textured multicrystalline and 24.4% monocrystalline silicon solar
cells. Appli~dPhysicsLerlerr. vol. 73, pp. 1991-1993, 1998.
A.M. Bamett, J.C. Checci, S.R. Collins, D.H. Ford R.B. Hall, C.L.
Kendall, S.M. Lampo, J.A. Rand and A.M. Traband, AP-225 silicon
-film solar cells for power applications. Proc. qfrhe 12 rh European
PholovoltoicSolar E n e r a Con/ Amsrerdom, pp.lI81-1814, 1994.
A.M. Bamen, S.R. Collins, E. Coller, D.H. Ford R.B. Hall and J.A.
Rand, Polycrystalline silicon film solar cells: present and future.

Progresr in Phorovol laics. vol. 2, pp. 163-170. 1994.

[14]. A.M. Bamen, J.A. Rand, R.B. Hall, D.U. Ford,E.J. Del1eDonne.A.E.
Ingram, J.S. Bisaillon, B.W.
Feyock, M.G. Mauk and P.E.
Simr,Highcurrenl thin silicon-on-ceramic solar cells. TeclWcol Digerr, Internoliono1 PVSEC-ll, Japan, p. 982, 1999.
[15]. K. Yamamota, M. Yashimi, Y. Tawada Y. Okamoto and A. Nalajima Cost effective and high performance thin film Si solar cells towards the 21 century. Technical Digesr. The lnrernorionol PVS&C-I I.
Japan, pp. 225-228. 1999.
1161. Y. Hamalawa. Recent advances of thin film solar cells and their
technologies, Proc. o/ rhe Firs1 IVorld Conference opt Pl~o~ovolroic
EnereComersron, USA, pp.34-41, 1994.
1171.
. . S . Guha. Anlomhous Silicon alloy technolonv fomhotavallaics.

_. .

Proc. o/rhe 6 rli lnreniorional Plmrovol laic Science m d Engineering

[IS].

1191.
. .
1201.
1211.
122).

1231.
1241.

1251.
I261.
1271.

1281.

1291.
[301.
1311.

WI.
1331.

PI.

CO</India. pp. 55-60,1992.

P.V. Mayers and R.W. Bikmire. The future af CdTc photovoltaics,
Progress i n Phorovolroorc$. vol. 8. pp. 393-402. 1995.
R.W.Bikmire and E. Eser. Photocnstalline thin filmsolar CCIIS: mesent status and fulure potential. Anrrii. Rev. Ahrer. Sci. vol. 27. pp.
625-653. 1997.
H.S Ullal. I(.Zueihcl and B. Von Roedem. Currcnl status of polyc r y s t a l k thin film PV Iecbnologies, Con/ Record qflhr 26 rh IEEE
P l ~ o r a ~ ~ 1 1 ~ ; c S p e cCon$
r o I i ~USA.
r ~ pp.301-305.1997.
J. Britt and C. Ferekiden. Thin film CdSICdTe solar cell nith 15.8%
etficicncy. Appl. P h p Lerr. vol. 6 2 . p ~2851-2852.
.
1993.
N. Amin. T. kaka. A. Yamada and M. Konagai. High eflicicnt I p,
thick CdTe solar cells with tenured TCOs, Teclmicol Digesr q f r k
Inienrorronul Plroiowl roic Science ond Enginering Cor$ Japm.
pp.837-838.1999.
Pltorovolroie E n e r a Progror, Owervrew, US Depanment of Energy,
p.3, 2000.
T. Neganii, Y. Hashinlolo and S . Nishiwaki, Cu(ln,Ga)Se2 thim film
solar cells with an efficiency of 18 W . Tecl~nicolDigrrr ofrhe 1111~
lnrernorionol Phoror~olrorcScience and Engineering Con/. Japan pp.
993. 1999.
S . Marsillae, I D Paulson, M.W. Haimhadi, R.W. Birlmirs and W.N.
S h a f m a n , High efficiency solar cells based on Cu(ln, AI) Se2 thin
films. AppL Phys. Lerr. vol. 81, pp. 1359-1352,2002,
T.M. Rarykov, A new conception on the passivation of the grain
boundaries of polycrystalline thin film solar cells, Absrrocrs o/rlz Inrernarionol Con/ EuroSun96,Germany, pp. V-91- V-92, 1996.
R.W. Birkmire, Compound polycrystalline solar cells: recent progress and Y2K perspectives, Technical DIgesr o/the I1 rh Iniernamml Phorovolroie Science and Engineering Cmprence, Japan, pp.
33-34, 1999.
V.S. Sundaram, L.M. Fraas, P.E. Gruenbaum, J.E. Awry E. Maloesay
and M.S.Kulyla, High efficiency tandem GaAs/GaSb concentrator
solar cells, Prm. of rhe 6 rh ImernPional Phorovoltoic Science and
Digineering Con/ India. p.p. 395-400, 1992.
V.M. Andreev, High efficiency (24.6%, AMO) LPE g r o w AlGaAsIGaAs concentrator solar cells and modules, Proc. of rhe First
WCPEC, USA. p.p.2OY6-2099, 1994.
M. Yamguchi, Present and future prospects of super high efficiency
tandem solar cells, Technical Digesr oflhe Internalional PVSEC-11,
Japan, pp.3740, 1999.
J.E. Rannels, The case for a 40 % efficiency goal for photovoltaic
cells in 2005 Technical Digesr ofrhe 11 rh lnlernalional PVSEC-11.
Japan, pp.9-12, 1999.
J. Nelson, Quantum well S ~ N C ~ U I Cfor
S photovoltaic energy conversion, Thin Films, vol. 2 l . p ~ 31
. 1-369. 1995.
M.A. Green, Prospects for photovoltaic efficiency enhancement
using low dimensional structures, Nanoiechnology, vol. 1 I, pp. 4U2405,2000.
D.E. Ashenford, A.W. Dweydare, D. Sands, C.G. Scott, M. Yousaf. E.
Aperathitis, Z. Hatzopoulos and P. Panagotator, Investigation of p-in solar cell efficiency enhancement by use of MQW structures in the
i-region, Journol ofcty~ralGrow~h,vol. 159, pp. 920-924, 1995

3rd World Conference on Phorovolraic Enrrgv Conversion

Muy 11-18, 2003 Osoh, Japan

3P-B5-I 1

SUPER RADIATION TOLERANCE OF CIGS SOLAR CELLS DEMONSTRATED IN

SPACE BY MDS-1 SATELLITE
Shirou Kawakita'.*, Mitsuru Imaizumi', Taishi Sumita', Katsumi Kushiya', Takeshi Ohshima', Masafumi Yamaguchi',
Sumio Matsuda', Shinichi Yoda'.' and Tomihiro Kamiya4
1. National Space Development Agency of Japan (NASDA), 2-1-1Sengen, Tsukuba, lbaraki 305-8505, Japan
2.Tokyo Institute of Technology, 4259 Nagatsuta-cho, Midori-ku, Yokoyama, kanagawa 226-8503, Japan
3. Showa Shell Sekiyu K.K, Central R&D Lab. (ARL), 123-1 Shimo-kawairi, Atsugi, Kanagawa 243-0206, Japan
4. Japan Atomic Energy Research Institute, 1233 Watanuki, Takasaki, Gumma 370-1292, Japan
5. Toyota Technological Institute, 2-12-1 Hisakata, Tempaku-ku, Nagoya, Aichi 468-85 1 I , Japan
*E-mail: kawakita.shirou@nasda.go.jp

ABSTRACT
High radiation tolerance of ClGS solar cells is
demonstrated in space for the first timc by monitoring the
performance of Cu(ln,Ga)Scz thin-film solar cells on
MDS-I since February 2002. Short-circuit current of the
ClGS cells did not degrade and open-circuit voltage of the
cells degraded only about 1 %. In contrast, the
performance of other solar cells on the satellite, including
Si and GaAs space solar cclls, degraded I ycar after launch.
The high recovery of radiation damage of CIGS solar cells
due to thermal annealing was found from ground test. We
prcdicted the degradation of the ClGS solar cells in space
using the relative degradation coefficient, the annealing
rates of Voc and Isc for protons irradiating the cells and the
radiation response of the cells without thermal annealing.
The results are in good agreement with flight data of the
CIGS solar cells on the satellite. We were thus able to
demonstrate high radiation tolerance of ClGS solar cells in
space for the first time.

1. INTRODUCTION
A copper indium gallium di-selenide (CIGS) thin-film
solar cell is a promising candidate for future thin-film
space solar cells since the cells havc demonstrated
conversion efficiencies exceeding 18%, which is
significantly higher than other thin-film solar cells [I].
Excellent radiation tolerance of CIGS thin-film cells has
also been reported [2]. In addition, CIGS thin-film solar
cells have better performance than conventional Si and
GaAs space solar cells, such as low cost and light weight,
because the cells can be formed on polyimide [ 3 ] or
stainless steel sheet [4] substrates.
Radiation damage studies for CIGS thin-film solar
cells have revealed that (I) electrical properties of the cells
are not degraded by high-energy electron irradiation.
However, cell performance is degraded by high-energy
proton irradiation similar to other types of solar cells [2],
and (2) radiation damage of the cells due to proton
irradiation gradually recovers when irradiated cells are
kept at room temperaturc [SI. The recovery rate of the cells
depends on the temperature [6]. Since ClGS solar cells
recover well from radiation damage, a prediction of
on-orbit performance of ClGS solar cells must fake into

account the recovery of radiation damage by thermal

annealing.
In this study, we measured the relative damage
coefficicnt (RDC) for thc ClGS solar cclls to predict the
performance of the cells in space. In addition. we measured
the recovery rate of the proton-irradiated ClGS solar cells
by thermal annealing. Furthermorc, we derived the
intrinsic damage coefficient of ClGS solar cells from the
10 MeV proton test at low temperature. We apple the
prediction model with thermal annealing to the flight data
of the ClGS cells on thc satellite using by these results. In
addition, we forecast the cell performance in low-Earth
orbit (LEO) and geostationary orbit (CEO).

2. EXPERIMENT
ClGS thin-film solar cell samples (effective area of
0.6 cm') were separated from a 30cmx30cm CIGS
thin-film integrated mini-module. The circuit was
fabricated by a selenizationisulfurization method [ 7 ] , and
the stmcture of the cells is ZnO/Zn(O,S,OH)x/Cu(ln,Ga)Se, iMoiGlass [XI. The solar cells have an
average efficiency of 10.0%. open-circuit voltage (Voc) of
570 mV, short-circuit current density (Jsc) of 40 mAIcm',
and fill factor (FF) of 60%, which were measured under
AMO, I-sun conditions. Thc cclls have lower FF than
typical ClGS thin-film solar cells reported because of their
relatively high series resistance due to their contact
StNCture [ I ] .

3. RESULTS AND DISCUSSION

3.1 Relative Damage Coefficient of CIGS Solar
Cells
We performed proton irradiation with the ion
implanter, the tandem accelerator and the AVF cyclotron
accelerator at the Japan Atomic Energy Research Institute,
Takasaki. The cells were irradiated with protons which
energies of 0.05, 0.1, 0.2, 0.38, I , 3 and 10 MeV to study
radiation damage dependency on proton energy. The
degradation of the cells by proton irradiation is shown in
Fig. I . Thc cell performance did not degrade below 0.2
MeV proton energy, because the protons may stop at the

Poster

693

3rd World Confirence on Pholovolmic Energy Conversion

May 11-18, 2003 Osaka, Jupupan

B = Bd exp(-At)

(1)

where B is the degradation after annealing, Bd is the

degradation before annealing, A is the annealing rate and t
is time.
We conducted the thermal annealing experiments for
ClGS solar cells irradiated with 3 M e V proton in dark
conditions. The proton fluence was I x I O m i 2 . The
remaining factor of Isc and VOC is 85 % and 5 5 %,
respectively. The relation of the annealing rates of Isc and
VOCon the cells and temperature is shown in Fig. 3 (a) and
(b), respectively.

2.5

3.5

1OOOfl (K-)

IO

IO

(a) Short-CircuitCurrent

IO

Proton Ftuence (cm.)

(h) Open-circuit Voltage

Fig. 1 Remaining factor or isc ana VOC on the
ClGS solar cells irradiated with 0.05, 0.1, 0.2, 0.38,
I , 3 and I O MeV protons.

-m21
::10
>

I , ,

YI

ZnO window layer on the ClGS absorption layer. The

RDC of the performance on the cells is derived from Fig. 1
The RDC of the short-circuit current (Isc) and the open
circuit-voltage (Voc) is derived from the proton fluence of
10 % degradation compared with that in the case of I O
MeV proton. The RDC is shown in Fig.2.

25

35

IOOOiT (K-)

(b) Opcn-Circuit Voltage

Annealing rates of Isc and VOC of the cells

irradiated with 3 MeV proton. The thermal annealing
was conducted in dark conditions.
Fig.3

3.3 I n t r i n s i c D e g r a d a t i o n on CIGS S o l a r C e l l s by
P r o t o n s a t Low Temperature

0. I
I
10
Proton Energy (MeV)
Fig. 2 Relative damage coefficient of the ClGS solar
cells for protons derived from Fig. 1 .
0.01

~~

3.2 A n n e a l i n g

Rates of Vac and Isc of

P r o t o n - I r r a d i a t e d CIGS S o l a r C e l l s
The annealing rate of irradiated ClGS solar cells is
expressed in the following equation:

694

Poster

We performed the proton irradiation test for CIGS

solar cells at low temperature to verify the intrinsic
degradation rate without thermal annealing. The cells were
cooled by a cry0 generator system in order to avoid
thermal annealing effect. The lowest temperature of the
sample without proton and light irradiation was below 200
K. Figure 4 illustrates the change of remaining factor of
Vac and Isc after irradiation by 10 MeV protons. We
measured the performance of the cells irradiated with
IOMeV protons for each fluence level. ?he temperature of
the cells was 250K to 270K during the irradiation test. The
recovery rate of the cell performance is less than I x 10. s-
according to the data in Fig. 3, and the experiment time

3rd World Conference on Phorovolmic Erterm Conversion

May 11-18,2003 Osaka,Jopon

Tab. 1 Degradation rate of ClGS solar cells irradiated

with 10 MeV protons.

Isc
Voc

$0

1x10"
1x10'2

0.15

1.1

(cm.7

0.03

0
L.

I 1

10
100
IO00
MET (days afler hunch)

10000

(a) Short-circuit Current

10 MeV Proton Fhmce (cm~')

(b) Open-circuit Voltage
Fig. 4 Remaining factor of Isc and VOC on the ClGS
solar cclls irradiated with IOMeV protons. The remaining
factor of Isc and VOCis measured at low temperature. The
solid lines are derived from Eq. (2).

was 5 hours. Therefore, the recovery of proton

induced-damage in the solar cells can he neglected because
it is lcss than 1 %.
The solar cell electrical performance degradation can
be expressed as follows:

D=D,-Clog

I+-

4 3

where D is the performance after irradiation, Do is the

initial performance, C is the constant value, $O is the fitting
parameter and the Q is radiation fluence. C and OO of the
cells with 10 MeV proton are derived from the fitting to
the experimental result using equation (2). Table I
indicates C and &, of Isc and Voc.

3.4 Validation of the Model by Observing the

Performance on the CIGS Cells in the Satellite
Thc model is validated through accurate prediction of
data from the satellite.
The on-orbit cell performance recovery of solar cells
due to thermal annealing can be described as follows:

where rp is the fluence rate of equivalent 10 MeV protons

derived from the RDC and the radiation environment in
space.
Mission Demonstration Satellite No. 1 (MDS-I) was
launched on February 4, 2002, by H2A rocket. The

IO

IO0

1000

10000

MET (days after launch)

(b) Open-circuit Voltage
Fig. 5 Remaining factor of Isc and Voc of the ClGS solar
cells with IOOpm cover-glass by the experiment on the
satellite and the prediction model.
satcllite was placed in a geostationary transfer orbit (GTO)
with an apogee, and a pcrigee, an inclination angle and a
period of 500km. 36000!un, 28.5"and about 11 hours,
respectively. The orbit cuts across the inncr and the outer
Van Allen Belts, so the satellite is exposed to a vcry severe
radiation environment. The satellite is equipped with a test
board composed o f six kinds of terrestrial solar cells
including the ClGS solar cells. Two of each type of cell
was prepared, and 500 pni and 100pm thick cover glasses
wcre atlached to them. The performance of the cells was
monitored [9].
The equivalent fluence rate of lOMeV protons
on-orbit is derived from the RDC of thc ClGS solar cells
and the radiation environment calculated by AP-8 model of
NASA Glenn Research Center [IO]. The equivalent
fluence rate for Isc and Vac of the cells is 2.lxlO' llcm21s
and 6.5* IOs Ilcmzlls, respectively. The average temperature
of the satellite is 7 0 T . The degradation of Isc and Voc of
the cells on the satellite is predicted both with and without
thermal annealing. The result is shown in Fig. 5.
The figure shows that the degradation of Isc and Voc
on the CIGS solar cells without recovery is 5% and 20 %
about 1 year after launch, respectively. In contrast, neither
Isc nor VOCis degraded with recovery. This result agrees

Poster

695

3rd World Conference on PhotovoNnic Energv Conversion

May 11-18. 2003

with the data of the experimental CIGS solar cells on the

satellite.

Osoka.Japan

Technology for many helpful suggcstions during this study.

3.5 Prediction of the Cell Performance on Orbit

In this section, we analyze the prediction of on-orbit
performance of CIGS solar cells using the RDC, the
annealing rate and intrinsic degradation coefficients of Isc
and VOC.
We calculated the predicted result for the cells in
LEO and GEO. An LEO altitude of 600 km and the
inclination angle of 98" were adopted for the prediction.
The temperature of the satellites is estimated to he 60C.
From Fig. 3, the annealing rate of Isc and VOCof the cells
~ ~and 2 ~ 1 0 - ~ s - ' ,
irradiated with protons at 60C is I X I O s-'
respectively. The radiation environment of the orbits is
calculated using the AP-8. The equivalent fluence rate of
lOMeV proton is derived from the RDC of the ClGS solar
cell and the radiation environment on orbit. The cells were
not attached with cover glasses.
Figure 6 shows the predicted Isc and VOCof the CIGS
solar cells on orbit with and without thermal annealing. For
LEO, the degradation of Isc and VOC without thermal
annealing 5 years after launch is 5 % and 15 %,
respectively. On the other hand, Isc and VOCdo not degrade
with recovery. In addition, Isc and VOC of cells in CEO
degraded 7 % and 45 % without recovery In contrast, the
cell performance is not degraded by thermal annealing.
This result suggests that the performance of ClGS solar
cells without a cover glass does not degrade by radiation in
space.

4. CONCLUSION
We investigated the on-orbit performance of ClGS
solar cells with thermal annealing. The ClGS solar cells
were irradiated with high-energy protons for the radiation
response. We measured the RDC of proton, the thermal
annealing rate of Isc and VOCof the proton irradiated cells
and the intrinsic degradation coeffiicicnts ofthe cells for 10
MeV proton. The thermal annealing rate of the cells
exceeds that of other types of solar cells.
A validation of thermal annealing on orbit is provided
through accurate prediction of data from the cxpcriments
of the MDS-I satellite. Consequently, the performance is
not degraded by thermal annealing. This result is in good
agreement with the flight data collected from cells on the
MDS-I satellite. Sincc the ClGS solar cells continuously
recover due to thermal annealing in space, we assume that
radiation degradation and recovery of ClGS solar cells
occur simultaneously in space. We predicted the
performance of ClGS solar cells in LEO and GEO using
the results and found that cell. Performance was not
dcgradcd at all.
We plan to demonstrate the cell performance of CIGS
solar cells without cover-glasses to validate of the model
by a small satellite in LEO.

5. ACKNOWLEDGEMETS
The authors wish to thank Dr. T. Negami of
Matsushita Electric Industrial CO., LTD., Dr. S. Niki of
AIST, Professor T. Wada of Ryukoku University, and
Professor A. Yoshida of Toyohashi University of

696

Poste1

1% and VOCwith a m a h s

.p.95

oc without a m a

0.9
0.85
,

0.8

, ,.

10

100

1000

10000

ME? (days afler launch)

(a) LEO
I.2

0.2

.,...

Isc and Vcc with ameahg

....

. , , . , ,I

10
100
1000
MET (days afler bunch)
(b) CiEO

I
10000

Fig. 6 Remaining factor of the performance of the

ClGS solar cell in LEO and GEO.

REFERENCES
[ I ] M.A. Contreras, B. Egaas, K. Ramanathan, J. Hiltner,
A. Swartzlander, E Hasoon, R. Noufi: Prog. Photovolt.
7 (1999) 31 I.
[2] 7. Aburaya , T. Hisamatsu and S. Matsuda, Second
WCPEC, 1998, pp.3568-3571
[3] S. Wideman, M. E. Beck, R. Butchcr, 1. Repins, N.
Gomer, B. Joshi, R. G. Wcndt and 1. S. Britt,
Twenty-ninth IEEE PVSC, 2002, pp.575-578
[4] J. R. Tuttle, A. Sralaj and J. Keane, Twenty-eighth
IEEE PVSC, 2000, pp. 1042-1045.
[ 5 ] A. Bohen, D. Briunig, J. Klaser, F. H. Kang, B.
Hosselharth and G. La Roche, Twenty-eighth lEEE
PVSC, 2000, pp.1038-1041
[6] S. Kawakita, M. Imaimmi, M. Yamaguchi, K. Kushiya,
T. Ohshima, H. Ito, and S. Matsuda, lpn. J. Appl. Phys.,
41 (2002), p. L797
[7] K. Kushiya, M. Tachiyuki, T. Kase, Y. Nagoya, I .
Sugiyama, 0. Yamase and H. Takeshita: Sol. Energy.
Mater. Sol. Cells 49(1997) 277.
[SI K. Kushiya, M. Tachiyuki, Y. Nagoya, A. Fujimaki, B.
Sang, D. Okumura, M. Satoh and 0. Yamase: Sol.
Energy Mater. Sol. Cells 61 (2001) 11
[9] S. Matsuda, M. Imaizumi, S. Kawakita, S. Taishi and K.
Aoyama, this proceeding.
[IO] http://nssde.gsfc.nasa.gov/space/model/

3rd World Conference on Phorovolraic Enrrgv Conversion

Muy 11-18, 2003 Osoh, Japan

3P-B5-I 1

SUPER RADIATION TOLERANCE OF CIGS SOLAR CELLS DEMONSTRATED IN

SPACE BY MDS-1 SATELLITE
Shirou Kawakita'.*, Mitsuru Imaizumi', Taishi Sumita', Katsumi Kushiya', Takeshi Ohshima', Masafumi Yamaguchi',
Sumio Matsuda', Shinichi Yoda'.' and Tomihiro Kamiya4
1. National Space Development Agency of Japan (NASDA), 2-1-1Sengen, Tsukuba, lbaraki 305-8505, Japan
2.Tokyo Institute of Technology, 4259 Nagatsuta-cho, Midori-ku, Yokoyama, kanagawa 226-8503, Japan
3. Showa Shell Sekiyu K.K, Central R&D Lab. (ARL), 123-1 Shimo-kawairi, Atsugi, Kanagawa 243-0206, Japan
4. Japan Atomic Energy Research Institute, 1233 Watanuki, Takasaki, Gumma 370-1292, Japan
5. Toyota Technological Institute, 2-12-1 Hisakata, Tempaku-ku, Nagoya, Aichi 468-85 1 I , Japan
*E-mail: kawakita.shirou@nasda.go.jp

ABSTRACT
High radiation tolerance of ClGS solar cells is
demonstrated in space for the first timc by monitoring the
performance of Cu(ln,Ga)Scz thin-film solar cells on
MDS-I since February 2002. Short-circuit current of the
ClGS cells did not degrade and open-circuit voltage of the
cells degraded only about 1 %. In contrast, the
performance of other solar cells on the satellite, including
Si and GaAs space solar cclls, degraded I ycar after launch.
The high recovery of radiation damage of CIGS solar cells
due to thermal annealing was found from ground test. We
prcdicted the degradation of the ClGS solar cells in space
using the relative degradation coefficient, the annealing
rates of Voc and Isc for protons irradiating the cells and the
radiation response of the cells without thermal annealing.
The results are in good agreement with flight data of the
CIGS solar cells on the satellite. We were thus able to
demonstrate high radiation tolerance of ClGS solar cells in
space for the first time.

1. INTRODUCTION
A copper indium gallium di-selenide (CIGS) thin-film
solar cell is a promising candidate for future thin-film
space solar cells since the cells havc demonstrated
conversion efficiencies exceeding 18%, which is
significantly higher than other thin-film solar cells [I].
Excellent radiation tolerance of CIGS thin-film cells has
also been reported [2]. In addition, CIGS thin-film solar
cells have better performance than conventional Si and
GaAs space solar cells, such as low cost and light weight,
because the cells can be formed on polyimide [ 3 ] or
stainless steel sheet [4] substrates.
Radiation damage studies for CIGS thin-film solar
cells have revealed that (I) electrical properties of the cells
are not degraded by high-energy electron irradiation.
However, cell performance is degraded by high-energy
proton irradiation similar to other types of solar cells [2],
and (2) radiation damage of the cells due to proton
irradiation gradually recovers when irradiated cells are
kept at room temperaturc [SI. The recovery rate of the cells
depends on the temperature [6]. Since ClGS solar cells
recover well from radiation damage, a prediction of
on-orbit performance of ClGS solar cells must fake into

account the recovery of radiation damage by thermal

annealing.
In this study, we measured the relative damage
coefficicnt (RDC) for thc ClGS solar cclls to predict the
performance of the cells in space. In addition. we measured
the recovery rate of the proton-irradiated ClGS solar cells
by thermal annealing. Furthermorc, we derived the
intrinsic damage coefficient of ClGS solar cells from the
10 MeV proton test at low temperature. We apple the
prediction model with thermal annealing to the flight data
of the ClGS cells on thc satellite using by these results. In
addition, we forecast the cell performance in low-Earth
orbit (LEO) and geostationary orbit (CEO).

2. EXPERIMENT
ClGS thin-film solar cell samples (effective area of
0.6 cm') were separated from a 30cmx30cm CIGS
thin-film integrated mini-module. The circuit was
fabricated by a selenizationisulfurization method [ 7 ] , and
the stmcture of the cells is ZnO/Zn(O,S,OH)x/Cu(ln,Ga)Se, iMoiGlass [XI. The solar cells have an
average efficiency of 10.0%. open-circuit voltage (Voc) of
570 mV, short-circuit current density (Jsc) of 40 mAIcm',
and fill factor (FF) of 60%, which were measured under
AMO, I-sun conditions. Thc cclls have lower FF than
typical ClGS thin-film solar cells reported because of their
relatively high series resistance due to their contact
StNCture [ I ] .

3. RESULTS AND DISCUSSION

3.1 Relative Damage Coefficient of CIGS Solar
Cells
We performed proton irradiation with the ion
implanter, the tandem accelerator and the AVF cyclotron
accelerator at the Japan Atomic Energy Research Institute,
Takasaki. The cells were irradiated with protons which
energies of 0.05, 0.1, 0.2, 0.38, I , 3 and 10 MeV to study
radiation damage dependency on proton energy. The
degradation of the cells by proton irradiation is shown in
Fig. I . Thc cell performance did not degrade below 0.2
MeV proton energy, because the protons may stop at the

Poster

693

3rd World Confirence on Pholovolmic Energy Conversion

May 11-18, 2003 Osaka, Jupupan

B = Bd exp(-At)

(1)

where B is the degradation after annealing, Bd is the

degradation before annealing, A is the annealing rate and t
is time.
We conducted the thermal annealing experiments for
ClGS solar cells irradiated with 3 M e V proton in dark
conditions. The proton fluence was I x I O m i 2 . The
remaining factor of Isc and VOC is 85 % and 5 5 %,
respectively. The relation of the annealing rates of Isc and
VOCon the cells and temperature is shown in Fig. 3 (a) and
(b), respectively.

2.5

3.5

1OOOfl (K-)

IO

IO

(a) Short-CircuitCurrent

IO

Proton Ftuence (cm.)

(h) Open-circuit Voltage

Fig. 1 Remaining factor or isc ana VOC on the
ClGS solar cells irradiated with 0.05, 0.1, 0.2, 0.38,
I , 3 and I O MeV protons.

-m21
::10
>

I , ,

YI

ZnO window layer on the ClGS absorption layer. The

RDC of the performance on the cells is derived from Fig. 1
The RDC of the short-circuit current (Isc) and the open
circuit-voltage (Voc) is derived from the proton fluence of
10 % degradation compared with that in the case of I O
MeV proton. The RDC is shown in Fig.2.

25

35

IOOOiT (K-)

(b) Opcn-Circuit Voltage

Annealing rates of Isc and VOC of the cells

irradiated with 3 MeV proton. The thermal annealing
was conducted in dark conditions.
Fig.3

3.3 I n t r i n s i c D e g r a d a t i o n on CIGS S o l a r C e l l s by
P r o t o n s a t Low Temperature

0. I
I
10
Proton Energy (MeV)
Fig. 2 Relative damage coefficient of the ClGS solar
cells for protons derived from Fig. 1 .
0.01

~~

3.2 A n n e a l i n g

Rates of Vac and Isc of

P r o t o n - I r r a d i a t e d CIGS S o l a r C e l l s
The annealing rate of irradiated ClGS solar cells is
expressed in the following equation:

694

Poster

We performed the proton irradiation test for CIGS

solar cells at low temperature to verify the intrinsic
degradation rate without thermal annealing. The cells were
cooled by a cry0 generator system in order to avoid
thermal annealing effect. The lowest temperature of the
sample without proton and light irradiation was below 200
K. Figure 4 illustrates the change of remaining factor of
Vac and Isc after irradiation by 10 MeV protons. We
measured the performance of the cells irradiated with
IOMeV protons for each fluence level. ?he temperature of
the cells was 250K to 270K during the irradiation test. The
recovery rate of the cell performance is less than I x 10. s-
according to the data in Fig. 3, and the experiment time

3rd World Conference on Phorovolmic Erterm Conversion

May 11-18,2003 Osaka,Jopon

Tab. 1 Degradation rate of ClGS solar cells irradiated

with 10 MeV protons.

Isc
Voc

$0

1x10"
1x10'2

0.15

1.1

(cm.7

0.03

0
L.

I 1

10
100
IO00
MET (days afler hunch)

10000

(a) Short-circuit Current

10 MeV Proton Fhmce (cm~')

(b) Open-circuit Voltage
Fig. 4 Remaining factor of Isc and VOC on the ClGS
solar cclls irradiated with IOMeV protons. The remaining
factor of Isc and VOCis measured at low temperature. The
solid lines are derived from Eq. (2).

was 5 hours. Therefore, the recovery of proton

induced-damage in the solar cells can he neglected because
it is lcss than 1 %.
The solar cell electrical performance degradation can
be expressed as follows:

D=D,-Clog

I+-

4 3

where D is the performance after irradiation, Do is the

initial performance, C is the constant value, $O is the fitting
parameter and the Q is radiation fluence. C and OO of the
cells with 10 MeV proton are derived from the fitting to
the experimental result using equation (2). Table I
indicates C and &, of Isc and Voc.

3.4 Validation of the Model by Observing the

Performance on the CIGS Cells in the Satellite
Thc model is validated through accurate prediction of
data from the satellite.
The on-orbit cell performance recovery of solar cells
due to thermal annealing can be described as follows:

where rp is the fluence rate of equivalent 10 MeV protons

derived from the RDC and the radiation environment in
space.
Mission Demonstration Satellite No. 1 (MDS-I) was
launched on February 4, 2002, by H2A rocket. The

IO

IO0

1000

10000

MET (days after launch)

(b) Open-circuit Voltage
Fig. 5 Remaining factor of Isc and Voc of the ClGS solar
cells with IOOpm cover-glass by the experiment on the
satellite and the prediction model.
satcllite was placed in a geostationary transfer orbit (GTO)
with an apogee, and a pcrigee, an inclination angle and a
period of 500km. 36000!un, 28.5"and about 11 hours,
respectively. The orbit cuts across the inncr and the outer
Van Allen Belts, so the satellite is exposed to a vcry severe
radiation environment. The satellite is equipped with a test
board composed o f six kinds of terrestrial solar cells
including the ClGS solar cells. Two of each type of cell
was prepared, and 500 pni and 100pm thick cover glasses
wcre atlached to them. The performance of the cells was
monitored [9].
The equivalent fluence rate of lOMeV protons
on-orbit is derived from the RDC of thc ClGS solar cells
and the radiation environment calculated by AP-8 model of
NASA Glenn Research Center [IO]. The equivalent
fluence rate for Isc and Vac of the cells is 2.lxlO' llcm21s
and 6.5* IOs Ilcmzlls, respectively. The average temperature
of the satellite is 7 0 T . The degradation of Isc and Voc of
the cells on the satellite is predicted both with and without
thermal annealing. The result is shown in Fig. 5.
The figure shows that the degradation of Isc and Voc
on the CIGS solar cells without recovery is 5% and 20 %
about 1 year after launch, respectively. In contrast, neither
Isc nor VOCis degraded with recovery. This result agrees

Poster

695

3rd World Conference on PhotovoNnic Energv Conversion

May 11-18. 2003

with the data of the experimental CIGS solar cells on the

satellite.

Osoka.Japan

Technology for many helpful suggcstions during this study.

3.5 Prediction of the Cell Performance on Orbit

In this section, we analyze the prediction of on-orbit
performance of CIGS solar cells using the RDC, the
annealing rate and intrinsic degradation coefficients of Isc
and VOC.
We calculated the predicted result for the cells in
LEO and GEO. An LEO altitude of 600 km and the
inclination angle of 98" were adopted for the prediction.
The temperature of the satellites is estimated to he 60C.
From Fig. 3, the annealing rate of Isc and VOCof the cells
~ ~and 2 ~ 1 0 - ~ s - ' ,
irradiated with protons at 60C is I X I O s-'
respectively. The radiation environment of the orbits is
calculated using the AP-8. The equivalent fluence rate of
lOMeV proton is derived from the RDC of the ClGS solar
cell and the radiation environment on orbit. The cells were
not attached with cover glasses.
Figure 6 shows the predicted Isc and VOCof the CIGS
solar cells on orbit with and without thermal annealing. For
LEO, the degradation of Isc and VOC without thermal
annealing 5 years after launch is 5 % and 15 %,
respectively. On the other hand, Isc and VOCdo not degrade
with recovery. In addition, Isc and VOC of cells in CEO
degraded 7 % and 45 % without recovery In contrast, the
cell performance is not degraded by thermal annealing.
This result suggests that the performance of ClGS solar
cells without a cover glass does not degrade by radiation in
space.

4. CONCLUSION
We investigated the on-orbit performance of ClGS
solar cells with thermal annealing. The ClGS solar cells
were irradiated with high-energy protons for the radiation
response. We measured the RDC of proton, the thermal
annealing rate of Isc and VOCof the proton irradiated cells
and the intrinsic degradation coeffiicicnts ofthe cells for 10
MeV proton. The thermal annealing rate of the cells
exceeds that of other types of solar cells.
A validation of thermal annealing on orbit is provided
through accurate prediction of data from the cxpcriments
of the MDS-I satellite. Consequently, the performance is
not degraded by thermal annealing. This result is in good
agreement with the flight data collected from cells on the
MDS-I satellite. Sincc the ClGS solar cells continuously
recover due to thermal annealing in space, we assume that
radiation degradation and recovery of ClGS solar cells
occur simultaneously in space. We predicted the
performance of ClGS solar cells in LEO and GEO using
the results and found that cell. Performance was not
dcgradcd at all.
We plan to demonstrate the cell performance of CIGS
solar cells without cover-glasses to validate of the model
by a small satellite in LEO.

5. ACKNOWLEDGEMETS
The authors wish to thank Dr. T. Negami of
Matsushita Electric Industrial CO., LTD., Dr. S. Niki of
AIST, Professor T. Wada of Ryukoku University, and
Professor A. Yoshida of Toyohashi University of

696

Poste1

1% and VOCwith a m a h s

.p.95

oc without a m a

0.9
0.85
,

0.8

, ,.

10

100

1000

10000

ME? (days afler launch)

(a) LEO
I.2

0.2

.,...

Isc and Vcc with ameahg

....

. , , . , ,I

10
100
1000
MET (days afler bunch)
(b) CiEO

I
10000

Fig. 6 Remaining factor of the performance of the

ClGS solar cell in LEO and GEO.

REFERENCES
[ I ] M.A. Contreras, B. Egaas, K. Ramanathan, J. Hiltner,
A. Swartzlander, E Hasoon, R. Noufi: Prog. Photovolt.
7 (1999) 31 I.
[2] 7. Aburaya , T. Hisamatsu and S. Matsuda, Second
WCPEC, 1998, pp.3568-3571
[3] S. Wideman, M. E. Beck, R. Butchcr, 1. Repins, N.
Gomer, B. Joshi, R. G. Wcndt and 1. S. Britt,
Twenty-ninth IEEE PVSC, 2002, pp.575-578
[4] J. R. Tuttle, A. Sralaj and J. Keane, Twenty-eighth
IEEE PVSC, 2000, pp. 1042-1045.
[ 5 ] A. Bohen, D. Briunig, J. Klaser, F. H. Kang, B.
Hosselharth and G. La Roche, Twenty-eighth lEEE
PVSC, 2000, pp.1038-1041
[6] S. Kawakita, M. Imaimmi, M. Yamaguchi, K. Kushiya,
T. Ohshima, H. Ito, and S. Matsuda, lpn. J. Appl. Phys.,
41 (2002), p. L797
[7] K. Kushiya, M. Tachiyuki, T. Kase, Y. Nagoya, I .
Sugiyama, 0. Yamase and H. Takeshita: Sol. Energy.
Mater. Sol. Cells 49(1997) 277.
[SI K. Kushiya, M. Tachiyuki, Y. Nagoya, A. Fujimaki, B.
Sang, D. Okumura, M. Satoh and 0. Yamase: Sol.
Energy Mater. Sol. Cells 61 (2001) 11
[9] S. Matsuda, M. Imaizumi, S. Kawakita, S. Taishi and K.
Aoyama, this proceeding.
[IO] http://nssde.gsfc.nasa.gov/space/model/

2186

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 51, NO. 12, DECEMBER 2004

Laser-Grooved Backside Contact

Solar Cells With 680-mV
Open-Circuit Voltage
Jiun-Hua Guo and Jeffrey E. Cotter

AbstractIn this paper, we demonstrate for the first time the

use of the laser-grooved solar cell technology, a proved commercial technology, for the implementation of the rear junction backside contact solar cells. Laser-grooved backside contact solar cells,
designated as interdigitated backside buried contact (IBBC) solar
cm wafers with a
cells, have been fabricated on planar, n-type, 1
single layer SiO2 anti-reflection coating, achieving 17% efficiency
with open-circuit voltage oc of more than 680 mV. Front and
rear surface recombination velocities of 350 cm/s and 4800 cm/s,
and more than 1 ms of post-processing bulk lifetime confirm that
commercial laser-grooved solar cell fabrication process is capable
of obtaining the efficiency advantages of the rear junction, backside
contact design. Moreover, this paper presents a side-by-side comparison between the more conventional double-sided buried contact solar cell and the IBBC solar cell. The advantages of higher
short-circuit current in the latter design due to no contact shading
loss, and higher oc due to inherently lower surface recombination
velocity of the IBBC structure are demonstrated.

( )

Index TermsBackside contact, laser grooved, n-type solar cells,

silicon solar cells.

I. INTRODUCTION

ACKSIDE contact solar cells have shown outstanding performance under both one-sun and concentrated illumination. Perhaps the most notable backside contact solar cells are
point contact solar cells [1], which have reached conversion efficiencies of 28.4% at 200-suns [2] and 22.3% at one-sun [3]
conditions. The point contact solar cell is similar to the interdigitated backside contact solar cell in that both electrical contacts
are on the back side [4][6]; however, there is a difference since
its contact metal only touches the silicon in an array of points.
The major advantage of restricting the contact coverage to small
points is that it significantly increases the cell output voltage.
These advantages, combined with the use of a high-resistivity
float-zone silicon substrate, and high-quality SiO passivation
on all surfaces, contribute to the high efficiency of the point contact solar cell. However, the rear contact side of point contact
solar cells often requires four to six mask levels to fabricate [7].
The advantages of the backside contact design include no
contact grid shading on the sunward side and potentially low
Manuscript received April 21, 2004; revised September 15, 2004. This work
was supported by the Australian Research Council, through The Center of Excellence for Advanced Silicon Photovoltaics and Photonics, University of New
South Wales. The review of this paper was arranged by P. Panayotatos.
The authors are with the Center of Excellence for Advanced Silicon Photovoltaics and Photonics, University of New South Wales, Sydney, NSW 2052,
Australia (e-mail: jiun-hua.guo@student.unsw.edu.au; j.cotter@unsw.edu.au).
Digital Object Identifier 10.1109/TED.2004.839870

wafer cost due to the use of a thin wafer, although high diffusion length and low front-surface recombination are required
to maintain high charge carrier collection needed for high efficiency. The contact side can be optimized for series resistance
whereas the sunward side can be optimized for light collection without the need to compromise. Moreover, the module assembly costs are potentially lower as both contacts are on the
rear, and the solar cells can be placed closer together in the
module because there is no need for a space between the cells
for interconnect tabs [8].
Significant cost reduction while maintaining the above-mentioned advantages have been the focus of research on the backside contact solar cells. The Pegasus cell, a simplified backside
point contact solar cell, achieved 21.9% AM1.5G front illumination efficiency with a reduced number of photolithography
masking steps from six to three [9], [10]. The trench-mesa type
of backside contact solar cell demonstrated 21.9% efficiency
with only two photolithographic steps [11].
Several methods have been utilized to fabricate interdigitated
backside contact solar cells commercially. Robert et al. used
the screen-printing technology and found that the cell performance was hindered by the limitations on the contact metallization geometry imposed by the screen-printing process [12].
EBARA Solar used a screen-printing process that incorporated
self-doping contacts, demonstrating an efficiency of 10.4% on
dendritic web silicon substrates [13]. The Back-OECO cell presented by ISFH has achieved a preliminary result of more than
18% efficiency [14]. More recently, SunPower demonstrated an
outstanding backside contact solar cell result with 20.4% efficiency by utilizing low-cost processing without photolithography [15].
Three designs of the backside contact cell have been developed for the implementation on lower lifetime substrates: 1)
emitter wrap-through (EWT) solar cell [16], [17]; 2) metallization wrap through (MWT) solar cell [18], [19]; and 3) metallization wrap around solar cell (MWA) [20]. The EWT cell uses
laser-drilled holes to wrap the emitter on the front surface to the
interdigitated contacts on the back surface. The MWT cell has
front collecting junction and front contact grid with the busbar
brought to the backside through laser-drilled vias in each front
contact finger, and the front grid of the MWA cell is connected
to the backside contact pads via two opposite edges of the cell.
A 41 cm screen-printed EWT cell with 15.3% efficiency under
AM 1.5G illumination has been demonstrated [21]. Efficiency
of more than 17% has been achieved by both the 25 cm MWT
and the MWT solar cells using solar grade CZ wafers [19].

0018-9383/04$20.00 2004 IEEE

GUO AND COTTER: LASER-GROOVED BACKSIDE CONTACT SOLAR CELLS

2187

TABLE I
PROCESSES FOR IBBC SOLAR CELLS

In this paper, we demonstrate the use of laser-buried grid

technology, currently utilized in high-volume commercial production, to fabricate and compare two solar cell designs-rear
junction, interdigitated backside buried contact (IBBC), and the
more conventional double-sided buried contact (DSBC) solar
cells. The fabrication process for these device designs on planar,
cm silicon FZ wafers is described. The impact of
n-type, 1
the bulk and the surface recombination on the IBBC solar cells
are also addressed to evaluate the limitations and potential of
the solar cell performance when commercial silicon wafers are
considered.
II. FABRICATION AND DESIGN
Schematic diagrams of IBBC solar cell and DSBC solar cell
are shown in Fig. 1. Fabricated solar cells are 8 cm , 175 m
-type float-zoned silicon
thick and are fabricated on
wafers. The process steps are summarized in Table I.
In the IBBC device, an n -n high-low junction is formed on
the sunward side by means of phosphorus diffusion. This phosphorus-doped layer provides front-surface-field passivation and
also ensures a good UV stability [22]. The p emitter on the
rear surface is formed by the boron liquid-source (boron tribromide) diffusion. Compared to the aluminum alloying and solid
diffuboron nitride sources, the p-n junction formed by
[23], and
sion has lower emitter saturation current density
can maintain high bulk lifetime [24]. A thick SiO grown by
a combination of dry and wet oxidation is used as the masking
oxide for subsequent heavy n and p diffusions in the laserand p
grooves are formed
scribed groove regions. The n
by laser scribing followed by phosphorus solid source and boron
liquid source diffusion. Groove doping and drive-in time are
critical since regions where phosphorus groove diffusion and
boron emitter diffusion overlap are one of the shunt routes in the

Fig. 1. Schematic diagram of (a) IBBC and (b) DSBC solar cells (not to scale).

IBBC structure. Longer groove diffusion drive-in time is preferable because it not only pushes the metallurgical junction far
away from groove walls, preventing nickel contact metal spiking
through the groove diffusion, but also increases the width of
the potential energy barrier to prevent junction tunneling [25].

2188

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 51, NO. 12, DECEMBER 2004

TABLE II
TYPICAL IMPLIED V OF DSBC AND IBBC SOLAR CELLS
GROOVE FORMATION, AND
AFTER OXIDATION, n
GROOVE FORMATION FABRICATION STEPS
p

Groove deglazing presents another important step in the process

that requires careful control over oxide deglazing [26].
DSBC devices are fabricated using the identical process sequence and conditions, except that the base-layer contacts of the
DSBC device are located within an n diffused layer on the rear
side and the collecting p -n junction and its metal contacts are
located on the front side, as shown in Fig. 1(b). In this case, the
rear n diffused layer acts as a back surface field (BSF). In addition, locating electrical contacts on the opposite faces of the
solar cell avoids many of the problems associated with the diffusion overlap regions of the IBBC device.
To ensure that there is no bulk lifetime degradation due to
contamination, photoconductance (PC) measurements [27],
[28] are applied as a process-monitoring tool after oxidation,
phosphorus groove diffusion, and boron groove diffusion in
the solar cells fabrication. The quasi-steady-state PC method
gives important information about how Shockley-Read-Hall
recombination and saturation currents of the diffused regions
affect the effective lifetime of the solar cells after each processing step. More detailed interpretation of the injection level
dependant lifetime spectroscopy can be found elsewhere [29].
Typical implied open-circuit voltage values [30], [31], extracted
from QSSPC measurements after the above-mentioned steps
for IBBC and DSBC solar cells (see Table II), indicate that high
bulk lifetime and low surface recombination are maintained in
the laser-groove DSBC and IBBC process.
The grid design of the IBBC device is especially important
for high efficiency. In the interdigitated comb design of the original IBC solar cells [4], the area between the n-type busbar and
the cell edge is largely inactive where the light-generated carrier collection probability is low. Further, in the case where the
two perimeter fingers form a base contact, the perimeter area becomes electrical isolated from the emitter contact, in which case
there is a total of three inactive perimeter areas around the solar
cell, resulting in the greatly reduced short-circuit current den. The contact design shown in Fig. 2 eliminates nearly
sity
all of the inactive areas by extending the p-type emitter contact
of this contact
around the n-type busbar. By comparison,
design is about 32.8 mA/cm , compared to 31.9 mA/cm with
one inactive perimeter area and 30.8 mA/cm with three inactive
perimeter areas. This value is close to 33 mA/cm as predicted
by PC-1D [32] simulation and reported in [9].
Alignment with laser lithography can be accurate to around
1 m; however, the optical spot size (around 2040 m) and
the plating metallization scheme limit the minimum achievable
finger-to-finger spacing to about 250 m in practice. In order
to obtain the maximum light-generated carrier collection in this
case, it is essential to use a planar diffused emitter on the back
surface in the IBBC solar cell design. Typically, IBBC solar

Fig. 2. Schematic diagram of the rear electrode pattern showing the two
interdigitated electrodes (not to scale).

cells fabricated with a planar diffused emitter show short-circuit

current densities that are 6 to 7 mA/cm higher than designs that
rely solely on the p diffused grooves for current collection.
is 6 to 8 mV lower with
Although open-circuit voltage
the planar diffused emitter included, the current gain outweighs
, with an absolute efficiency difference of more
the loss in
than 3%.
III. RESULTS AND ANALYSIS
The results of a split batch of DSBC and IBBC solar
cells are listed in Table III, which were measured at UNSW
under standard testing conditions (one-sun illumination, 25 C).
The 8 cm solar cell has been cleaved from its wafer, so the effect of the edge recombination loss is included in the measurement [33]. The front and rear illumination definitions are reverse
to those of IBBC solar cells since DSBC device is designated as
collecting-junction illuminated. The IBBC solar cells show ex, and efficiencies of 17.0% and 15.7% are achieved
cellent
under front and rear illumination, respectively. The DSBC solar
cells show lower performance compared to that of IBBC solar
cells; however, any given batch of the DSBC solar cells tend to
have a narrower distribution of the electrical characteristics.
Fig. 3 shows reflectance, external quantum efficiency (EQE)
and internal quantum efficiency (IQE) of an IBBC solar cell and
a DSBC solar cell under front and rear illumination. The near
unity of IQE for short and long wavelength light under front
illumination clearly demonstrates that the fabrication process
of the laser-grooved solar cells is capable of maintaining excellent post-processing lifetime. The reduction of IQE beyond
950-nm wavelength suggests that light trapping reflector, such
as pigmented dielectric reflector (PDR), can be added to enhance the rear surface reflectance [34]. The effective front and
rear surface recombination velocities of the IBBC device extracted by PC-1D simulation are 350 and 4800 cm/s; and 4200
and 2500 cm/s for the DSBC device. These values are excellent
considering there is no aluminum and forming gas anneals, nor
has hydrogenation treatment been applied to further improve the
surface passivation.

GUO AND COTTER: LASER-GROOVED BACKSIDE CONTACT SOLAR CELLS

2189

TABLE III
RESULTS OF A SPLIT BATCH OF DSBC AND IBBC OLAR CELLS

3 Solar cells are 8 cm and have been cleaved from their host wafers.

Fig. 4. Effect of the wafer thickness on the performance of IBBC solar cells.

high post-processing lifetime required to achieve high efficiency

is maintained for the
in the backside contact design. That
thinnest devices confirms the good quality of passivation provided by the diffused phosphorus front surface field. The simulation also reveals that the optimal thickness for IBBC solar cell
is between 175 and 195 m for the particular processing and
device design used for these solar cells.
IV. COMPARISON BETWEEN IBBC AND DSBC SOLAR CELLS
Fig. 3. Reflectance, external quantum efficiency (EQE), and internal quantum
efficiency (IQE) of (a) an IBBC solar cell and (b) a DSBC solar cell under front
and rear illumination.

The summary of a batch of IBBC solar cells fabricated

with wafer thickness ranging from 150 m to 230 m is shown
in Fig. 4. If the process and the device parameters, such as
the doping profiles of the diffused layers, surface reflectance,
and front and rear surface recombination velocities can be assumed to be the same across the solar cells, then the wafer thickness, or alternately the bulk recombination, will be the main
. The lines decorated with open symbols
factor affecting
in Fig. 4 were generated from PC-1D using a 1-ms post-processing bulk lifetime, and were corrected for the proportion of
the inactive perimeter area associated with the finger pattern
confirms
that was used for these solar cells. The excellent
that the laser-grooved technology is capable of supporting the

of the DSBC
As shown in Table III, front illumination
solar cells is about 310 mV lower, and
is about 0.6 mA/cm
smaller compared to those of IBBC solar cells. It is found that
in DSBC solar cells mainly results from higher
smaller
front-surface recombination velocity of the boron diffused selective emitter. PC-1D simulation using the surface recombination velocities extracted in the previous section confirms this
as well. The DSBC solar cells exhibit
of
difference in
32.2 mA/cm due to contact shading loss (approximate 2.5%)
compared to the value obtained from the simulation of the IBBC
solar cells under front illumination (33 mA/cm ). The main reathan front illumination
sons for smaller rear illumination
in DSBC solar cells are: 1) the collecting junction is located
on the front surface and 2) the contact shading of the rear base
contacts is larger. The electroless nickel-plating rate is higher in
the phosphorus-doped grooves, and the subsequent electroless
copper plating is always thicker, leading to a naturally higher

2190

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 51, NO. 12, DECEMBER 2004

TABLE IV
CELL PARAMETERS USED FOR PC-1D SIMULATION

Fig. 5. Effect of the bulk lifetime on the performance of IBBC and DSBC solar
cells. Modeling parameters are listed in Table IV.

metal contact shading on the rear side contacts. Fill Factor in the
DSBC and IBBC devices is relatively low due to both shunting
and series resistance. The detailed cause of the parasitic effects
is currently under investigation.
High bulk lifetime and low front surface recombination velocity are essential for high efficiency in IBBC and DSBC solar
cells. While these conditions can be maintained in a laboratory,
it is much more challenging to do so in a commercial environment. To examine the impact of the bulk lifetime and the front
surface recombination on the performance of the IBBC and the
DSBC solar cells, we extended our experiment using the PC-1D
simulation based on measured and extracted device parameters.
Fig. 5 shows the influence of the wafer bulk lifetime on the
performance of the IBBC and the DSBC solar cells. Modeling
parameters are listed in Table IV. Here, it is clear that the backside contact design relies on high quality wafers to obtain the
benefit of no grid shading. In the case of low wafer lifetime, the
DSBC device is superior, as expected, owing to its front-side
collecting junction. Alternatively, a very thin wafer can be used
in order to maintain a high ratio of minority carrier diffusion
length to wafer thickness, although thin wafers tend to introduce higher mechanical yield losses.

Fig. 6. Effect of the front surface recombination velocity on the performance

of IBBC and DSBC solar cells. Modeling parameters are listed in Table IV.
The circles overlaid on the V curves indicate the extracted effective surface
recombination velocities of experimental devices.

Fig. 6 shows the effect of the front surface recombination velocity on the IBBC and the DSBC solar cells. Modeling parameters are listed in Table IV. It is expected that the decrease in
of the IBBC design with increasing front surface recombination is much more significant than that of the DSBC design.
However, one must keep in mind that the IBBC design has a
lower inherent front surface recombination velocity because it
uses a phosphorus diffused front surface field, which results in
a substantially lower effective surface recombination velocity
compared to the boron diffused emitter on the front surface of
the DSBC design [35], [36]. The overlaying circles on the
curves indicate the extracted effective surface recombination velocities of the experimental devices made in a split batch, and
this highlights the importance of the inherent differences in the
front surface recombination on the performance of the IBBC
and the DSBC designs.
Two important observations can be made from the simulation
results presented in Figs. 5 and 6. Firstly, the DSBC design is superior to the IBBC design in that it is more tolerant to variations
in wafer lifetime and in front surface recombination velocity,
which is a significant consideration for commercial production.

GUO AND COTTER: LASER-GROOVED BACKSIDE CONTACT SOLAR CELLS

However, the IBBC device is superior where high lifetime and

low front surface recombination velocity can be maintained.
Secondly, front and rear surface recombination velocities of
350 and 4800 cm/s are typical for thermal oxidations without
the hydrogenation treatment. Therefore, post-processing lifetime becomes the dominant factor when selecting the optimal
combination of wafer and device design. Commercial substrates, such as n-type Czochralski (CZ) and multicrystalline
wafers, can be suitable for the IBBC design if post-processing
can be maintained.
bulk lifetime of greater than about 300
However, the DSBC design is likely to be most suitable for
multicrystalline and boron-doped CZ wafers, owing to the naturally lower bulk lifetime of these wafers. Phosphorus-doped
CZ wafers, on the other hand, are commercially promising for
the IBBC design, since high post-processing lifetimes can be
maintained. For example, 21.1% efficiency has been achieved
on passivated emitter, rear totally diffused (PERT) solar cells
made on phosphorus-doped CZ wafers [37].
V. CONCLUSION
Laser-grooved DSBC and IBBC solar cells have commercial
potential. Here, we have demonstrated IBBC solar cells, fabricated on untextured, SLAR-coated, float-zoned silicon wafers,
with 17.0% front illumination efficiency and more than 680-mV
open-circuit voltage. The laser-grooved processing is shown to
be compatible with high post-processing bulk lifetime, demonstrating the carrier collection efficiency required for the backside contact design. The IBBC design is superior to the DSBC
and
on high-quality silicon
design in terms of both
wafers, due to smaller effective front surface recombination velocity and no grid shading loss. However, a choice between
IBBC and DSBC designs would be largely governed by wafer
bulk lifetime and the quality of the front surface passivation in
a commercial production environment.
We believe that improving the optical performance of the
IBBC design by front surface texturing and internal light trapping, both of which are compatible with commercial silicon
solar cell production requirements, will lead to significant performance improvement of the laser-grooved IBBC and DSBC
solar cells presented here. Further, we suggest that the lasergrooved IBBC process with some modifications is a commercially potential and cost effective technology for the fabrication
of the commercial backside contact solar cells.
ACKNOWLEDGMENT
The authors would like to thank M. Abbott for the collaboration of the fabrication of the IBBC solar cells presented in Fig. 4,
and K. Fisher for preparing the wafers for solar cell fabrication.
REFERENCES
[1] R. M. Swanson, Point-contact solar cells: Modeling and experiment,
Solar Cells, vol. 17, pp. 85118, 1986.
[2] R. A. Sinton, P. Verlinden, D. K. Kane, and R. M. Swanson, Development efforts in silicon backside-contact solar cells, in Proc. 8th E.C.
Photovoltaic Solar Energy Conf., 1988, pp. 14721475.
[3] R. R. King, R. A. Sinton, and R. M. Swanson, Doped surfaces in one
sun, point-contact solar cells, Appl. Phys. Lett., vol. 54, pp. 14601462,
1989.

2191

[4] M. D. Lammert and R. J. Schwartz, The interdigitated back contact

solar cell: A silicon solar cell for use in concentrated sunlight, IEEE
Trans. Electron Devices, vol. ED-24, pp. 337342, 1977.
[5] C. M. Garner, R. D. Nasby, and F. W. Sexton, An interdigitated back
contact solar cell with high-current collection, IEEE Electron Device
Lett., vol. EDL-1, pp. 256258, 1980.
[6] F. W. Sexton, C. M. Garner, and J. L. Rodriguez, Process for high
photocurrent in IBC solar cells, J. Electrochem. Soc., vol. 129, pp.
26242628, 1982.
[7] P. Verlinden, F. Van de Wiele, G. Stehelin, F. Floret, and J. P. David,
High efficiency interdigitated back contact silicon solar cells, in Proc.
19th IEEE Photovoltaic Specialists Conf., 1987, pp. 405410.
[8] A. Kress, O. Breitenstein, S. Glunz, P. Fath, G. P. Willeke, and E.
Buncher, Investigations on low-cost back-contact silicon solar cells,
Solar Energy Material and Solar Cells, vol. 65, pp. 555560, 2001.
[9] C. Z. Zhou, P. J. Verlinden, R. A. Crane, R. M. Swanson, and R. A.
Sinton, 21.9% efficient silicon bifacial solar cells, in Proc. 26th IEEE
Photovoltaic Specialists Conf., Anaheim, CA, 1997, pp. 287290.
[10] M. J. Cudzinovic and K. R. McIntosh, Process simplifications to the
Pegasus solar cell-sunpowers high-efficiency bifacial silicon solar cell,
in Proc. 29th IEEE Photovoltaic Specialists Conf., New Orleans, LA,
2002, pp. 7073.
[11] R. A. Sinton and R. M. Swanson, Simplified backside-contact solar
cells, IEEE Trans. Electron Devices, vol. 37, pp. 348352, Feb. 1990.
[12] S. Roberts, K. C. Heasman, T. M. Bruton, R. W. Russell, and D. W. Cunningham, Interdigitated-contact silicon solar cells made without photolithography, in Proc. 2nd World Conf. and Exhib. Photovoltaic Solar
Energy Conversion, Vienna, Austria, 1998, pp. 14491451.
[13] D. L. Meier, H. P. Davis, A. Shibata, T. Abe, K. Kinoshita, C. Bishop, S.
Mahajan, A. Rohatgi, P. Doshi, and M. Finnegan, Self-doping contacts
and associated silicon solar cell structures, in Proc. 2nd World Conf. and
Exhib. Photovoltaic Solar Energy Conversion, Vienna, Austria, 1998,
pp. 14911494.
[14] J. W. Mller, A. Merkle, and R. Hezel, Self-Aligning industrially
,
feasible back contacted silicon solar cells with efficiencies
in Proc. 3rd World Conf. on Photovoltaic Energy Conversion, Osaka,
Japan, 2003, pp. 14031406.
[15] K. R. McIntosh, M. J. Cudzinovic, D. D. Smith, W. P. Mulligan, and E.
M. Swanson, The choice of silicon wafer for the production of low-cost
rear-contact solar cells, in Proc. 3rd World Conf. Photovoltaic Solar
Energy Conversion, Osaka, Japan, 2003, pp. 971974.
[16] J. M. Gee, W. K. Schubert, and P. A. Basore, Emitter wrap-through solar
cell, in Proc. 23rd IEEE Photovoltaic Specialists Conf., Louisville, KY,
1993, pp. 265270.
[17] C. B. Honsberg, F. Yun, M. A. Green, and S. R. Wenham, Mechanically
grooved, multi-junction, interdigitated rear contact silicon solar cells for
low lifetime substrates, in Proc. 12th European Photovoltaic Solar Energy Conf., Amsterdam, The Netherlands, 1994, pp. 6366.
[18] E. V. Kerschaver, R. Einhaus, J. Szlufcik, J. Nijs, and R. Mertens, A
novel silicon solar cell structure with both external polarity contacts on
the back surface, in Proc. 2nd World Conf. and Exhib. Photovoltaic
Solar Energy Conversion, Vienna, Austria, 1998, pp. 14791482.
[19] W. Jooss, K. Blaschek, R. Tolle, T. M. Bruton, P. Fath, and E. Bucher,
17% back contact buried contact solar cells, in Proc. 16th European
Photovoltaic Solar Energy Conf., Glasgow, U.K., 2000, pp. 11241127.
[20] D. Michaels, N. Mendoza, and R. Williams, Large area, low cost space
solar cells with optional wraparound contacts, in Proc. 15th IEEEE
Photovoltaic Specialists Conf., 1981, pp. 225227.
[21] D. D. Smith and J. M. Gee, Review of back contact silicon solar cells
for low cost application, in Proc. 16th European Photovoltaic Solar
Energy Conf., Glasgow, U.K., 2000, pp. 11041107.
[22] P. E. Gruenbaum, R. R. King, and R. M. Swanson, Photoinjected hotelectron damage in silicon point-contact solar cells, J. Appl. Phys., vol.
66, pp. 61106114, 1989.
[23] A. Cuevas, C. Samundsett, M. J. Kerr, D. H. Macdonald, H. Mackel, and
P. P. Altermatt, Back junction solar cells on N-type multicrystalline and
CZ silicon wafers, in Proc. 3rd World Conf. Photovoltaic Solar Energy
Conversion, Osaka, Japan, 2003, pp. 936966.
[24] F. W. Chen, B. S. Tjahjono, J. H. Guo, P. J. Cousins, and J. E. Cotter,
Evaluation of boron solid source diffusion for high efficiency silicon
solar cells, in Proc. Solar 2003Australian and New Zealand Solar
Energy Society, Melbourne, Australia, 2003.
[25] J. H. Guo and J. E. Cotter, Interdigitated backside buried contact solar
cells, in Proc. 3rd World Conf. Photovoltaic Solar Energy Conversion,
Osaka, Japan, 2003, pp. 14631466.

>18%

2192

[26] J. H. Guo, P. J. Cousins, C. B. Honsberg, and J. E. Cotter, Selective

groove deglazing etchant for DSBC and IBBC solar cells, in Proc. Solar
2002Australian and New Zealand Solar Energy Society, Newcastle,
Australia, 2002.
[27] D. E. Kane and R. M. Swanson, Measurement of the emitter saturation
current by a contactless photoconductivity decay method, in Proc. 18th
IEEE Photovoltaic Specialists Conf., Las Vegas, NV, 1985, pp. 578583.
[28] R. A. Sinton and A. Cuevas, Contactless determination of current-voltage characteristics and minority carrier lifetimes in semiconductors from quasisteady-state photoconductance data, Appl. Phys.
Lett., vol. 69, pp. 25102512, 1996.
[29] S. Rein, T. Rehrl, W. Warta, and S. W. Glunz, Lifetime spectroscopy
for defect characterization: Systematic analysis of the possibilities and
restrictions, J. Appl. Phys., vol. 91, pp. 20592070, 2002.
[30] A. Cuevas and R. Sinton, Prediction of the open-circuit voltage of solar
cells from the steady-state photoconductance, Prog. Photovoltaics: Res.
Applicat., vol. 5, pp. 7990, 1997.
[31] R. A. Sinton and A. Cuevas, A quasisteady-state open-circuit voltage
method for solar cell characterization, in Proc. 16th European Photovoltaic Solar Energy Conf., Glasgow, U.K., 2000, pp. 11521155.
[32] P. A. Basore, Numerical modeling of textured silicon solar cells using
PC-1D, IEEE Trans. Electron Devices, vol. 37, pp. 337343, Feb. 1990.
[33] R. A. Sinton, P. J. Verlinden, R. M. Swanson, R. A. Crane, K. Wickham,
and J. Perkins, Improvements in silicon backside-contact solar cells for
high-value one-sun application, in Proc. 13th European Photovoltaic
Solar Energy Conf., Nice, France, 1995, pp. 15861589.
[34] J. E. Cotter, R. B. Hall, M. G. Mauk, and A. M. Barnett, Light trapping
in silicon-film solar cells with rear pigmented dielectric reflectors, Prog.
Photovoltaics: Res. Applicat., vol. 7, pp. 261274, 1999.
[35] R. R. King, R. A. Sinton, and R. M. Swanson, Studies of diffused
phosphorus emitters: Saturation current, surface recombination velocity,
and quantum efficiency, IEEE Trans. Electron Devices, vol. 37, pp.
365371, Feb. 1990.
[36] R. R. King and R. M. Swanson, Studies of diffused boron emitters:
Saturation current, bandgap narrowing, and surface recombination velocity, IEEE Trans. Electron Devices, vol. 38, pp. 13991409, June
1991.

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 51, NO. 12, DECEMBER 2004

[37] J. Zhao, J. Schmidt, A. Wang, G. Zhang, B. S. Richards, and M. A.

Green, Performance instability in N-type PERT silicon solar cells, in
Proc. 3rd World Conf. Photovoltaic Solar Energy Conversion, Osaka,
Japan, 2003, pp. 923926.

Jiun-Hua Guo received the B.Sc. and M.Sc. degrees

in agricultural machinery engineering from National
Taiwan University, Taipei, Taiwan, R.O.C. in 1993
and 1995, respectively. He is currently pursuing the
Ph.D. degree in electrical engineering at the University of New South Wales, Sydney, Australia.
He was a Research Engineer at the Automotive
Research and Testing Center, Taiwan from 1997 to
1999, having served the military from 1995 to 1997.
His research focuses on n-type high efficiency buried
contact silicon solar cells.
Mr. Guo was awarded the Presidential Award in 1992 and 1993. He was
elected as an Honorable Member of the Phi Tau Phi Scholastic Honor Society
in 1993. Part of his masters thesis won the Paper of the Year Award in 1998,
presented by the Chinese Institute of Agricultural Machinery.

Jeffrey E. Cotter received the B.S., M.S., and Ph.D.

degrees from the University of Delaware, Newark, in
1987, 1989, and 1997, respectively.
From 1989 to 1996, he was with AstroPower Inc.,
Delaware, working on the development of thin-film
crystalline silicon solar cells. In 1997 he joined the
Photovoltaics Special Research Centre, University of
New South Wales, Sydney, Australia as a Postdoctoral Fellow working on crystalline silicon solar cell
research and development. He is currently a Senior
Lecturer with the Centre of Excellence, Silicon Photonics and Photovoltaics, University of New South Wales where he leads the
development of commercial silicon solar cells.

INVESTIGATION OF PLANAR ANTENNAS WITH PHOTOVOLTAIC SOLAR CELLS FOR

MOBILE COMMUNICATIONS
Norbert Henze', Martin Weitz' , Pascal Hnfmann', Christian Bende13,Jorg Kirchhof' and Henning Friichtingl

' University ofKassel, Wilhelmshoeher Allee 73, D-34121, Kassel, Germany, nhenrel~uni-kassel.de
' DaimlerChrysler AG, Wilhelm-Runge-Strasse 11, D-8908 I Ulm, pascal.hofmann@daimlerchrysler.com

' Institute for Solar Energy Supply Technology, Koenigstor 59, D-34119, Kassel, Germany, cbendel@iset.uni-kassel.de

Abstract - This paper describes the application of photovoltaic (PV) solar cells in planar antenna stmctures. The
radiating patch element of a planar antenna is replaced by a
solar cell. Furthermore radiating slots are built due to the
cell spacing in a solar cell array. The original feature of a
solar cell (DC current generation) remains, but additionally
the solar cell is now able to receive and transmit electromagnetic waves. Both single solar cells as well as solar cell
arrays can be used as antennas. At first some basics regarding the RF properties of solar cell patches are discussed.
The efficiency of solar cell antennas is measured by means
of the Wheeler Cap method, which is investigated by means
of numerical simulations. The lossy substrate of the feed
lines has the strongest influence on the antenna efficiency.
Finally two applications are presented: a solar cell GPS antenna for vehicular applications and a solar cell slot antenna
for mobile communications (GSM). The aim is to integrate
these antennas into vehicular glass roofs which are covered
with photovoltaic solar cells in order to deliver the electric
power for the indoor ventilation of the car. The GPS antenna
provides circular polarisation and a main lobe in zenith direction whereas the GSM antenna is vertically polarized and
has a monopole-like radiation pattern. Both antennas are
built up with commonly used solar cells. The comparison of
measured and simulated antenna properties shows a good
agreement.
Keywords - planar antenna, solar cells, simulation, mobile
communication, vehicular application, Wheeler Cap

I. lNTRODUCTION
Wireless communication systems require electric energy for
their operation and their use. Stand alone applications like
environmental monitoring systems, vehicular communication systems or satellite systems need a net-independent
power supply which is preferably realisable by photovoltaics, an advanced technology distinguished by reliability,
longevity and eco-friendliness. Besides that antennas are
needed in order to receive or transmit electromagnetic
waves.
At present, photovoltaic generator and antenna are two separate devices. They compete for the available space on mobile and stand alone systems, which are generally limited in
size. Furthermore they may be bulky and expensive and they

0-7803-8523-3/04/$20.00 02004 IEEE.

limit the capabilities of product designs. To overcome these

restrictions a combination of antenna and solar cell in one
device is desired. Thus, new product designs and cost reduction becomes possible [I], [2].
A new approach, the Solar Planar Antenna - SOLPLANT ',
avoids these disadvantages. It consists of an appliance for
the conversion of solar radiation energy into electric energy
with at least one solar cell. The electrically conductive contacts of the solar cell are used simultaneously as antenna for
radiation or to the reception of electromagnetic waves. In
this case the backside contact of a solar cell is used as radiating element in terms ofradio frequency.
Based on these considerations, a product development concept was originated at the Institute for Solar Energy Supply
Technology (ISET e.V.) whose hasic idea has been registered as a patent in Germany, Europe, Japan and USA [3].
The invented appliance implies that the solar cell and antenna are combined as an integrated unit and that the conductive parts which are available anyhow for operational
reasons in a solar cell (or a solar module) are used simultan.eously as antenna elements.
11. USAGE OF PHOTOVOLTAIC SOLAR CELLS IN
PLANAR ANTENNA DESIGNS

A . RFImpedance of solar cellpatches

In order to provide high gain and a good efficiency the materials used in planar antenna designs should be as lossless
as possible. Numerous measurements have shown, that fiom
an RF point of view the solar cell acts nearly like a metallic
plate.

NA

NA

Figure 1. Test setup for the RF impedance measurement

between the DC contacts of a PV cell. The measurement is
performed with a network analyzer.
For this purpose comparative measurements of different
solar cells and metallic plates with identical dimensions

622

were performed. The test setup is shown in Figure 1. The

backside contact of a solar patch is connected to the ground
on one side whereas the upper side contact is connected to a
coaxial port. This setup is comparable to a Planar Inverted F
Antenna (PIFA) which is fed at the open end. The input reflection coefficient is measured with a network analyzer.
The resultant impedance is depicted in Figure 2 .

, , .,..I

radiation. Thus the free space S-parameter SiiFs contains

both the radiation resistance RRA, as well as the loss resistance RL, whereas the Wheeler Cap S-parameter SiiKT
contains the loss resistance of the antenna only.
The antenna efficiency can be expressed by the ratio [4]

....,....I

RRAi~

I=

RR4D + RLOSS

With the free space Sparameter

and the Wheeler Cap S-paramteei

1W,QE*3 100.1E16

1,0E+9

1.5E+9

2.E-S

Z.5PO

W E 4

FmuencylHZ
L-Me.llbpl.te,

Re,Z)-Solarse,,,

RelZ)--Me~ll,Cplafe. Irn,Z) -Solarcel,,

SIIWC =
IrniZ)J

Figure 2. Complex impedance of a poly crystalline solar cell

compared to metallic patches.
In a wide frequency range both impedance curves are nearly
identical. Primarily the depletion layer of the solar cell
semiconductor provides a low impedance between its DC
contacts. From an RF point of view the solar cell acts like a
metallic plate.

B. Ef$ciency considerations
Efficiency measurements are performed by means of the
Wheeler Cap method. The method described in [4] and [ 5 ]
was expanded to measurements of aperture coupled patch
antennas. The test equipment is shown in Figure 3.

RLOSS

- R.s

RLOSS + Rs

(3)

the antenna efficiency i s

(4)
As can be seen from (2) and (3) the used S-parameter in (4)
are real. That means the measurements must be made at
resonance. In this study the efficiency of aperture coupled
patch antennas is measured. The aperture is excited with a
microstrip line of length 1. Due to the transformation properties of microstrip lines the real feed point impedance Ge,at
resonance (which is the impedance at the aperture location)
is transformed to the complex port impedance I&, according to

The propagation constant p i s

Figure 3. Wheeler Caps in different sizes for efficiency

measurements. The reflection coefficient is measured with a
vector network analyser.
The method is based on two consecutive measurements of
the antenna input reflection factor S,i in free space and inside the Wheeler Cap respectively. The Wheeler Cap represents a metallic cover around the antenna in order to prevent

where

I , is the wavelength on the microstrip substrate and

I/&

is the velocity factor. In order to yield accurate

results the measured S-parameters at the antenna port must

be transformed backward to the aperture location of the antenna. This can be easily achieved by a proper port extension setting of the network analyser. For FR-4 substrate the
velocity factor is about I/&
= 0.55, With this specification the port extension can be set to the actual length of the
microship line.

In order to verify the Wheeler Cap measurement the complete test setup was simulated by means of numerical simu-

623

lations with Microwave Studioa. The antenna substrate (FR4) was modelled with a loss tangent of tan&O.OIJ and a
permittivity of ~ , . = 4 . 3 .

results from the simulation with Microwave Studio which is

the ratio of radiated to stimulated power. The efficiency of a
solar cell patch antenna is slightly smaller compared to the
copper patch antenna.

1
0.8

111. SOLPLANT - GPS ANTENNA

0.6

0.4

!i .,
5

-0.2

p -0.4

5 -0.6
48

13W

1400

l500

1200

1700

16W

la00

Frequency IMHz
~--ReallSl1). Sim. -1rnaglS11).

Sim.-Resl(Sl1/.

Meas. --Imsg(S11l, Mear.1

Figure 4. Measured and simulated reflection coefficient (real

and imaginary part) of an aperture coupled patch antenna in
free space. The antenna patch is made of copper.

2 -0.4
e -0.6
-0.8
1
lZoo l 3 O o

1400

1700

l6O0

1500

l6O0

Frequency IMHZ

~-Rsal(Sl1).Lm.-lmag(Sll).S~rn.

-Real(Sll).Mear.

-Imag(Sli).Mess.l

The desired antenna should consist of a solar cell array

where several cells are used as radiating elements. The GPS
antenna is based on the SSFIP (Strip-Slot-Foam-Inverter
Patch) principle [8]. The basic design is depicted in Figure
6. The aim of this study was to investigate the feasibility of
using a solar cell array as GPS antenna. In a later stage of
this project an integration of a SOLPLANT@GPS antenna in
vehicular glass roofs is considered. Due to considerations
concerning production the PV antenna module should be
manufactured with usual manufacturing techniques. This
involves a layered construction of the module as shown in
Figure 6. The solar cells are embedded in EVA films
(Ethylen-Vinyl-Acetat) which glues the solar cells on the
glass substrate. The backside of the solar cell-EVA layer is
covered with robust Tedlar@ film which serves as a
protective coating. The RF parts of the antenna are located
underneath the photovoltaic layer. At first there is an
antenna substrate (air or Rohacell foam). The aperture and
the microstrip feedlines are etched on a FR-4 substrate. The
overall thickness of this structure is less than 12mm. In
Figure 7 the complete module is shown. For tests with
several commercial GPS receivers a low noise amplifier was
integrated in the feeding network on the backside of the
antenna. However a reception of the GPS signal is also
ossible with the passive solar cell antenna.

Figure 5 . Measured and simulated reflection coefficient (real

and imaginary part) of an aperture coupled patch antenna in
the Wheeler Cap. The antenna patch is made of copper.

As indicated in Figure 4 and Figure 5 the de-embedding

of the measured S-parameter provides a minimum in the real
part at the resonance frequency (imaginary part is zero). The
efficiency obtained with these values and equation (4) is
summarized in Table 1. Additionally the results of measurements with an aperture coupled antenna with poly crystalline solar cell patch are listed.
T^LI^ 1
Id1S 1

Measured and simulated antenna efficiencies

r

Slips

Sllwc q

Direct calculation

Copper patch

-0.2
-0.18

-0.75 79%
-0.76 80%

--

Copper patch simulation

Poly crystalline solar cell

-0.14 -0.67

74%

--

78%

igure 6. Antenna with 36 solar cells. Four cells are excite

y means of apertures in a ground plane.
The solar cells type used in this module is RWE TE 16
SFTTAP. The material is mono cwstalline silicon. In order
to get the desired cell size of 49mm in square the cells with
an original edge length of l O O m m were cut in quarters. The
cell voltage in the maximum power point of VMpp= 0.49 V
remains the same since only the current depends on the cell
size. Thus the voltage of the whole module is 17.6 V. The
short circuit current is 0.87 A. A detailed discussion of this
antenna can be found in [7]

Simulated and measured efficiencies show a good agreement. The direct calculation is the total efficiency that

624

to the desired frequency. The coupling patch and the solar

cell have the same size (loomm x 100mm). They are separated by a thin film which isolates the solar cell from the
metallic structure.
lm

Figure 7. Active GPS solar cell antenna feeding a fan and a

commercial automotive navigation system.

Figure 9. Cross section of the planar solar cell antenna for

GSM 1800. Dimensions in m.

IV. SOLPLANT- GSM ANTENNA

In order to provide vertical polarisation the electric field

along the slot must be constant. This is achieved by a dual
feed at opposite sides of the coupling patch which forces a
nearly constant electric field along the slot.

Similar to the GPS antenna in the previous section this antenna is also designed for the integration into vehicular glass
roofs in order to provide a dual use of the solar cells as
electric power generator and antenna respectively. The aim
was to design antennas both for the GSM 900 and GSM
1800 band. The fixed size of the solar cells complicates the
construction of a dual mode antenna therefore a single hand
antenna for the GSM 1800 band is presented here. This integrated planar antenna device should substitute conventional
rod antennas on the car roof. Thus the requirements of the
antenna are vertical polarisation in the horizontal plane and
a monopole-like gain and radiation characteristic.

The measured and simulated return loss is depicted in Figure

IO. Due to uncertainties of the permittivity of the glass cover
the resonance frequency is slightly shifted. However the
bandwidth is nearly 200 MHz. The simulation are performed
with Microwave Studio which is a powerful tool for the
calculation of 3D electromagnetic fields.

The concept of the present antenna ties up to flush disc antennas as presented in [7].Basically it is a rectangular slot
antenna as can be seen in Figure 8 and Figure 9. Since standard solar cells are used with an edge length of IOOmm the
dimension of this patch and the arising slot around the patch
doesnt correspond to a operating frequency of 1800 MHz.

Figure 10. Measured and simulated return loss

The radiation patterns are shown in Figure 11 - Figure 13.
The vertical cut in Figure 11 is similar to the radiation pattern of a monopole on a small ground. There is only little
radiation in zenith direction. Figure 12 and Figure 13 show
the radiation pattern in an angle of 20 above the horizontal
plane. Due to the strong E-8 component the field is mainly
vertical polarized.

Figure 8. Expanded model of the planar solar cell antenna

for the GSM 1800 band.
The antenna consists of a rectangular slot with a length of
lOOmm in square. A coupling patch is connected to ground
by means of a large short circuit in order to tune the antenna

V. CONCLUSIONS
These studies have shown that a fully integration of planar
antennas and solar cells in only one device is possible. The
investigated antennas are thin and suitable for vehicular applications. However solar cell antennas can be utilized in
many other areas. The Solar Planar Antenna - SOLPLANP

625

opens a very wide field of application. Stand alone wireless

monitoring systems, mobile communication systems or satellites may benefit from integrated solar cell antennas, due
to weight reduction and optimal usage of the available
space. Applications of SOLPLANT@arein the fields of bureau and laboratory communication (Bluetooth), wireless
environmental monitoring, online building observations and
others, as well as in the extraterrestrial field of application.
Generally the RF properties of solar planar antennas can be
designed using commercial EM software packages. However the DC current path should be taken into account in the
simulation.

Gain

Gain

27V
Maaweman. Max :dBilF 3.m

Sirnuwon. Max [dBiI- 4.41

Figure 13. Radiation pattern 20 above the horizontal plane,

E-0 component.

REFERENCES
N. Henze, C . Bendel, J. Kirchbof and H. Friichting:
Application of Photovoltaic Solar Cells in Planar Antenna Structures, 12. International Conference on Antennas and Propagation ICAP 2003, Exeter UK, 3 Ird
March - 3 April 2003, pp.731-734.
[2] C. Bendel, N. Henze and 1. Kirchhof: Solar Planar Antenna - SOLPLANT, 17th European Photovoltaic Energy Conference, Munich 22-26 Oct. 2001.
[3] C. Bendel, N.Henze and J. Kirchhof: Apparatus for
converting solar e n e r a into electrical energy andfor
radiating and/or receiving highfrequency electromagnetic waves,United States Patent 6,395,971, Aug. 2000
(US), Deutsches Patent- und Markenamt, Patentschrift
DE 199 38 199C1, Aug. 1999 (DE). .
[41 D. Agahi and W. Domino: Efficiency Measurement of
Portable Handset Antennas using the Wheeler Cap, AppliedMicrowave & Wireless, Vol 12, No. 6, pp. 34-42,
June 2000.
[ 5 ] H. Wheeler: The Radiansphere around a Small AIlenna, Proceedings o f t h e l m , Vol. 47 pp. 1325-1331,
August 1959.
[ 6 ] C. Bendel, J. Kircbhof and N.Henze: ,,Application of
Photovoltaic Solar Cells in Planar Antenna Structures,
3. World Conference on Photovoltaic Energv, Osaka,
12.-16. Mai 2003.
[7] N. Henze, A. Giere, H. Friichting and P. Hofmann:
GPS Patch Antenna with Photovoltaic Solar Cells for
Vehicular Applications, 58th IEEE Vehicular Technologv Conference Fall 2003, Orlando, US, pp. October 69,2003.
[8] Ziircher, J.-F. and Gardiol, F. E.: BroadbandPatch Antennas, Artech House, Boston-London, 1995.
[9] K. Kaneta, T. Kondo, M. Ando and N. Goto: A Flushmounted Antenna for Mobile Communications, IEEE
AP-S. Syracuse, N.Y., pp. 1323-1326, June 1988.
[I]

1W
Sirnukhan, Mar IdBC =-0.91 M B B S U ~ ~6 aTx~rail
~ . =-0.74

Figure 1 1. Vertical radiation pattern, E-0 component. The

zenith direction is at 0.
Gal

90

-042

300
210

Sirnulaion, Msx LdBil = 4.81 Ueasui~ne?t.

Mar [dSii = 4 32

Figure 12. Radiation pattern 20 above the horizontal plane,

E-q component.

AKNOWLEDGEMENTS
This work has been carried out in cooperation with DaimIerChrysler AG, Ulm, Germany. Special thanks to Pascal
Hofmann for his valuable support of this work.

626

High Efficiency Organic Solar Cells by Screen Printing Method

Jun Sakai , Eisuke Fujinaka , Taisuke Nishimori , Norihiro It0 and Junji Adachi
Matsushita Electric Works, Ltd. , Kadoma Osaka , Japan
Shoji Nagano and Keiichi Murakami
NODA SCREEN CO ., LTD. , KomaMAichi , Japan
ABSTRACT
This study relates to a bulk heterojunction organic solar
cells, more specifically, it relates to the fabrication method
of a bulk heterojunction organic solar cells employing
conjugated polymers as a donor and the derivatives of
fullerenes as an acceptor. [I
-31
For the low cost production of the organic solar cells,
some printing methods has been studied. [4-51
In order to verify whether a highly efficient organic solar
cells can be obtained also in the printing method or not.
we chose a screen printing method for making a MDMOPPVPCBM bulk heterojunction type organic solar cells
and investigated the relations among printing conditions,
solar cell characteristics, and cell structures.
As a result, by optimizing printing conditions, we obtained
excellent characteristics as an organic solar cell wherein
its active layer was deposited by the screen printing
method. We attained 2.35% of conversion efficiency under
AM 1.5 100mW/cm2,5.06mA/cm2 of Jsc, 0.821V of VOC,
0.566 of FF and 82nm of film thickness in 6mm2 of photo
active area. Consequently, we could have confirmed that
the screen printing is useful as a fabrication method for the
this solar cells also in the solar cell characteristics.

very homogeneous film and available to investigate cell

performance for higher efficiency.
However, for the future product strategy, It's
considerable to investigate whether the printing method is
available or not as the organic solar cells manufacturing
method in the solar cell characteristics, large area
deposition, low cost and so on.
The screen printing method is made with simple and
easy facilities, and can form a large area pattern film with
high precision, therefore, it is considered to be an
advantageous method, and there has been so far reports
on the studies for making of a bulk heterojunction organic
solar cells employing the screen printing method . 14-51
The objective of the present study is to verify whether a
highly efficient bulk heterojunctionorganic solar cells can
be obtained by use of the screen printing method like spin
coating. For this purpose, the relations among printing
conditions, solar cell characteristics, and morphology are
evaluated, and the influence upon the solar cell
characteristicsand the mechanism of film deposition by
the screen printing are verified. And through these studies,
it is clarified that the screen printing method is a useful
method for the manufacturingof organic solar cells.
2. EXPERIMENTAL

1. INTRODUCTlON

A bulk heterojunction organic solar cells that has a

blends of conjugated polymers (donor) and an acceptor is
made by wet coating process, and it can be manufactured
with rather simpler facilities as a printing method than
those for manufacturing the conventional inorganic solar
cells.
And with regard to advantages of this kind of solar cells,
in comparison with a mnventional solar cells employing
the semiconductor process and high vacuum process,
significantly lower cost is expected since it's possible to
use printing process. And recently a report of a conversion
efficiency of 3% or higher under AM 1.5 has been already
made. [6]
The present study relates to a bulk heterojunction
organic solar cells as a blend of poly(2-methoxy-5.(3',7'Dimethyloctyloxy)-l4-phenylenevinylene) (MDMO-PPV)
and 1-(3-methoxycarbonyl)-propyl-l-phenyl-(6.6).C6,,
(PCBM) , more specifically, it relates to the method for
fabricating of a bulk heterojunction type organic solar cells.
As for film deposition method in the current activities of
organic solar cells employing conjugated polymers, the
spin coating method is mainly employed. It's easy to form

0-7803-8707-4/05/$20.00 02005 IEEE.

The organic solar cell used in the present study is a

typical bulk heterojunction type organic solar Cells, and the
basic structure is shown in Fig. 1.
As for the photovoltaic operation , the exciton generated
by photo absorption in the conjugated polymer dissociates
at the interface between polymer and fullerrene ,and
electron is transported to AI electrode via fullerrene
network, and hole is transported to IT0 electrode via
polymer due to built in potential by the work fundon
difference of both electrode.
In this structure, IT0 thickness is 150 nm and sheet
resistance is 10 ohdsheet. The active layer (MDMOPPVPCBM blend) has been deposited by use of the
screen printing method. Herein, it is forecasted that the
composition and viscosity of solution employed for
depositing the active layer will give a influence upon film
thickness and morphology, while in the present study, with
regard to solution, a standard solution that has
achievements where high efficiency (more than 2% of
conversion efficiency ) has been so far attained in the
'

I25

Glass substrate
Substrate
Holder

Fig.2 Semi automatic type screen printer

Fig.1 (a)MDMO-PPVIPCBM bulkheterojunction

solar cell basic structure (b) Charge carriers
dissociation and transport (c) MDMO-PPV
(d) PCBM

Max substrate size : 150"

square
(a) The whole (6)Substrate holder

spin coating has been employed. And the relation

between cell characteristics and printing condition has
been investigated.
The screen printer is shown in Fig. 2. The present
device is a semi automatic type desk-top printer. The
printing conditions that can be examined are screen mesh,
screen material, squeeze material, squeeze shape,
squeeze angle, squeeze speed, squeeze.pressure, and
the clearance between the screen and the substrate.
The sample making procedure is shown below. The size
of the glass substrate with I T 0 film was 16"
x 14mm,
and a cell with an effective area 6mm2 was formed
therein. The film thickness of IT0 was 150nm, and the
sheet resistance was 10 ohm.
PEDOT : PSS ( Bayer ) film is useful in hole transpoh
and the smoother of ITO. We deposited this film on
IT0 film by spin coating. The thickness was around 40nm.
After that we deposited active layer by the screen
printing method. The blend solution consists of MDMOPPV as conjugated polymers, PCBM as an acceptor, and
chlorobenzene as a solvent. The weight ratio of MDMOPPV and PCBM was 1:4. By use of this solution, the active
layer was formed by the screen printing. Usually
Conjugated polymer is coated in inert atmosphere
because it is affected by oxide and moisture. But in this
study we deposited it by the screen printing in air.
Screen printing has a lot of parameters. Mainly they can
be divided screen condition, solution ( ink ) condition and
screen type. ln this work we fixed solution and screen
condition, and adjusted printing condition : squeeze ,
shape, squeeze angle, squeeze speed, squeeze pressure
and gap between screen and substrate. As for screen ,we .
used 420 mesh screen. After printing the device was
baked at 50 C. for 1Smin.
In the nexl place, by vacuum deposition, a film of AI of
150nm was deposited and a solar cell was formed.
As for the light source of the solar cell characteristics
measurement, an AM 1.5G solar simuiator ( Oriel ) was

I26

employed, and it was adjusted to 100mW/cm2 by use

of a crystal Si reference solar cell calibrated according to
JIS8911 in Japan Electrical Safety 8 Environment
Technology Laboratories (JET). And the prepared solar
cell was put in an airtight package filled with Ar gas, and
was measured at room temperature (approximately23C).

3. Results and discussion

The I-V characteristics of the solar cell under AM1.5G
including screen printed active layer are shown in Fig. 3.
We have already made the MDMO-PPV I PCBM blends
solar cells by spin coating which has the conversion
efficiency : 2.0 - 2.3 %. In this study we used same
solution as spin coating. This figure indicates the screen
printing solar cell has excellent and same performance as
spin coating cells.

-.-EE

2.0
0.0

v)

-2.0

-0.4

-0.2 0

0.2

0.4

0.6

0.8

1.0

Voltage [ V ]
Fig.3 I-V characteristics of the solar cell in the dark
(a) and under AMI .5G 100mWlcm2 (b).The
conversion efficiency : 2.35% , short-circuit current :
5.06mA , open xircuit voltage : 0.821V , Fill factor :
0.566 (c) Spin coating cell I-V curve , Efficiency 2.3%

MDMO-PPVIPCBM
blend byer

(a) Scren printing

(b) Spin coating

10 urn

Blend laver

Fig.4 Cross section TEM image of the devices

(a) low magnification image of the screen printing cell
(b) low magnification image of the spin coating cell
(c) high magnification image of the screen printing cell
(d) high magnification image of the spin coating cell

(b) Spin coating

(a)
. . Scren printing

The screen printing solar cells were made 8 pieces at

same condition. Average conversion efficiency was
2.21 % , maximum efficiency was 2.35% and minimum
efficiency was 1.83%. These values are the highest
conversion efficiency as an organic solar cell whose active
layer has been deposited by screen printing has been
obtained.
As for printing condition , we used a screen of 420 mesh
with polyester wire diameter 27pm.and adjusted
conditions including squeeze shape, squeeze angle.
squeeze speed, squeeze pressure, gap between screen
and substrate, and so forth.
A cross sectional TEM images of screen printing cell are
shown in Fig. 4 (a) and (c). For the purpose of
comparison, a cross sectional TEM images of the spin
coating cell by use of the same solution is shown in fig. 4
(b) and (d). We have obtained the following
characteristics. The film thickness and the distribution in
the substrate was 82nm*10%. The film thickness and
morphology was almost same as spin coating. Generally
its necessary to form the film that has 80 - 100 nm
thickness for enough photo absorption in the present solar
cells.
Microscope photograph and AFM images are shown in
Fig.4. Average surface roughness Ra of both types are
almost same. But the peak to peak distance of Spin
coating is shorter than screen printing. This is due to the
differenceof film forming mechanism. And a little uneven
color can be seen in the blend film surface of the screen
printing cell. This shows that screen printing cell has
inhomogeneity of film thickness in the macro scale.

10 Bm

. (d) Spin coating

(c) Screen printing
Fig.5 Surface morphology by Microscope and AFM images
(a) Microscope photograph of the screen printing cell
(b) Microscope photograph of of the spin coating cell
(c) AFM images of the screen printing cell
Average Surface roughness Ra : 2.83 nm
(d) AFM images of the spin coating cell
Average Surface roughness Ra : 2.39 nm

In the nexl place, the film formation mechanism by the

screen printing is taken into examinations. In the spin
coating, solution is dropped onto a substrate, and then
rotated at high speed, thereby most solution is splattered
and solvent is evaporated immediately. Consequently , a
target film thickness 100nm is obtained.
In the screen printing, as shown in Fig. 6, At first
solution is filled in the mesh opening portion by squeeze
pressure and adhered to the substrate. And screen is
releasedfrom substrate as soon as squeeze travels. Then
solution is left on the substrate by adhesion with the
substrate. The shape of left solution is nearly equal to the
mesh opening portion right after screen releasing.

127

The screen printing has so far been considered to be

advantageous as the low cost mass production method of
organic solar cells because of availability of formation of
the accurate patterning in a large area.
Through this study, we have demonstrated that the
screen printing method has advantages for not only
productivity and accurate patteming but also solar cell
performance.
From now on It's necessary to investigate the effect of
solution and screen conditions, and making of large area
organic solar cells by screen printing.

Initial

0000000000

Clearance between screen and substrate

a
Squeezing

around 20 micron

Acknowledgements
We are grateful to Y. Kim and J. R. Durrant from Imperial
College for the fundamental instruction of MDMOPPV/PCBM solar ceIls.

Leveling

a
Solvent Evaporation

References

[I]
N.S.Sariciftci, LSmilowitz, A.J.Heeger, and F.Wudl,
'Photoinduced electron-transfer from a conducting
polymer to buckminsterfullerenre",
Science 258, 1474 (1992)
121 J.J.M.Halls, C.A.Walsh, N.C.Greenham. E.A.Maeseglia.
R.H.Friend, S.C.Moratti and A.B.Homes, "Efficient
photodiodes from interpenetrating polymer networks",
Nature 376, 498(1995)
[3] C.J.Brabec, N.S.Sariciftci and J.C.Hummelen,
"Plastic solar cells",
Adv. Func. Mater.11, 15, (2001)
141 S.E.Shaheen, R.Radspinner, N. Peyghambarian,
G.E.Jabbour. "Fabrication of bulk heterojunction plastic
solar cells by screen printing".
Appl. Phys. Lelt.79,2996 (2001)
[5] T.Aemouts. P.Vanlaeke, J.Poortmans and P.
Heremans , "Polymer solar cells: screen printing as a
novel deposition technique" , 2004 MRS fall Meeting
[6] A. Zakhidov, K. lnoue, S. Lee, Mike Sampson,
CPallavi and R. Ulbrich ,"Temperature and Time
Dependence of Heat Treatment on P3HTIPCBM
Polymer 3- Hexalthiophine soluble C60 Solar Cells"
2004 MRS fall Meeting

~ ~ u n d
\

Fig.5 Schematic drawing of screen printing deposition

model , Screen mesh : 420 mesh , wire
diameter : 27 micron in this study
So initial thickness of solution on the substrate is nearly
equal to the diameter of screen wire ( 27micron).
The nexl step is solution leveling. The solution flows on
the substrate and become uniform thickness film. Then
solvent evaporates and the final film is formed.
Generally, Screen printing condition consists of solution,
screen and printing condition. In this study we fixed
solution and screen, and aimed homogeneous and
appropriate thickness film (300nm) by controlling of
printing condition. As a result we found it is possible to
form above characteristics film by optimizing the printing
parameter including clearance, squeeze speed, squeeze
pressure and squeeze angle. Our understanding is that
the.amount of solution filled in the opening portion of
screen, screen releasing speed from substrate elastic
deformation of screen wire and solvent evaporation speed
have an effecton film thickness and morphology and
contribute to form thin and homogeneous film.
4. Conclusion

We could make an excellent performance organic solar

cells by screen printing method. And we confirmed that
the screen printing cell has almost same characteristicsas
spin coating cell in solar cell performance and film
morphology. This film morphology and thickness were
gotten by controlling printing conditions including
clearance, squeeze speed, squeeze pressure and
squeeze angle and so forth.

128

ANALYSIS OF THE RADIATION HARDNESS OF TRIPLE- AND QUINTUPLEJUNCTION SPACE SOLAR CELLS
C. Baur, M. Meusel, F. Dimroth, A.W. Bett, M. Neil', G. Strobl, S. Taylor3, C. Signorid
Fraunhofer Institute for Solar Energy Systems, D-79110 Freiburg, Heidenhofstr. 2, Germany
RWE Space Solar Power GmbH, 74072 Heilbronn, Germany
European Space Research & Technology Centre. Keplerlaan 1, NL-2200 AG Noordwijk, The Netherlands
email: carsten.baur@ise.fraunhofer.de
1

cells but higher EOL efficiency. First results in comparison

with a standard 35 cell are presented.

ABSTRACT
The Fraunhofer ISE has developed a characterization

RESULTS

tool called spectrometric characterization. In this paper we

discuss that this tool is extremely powerful to characterize
triple-junction (3J) EOL cells. The current-mismatch of the
subcells can be determined and the performance for
current-matched subcells can be predicted.
Additionally,
lattice-matched
AlGaInPlGalnPl
AIGalnAs/GafnAs/Ge quintuple-junction (54 cells are
being developed at Fraunhofer IS as a possible nextgeneration of space solar cells. This material combination
aims at the same BOL efficiency as stateof-theart 3J
cells but higher EOt efficiency. Based on spectral
response measurements it will be demonstrated that the
5J cells do in fact show a higher radiation hardness
compared to the 3J cell.

35 cells
It is well known that the GalnP and Ge layers show a
superior radiation hardness compared to the GalnAs
middle cell [5,6]. This can also be seen in Fig. 1 (EQE)
and Fig. 2.
1001,
.
,
, .
,
, .
. , .I
1

80

-E

INTRODUCTION

60

Solar cells in space are exposed to very harsh

environmental conditions. High energy electrons and
protons impinging on the solar cells cause defects in the
semiconductor materials leading to a degradation of the
diffusion length and a reduced cell performance.
Currently, GalnPlGalnAslGe triple-junction (35) solar cells
are mainly used in space due to the high efficiency and
radiation hardness. Averaged begin-of-life (BOL)
efficiencies of about 28 % and remaining factors of about
85% have been demonstrated not only by US
manufacturers [I
,2] but recently also in Europe [3].
Studies are currently being performed at Fraunhofer
ISE to optimize the cell design for end-of-life (EOL)
performance. Thereby, the spectrometric characterization
[4] is used as an additional characterization tool in order
to verify the results obtained by spectral response and I-V
measurements. Furthermore, this technique gives a better
understanding of the degradation behavior of the subcells
and their influence on the whole device. The latest results
on the development of radiation hard 3J cells are given.
Additionally,
lattice-matched
AlGalnPlGalnP
/A!GalnAs/GalnAs/Ge quintuple-junction (5J) calls are
being developed at Fraunhofer ISE as a possible nextgeneration of space solar cells. This material combination
aims at the same BOL efficiency as state-of-the-art 35

0-7803-8707-4/05/$20.00
02005 EEE.

40

20

Wavelength [nm]
Fig. 1: External quantum efficiency (EQE) of top and
middle cell of the 35 cell 15852-9 before (BOL) and after
irradiation (EOL) with 1 MeV electrons at a fluence of
1-105cmS. Only the top and the middle cell EQEs are
shown since the bottom cell has a lot of excess current
and is known to be very radiation hard.
In Fig, 1 the external quantum efficiency (EQE) of the
top and the middle cell of a 35 cell before (BOL) and after
irradiation with 1 MeV electrons at a fluence of l.1015cmh2
(EOL) are shown. White the EQE of the top cell remains
almost constant the EQE of the middle cell shows a high
degradation. The remaining factors of the photocurrentsof
top and middle cell calculated from spectral responsesare
0.97 and 0.78, respectively. In Fig. 2 the spectrometric
characterization of the same cell is shown. tn the course
of a spectrometric characterization several I-V
measurements are performed under different spectral

548

conditions. Thereby, the photocurrents of the subcells are

varied in a defined way meeting the following relation:

measurements. However, due to the high differences in

degradation behavior the middle cell becomes current
limiting after irradiation. The current matching is then at
0.94. Thus, the middle cell generates now 12.8 % less
current compared to the top cell generation under AMO.
From the shin of the current matching of 0.1 it can be
deduced that the middle cell degrades about 20 % more
than the top cell. This is consistent with the data obtained
from the spectral response measurement (Fig. 1).
However, the spectrometric characterization gives not
only information about the degradation behavior in current
but also of all other cell parameters. One interesting point,
for example, is the behavior of the fill factor (FF) that
reaches a minimum when the cells are current matched
and increases with higher mismatch. Thus, changes in ISC
are partly cancelled out.
Furthermore, spectrometric characterization gives
very descriptive information about how to improve the cell
with regard to radiation hardness. And it allows the
performance of a given cell design that is optimized for
EOL to be predicted. Since the power has a maximum
there, the optimum structure would be current-matched at
EOL under AMO, In the given case the cell could have
reached 21.9 % EOL efficiency when the cells would have
been current-matched at AMO.
In Table 1 the BOL, EOL and remaining factors R for
different celt designs are given. Results of structure
1585-2-9 are already shown in Fig. 1 and 2. The cells
1701-1-9 and 1702-3-5 used modified cell stnrctures. The
BOL efficiencies are comparable, lying in the range of
26.5 and 27 %. Afl cells show a very high radiation
hardness in VOcwith the same remaining factors of 0.95.
However, with regard to JSC. large improvements have
been achieved starting from 0.88 to a remaining factor of
0.96.

JZ

JL

where
and
are the photocurrents of the top
and middle cells and E is the spectral irradiance,
According to Eq. (1) the reference point at 1.0
corresponds to an I-V measurement under AM0
conditions. Moving from left to the right on the x-axis the
photocurrent of the top cell is increased while the
photocurrent of the middle cell is decreased at the same
time thereby always satisfying Eq. (9).

= =

0.95

1.00

-=

0.90

J"tE(h.))

1.05

1.10

J*"F,tU)

Fig. 2: Spectrometric characterization of a 3J cell before

and after irradiation with 1 MeV electrons. At BOL the
current is limited by the top cell - the cell is matched at a
more blue rich spectrum (x values higher than 1). At EOL
the middle cell becomes the current limiting cell -the cell
is matched at a more red rich spectrum (x values lower
than 1).
At BOL the top cell is current limiting. The differences
in the photocurrentsof the top and middle cell under AM0
conditions can be read from the position of the current
matching (here: 1.04 at 80L):

different cell designs,

In Fig. 3 and 4 the spectrometric characterizations of

all three cells are shown at BOL and EOL, respectively.
The modificationsof the cell design and improvements of
the middle cell result in a higher current mismatch at BOL
under AM0 conditions for structures 1701-1-9 and
1702-3-5. A current matching of 1.05 and 1.06 can be
seen from Fig. 3.
At EOL cell 1701-1-9 is matched at 0.96 while the
aim of cell 1702-3-5, which has been achieved, was that
the cell be current matched at EOL. This is not only the

with XCM being the x value where the cells are current
matched. Thus, the middle cell generates 7.3 % more
photocurrent at AM0 compared to the top cell. This is in
perfect agreement with the results obtained from EQE

549

matched at 0.94 the structures 1701-1-9 and 1702-3-5are

current matched at 0.96 and 1 respectively.

result of the higher current mismatch at BOL but also an

improved radiation hardness of the middle cell, which can
be seen from the lower shift of the current matching point
of the improved structures - only 0.09 and 0.06 compared
to 0.1 for the cell 1585-2-9.

5J cells
The 55 cell was developed to improve the radiation
hardness. The diffusion length in the tayers is the
parameter mainly affected by particle irradiation. The
radiation hardness in a 55 cell should be improved by
using thinner absorber layers. Thus, the diffusion lengths
may be lower without losses in photocurrent.

T38

5 %
E34
32

7-

(,J = 7.3 w v c m 2 )
( FF = 86.3 % )
( eta = 24.0 % )

U)

U
c.

1
E2.641%E=+u=q9p%5

-n-

'

2.62
' V

0.90
Jpm

c
0.95

;
1.00

AMO, 136NV/m2.T=25"C

0
,

1.05

(EIW JpmlOPIE,(M

= 5.196 V

V,

.-

Voltage

1.10

Fig. 5: IV-curve of a 5J cell. A remarkable high voltage of

almost 5.2 V has been reached.

.''r.

Fig. 3: Spectrometric characterizations of the three

investigated cell structures at BOL. While structure
1585-2-9 is current matched at 1.04 the structures
1701-1-9 and 1702-3-5 are current matched at 1.05 and
1.06 respectively.

2nd

100

-O--m-

400

600
700
Wavelength [nm]

500

800

900

Fig. 6: Quantum efficiency of the upper four junctions of

the 5J cell at BOL and afier irradiation with ?MeV
electrons at a fluence of 1-10'5cm-' (EOL). Since the Ge
bottom cell generates a large excess photocurrent and is
also very radiation hard it is not shown here. Also the sum
of the quantum efficiencies is shown.
2.52

In a 55 cell the upper two subcells of a 3J cell are

each divided into two [7]. This results in half the current
while the voltage is nearly doubled compared to a 35 cell.
Therefore, BOL efficiencies can reach similar values as 3J
cells. In Fig. 5 an I-V characteristic of a 56 cell is shown.
Since the measurement was performed with a 3 source
solar simulator the results should not be taken as
calibrated values. Only the remarkably high opemcircuit
of 5.2 V is a reliable value. A correct
voltage (V,)

E
> 2.50
0.90

0.95

1.00

1.05

1.10

Jpm""(E(m I JW""(E,,(U

Fig. 4: Spectrometric characterizations of the cells after

irradiation. While structure 15852-9 is now current

550

measurement could be performed with 5-source solar

simulators but these are not yet available. Consequently,
analysis of 5J cells is limited to VOC and quantum
efficiency measurements.
Fig. 6 shows the quantum efficiency of the first four
junctions of the 5J cell at 3 0 L and after irradiation with
1 MeV electrons at a fluence of 1-10'5cm4 (EOL). Since
the Ge bottom cell generates a large excess photocurrent
and is very radiation hard it is not shown here. Also the
sum of the quantum efficiencies is shown. It is noteworthy,
that the blue response of the 5J cell is still rather low due
to absorption losses in the upper tunnel diodes. This
problem can be solved in the future by using materials
with higher bandgaps. However, a Hgh radiation hardness
can be observed. Only the 4th junction shows significant
degradation of the quantum efficiency.
.

3J
5J

Imv]

[mq

2638
5196

2516
4937

an improved structure. This yields a remaining factor of

0.88 and an EOL efficiency of 23.3%. Spectrometric
characterization has been demonstrated as a useful
characterization tool for analyzing the radiation hardness
O f 35 cells giving additional information about the behavior
of all cell parameters under dqradation.
In addition, latest results of 5J cells irradiated with
1 MeV electrons have been shown. Due to the thinner
layers the 5J cell outperforms the 3J with regard to
radiation stability. Thus, 5J cells show their potential to
become the next generation space solar cell.

ACKNOWLEDGEMENTS
This work was partry supported by the European
Space Agency (ESA) through the funding of the project
"Development of next generation GaAs based multi
junction solar cells" (Ref. N.: A0/3-1f053/04/NUFM).

[%I
95.4
95.0

REFERENCES
[ I ] M. A. Stan, D. Aiken, P. R. Sharps et al., "The
Development of >28% Efficient Tnple-Junction Space
Solar Cells at Emcore Photovoltaics", froc. of 3rd Wudd
Conference on Photovofiaic Energy Conversion,2003, pp.
662-665.

regard to the iemaining factor of the Vm

Fig. 7 compares the total quantum efficiencies of a 35
and the 55 cell. The remaining factor of the VOCof both
cells is comparable and about 95 % (see Table 2). In
Fig. 7 it can be observed that the quantum efficiency of
the 5J cell at EOL is significantly higher than that of the 3J
cell in the wavelength range between 640 to 830 nm. This
can be attributed to the high radiation hardness of the thin
AlGalnAs subcell of the 5J cell and demonstrates the
superior radiation hardness of 5J cetls compared to 3J
solar cells.

i
400

[Z] J. E. Granata, C. Fetzer, J. H. Ermeretal., "UltraTriple

Junction GalnPIGaAslGe Solar Cells: Cell Design and
Qualification Status", Proc.of 3rd Wodd Conference on
Photovoltaic Energy Conversion, 2003, pp. 654-657.

I
I

[3] G. Strobl. R. Dietrich, W. Kostler et al., "Dwelopment

and Qualification Status of EuropeanTripleJunction Solar
Cells for Space Application", Pmc. of 19th European
Photovoltaic Solar Energy Conference, 2004. pp. 36143617.

[4] M. Meusel, R. Adelhelm, F. Dimroth et al.. "Spectral

Mismatch Correction and Spectrometric Characterization
of Monolithic Ill-V MultiJunction Solar Cells", ffvgressin
Photovoffaics: Research and Applications 10, 2002, pp.
243-255.

I
.

6M3

[5]R. J. Walters, G. P. Summers, S. R. Messenger,

"Analysis and modeling of the radiation response of
multijunction space solar cells", Pmc. of 28th I
Photovokaic Specialists Confemnce, 2000, pp. 10921097.

8M)

3-"3

[6]S. R. Messenger, R. J. Walters, G. P. Summers et al.,

"Radiation Response Analysis of Triple Junction
InGaPIlnGaAsIGe Solar Cells", P m . of 17th European
Photovofiaic Solar Energy Conference and Exhibition,
2001, pp. 2251-2254.

F]M.

Meusel, F. Dimroth, C. Baur et al.. "European

roadmap for the development of Ill-V mufti-junction space
solar cells", P m . of 19th Eumpean PhotovoFfaic So!ar
Energy Conference,2004,pp. 3581-3586.

SUMMARY

Results of new investigations aimed at improving the

radiation hardness of space solar cells are shown. Current
matching at EOL under AM0 conditions was achieved with

55 1

PERFORMANCE OF HIGH-EFFICIENCY ADVANCED TRIPLEJUNCTION SOLAR PANELS FOR THE LILT MISSION
DAWN
Navid S. Fatemi, Surya Shana, Oscar Buitrago, John Crisman & Paul R. Sharps
Emcore Photovottaics, 10420 Research Rd., SE, Albuquerque, NM 87123, USA
Ron Blok, Martin Kroon 8r Cees Jatink
Dutch Space, P.O. Box 32070,2303 DB, Leiden, The Netherlands
Robin Harris
Orbital Sciences Corporation, 21839 Atlantic Blvd., Dulles, VA 20166, USA
Paul Stella & Sal Distefano
Jet Propulsion Laboratory (JPL), 4800 Oak Grove Dr., Pasadena, CA 91 109, USA

ABSTRACT

distance between the asteroid belt and the sun, in the

order of 2-3 astronomical units (AU), necessitates
efficient operation of the solar arrays under the low
intensity, low-temperature (LILT) conditions.

NASA's Discovery Mission Dawn is designed to

operate within the solar system's Asteroid belt, where
the large distance from the sun creates a lowintensity,
low-temperature (LILT) condition. To meet the mission
power requirements under, LILT conditions, very highefficiency multi-junction solar cells were selected to
power the spacecraft to be built by Orbital Sciences
Corporation (OSC) under contract to JPL. Emcore's
1nGaPllnGaAdGe advanced triplejunction (ATJ) solar
cells, exhibiting an average air mass zero (AMO)
efficiency of greater than 27.6% (onesun, 28"C),were
The two solar
used to populate the solar panels [l].
array wings, to be built by Dutch Space, with 5 large
area panels each (total area of 36.4 m2) are projected to
produce between f0.3 kWe and 1.3 kWe of end-of-life
(EOL) power in the 1.0 to 3.0 AU range, respectively.
The details of the solar panel design. testing and power
analysis are presented.

Fig. 1. Typical Dawn Trajectory Toward Ceres 8 Vesta

BACKGROUND

Lavncl?mn8@"wl

In mid 2006, NASA's Discovery Mission Dawn is

scheduled to launch its mission to characterize the
conditions and processes of the solar system's earliest
epoch by investigating in detail two of the largest
protoplanets remaining intact since their formations,
Ceres and Vesta. The spacecraft's trajectory toward the
asteroid belt is shown in Fig. 1. The mission life is
envisioned to be 10.5 years. Under contract to JPL,
Orbital Sciences Corporation is responsible for design,
manufacturing, integration and testing of the Dawn
spacecraft and support its launch and flight operations
Dutch Space is responsible for the
(Fig. 2).
manufacturing and testing of the solar arrays, while
Emcore provides the photovoltaic assembly for the
mission. Ceres and Vesta reside in the extensive zone
between Mars and Jupiter called the asteroid belt. The

0-7803-8707-4/05/$20.0002005 IEEE.

pr

Fig. 2. Dawn Spacecraft Schematic

618

I":-h

0C temperatures in a manufacturing line. The same

testing and grading principals were applied to the
production of 11,480 largearea (27.5 cm2) ATJ CIC
(coverglass-interconnected-cell) assembIies for the
solar panel build. The photograph of a Dawn ATJ CIC
is shown in Fig. 3.

Historically, the need for power in space has been

dominantly provided by silicon solar cells. In the past
several years, however, high-volume manufacturing of
high-efficiency multijunction solar cells has enabled the
use of this altemative technology for space power
generation [14].
The solar array for the Dawn mission is planned to
provide power under greatly varying illumination and
temperature conditions. The sun intensity will vary
approximately from 136.7 mW/cm2to 15.2 mW/cm2(i.e.,
1 - 3 AU), while the calculated operating temperature of
the solar array varies from +60"Cto -88%. Solar arrays
populated with Emcore's InGaPllnGaAslGe advanced
triple-junction (ATJ) solar cells are selected to provide
primary electrical power for the Dawn mission. These
cells, while exhibiting >27.5% beginning-of-life (BOL)
minimum average efficiency under 1 AU (one sun), airmass zero (AMO) illumination conditions at 28"C, are
projected to demonstrate >30% conversion efficiency
under LILT conditions.

Fig. 3. ATJ production CIC -area of 27.5 cm2

ATJ SOLAR CELLS

It is well known that typical space solar cells

designed for one sun operation may behave quite
differently under LILT conditions. Historically. some
space silicon cells have ehibited an illuminated currentvoltage (I-V) characteristic with an unusual shape under
LILT testing. This is referred to as the "flat spot"
phenomenon [5]. This phenomenon causes these cells
to exhibit a lower performance, at or near the maximum
power point, than expected under LILT conditions. A
similar effect has been &served in some of the lowest
grade multi-junction GaAs cells, within the normal
distribution of efficiencies in the cell production. This
seems to be correlated to the shunt resistance (Rsh), in
that cells with lower fill factor (FF) values are the only
cells that have a potential to show this effect. One of
the factors that affect Rsh is defect sites within or at the
edge of the semiconductor layers.

Solar cells with higher FF values when tested under

AMO, l x intensity, maintained or showed improvement
in FF under low intensity testing. On the other hand,
low FF cells (Le., 4 0 % at 0.277~
intensity) tested under
l x intensity, sometimes ehibited a greater degree of FF
degradation under lower intensities. The low intensity
measurement of all cells and ClCs was paformed under
0.277~AM0 illumination. It was concluded from this test
that even the lower-grade cells exhibit an acceptable
performance under low-intensity conditions, and that the
power demand of the Dawn spacecrafl could be met
using the fypical range of the ATJ cell performance
distribution. This typical performance range under l x
-,.-

It is welt known that under I x AM0 intensity, the

solar cell performance is enhanced at lower operating
temperatures. As might be expected from cells having
below nominal values of Rsh, the Lower-thanexpected
performance observed with some cells under LILT is
emphasized under the conditions of low intensity flux,
and not low junction temperatures. The LILT behavior
of the solar cells can therefore be well characterized by
only measuring the cells under low intensity light
illumination levels at room temperature and
subsequently applying temperature coefficient values
which have been measured by JPL under LILT
conditions. Consequently, the Emcore ATJ solar cetls,
manufactured for the Dawn mission, were all tested and
electrically graded under lowintensity conditions at
28C (instead of at l x AMO).

>l"Cetk

n.0

W.0

A0

-ny

(W

AM0 is shown in Fig.4.

Fig. 4. ATJ production cell AM0 efficiency histogram

In the following sections, the power analysis
methodology and the results of flight panel testing under
Dawn illumination conditions are presented.

This test methodology also removes the onerous

condition of measuring large number of cells at below

619

SOLAR PANEL CONFlGUMTlON

POWER REQUIREMENTS & ANALYSIS

As mentioned earlier, the Dawn solar array consists
of two wings, with 5 solar panels per each wing. One
wing configuration is schematically shown in Fig. 5.

Sun Distance
(AU1

Fig. 5. Schematic solar wing configuration (circled

numbers denote circuit numbering)

'1.oo

As shown in the figure, each panel is comprised of

two circuits. Each circuit contains 14 strings. Total of
28 strings are bonded to every panel, with 41 cells
connected in series per each string, bringing the total
number of cells in each panel to 1,148 cells. All
rectangular panels have the same dimensions: 2.265 x
1.608 m, resulting in a panel area of 3.64 m2. A
photograph of one representative panel is shown in Fig.
6.

Intensity
(MO)

'

Required
Fowl3 (W)

Calculated
POWW (W)

1.000

3,600

f0,3f7

0.319

3,650

3,804

0.197

2,350

0.148
0.111

1,770
1.300

1,781
1.323

Other requirements included array maximum power

voltage (Vmp) to be greater than >85 V & cl40 V 8 BOL
short-circuit current (Isc) c 7.5 A. The calculated power
values shown in Table I were the result of a
comprehensive power analysis that was performed to
predict the array level power. The solar array power
analysis predicts the array electrical performance at all
relevant stages during the mission by simulating the I-V
curve generated over the required temperature range,
sun intensity, and environmental conditions and loss
factors, i.e. radiation, rnicrom eteorites, etc. The solar
array model allows for the generation of an I-V c m e at
the cell, string. circuit. panel. and array levels.
Performance parameters such as Isc, VOC, maximum
power point current (Imp), maximum power point power
(Pmp), and power or current at any given voltage may
be determined from these generated I-V plots.
In the model, as in the solar panels, the cells are
connected in series to form strings and the strings are
In order to
connected in parallel to form circuits,
calculate the array power, the individual solar cell power
outputs at any given stage during the mission is
multiplied by the number of cells in the array and all
pertinent current and vottage drop losses are applied.
The voltage output of a circuit is calculated by
multiplying the number of cells connected in series in a
single string by the voltage output of a single cell. The
current output of a circuit is calculated by multiplying the
number of strings connected in parallel by the current
output of a single string. Additional voltage losses due
to wiring, connectors, and blocking diodes are then
applied to the final results.

Fig. 6. Photograph of a Dawn panel (one of 10)

populated with 1,148 large-area (27.5 cm2)ATJ cells

The results of the analysis rely on the actual

measurement of the electrical performance parameters

620

(i.e.. VOC, Isc. Vmp. and Imp at 0.277~sun intensity and

28OC for cells manufactured for the Dawn program. The
ClCs were electrically graded at a load voltage (Vspec.)
of 2.205 V. The minimum average illuminated CIC
current at this load voltage (lspec.) was slightly greater
than 121 mA, and the bin width was set at ImA.

2.00

I
9 1.50
c

The Dawn solar array power analysis was therefore

performed for the 0.277~case (I
-9 AU) using the actual
CIC test data, while the performance parameters
(including temperature coefficients) for other lowerintensity AU cases of interest (i.e., 2.25, 2.6, and 3.0
AU) were extrapolated from the 1.9 AU case. The
extrapolationwas performed using a sampling of actual
ATJ CIC measurement data taken under LILT
conditions. Several representative ATJ ClCs were
measured at JPL under low intensity and low
temperature conditions corresponding to the Dawn
mission profile. enabling both BOL & EOL
characterization of the ClCs under operational
conditions.

10.50
U

0.00

WMO)

t .oo

0.277

0.20
0.125

5,568

1,057
622

5,852
1.551

SUMMARY

Solar panels were built and tested for the Dawn

spacecraft for a LILT mission to the Asteroid bett. Highefficiency ATJ cells were used to populate the panels.
All cells were binned under a 0.277~AM0 lowintensity
8 28C conditions at Emcore. A sampling of ClCs was
also tested at J P l under the LILT mission profile, The
results were used to perform a comprehensive power
analysis of the mission incident intensity and
temperature conditions. The electrical performance
measurements of the completed solar panels
demonstrate that the power requirements of the Dawn
mission wiil be met.
REFERENCES
[l]M.A. Stan, D.J. Aiken, P.R. Sharps, N.S. Fatemi.

F.A. Spadafora, J. Hills, H. Yoo, and B. Clevenger.

"27.5% Efficiency InGaPllnGaAslGe Advanced Triple

Junction (ATJ) Space Solar Cells For High Volume
Manufacturing," 29th PVSC, New Orleans, May 2002.

I21 N. S. Fatemi, et al.. "HigbEfkiency Advanced

TripleJunction Solar Cells With Monolithicafly
Integrated Bypass Diode," 19'" European Photovoltaic
Conference (PVSEC), June 7-1 1,2004, Paris, France.

1,082
663

[3] N. H. Karam, et at., 'High efficiency GalnPdGaAslGe

dual and triple junction solar cells for space
applications," 2"d World Conf. & Exhibition on
Photovoltaic Solar Energy Conversion, 6 10 July, 1998,
Vienna, Austria, p.3534.

based on analysis.

[4] Y. C. M. Yeh, et al., "Advances in production of

cascade solar cells for space," 26th PVSC, Sept. 3 s
Oct. 3, 1997, Anaheim, CA, p.827.

In addition to electrical health check and visual

inspection, several other functional tests were
performed on all panels. These included:

150.0

Fig. 7. BOL I-V plot of one Dawn circuit at 0.277x, 28C

As shown in Table II, the measured BOL power'at

all intensities meet or exceed the power predictions

200.0

VOLTAGE (VOLTS)

Measured Power per LAPSS (W)

1.495

50.0

Table II. Summary LAPSS results for +Y Wing

per Analysis (W)

0 .o

The BOL electrical performance of the panels were

measured with a large-area pulsed solar simulator
(LAPSS) at 28C under four different AM0 sun
0.20~
& 0.125% A summary of
intensities: 1.Ox, 0.277~.
the LAPSS test results for 5 panels (+Y Wing) are
presented in Table II. The values in thesecond column
represent predicted power of one wing calculated at
28C under the indicated sun intensities, such that they
could be directly compared to the actual measured
values under the same conditions shown in the last
column. This is followed by a representative example of
an I-V plot of one circuit at 0.277~intensity in Fig. 7.

Predictad Power -

1.00

PANEL LEVEL ELECTRICAL RESULTS

Intensity

2.50

[5] V.G. Weirer 8 J.D. Broder, 'On the cause of the flatspot phenomenon observed in silicon solar cells at low
temperatures and low intensities," J. Appl. fhys. Vol. 53
(8),Aug. 1982, pp. 5926-5930.

Mass measurements
Protection diode functionality lnfra-red (IR) test
Insulation resistance
Functionality of bleed resistors, blocking
diodes & thermaf sensors

62 1

Si-BASED TlilN-FILM SOIAR CELLS

: PROCESS AND DEVICE PERFORMANCE ANALYSIS
Jeong Chul Lee, Sang Won Jun, Jae Ho Yun, Seok Ki Kim, Jinsoo Song and Kyung Hoon Yoon
Solar Cells Research Center, Korea Institute of Energy Research,
71-2, Jang-dong, Yuseong-gu, Daejeon 305-343,Korea

ABSTRACT
Ttis paper briefly introduces silicon based thin film solar
cells: amorphous (a-Si:H). microcrystalline (pc-Si:H)
single junction and a-Si:H/pc-Si:H tandem solar cells. The
major difference of a-Si:H and pGSi:H cells comes from
electro-optical properties of intrinsic Si-films (active layer)
that absorb incident photon and generate electron-hole
pairs. The a-Si:H film has energy band-gap (Eg) of 1.71.8eV and solar cells incorporating this wide Eg a-Si:H
material as active layer commonly give high voltage and
low current, when illuminated, compared to pc-Si:H solar
cells that employ low Eg (l.leV) material. This Eg
difference of two materials make possible tandem
configuration in order to effectively use incident photon
energy. The a-Si:H/paSi:H tandem solar cells, therefore,
have a great potential for low cost photovoltaic device by
its various advantages such as low material cost by thinfilm structure on low cost substrate instead of expensive
c-Si wafer and high conversion efficiency by tandem
structure. In this paper, the structure, process and
operation properties of Si-based thin-film solar cells will
be discussed.

1. INTRODUCTION
Silicon based thin-film solar cells (Si-TFSC) have been
promising candidate for low cost and high efficiency
photovoltaic devices. Although wafer-based crystalline
silicon (c-Si) solar cells have possessed more than 90%
of world PV market over the fast few decades['], high
material cost and feedstock shortage of c-Si substrates
have been still obstacles for wide spread use of c-Si solar
cells. On the contrary to c-Si celts, Si-TFSC has a great
potential for reducing material cost since it use very thinfilms on low cost substrate such as glass, stainless steel
sheet, plastic, etc, instead of using expensive thick (250350 pm) c-Si wafer. Regardless of this low cost potential
of Si-TFSC, the market possession is still low. The
amorphous silicon (a-Si:H) thin-film solar cells has been
introduced early in the 1970s and market share has been
more than 25% in 1990 but drastically decreased to about
6% in 2003[']. The main reason for this reduced market
possession of a-Si:H solar cells might be resulted from
low module efficiency and stability problem (StaeblerWronski Effect)121,
by which the production cost ($Iwp) of
a-Si:H solar cells has not been competitive with c-Si solar
cells. Although a-Si:H thin-film modules have similar or
slightly less production cost over the c-Si cells, the
installation area should be more than double in order to

0-7803-8707-4/05/$20.00
02005 IEEE,

generate the same electricity because of its low efficiency.

During the 1990% microcrystalline silicon (pc-Si:H) thinfilm and solar cell have a great attention since they have
problem and high potentials for improving
The pc-Si:H layer is known to be a complex
material consisting of crystalline and amorphous silicon
phases plus grain boundaries'". The band-gap energy
(Eg) is the same with c-Si (1.leV). However, the pc-Si:H
thin-film solar cells have shown some drawbacks and the
efficiency has stayed near 10%. This low efficiency of p
Si:H cells is related to material properties of pc-Si:H layer,
its low light absorption coefficient and imperfect material
quality. The world record &i:H solar cells give current
P
densityusc) of 26mAlcm and voltage (Voc) of 0.5W
which is quite low compared to commercially available 6
Si solar cells". The low jsc of pc-Si:H cells mainly results
from incident photon loss since thickness of intrinsic pci
Si:H layer is just a few JTI.I that is not enough to fully
absorb incident photon; at least 50pm silicon needed to
absorb 90% of photon. In order to effectively utilize
incident photon, tandem concepts are introduced'']. The
a-Si:H/pSi:H tandem solar cells consist of wide Eg
(1.7eV) a-Si:H top-cell and narrow Eg (1.leV) pc-Si:H
bottom-cell. The high photon energy is absorbed at a-Si:H
topcell and low energy at pc-Si:H bottom-cell. The series
interconnection of two cells produces high voltage of solar
cells, which is beneficial to reduce series resistance loss
by transparent conducting oxide (TCO) in a large area
module. The Kaneka c o p has already achieved 14.7%
efficiency in a lab-scale small area tandem cells and
about 10% in mass-produced meter-scale modules[g1.
However, the module efficiency more than 15% and
production cost of 50$/m2 should be realized for
photovoltaics with Si-TF SC to have competitiveness with
conventional power source.

L:!.i

II. Solar Cells Structure and Process

2-7. Device structure

Figure 1 shows a cross sectional stmcture of
supersfrate pin a-Si:HlpGSi:H tandem cells. All single
junction and tandem solar cells are prepared on glass
substrate. As a front transparent conducting oxide(TCO),
SnOo:F(Asahi-U glass) or Zn0:Al with rough surface are
used. The Zn0:Al films with textured surface are
prepared by chemical etching(1% HCr) of rf-sputtered
Zn0:Al that has originally flat surface morphology. The
front TCOs with textured surface has been reported to
give improved solar cell performances by their advanced
optical properties such as low reflectance and high
scattering of incident light. As a p-type window materials,

I552

a-SiC:H, pc-Si:H single layer and pc-Si:H/a-SiC:H bylayer are employed. The effectof different p-type window
layers on a-Si:H solar cells will be discussed in the
following section. The tandem solar cells composes of
series connection of pin a-Si:H top-cell and pc-Si:H
bottom-cell. The ZnO buffer between top and bottom cells
reportedly gives higher current to topcell by intemal
reflection of incident light at ZnQ buffer, by which
degradation (SW-effect) of a-Si:!+ top cell can be reduced
by making a-Si:H intrinsic layer thin. The Ag and ZnO/Ag
are used as a back contact.

2-2. Solar cell process

The a-Si:H, pc-Si:H and a-Si:H/pc-Si:H thin-film solar
cells were fabricated with multi-chamber cluster tool
system. The system consists of five process, two central
and load/unload chambers. The process chambers
contain 2 PECVDs, VHFCVD. HWCVD and rf sputtering
systems. Robot arm is located inside the central chamber
in order to transfer substrate into each process chamber
without breakingvacuum. With this system, the solar c8lls
can be produced without any impurity contaminations
during deposition of pin thin-films and sample transfer.
The intrinsic a-Si:H and pc-Si:H thin-films were deposited
by 6OMHz VHFCVD with SiH4 and HP gas mixture. The p
pc-Si:H and n a-Si:H layers were prepared by
conventional PECVD method. The ZnO intermediate
buffer in a tandem cell was deposited by rf sputtering
method with ZnO:A1203 target at m m temperature. The
low rf power was employed to prevent sutface damage by
ion collisions during sputtering of intermediate ZnO film.
The solar cell area was defined to be 0.25cm2 by shadow
mask during deposition of 2nO:Al and Ag back contact.

SnOz:F and Zn0:Al TCO. The Ag(5OOnm) film was used

as a back contact. The a-Si:H cell prepared on Zn0:Al
has higher open circuit voltage(Voc) but low short circuit
current densi!y(jsc) and fill factor(FF). The low FF of solar
cells on 2nO:Al results from hi h contact resistance
between ZnO-A and p a-SiC:H[". It has been reported
that the contact problem of Zn0:Al with p a-SiC:H can be
removed by inserting thin p p S i : H between Zn0:Al and
p a-SiC:H[ 'I. Figure 3 (c) shows the effect of p pc-Si:H
buffer on solar cell performances. By inserting p Wc-Si:H
thin-layer, the FF and VOCimproved as the cost of slight
decrease of jsc. This decrease of jsc was caused by
increased absorption at p-layer by inserting p p-Si:H
thin-layer as shown in external quantum effciency(EQE).

Fig. 2. The effect of transpatent conducting oxide on

performances of a-Si:H thin-filmsolar mils.

As a wide-gap top cell in a a-Si:H/pc-Si:H tandem cell, the
thickness of a-Si:H top-cell should be important
parameter in a view of current matching and device
stability. Although the thin a-Si:H top-cell reportedly gives
low light induced degradation, its jsc should be matched
with series connected pc-Si:H bottomcell. Fig. 3 shows
AM 1.5 I-V cumes and EQE of a-Si:H topcells
incorporating intrinsic a-Si:H layer with different thickness.
SnOz:FTCOs are used as front TCO and the other layers
are prepared at the same conditions. The solar cells with
thicker i a-Si:H layer produce higher jsc but lower VOC.
The higher jsc with thicker i a-Si:H results from increased
absorption of long wavelength photons as shown in EQE
curves.

'"

Fig. 1. Structure of a-Si:H/pc-Si:H tandem solar cells.

111. Results and Discussions

3-1. Wide-gap a-Si:H topcells

Fig. 2 (a) and (b) show the effect of transparent TCO on
performances of aSi:H solar cells. Device (a) and (b) we
prepared by the same process run. The p a-SiC:H(ZOnm),
i a-Si:H(XOnm) and n a-Si:H(30nmf are deposited on

1553

3.2 Narrow-gap pc-Si:H bottom-cells

The pc-Si:H sofar cell as a narrow-gap bottom cell in a
tandem configuration gives higher jsc and better stability
compared with a-Si:H top-cell. The pc-Si:H thin films used

front ZnO morphology and light scattering is caused by

enhanced collection at blue region (40G600nm).
Although the total and diffused transmittance in type2
ZnO films is much higher than that of type1 in whole
interesting wavelength as shown in figure 2, the
improvement of collection in a solar cell dominantly
occurs at blue region. This means that the light scattering
at surface of front ZnO films contributes to the
improvement of blue response. Since the slight increase
of collection in 600-100nm range resulted both from light
scattering and from increased total transmittance, the
contribution of scattering at ZnO surface to the carrier
collection at red light is quite weak compared with blue
region. On the other hand, the enhancement of red
response is clear in a solar cell with ZnOlAg back
reflector as shown in figure (b) and (c). The ZnOlAg back
reflector gives higher collection at 600.1000nm while
slight lower response at 400-550nm. From the above
results, it is concluded that surface morphology and light
scattering at ZnO front contact in pc-Si:H thin-film solar
cells are quite important for improving Voc and FF as well
as jsc. While the textured front ZnO films dominantly
contribute to the improvement of carrier collection at blue
region, back reffectors are effective for increasing red
response.

as an intrinsic active layer in a solar cell has been known

to be a complex material consisting of crystalline and
amorphous
silicon
phases.
Changing
Silane
Concentration (SC) in the plasma gas phase, SC =
SiHd(SiH4+H& during deposition of Si-films is most
effective in changing the structure of the deposited layer
from amorphous material (at high SC-value) to
microcrystalline material (at low SC). It has been reported
that pc-Si:H solar cells prepared at this phase transition
point has best performances['*'13].

U,

20

Fig. 3. AM 7.5 CV curves(/efi) and Q(righi) of a-Si:H

solar cells with differenfthickness of intrinsic a-Si:H light

absorbing layer.
Fig. 4 shows the illuminated I-V curves and spectral
response of pin pc-Si:H solar cells with different light
trapping scheme by changing front ZnO roughness and
structure of back reflector. The figure (a) through (c) show
the effect of front ZnO films, (c) and (d) represent the
effect of back reflector. The solar cell without any light
trapping. flat ZnO and Ag back contact, show low jsc
(14mAlcm2) and VOC(490mV). On the other hand, solar
cell performances linearly improve by just enhancing light
scattering at a surface of ZnO films. The increase of jsc
with surface texturing of ZnO films resulted from the
longer travel length of incident light within the i-layer.
However, the improvement of VOCand FF has not we[l
understood until now. The results by Nasuno el all4] is
opposed to our experimental results. They have observed
the reduction of VOCand FF in solar cells deposited on
textured ZnO films, which is caused by increased
crystalline defects within the poSi:H films deposited on
textured ZnO and Asahi-U glass. The insertion of ZnO
thin-layer behnreen n a-Si:H and Ag back metat is very
effective for improving jsc of solar cell as shown in figure
(c) and (d). It should be noted that high jsc (22mAlcm2)
and efficiency (8.5%) can be achieved in a solar cell with
thickness of only lpm i-layer by enhancing light trapping
scheme. The 4mAlcm' increase of jsc with optimizing

.
.
.

3(D

54m

600 700

900

900

1OOO1IM

WaVeh$th(ma)

Fig. 4. AM 1.5 l-V curves and EQE of pc-Si:H thin-film

solar cells with dflerent Iigbt trapping schemes.

3.3 a-Si:Wpc-Si:H tandem cells

Fig. 5 shows the effect of a-Si:H top-cell thickness on
the operation properties of a-Si:H/pc-Si:H tandem cells.
The thickness of peSi:H bottom cells was fixed to be
1 . 8 and
~ ZnO:AI/Ag used for back reflector . With
reduced thickness of a-Si:H top cell, jsc of top cell
decreases from 14.8Wcm2 at 380nrn thickness to
12mA/cm2 at 190nm. Contrary to this decrease of togcell
current, jsc of a-Si:WpcSi:H tandem cells continuously
increases with reduced topcell thickness. The jsc
discrepancy between a-Si:H top-cell and a-Si:H/pSi:H

1554

and A. Nakajima, Technical Digesf of the fntemafional PVSEC12, Jeju, Korea. 547,2001
[g] J. Meier et al. Solar Energy Mafe!fa/s & Solar Cells 49, 35,
1997
[9] Y. Tawada. H. Yamagishi, K. Yamamoto. Solar Energy
Materisls and Sdar Cells 78, 647-662,2003
[IO] Franc Smote, Marko Topic, Jore furlan. J. of Non-Crystalline
Sdids 194,312-318,1996
[ll] M. Kubon. E.Boehmer, F. Siebke, B. Rech, C. Beneking and
H. Wagner, Sdar Energy Materials and Solar Cells 41-42, 485492,1996
[ 121 Corinne Dror, Ph. D. Thesis, Universityde Neuchatel, 2003
[I31 H. Stiebig. T. Brammer, J. Zimmer, 0. Vetted. H. Wagner, J.
ofNon-CrysfaI/imSdaS 266-269, 1104-1 108,2000
[I41 Y. Nasuno, M. Kondo and A. Matsuda, Mat. Res. Symp.
P~oc.W. 664, 2001, A15.5.1-A15.5.11.

tandem cell, therefore, reduces from 4.5mAlcm2 at 380nm

top-cell thickness to fess than lmA/cm2 at 190nm. This
improved current matching at reduced top-cell thickness
means that current of tandem cells is limited by pc-Si:H
bottom cell. As previously explained, the pc-Si:H bottom
cell absorbs extra part of incident photons penetrating aSi:H top cell, it is necessary to increase thickness of pcSi:H bottom cell and to improve light trapping for better
current matching. The important things that should be
noted from fig. 5 is a low fill factor (FF) of tandem cells
prepared both on 2nO:Al TCO compared to Sn02:F and
on thin a-Si:H top cells. In case of SnO2:F TCO, the FF of
a-Si:H top cell increases with reduced thickness.
However, FF of tandem Celts decreases from 73% at
380nm to 67% at 190nm. The reason for this inverse
proportion of FF of top and tandem cells with thickness is
not yet understood. The further increase of jsc and FF
should be done to increase efficiency of tandem cells.
Possible approach to increase of jsc of pGSi:H bottom
and tandem cells is to enhance light trapping by using
front TCO and back reflector.

1A

13 12

- - - - -I- - - - - __ _ _ -

A-)-A
,

-.-

E :

Augbss (tandem)

1.1

-19

-+&ON

$11:

03:

tV. CONCLUSION

-.1

OJ-

this paper, device structure, process and

performance of Si-based thin-film solar cells (Si-TFSC)
are briefly introduced. The a-Si:H/pc-Si:H tandem cells,
composing of wide gap a-Si:H top cell and narrow gap pcSi:H bottom cell, has a great potential to overcome the
obstacles of present PV technology that uses high cost o
Si solar cells. The outstanding merits of Si-TFSC over the
conventional c-Si solar cells are reduced materiaf cost by
thin-film (less than 2% material usage compared to c-Si
wafer), high throughput by in-line process, possible usage
of a variety of substrates such as low cost glass, plastic.
metal foil or sheet, etc by low temperature process.
However, for these thin-film solar cells to be competitive
with conventionat technology, the following technical
obstacles should be solved;
- low efficiency (currently, small area cells 14.7% and
production modules 10%)
- light degradation of a-Si:H solar cells
- low deposition rate of silicon layers, especially pc-Si:H
large efficiency gap between small area solar cells and
large area modules
- law insight on material and device physics.
In

a7
M

.,
4

RETERENCES
[l] Paul Maycock, Renewable Energy World, July-Augusf 2003,
pp. 84-101.
[2] D.L. Staebler and C. R. Wronski, Appl. Phys. Lett. 31, 292294,1997
[3] .I. Meier, S. Dubail, R. Fluckiger, D. Fischer, H. Keppner, A.
Shah, I" WCPEC, Dec. 59,409412, 1994
[4] 0.Velterl, F. Finger, R. Carius, P. Hakpe, L. Houben, 0.
Kluth, A. tambertz, A. Muck, 6.Rech, H. Wagner, Solar Energy
Materia/$&Solar Cells 62, 97-108, 2000
[5] N. Wyrsch, C. DOE, L. Feitknecht, M. Goerlitrer. U. Krolf, J.
Meier, P. Tortes, E. Vallats-Sauvain, A. Shah, M. Vanecek, MRS
Sping Meeting 2000, Vol. 609
[6] R. W. Collins, A. S. Ferlauto, G. M. Ferreira, Chi Chen,
Joohyun Koh, R. J. Koval, Yeeheng Lee, J. M. Pearce, C. R.
Wronski. Solar Energy Materials & Solar Cells 78, 143-180,2003
[7] K. Yamamoto. M. Yoshimi, Y. Tawada, S.Fukuda, T. Sawada.
T. Msguro, H. Takata, T.Suezaki. Y . Koi, K. Hayashi, T. Suzuki

Fig. 5. IV parameters under AM 1.5 illumination of aSi:H/pc-Si tandem solar cells as a funcfion of
a-Si:H top-cell thickness.

1555

FAST QUANTUM EFFICIENCY MEASUREMENT AND CHARACTERIZATION

OF DIFFERENT THIN FILM SOLAR CELLS
BY FOURIER TRANSFORM PHOTOCURRENT SPECTROSCOPY
A. Poruba 1,2, L. Hodakova 1 A. Purkrt 1, J. Holovsky 1 and M. Vanecek 1
1

Institute of Physics, Academy of Sciences of the Czech Republic, Cukrovarnicka 10, 162 53 Prague 6, Czech Republic
2
Solartec s.r.o., Televizni 2618, 756 61 Roznov pod Radhostem, Czech Republic

ABSTRACT
Fourier-transform Photocurrent Spectroscopy (FTPS)
was introduced four years ago [1] as a method for fast and
very sensitive evaluation of the spectral dependence of
the optical absorption coefficient of photoconductive thin
films and recently also for the quality assessment of thin
film silicon solar cells [2]. In this contribution we refer
about the FTPS characterization of different thin film solar
cells (amorphous silicon, microcrystalline silicon,
micromorph tandem, polymer solar cells) with the main
target of a fast quantum efficiency (QE) measurement.
Further, we study the possibility of the FTPS-QE
measurement of single cell incorporated in the solar
module when only the module terminals can be used for
contacting. The FTPS measurement has proved to be very
fast (time scale of seconds), enabling a quick verification
of quantum efficiency and subgap absorption
measurement of absorber of different thin film solar cells.

interface for the external detector. Our solar cell

connected into an electrical circuit with current preamplifier
(and alternatively with an ultra low noise voltage source)
serves as the external detector.
We normalize the FTIR signal from the sample with
the signal from spectrally independent detector
(pyroelectric detector from deuterated triglicine sulfate,
DTGS, or Si calibrated photodiode). Of course, a different
frequency dependency of measured samples and
detectors have to be found (e.g., from the measurements
at different scan velocity) and taken into account.

sapphire window
mirror

FTIR
spectrometer
sample
(thin film cell)

INTRODUCTION
Quantum efficiency measurement belongs to the group
of solar cell characterization providing complex information
about its quality. It is used both in the research area of a
new cell structure testing as well as for the reproducibility
verification in production of solar cells and modules.
Relatively long measurement of QE curve by standard
QE set-up (light sources, monochromator, chopper, lock-in
amplifier) is disadvantage of this method. In this paper we
refer about the new possible QE set-up which affords to
obtain the data within few seconds and thus it gives the
potential for its integration into the fast and easy-to-use
method for global cell characterization in the R&D area or
production process.
Basic idea of the new method utilizes time-variables
spectrum of the illumination performed by FTIR
interferometer. Response of the measured solar cell is
then compared with the background spectrum from the
absolute detector or calibrated solar cell. A scheme of
the utilization of the Fourier transform infrared (FTIR)
spectrometer in the FTPS set-up is shown in Fig.1. FTIR
spectrometer equipped for the near infrared and visible
region (quartz beamsplitter and white light source) has
possibility of an external light beam output and an

1-4244-0016-3/06/$20.00 2006 IEEE

Current
preamplifier

A/D convertor
and interface
Fig.1. Scheme of the Fourier Transform Photocurrent
Spectroscopy (FTPS) set-up used for the QE
measurement.

EXPERIMENTAL DETAILS
We have used the FTPS method for the QE
measurement to characterize different thin film solar cells
(amorphous, microcrystalline and micromorph silicon and
organic solar cells). First of all, QE-FTPS results were
carefully compared with the independent measurement
obtained by standard QE set-up. This comparison was
done for the p-i-n structure of a small area microcrystalline
silicon solar cell having conversion efficiency of 8.3 %.

1525

f F = 2vs

disturbed by different Fourier frequency corresponding to

the wavenumber range of interest. Moreover, to check the
linearity of the FTPS signal spectra were also measured
with ten times lower light intensity as well as under white
light bias. The shape of the all three curves remains
identical over the whole measured region confirming the
correct conditions for the QE data interpretation.
FTPS of reference c-Si detector
FTPS of mc-Si:H p-i-n solar cell

100

relative FTPS signal (a.u.)

Measured FTPS spectrum of this solar cell has been

normalized (after its frequency dependence correction) by
the FTPS spectrum of the commercial calibrated silicon
detector.
After this verification of the correct QE evaluation, the
same approach has been used for testing the possibilities
of QE measurement of tandem micromorph structures [3]
and of a solar sub-cell incorporated into the solar module
when only module terminals are available. The same
experiment was then applied for the homogeneity study of
the large area solar module (30 x 30 cm) consisting of 27
series connected a-Si:H p-i-n single junctions.
The trial of the QE measurement of organic solar cells
was the last experiment of a thin film cell characterization.
The result of this spectroscopic study is presented for
blends of organic donor (P3HT) and acceptor (PCBM)
materials, used as absorber in bulk heterojunction solar
cells.
For all the measured FTPS spectra their light intensity
dependence have been studied using different neutral
density (mesh) filters varying the light intensity over two
orders of magnitude, to verify the signal linearity.
In a continuous-scan FTIR spectrometer, the moving
mirror moves continuously at a constant scan velocity, v
(cm/s), and the optical path difference at time t (s) is given
by d = 2vt (cm). The interferogram data points are digitized
at the zero crossings of a HeNe laser signal on the fly. The
use of laser signal ensures that I(d) is measured at
precisely equal intervals of mirror positions and provides
an internal wavelength calibration for every scan. Because
of the continuous movement of the mirror, the
interferogram I(d) becomes an explicit function of time.
The Fourier frequency (fF) of the interacting light at
wavenumber s is given by:

80

60

40

20

400

1000

1200

Fig.2. Single beam FTPS signals of mc-Si:H cell and c-Si

detector obtained using blue optical filters.

Figure 3 shows FTPS spectrum of mc-Si:H cell from

Fig.2 divided by baseline measurement of c-Si detector
(also shown in Fig.2) and finally multiplied by QE curve of
this detector.

QE of mc-Si:H cell measured by:

standard QE set-up
FTPS set-up

0.8

Of course, the frequency dependence of the sample

has been analyzed from the FTPS data measured with
four mirror velocities. Due to the relative small area of the
sample and therefore also relative small capacity its
frequency dependence is absolutely negligible up to the
frequency about 10 kHz. It means that FTPS spectra
measured at low mirror velocity of 0.158 cm/s are not

800

wavelength (nm)

quantum efficiency (-)

EXPERIMENTAL RESULTS
For the all FTPS measurements in the range 400
1200 nm the set of various optical filters has been used to
optimize the light intensity in the whole spectrum and by
this way to limit the necessity to measure over the high
dynamic range. The baseline measurement has also been
taken with the same filters, but due to the sensitivity of the
c-Si reference detector (contrary to the thin film solar cells)
also in the near IR region, special care had to be applied
for the blue/violet region of a low light intensity.
Figure 2 shows the single beam FTPS signal of a
microcrystaline silicon p-i-n cell together with the FTPS
result of the calibrated c-Si detector. As it was mentioned
above the spectra were scanned with the same blue
filters reducing the red light intensity and thus having the
relatively flat response for the measured thin film cells.

600

0.6

noise from
He-Ne laser

0.4

0.2

0.0

400

600

800

1000

wavelength (nm)
Fig.3 Quantum efficiency results evaluated from standard
QE set-up and using fast FTPS measurement.
Final curve is compared with the independent QE
result measured by standard QE set-up and very good
agreement can be seen except the slightly noisy result
close to the UV range due to the extremely low light
intensity from halogen lamp inside the spectrometer.

1526

In Fig. 3 we can also observe a noisy signal around

the wavenumber of HeNe laser beam which cannot be
fully removed but fortunately this very narrow peak does
not influences the shape of the measured curve.

These data highlights the possible troubles of thin film

growth performed close to the amorphous-microcrystalline
conditions for the large area deposition.
100

QE - FTPS spectra of the a-Si:H modules

90

60
50
40

20
10
0

400

500

100

800

QE results (in log. scale)

measured in 6 different spots
of the a-Si:H module

10

50
40

20

700

Fig.5 Relative QE results of a-Si:H module grown close to

amorphous-microcrystalline regime measured in 6
different places over the module area.

60

30

600

wavelength (nm)

QE-FTPS measurement when:

single cell contacted
module terminals used

70

70

30

relative QE

relative QE results (%)

80

relative QE

Demonstration of the FTPS capability for the QE

measurement of the solar cells incorporated into a solar
module when only the module terminals are available was
the second target of this paper. We chose the a-Si:H minimodule where all sub-cell contacts are available. We
started with QE diagnostic of an individually contacted cell
which then was compared with the result of QE of the
same sub-cell when the module terminals were used. In
this case all other sub-cells have to be white light biased
to generate 10-15% higher DC current compared to that of
pseudo-DC current from the probe beam of the FTIR
spectrometer. The measured cell has been shadowed
from the DC bias light by black sheet with the aperture just
for the probe beam. From Fig.4 we can observe excellent
agreement of both spectra. The spectrum measured using
module terminals exhibits only slightly enhanced noise
level.

80

QE results measured
in 6 different spots
of the a-Si:H module

90

a-Si:H mini-module
(5 sub-cells)

"partially"
microcrystalline
region

0.1

10
0

400

500

600

700

0.01

800

wavelength (nm)
Fig.4 Relative quantum efficiency results of the measured
spot of one a-Si:H solar cell incorporated into a solar
module. Heavy line represents result when directly
measured cell has been contacted, dotted line shows
result where the module terminals and white light bias for
non-active sub-cells were used.
The experiment described above was used for
homogeneity study of the large area a-Si:H solar module
(30 x 30 cm) grown close to the amorphousmicrocrystalline conditions consisting of 27 series
connected single junction cells. Non-homogeneity in
thickness of the active layer or of the top TCO is usually
well visible by a naked eye due to the light interference
effect. Here we investigated influence of nonhomogeneities of three sub-cells measured in two different
spots on the local quantum efficiency. Fig.5 shows these
results in the linear scale, while Fig. 6 shows the same
data plotted in logarithmic scale of QE from which
differences in the material properties are better visible.

1E-3
400

"fully" amorphous
region
500

600

700

800

900 1000 1100 1200

wavelength (nm)
Fig.6 Relative QE results of a-Si:H module (data from
Fig.5) in logarithmic scale grown close to the amorphousmicrocrystalline regime measured in 6 different places
over the module area.

QE - FTPS spectra of micromorph tandem cell

The principle of the QE measurement of cells
incorporated into the stack (tandem) is similar with the
concept of the experiment where we measure response of
a sub-cell using the module terminals. One cell in the
tandem has to be light biased to make it conductive but
this bias light must not be absorbed by the measured cell,
so low or high pass filter are used to choose the spectrum
for the correct bias light.

1527

120

-6.4 V

80

effect of reverse bias on QE

of P3HT:PCBM solar cell

relative QE (a.u.)

relative QE (-)

100

60

a-Si:H cell

40

20

-4.8 V
80

40
20
0

600

800

0V, QE curve multiplied

by the factor of 5

-1.6 V

mc-Si:H cell

400

-3.2 V

60

1000

1200

400

500

600

700

800

wavelength (nm)

wavelength (nm)

Fig.8. Relative QE results of P3HT:PCBM (encapsulated)

solar cell measured in the Isc conditions and reversed
voltage biased (-1.6, -3.2, -4.8, -6.4 V). Curve measured
in the 0 V bias is shown multiplied by the factor of 5.

Fig.7. Relative QE-FTPS of the micromorph tandem solar

cell measured sequentially by low and high pass optical
filters

QE - FTPS spectra of organic cell with the bulk

heterojuction
With the discovery of the ultra fast electron transfer
between excited conjugated polymers and fullerenes, a
promising
destination
was
found
for
organic
semiconductors, using them as low-cost, lightweight and
easy-to-produce solar cells [4]. The combination of
regioregular poly (3-hexylthiophene) (P3HT) as a donor in
conjunction with (6.6)-phenyl-C61-butyric-acid (PCBM) in
a 2:1 weight ratio lead to achieved conversion efficiencies
from 3.5% to 5% [5, 6] opening up commercial
perspectives for organic solar cells, on condition that these
efficiencies continue to improve substantially.
Recently, FTPS method has been applied for such
type o solar cell to highlight the electronic structure of the
composite [7]. Here, we report about the quantum
efficiency measurement for the same type of organic solar
cell after its temperature treatment (80C, 1 hour). QE
spectra were measured in Isc condition and, to enhance
the charge transport mechanism and collection efficiency,
also using the reverse voltage bias [8].
Fig. 8 shows the relative QE-FTPS measurements
performed at the external voltage bias 0, -1.6, -3.2, -4.8
and -6V. Data for the QE measured with no voltage bias
are multiplied by factor of 5.
CONCLUSIONS
This paper reviews the possibility of the fast QE
measurement of thin film solar cells using the FTPS setup. Measured and normalized FTPS spectra were
compared with results from standard QE set-up and very
good agreement has been found. Moreover, we have
demonstrated the suitability of this method for
characterization of solar sub-cells incorporated into the PV
module when only the module terminals are available and
also for QE measurement of micromorph tandems and
solar cells based on organic bulk heterojunctions.

Acknowledgement
This work was financially supported by the Czech Ministry
of Environment (project VaV-SN-3-172-05), EU projects
LPAMS (INCO-CT-2004-509178) and ATHLET (019670),
BIL Cooperation Flanders - Czech Republic BOF04B03
and the Institutional Research Plan AVOZ10100521.
REFERENCES
[1] M. Vanecek, A. Poruba, Appl. Phys. Lett. 80 (2002)
719
[2] A. Poruba, J. Springer, L. Mullerova, M. Vanecek, T.
Repmann, B. Rech, J. Kuendig, N. Wyrsch, A. Shah, Proc.
3rd World Conference on Photovoltaic Energy Conversion,
Osaka, Japan, 2003
[3] J. Meier, S. Dubail, J. Cuperus, U. Kroll, R. Platz, P.
Torres, J.A. Anna Selvan, P. Pernet, N. Beck, N. Pellaton
Vaucher, Ch. Hof, D. Fischer, H. Kepner and A. Shah, J.
Non-Crystal. Solids 227-230 (1998) 1250
[4] N. S. Sariciftci, L. Smilowitz, A. J. Heeger, and F. Wudl,
Science 258, 1474 (1992)
[5] F. Padinger, R. Rittberger, and N. S. Sariciftci, Adv.
Funct. Mater. 13, 85 (2003)
[6] C. J. Brabec, Sol. Energy Mater. Sol. Cells 83, 273
(2004)
[7] L. Goris, A. Poruba, L. Hodakova, and M. Vanecek, K.
Haenen, M. Nesladek, P. Wagner, D. Vanderzande, L. De
Schepper, and J. V. Manca , Appl. Phys. Letters 88,
052113 (2006)
[8] W.D. Mihailetchi, L.J.A. Koster, J.C. Hummelen, and
P.W.M. Blom, Phys. Rev. Letters 93, 21 (2004)

1528

A New Solar Cell Model for Evaluating Degradation

of various il icon Cells at LEO and eO r its
A.R.

Fasooniehchil, M. Taherbaneh23

'Iran Telecommunication Research Center, A.Fasooniehchigirost.org

2

Iranian Research Organization for Science and Technology

Electrical & Computer Department, M.Taherbanehgirost.org
3Amir Kabir University of Technology
Electrical Engineering Department, Taherbanehgcic.aut.ac.ir

Abstract-In this paper a new solar cell model is presented. The

We have simulated the effects of space environment on
model is used for evaluating the performance of various silicon silicon solar cells behavior. For this purpose we have modelled
solar cells in LEO and GEO. Irradiance, temperature, and the cell. A solar cell model is introduced using radiation as
accumulated dose of radiation are considered as the model's input well as irradiance and ambient temperature as the inputs. The
parameters. In this research, some types of terrestrial and space- model is a modified version of general solar cell "One Diode
qualified silicon solar cells are investigated and evaluated. N- Model" [3]

type-base and poly crystalline silicon are selected as terrestrial

solar cells. High Efficiency Silicon (HES) cells are selected as

ModeL"i[3].

e/cm, respectively. It is finally concluded that radiation

sensitivity of poly crystalline silicon solar cell is very close to HES
cell, but it has lower generated power. Also, N-type-base silicon
solar cell is the most radiation sensitive cell with the lowest
generated power at EOL, while HES solar cell is the least
radiation sensitive cell with the highest generated power at EOL.

simulation results. Section V concludes the paper.

imuATLAB/SimulWnk environment iS used for the purpose of

space-qualified solar cells. The selected HES cells are bulk single- simulation. We evaluated solar cells performance based upon
crystal silicon solar cells with 2 ohm-cm base resistivity and the maximum generated power at various conditions.
different cell thicknesses (2 mil and 4 mil). End Of Life (EOL)
The rest of paper is organized as follow:
conditions for LEO and GEO orbits are assumed to be equivalent
Section II descirbes the modeling issues. Simulation issues
to degradation due to 1MeV electron fluences of 1e15 and 5e14 are discussed in section III. Section IV devoted to presented
2
II. MODELING ISSUES

in this research, we were involved in two main tasks: space

environment simulation and solar cells modeling.

I. INRODUCTION
A. Space Environment
In this paper, standard conditions are assumed Air Mass Zero
Solar cells, in space, are mainly affected from irradiance
level, temperature and particles radiation [1,2]. Fig. 1 shows or AMO, with irradiance and temperature values of 1367 w/m2
the electrical characteristics of solar cells that we have used to and 28 C [2,4]. Variation of space environmental parameters
in two orbits is determined as follow:
describe cell behavior in space applications.
Cyclic profile for irradiance in LEO and GEO is used,
because of daylight and eclipse times [1,2]. Ambient
I
MZIXifllU]fl
temperature is calculated using irradiance variation, earth
_
LV cumte XorXot X
X
=albedo and IR emissions. Earth albedo and IR emissions have
no significant effect on GEO, so they are only considered for
p onsiant

Volag
X~ ~~~Cl 11|(V)

Radiation is the main environmental factor in space that

affects solar cell performance [4]. We only considered its
cumulative effects over a long time, because it is the main
reason of solar cell degradation. Radiation mainly includes UV
_~
\
_radiation, and energetic electrons and protons. Ionizing
radiation, containing UV and energetic electrons, reduces
transmittance of solar cell cover glass. Atomic displacement,
caused by energetic electrons and protons, is the basis of cell
damage. Displacements form stable defects in cell base region,
/
which reduce the life time and diffusion length of minority
S~~~~~~~~~~~~~carriers, and permanantly degrade solar cell [5,6]. LEO and

1~~~~~
Soa
celIVcaatrsi.uveadmxmmotu
Fig.

1-4244-1057-6/07/$25.OO )2007 IEEE.

oe

types of energetic particles as shown in

protons and inner zone electrons for LEO, and
~ GEO
~ ~ ~ ~ ~ I.~tinclude
ableTrappedvarious

768

solar flare protons and outer zone electrons for GEO are the
most common particles that affect cell performance.

G
The radiation input ((O) of the cell model, is the "equivalent
D
Ta
1P
1MeV electron fluence" [4,5,6]. The equivalent fluence is
determined by applying Relative Damage Coefficients (RDCs)
of the specified cells to the omni directional proton and
electron fluxes in the specified orbit, which are output data of
AE8 and AP8 models [4]. The radiation data is also verified
Fig. 2. New solar cell electrical model with radiation input
with the certified European Space Agency's web-based
software, "Space Environment Information System D.
and Temperature
D. Equationsfor Irradiance and
Irradiance has utiize
a cyclic pattern, depending on the orbit [2,8].
(SPENVIS)",
(SEVS1 which utilizes AE8, AP8 and EQFLUX models
[4,12].
The values of eclipse time and orbital period are given by (1),
(2).
B. Solar Cell Modeling
T T -(I/180o) -cos1 [(I-(RE )2)1/2 /COSfi
Radiation, as a new input ( O), is added to the general "one
(1)
E~~~~~~
a
diode model" of solar cells, to form a new cell model as shown
in Fig. 2. To avoid additional complexity of mathematical
T=2f afu
S
modeling, the simplest solar cell model is used [9,10]. Since
(2)
the results are used for a comparative evaluation of cells,
Where RE=6372 km, a=RE+Altitude, and ,u =3.986*10'
possible inaccuracy of the model is not a very critical subject.
Complete models with more accurate results will be presented kM3/S2. / is the angle between sun vector and orbital plane.
in future works.
Equation (1) gives eclipse time, only if ,6 < 6 , where

AE8,lc andmperatureel

whichon

Table 11 shows the specifications of selected si solar cells [24]. Selected cells include: n-type-base silicon cell, polycrystalline silicon cell, and High Efficiency Silicon (HES) solar
cell.
With cover glass thickness of approximately 100 ,um, only
protons with the energies higher than 4MeV can penetrate into
the cell base region [4].
There are essential differences between selected terristerial
and space-qualified solar cells.
Differences include: p-type vs n-type base , textured front
surface, thin cells, and Back Surface Reflectors (BSRs) in
space cells, especially H-ES cell. Space cells also have an
additional layer over the cover glass that deceeases glass
darkening and increases absorptance ofthe cells [3].

PARTICLEs ENERGY
Electrons Electrons Trapped
Protons
(LEO)
(GEO)
(LEO) PaGEicl) Eey (LEO)
0 - 100
-7 0 Ener1 00.
Pt(MeV)

Type of Particle

Solar
Flare
Protons
(GEO)
1 -200

TABLE II
SPECIFICATIONS OF SELECTED SILICON SOLAR CELL TYPES
silicon solar cell types (CG 100 urn)
Efficiency 7 (%)

(1,/7) dl

(/O

C)

Absorptance
Thickness( ,Umn)

T C=Tell =Ta +[G.(noct-20)/800]

(3)

Maximum cell temperatures for LEO and GEO are assumed

+80 C and +60 C, respectively [2,8].

E. Cell Equations

TABLE I

Parameter

ff = sin'(RE / a). For , > ,* the eclipse time is zero.

Irradiance maximum value for GEO is assumed Gr-1367
W/m2. Albedo effects cause some changes in the irradiance
value for LEO as shown in table III.
Ambient temperature (Ta) for GEO is proportional to
irradiance. Temperature for LEO is a function of irradiance,
albedo, and earth JR [3,4,15]. Cell temperature is related to
ambient temperature and irradiance as (3), which uses nominal
operating condition temperature (NOCT) [3].

n-type-base

poly-Si

13.2

14.8

HES
(2 Q - cm)
16.6

-0.55

-0. 46

-0. 35

0.75
150

0.8
100

0.85
50, 100

Main electrical characteristics of solar cells are short cicuit

current (Isc), open circuit voltage (Voc), maximum power
(Pmax), and voltage and current of maximum power point
(Vm,lm). Cell model parameters are Iph, Isat, and Rs [9,10].
All of the parameters are functions of irradiance (G), cell
temperature (T) and cumulative radiation fluence (y). Cell
model equations are presented below. Initial modeling

equations are adapted from [5,9,10]. The photo generated

current is almost the same as the short cicuit current (4). It has
a strong dependency on G, T, and (0.
TABLE III
IRRADIACE, ALBEDO, AND IR EMISSION FOR LEO

|Parameter
DEnsity

769

|Sun Irradiance |Albedo

15+5(Wm)

30 5(0%)

Earth JR

(W/m2

(4

I
'SC

:-

'ph

VD VOCnVTln(1+ISC/ISAT)
.

'SC

'SAT

(5)r

loo

FF(

' FFO)

Irradiance and temperature profiles for LEO and GEO at

worst case are shown in Fig. 3.
At the Begining Of Life (BOL) the cumulative radiation qo
is equal to zero. Model parameters JPH, JSAT, and Rs are
calculated as functions of G and T as stated in (9) to (12).
dT BOL (T-T,)]*GIG

JSAT (T) =EJ(A(Tl

EXP( KT
T,r Exp(KTr
~SATh
TITrY

Eg (T)
KT

ph

x10

(13)

/nVT)1)

(14)

ID = ISAT (EXP(VOCI V)1)

(4

VT * ln(I + Jsc (G,T)IJSAT (T))

Jm (GCT)[ Jmr + dsc

T

BOL

(15)

*(T-T,)]*G/G,

(16)

And, Vm and Pmax are calculated from (17) and (18).

Vm (G, T)

sVc
-

ds

'ph

Vo(G, T)

currents.

(9)

1 1

.sc (G, T)+- Im (G, T) (Ex VOc (G, T)

T()

nVT

1)]
1

(17)

RS I,, (G, T)

Where ISc =Jsc*A and Im=Jm*A.

Pmax(G,T)=Vm(G,T)* J,(G,T) *A

(10)

(18)

After some time in orbit operation, cell is affected by

absorption ofthe cumulative radiation y0. Short circuit current

Where ISATr is given in (6).

0.0004.73xT2

Eg(T)=Eg (0)~ 636+T

+

Time (Seconds)

Jsc (G,T)> JPH(G,T)

(8)

Cell electrical characteristics for AMO (with G, and T, values

1367W/m2 and 280C) are Vocr, Jscr , Vmr , Jmr , Pmax,r dJsc/dT ,
dVoc/dT. All of these parameters are accessible from the data
sheets of the cells. They are used in calculating cell's "in orbit
characteristics" such as Voc(0,G,T), Jsc( (0,G,T), etc. In the
modeling we have used densities of currents instead of

Rs (G, T)= Rs

LEO

Electrical characteristics Jsc, Voc, Jm of the cell are then

calculated
as functions of G and T as stated in (13) to (16).
(7)

[(Voc /n VT) - ln(Voc In VT + 0.72)] /[1 + (VOC /nVT)]

-200

0o

n/-

Fig. 3. Irradiance and Cell Temperature Profiles for LEO and GEO

(VmJm)A(VOCJIS() RS=0-

JPH(G,T)I[JsCr

Timc (Scconds)
n/- /n-f=

---

resistance is a function of G and T.

A)- Pma. /(J2

f-

(6)

Series resistance can be calculated from (7), in which FFo is

the cell's fill factor, when Rs is zero, given by (8). Series

R = Voc I(JAsc

14:.

5100

Diode voltage is the cell open circuit voltage as stated in (5).

(6). It is a function of T and

6EO

Gn lOf i --n- n

()1500

a(T T) + 3(G
-

Gr )

density isc (o, GCr, Tr)

(11)

be written as (1 9). A similar

relation can be used for calculation of J (q, Gr, Tr) Cell
Cr ) can be obtained by
open circuit voltage VOC ((o TrG

(12)

substitution of Jsc with Voc, Ki with Kv, and Q with 91/ in

(19). Then JSAT((,Gr,T) is calculated from (6).

Where ar, ,1 are positive constants.

can now

JSC (fJ9Gr,nTr ) =JSC (0GCr,nT, )-KJ .1og1j1 +qjq2)

(19)

V0c (f' C,G Tr) is calculated from (15) Using (20) and (21).

770

J~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~(~ ~ ~ ~ ~ ~ ~72,G,Tr) J~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ (V2~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ GrJr) C/Cr (20)~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . . . .

JSAT(V~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~G~ ~ ~ ~ ~ ~ ~ ~Tr>JSAT(V2~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ GrJr) (21)

~~~~~.....................................................

~~~~~~~(~ ~ ~ ~ ~ ~ ~ ~ ~ o~ ~ ~ ~ ~ C~ ,T)canbecalculated
(2 )from
.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . .
~~~~~~~~~~~
dJ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~*(T~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Tr)]*C/Cr
~ ~dT~ FOL
(2 )~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...................................................
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . .

Similarrelationcanbeusedfor calculation
~of~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ m(...............o,C,T). . . . . . . . . . . . . . . .
AndV0~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ (~ ~ ~ ~ ~ ~ ~ ~ o(~2~3)~.~C,~ ~T~c)~anbe~ca~lc~ul~fromat~ed~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . .
CelL
odel ain.........................................

dV

nV

Fig.
MATLAB/Simulink.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
inmodel
4. Cell

Voc(~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ;9,C,T)~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Voc(~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~;~9,C~,T~r)~+ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . .

dTFOLr (23) Firatos

sbti;lithioetwshe

ofpresented

inpredictimodel
ng~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . .

solar...cell..performance,....and
G1nscJG
G(G,p2C,T)..the..effects...of.space...environment....on.
second; to com pare various teristerial and space solar cel s in~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
space ap lications.
tLhEeO spaacned apEOlicationo,rbits,
5-year-period~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Then~~ ~ ~ ~ proper
~ ~ ~ ~ ~ ~Syp,Aform
TCT,i)caslfromculated of (6). in-orbit eration at Cells
p~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~is~ ~ ~c~on~si~de~re~d.~~ ~~ ~~ ~~ ~~ ~~ ~~~~~~~~~~~~.~.~.~.~.~. ~. ~. ~. ~. ~. ~. ~. .~ .~ .~ .~ .~ .~ .~ .~.~.~.~.~.~.~.~.~.~. ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
AndfinallyVm(~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~C,o, T)and~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ (to,
maRemaining
x C,Ta)recalculated Factor(PRF)ofthecel s.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
from(24)and(25).~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . . .

Th~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . .

Vm(~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ fl,G,=T)
EZD

G~~~~~~~~~~~~~~~~~~~~~~~................................................

Th(C)
(Fi,G,T)
J~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..(ID..........................................
V0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~(q9,G,T) (24)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.....................................................

flVT ln(1+ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~'~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~T

~ ~ (e~SC

-1))

D
C Fi(~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.............................,...2).......HHH

R~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ JmQ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~
~ ~ ~ ~ ~ ~ fl,G~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.............................................
,T)~ ~ ~
Pmax(G,T>~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Vm(G,T) Jm(G,T) (25)~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . .

AloadresistanceRisconnectedtotheoutputportofFig.2.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . .
fordrawingI-Vcharacteristiccurveofthecell. Socellcur ent~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . . .
andvoltage(I,V)arerelatedtothecellmodelparametersas~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . . .
statedin(26).WhereV~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . .RI. . . . . . . . . . . . . .
J(p,G,T) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ J5~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~@p,G,T) -~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . .
'SAT (~~~~~~~~~o, T)(((

P~~~~~~~~~~~~~~~ C, T)

R5 J(tp, C, T)

(26)

FV(eI~ ~ FiP~~~~~~~~~~~~~~~~~~....................................................
~ ~(1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

FiP(ek~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~n2)

ThecellmodelisimplementedinSimulinkashownFign.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . . .
includes
4.Themodel
some in~ ~ p~ar~am~et~e~rdse~pe~nd~specifications.
nd. . . . . . . . . . . .internal
inJogSTnceerlnasl y. . . .defined. . . .constants. . . . aMain
VP~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ In,..........................i,G,T)G. . . . .D. . . . . . .
proces eas reshowniTheFnig.5.
model
is capable
of~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . .
applyingvariousvaluesotemperature,
fir adiance,
and~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . . .
radiation.Simulationfollowstwomaingoals.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~. . . . . . . . . . . . . . . . . . . . . . . . . . .
RI~~~~~~~~~~~~~~~~~~~~~~~...................................................

7 1~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ . . . . . . . . . . . . . . . . . . . . . . . . .

For each of the selected cells, the model is used with

different settings.
Settings include: cell's electrical
characteristics at AMO (y = 0,G,, Tr) condition, transmittance
of cover glass, and degradation of the cell thermal coefficients
(e.g. dJsc/dT) due to the radiation. Irradiance and Temperature
are applied to the cell model with the profiles shown in Fig. 3.
The value of (o after 5 years in-orbit operation (end of life) is
supposed as lel5 and 5el4 for LEO and GEO, respectively
[3,13].
Simulation is done for various environmental
conditions from BOL to EOL. At first for evaluation of
degradation of the selected cells in AMO condition due to the
radiation. Secondly for comparison of degradation of the
selected space qualified cells in AMO, GEO, and LEO
conditions. And finally for continuous variation of G and T at
some constant values of (O. The last simulation evaluates the
dependancy of cell maximum power on G and T after specified
time periods.

1
0.9

BSR2ohm-cm2mil
er.8uR
PolySi

0.7
N

n-type base Si

0.6
0.5
0.4

10

12
10

10

13

10

14

10

Equvalent IMeV Electron Fluence (e/cm2)

10

Fig. 6. Radiation Effects on Various Solar Cells at AMO

0.02

IV.REEsULTS

E 0.018-

C.)l_

Fig. 6. shows the degradation of normalized maximum

power of the irradiated cells, in AMO conditon. All cell types
are participated in this simulation. Results show that the PRF
for n-type base and poly Si cells due to the radiation fluence
value of 0= lel5 e/cm2, is 0.45 and 0.63, respectively.
Fig. 7. shows the simulation results for Pmax degradation of
two types of space-qualified silicon cells in LEO and GEO
conditions. As shown, two selected BSR cells both with base
resistivities of 2Q -cm differ significantly in Pmax and PRF

BSR 4 mil at AMI-O

' 0.016

S 0.014 .......
E

0.012-

0.01-

v 0.008

because of their different thicknesses ._

EOL conditions in LEO and GEO orbits are due to the
radiation fluences of leI5 e/cm2 and 5el4 e/cm2 respectively.
Simulation data are summerized in tables IV, V. As shown
the selected 50,um and 100,um BSR cells has PRF values of
0.62 and 0.58 in LEO and 0.68 and 0.65 in GEO. It is also
seen that both the generated power and PRF for the cells, in
GEO is greater than LEO.

BSR 2 mil at GEG

BSR4ilaIGEO

l_
l_
l_
l_
l11
0
10 12
1013
l104
Equivalent IMevElectronFluence (e/cm2)

0.006

l0 5

0.02

|
0.015

BSR 2 mil at AMO

BSR 4 mil at AMO

TABLE IV
SIMULATION DATA SUMMAvRY FOR LEO

LEO
Solar Cell

Type

BOL
BOLat

EOL

(mW/cm2)

(mW/cm2)

2Q-cm BSR2mil

10.5

6.5

2Q-cm BSR4mil

9.2

5.3

Pmax

PR_BSR

mil

0.62

at

LEO/

BSR4 mil atLEO

0.58

0.005

1010

1011

1012

1013

1014

1015

Equivalent IMevElectron Fluence (e/cm2)

TABLE V
SIMULATION DATA SUMMARviiu~Y FOR GEO

Fig. 7. Comparison of Radiation Effects in AMO, GEO (up), and LEO (down).

GEO

Solar Cell Type

L_______

2 Q-cm BSR 2 mil

2 Q -Cm BSR 4 mil

Pmax at
BOLT
(mW/cm2)
13.5
12.5

Pmax at
Pmax
EOL,
(mW/cm2) |____
9.2
0.68
8.1
0.65

The simulation results for the maximum generated power of

5O0,um BSR silicon solar cell in GEO and LEO orbits are
shown in Fig.8. Simulation is done for different values of
radiation fluence. The results show that for radiation fluence
values lower than tel 2 e/cm2 the degradation of cell is

772

We used MATLAB/Simulink for implementation of the

model. Then we used the model in the evaluation of the
41el3
|i
performance of 4 different types of silicon solar cells. For
12
evaluation
0
, lel4 of the degradation we used maximum power and
b 10
of the cell.
~~~~~~~~~~~~~PRF
8
s
'IX
' " | < p <$ 1e16 The results show that n-type base (p over n) silicon solar cell
g6 s " "' " Xi Xk X
is the most degradation sensitive cell, and for this reason has no
heritage in space. To mitigate the degradation, the cell requires
very thick cover glass that increases the weight and volume and
adds cost to the cell. Poly Si cell, as an another type of
1500
teristerial
cells, has a power remainin factor very colse to space
0
cells, but it has lower generated power.
500
-200
Maximum generated power for space qualified cells in GEO
Temperature (C) Irradiance(W/m2)
is greter than LEO. This is consistant with PRF too.
The results also show that PRF for the 50,um BSR cell at
(a) GEO
various conditions is higher than the 1OO,ym cell. The former
lo\cellhas higher generated power too.
x0
1e12

b8

s-

\
>

100

<
2
~
~~
50

-1e12

10

ss T 6 -\ | /1
+
-IeI4

+- | ---

-----lel5
<---jl
-lel6

-:.

~ ~ ~ ~~~~~~0
<

0
-3

Temperature(C)

I
J
I
\REFERENCES
[1] J.R. Wertz and W.J. Larson, Space Mission Analysis and Design, 3rd ed.,
--leI3 KLUWER Academic Publications, 2005, pp.413-417.
[2] M.R. Patel, Spacecraft Power Systems, CRC Press, 2005, pp.135-180.
[3] A. Luque, S. Hegedus, Handbook of Photo voltaic Science and
Engineering, John Wiley, 2003, pp.413-446.
V.L. Pisacane, Fundamental of Space Systems, 2nd edition, Oxford
[4]< 1>500

University Press, 2005, pp.345-358.

[5] B.E. Anspaugh, "GaAs Solar Cell Radiation Handbook", JPL Publication
96-9, July 1, 1996, pp.5.1-5.5.
500
[6] R.J. Walters, T.L. Morton, "Displacement Damage Effects in Solar CellsMining Damage From the Microelectronics and Photonics Test Bed

1000

Irradiance(W/m2)

Space Experiment", NASA/TP-2004-213338.

[7] J.W. Howard, D.M. Hardage, "Spacecraft Environments Interactions:

(b) LEO

Fig. 8. Maximum Generated Power Due to Cyclic Variation of Irradiance and

Temperature, a) at GEO, and b) at LEO.

negligible. It is also seen that generated power and PRF of the

Space Radiation and its Effects on Electronic Systems", NASA/TP-1999209373.

D.G.
[8]
Gilmore, Spacecraft Thermal Control Handbook, 2nd ed., VOL.1,
AIAA 2002, pp. 21-46.
[9] 14- L. Castaner, S. Silvestre, Modeling Photovoltaic Systems Using

cell for GEO is greater than LEO.

PSPICE, John Wiley,2002, pp.41-72.
Fig. 8. also shows that Pmax increases at first as irradiance [10] A. Goetzberger, J. Knobloch, B. Voss, Crystalline Silicon Solar Cells
John Wiley,
1998, pp35-63.
increases, but after a while cell temperature increases too and [11] B.E.
Anspaugh, "Solar Cell Radiation Handbook, Addenduml:1982decresaes the generated power.
1988", JPL82-69, Add.1, February 15, 1989.

V.CONCLUSION
In the paper we presented a new model for space solar cells.

[12] L.B. Rainey, Space Modeling and Simulation Roles and Applications
Throughout the System Life Cycle, AIAA,2004, pp.534-541.
[13] 17- N.S. Fatemi, H.E. Pollard, H.Q. Hou, and P.R. Sharps, "Solar Array
Trades Between Very High-Efficiency Multi-Junction and Si Space Solar
Cells", presented at the 28th IEEE PVSC, Anchorage, Alaska, September

The model
model uses
uses cummulative
cummulative radiation
radiation fluence
The
fluence as
as aa new*input
new input
2001
(with irradiance and temperature). The presented solar cell [14] T.V. Torchynska and G.Polupan, "High efficiency solar cells for space
model is capable of predicting the behavior of various solar
applications", Superficies y Vacio 17(3), pp.21-25, September 2004.
cells in space applications. The model requires a complete test
data of the manufacturer (data sheet ofthe cell).

773

706

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 3, MARCH 2008

Nanostructured Absorbers for Multiple

Transition Solar Cells
Michael Y. Levy and Christiana Honsberg, Senior Member, IEEE

AbstractThis paper identifies absorbers for multiple transition solar cells that are implemented with nanostructured heterojunctions [e.g., quantum well solar cells with quasi-Fermi-level
variations and quantum dot (QD) intermediate-band solar cells].
In the radiative limit, the solar cells implemented with these absorbers are capable of achieving a conversion efficiency 50%
with a geometric solar concentration of at least 1000. The technical approach enumerates a set of quantitative design rules and
applies the rules to the technologically important IIIV semiconductors and their ternary alloys. A novel design rule mandates a negligible valence band discontinuity between the barrier material and confined materials. Another key design rule
stipulates that the substrate have a lattice constant in between
that of the barrier material and that of the quantum-confined
material, which permits strain compensation. Strain compensation, in turn, allows a large number of QD layers to be incorporated into the solar cell because each layer is free of defects.
Four candidate materials systems (confined/barrier/substrate)
are identified: InP0 .8 5 Sb0 .1 5 /GaAs/InP, InAs0 .4 0 P0 .6 0 /GaAs/InP,
InAs/GaAs0 .8 8 Sb0 .1 2 /InP, and InP/GaAs0 .7 0 P0 .3 0 /GaAs. Resulting from the design features, the candidate systems may also find
use in other optoelectronic applications.
Index TermsHeterojunction, intermediate band, quantum dot,
quantum well, solar cell.

I. INTRODUCTION
HYSICAL phenomena that may increase the efficiency of
a single-junction solar cell above the ShockleyQueisser
limit [1] include proposals such as multiple transitions [2], multiple electronhole pair generation [3], and hot carriers [4].
Identifying materials that may be used to implement these
phenomena is of significant interest. For example, multiple
electronhole pair generation is demonstrated in several materials [5], [6]. Several approaches are proposed to implement
multiple transition solar cells, all of which require the introduction of valid quantum states deep into the previously forbidden
bandgap of a host semiconductor [2]. Two of these, which rely
on nanostructured heterojunctions, are the quantum well solar

Manuscript received May 22, 2007; revised September 25, 2007. This
work was supported by the U.S. National Renewable Energy Laboratory under Subcontract XAT-5-44277-01, administered by M. Symko-Davies and R.
McConnell. The work of M. Y. Levy was supported by the U.S. National Science
Foundation under a Ph.D. Dissertation Enhancement Award OISE-0509001,
administered by J. Hudson. The review of this paper was arranged by Editor
P. Panayotatos.
M. Y. Levy is with the University of Delaware, Newark, DE 19716 USA.
He is also with the Georgia Institute of Technology, Atlanta, GA 30332 USA
(e-mail: mlevy@ece.gatech.edu; qb@UDel.edu; mail2MYL@gmail.com).
C. Honsberg is with the University of Delaware, Newark, DE 19716 USA
(e-mail: honsberg@ece.UDel.edu).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TED.2007.914829

cell with quasi-Fermi-level variations [7] and the quantum dot

(QD) intermediate-band solar cell (IBSC) [8]. It is important to
note that both of these approaches may be shown to be thermodynamically consistent [9]. Until now, material systems that
are most suitable to implement these two solar cells are not
identified.
This paper identifies absorbers for solar cells that utilize
nanostructured heterojunctions to generate multiple transitions
[7], [8], and that are theorized [7], [8] and experimentally inferred [10], [11] to maintain more than one quasi-Fermi-level
separation. Specifically, the material systems identified here
may lead to devices capable, in principle, of conversion efficiencies 50% with a geometric solar concentration of at least
1000. The technical approach taken here identifies candidate
material systems by enumerating quantitative design rules and
by applying these design rules to the technologically important
IIIV semiconductors and their ternary alloys. Though the remainder of this paper gives pronounced emphasis to the QD
IBSC, the paper may find use in developing related photovoltaic
technologies and in other optoelectronic communities.
In the QD IBSC [8], [12], [13], the intermediate states within
the formerly forbidden bandgap are introduced by the threedimensional confining potential created by the QD heterojunctions. The states introduced by the QD heterojunctions may
either form a band of extended states, which would yield beneficial transport properties, or a collection of localized states,
as shown in Fig. 1. The material systems identified here would
be appropriate regardless of whether a band of extended states
is formed or not. With respect to the former case, a band of
extended states with well-defined band edges and bandwidth
requires some measure of order [14] (for a general discussion,
see [15] and references therein). Allowing that the required order exists, then one may consider a QD crystal with miniband
formation resultant from coupling between QD [16], and next
consider the design constraints of a QD IBSC [17]. With respect to the latter, a collection of localized states may result if
there would exist nonuniformities in the QD matrix or if the
inter-QD distances are large. If the localized intermediate states
within the heterojunctions confining potentials exist energetically close to one another, then they will create an energetic
interval of allowed intermediate states IB (see Fig. 1). With
respect to either of the two cases, assuming that all the carriers
in the intermediate states are in electrochemical equilibrium, the
population of carriers at intermediate states may be described
by a single quasi-Fermi level FI . If FI is energetically close
to the intermediate states, which may require the inclusion of
dopants [12], [17], [18], then there is strong likelihood that absorbed photons may induce electronic transitions between the

0018-9383/$25.00 2008 IEEE

LEVY AND HONSBERG: NANOSTRUCTURED ABSORBERS FOR MULTIPLE TRANSITION SOLAR CELLS

Fig. 1. Energetic structure of an array of nonuniform QD heterojunctions that

would be favorable for a multiple transition solar cells absorbing medium.
Here, the valence band discontinuity is negligible, which precludes the existence of hole states in the formerly forbidden bandgap. Bound intermediate
states within the heterojunctions confining potentials, which are formed from
the conduction band discontinuities CB , exist energetically close to one another. Inhomogeneities (e.g., nonuniform dot sizes) broaden the interval of allowed intermediate states (gray hue) and create an energetic interval of allowed
intermediate states IB . The minimum energetic distances between the intermediate states and the conduction band continuum and valence band continuum
are CI and IV , respectively. Assuming that all the carriers in the intermediate states are in electrochemical equilibrium, their occupancy statistics may be
described by a single quasi-Fermi level FI .

intermediate states and the conduction and valence band continua, respectively.
Previous work on the QD IBSC revolves around the
InAs/GaAs (confined/barrier) [11], [19][21] and InGaAs/GaAs
[22] material systems, both grown on GaAs substrates. These
choices derive from the systems being relatively well understood. In addition, the bandgap of a GaAs barrier is near to
the optimum conduction-to-valence bandgap associated with
the upper limiting efficiency of an IBSC [23]. These materials
systems have two characteristics limiting their usage for development of an QD IBSC: they have valence band discontinuities
and they accumulate strain. The first characteristic (type I heterojunction) enables the luminescent spectra of QD IBSC samples [11], [19] to illustrate narrowly separated and overlapping
peaks, rather than the luminescent signature of an intermediate
band material. The most rudimentary signature is the appearance of three [24] (or four [25]) strong luminescence peaks
separated across a wide energy range; each of the three peaks
corresponds to one of the three allowed interband electronic
transitions (whereas the fourth peak corresponds to an intraband
transition [25]). The second characteristic (strain accumulation)
causes the open-circuit voltage to degrade as the number of QD
multilayers increases [22]. This is a problem to the design of a
high-efficiency QD IBSC, for a large number of multilayers is
essential for a high absorptivity of subbandgap photons [26]
the high absorptivity of subbandgap photons may be viewed
experimentally by an increase in their external quantum efficiencies with an increase in the number of QD layers [22]. This
papers design rules identify candidate material systems that
may overcome these two characteristics.
The candidate material systems identified here yield heterojunctions at the transition between type I and type II heterojunc-

707

tions and the potential for strain compensation. Heterojuntions

at the transition between type I and type II exclusively permit
bound states due to either hole(s) or electron(s), but not both.
This choice is meritorious for several reasons. First, it abates
the difficulty in interpreting data from interband, optical experimental methods that exists if both holes and electrons are
involved [27]. Second, it may limit the extension of the band
edge into the previously forbidden bandgap [13], which would
reduce the open-circuit voltage when compared to a control cell
without QD [28]. Third, it may limit the accumulation of excess
miniband formation [28]. Further, material systems that compensate strain accumulation permit the realization of longer,
optically active, multilayered QD structures fabricated without
generating [29][33] and propagating large number of defects,
which decrease the open-circuit voltage. The literature presents
several techniques to introduce strain compensating epitaxial
layers [34][36], which permit fabricating optically active QD
structures comprising 100 or more QD multilayers [37], [38].
Beyond the quantum well solar cell with quasi-Fermi-level
variations [7] and the QD IBSC [8], the design features and
candidate material systems identified here may find use in developing related photovoltaic technologies and other optoelectronic devices. In addition, the candidate systems are of importance to the fundamental physical study of heterojunctions.
Within the photovoltaic field, all the operating principles of
several devices (i.e., multiband solar cells [39], and up- [40]
and down-converters [41]) rely on absorbers admitting multiple photo-induced electronic transitions. With respect to the
physical importance, type II heterojunctions are theoretically
predicted [42] and experimentally verified [43] as having relatively long radiative lifetimes resulting from the spatial division
of confined electrons and holes. In addition, the nature of electronic conduction is shown to differ in type I and type II heterojunctions [44]. Therefore, the study of systems at the transition
between type I and type II systems warrants attention, and may
eventually lead to novel devices and device applications.
The remainder of this paper is organized as follows. Section II
enumerates and details the quantitative design rules. Section III
explains how these rules are applied to the technologically IIIV
material systems. Section IV follows with concluding remarks.
II. DESIGN RULES
Seven quantitative design rules are specified to identify candidate material systems for nanostructured absorbers that are
suitable for solar cells with multiple transitions. The first and
second design rules are discussed extensively in Section I. The
third and fourth design rules pertain to the placement of the intermediate states to permit, in principle, 50% conversion under a
solar concentration of at least 1000 Suns. The remaining rules
assist in identifying candidates that offer an advantage in terms
of minimizing the technological complexity of fabrication and
characterization.
A. Negligible Valence Band Discontinuity
The first design rule specifies that the upper valence band
edge of the confined material must be continuous with the upper

708

valence band edge of the barrier material (see Fig. 1). The absence of a valence band discontinuity eliminates the possibility of introducing intermediate states resulting from quantumconfined holes. There are two reasons why the valence band
discontinuity is set to a negligible value rather than the conduction band discontinuity. First, in the technologically realizable
interval of QD radii, for even a small valence band discontinuity,
many hole states appear; however, fewer electron states appear
with a similar magnitude conduction band discontinuity. This
result stems from the fact that the technologically important
IIIV semiconductors have larger hole effective masses than
that of electron effective masses [45]. Thus, within technologically realizable QD radii, it is possible to provide a single electron state within the confining potential, whereas, at those same
radii, many hole states would exist. Second, in general, for the
technologically important IIIV semiconductors, the hole mobilities are less than the electron mobilities. In an actual device,
the higher the mobility of the carriers, the higher is the diffusion
length for a given carrier lifetime. Thus, the mobility, through the
diffusion length, enhances the collection of generated electron
hole pairs, and hence, the performance of the device.
B. Substrates Lattice Constant Between Those of the Confined
Material and the Barrier Material
The second design rule specifies inequalities relating the lattice constants of the confined, barrier, and substrate materials.
Specifically, the lattice constant of the substrate must be between the lattice constants of the barrier and the confined materials. Material systems complying with this rule may allow the
incorporation of tensile strain to compensate for compressive
strain accumulated during the growth of the multilayer quantum
well [46], [47] and QD structures [34]. Strain compensation reduces the density of defects in multilayer QD structures [48].
Thus, strain compensation is relevant to the number of QD
layers, which has a direct impact upon the absorptivity of subbandgap photons [26], that can be incorporated into multilayer
structures before relaxation occurs. Further, the reduced density of defects in multilayer QD structures [48] abates large
open-circuit voltage losses in experimental samples [19], [22].
C. Barriers Bandgap Between 1.25 and 3.13 eV
The third design rule specifies the required interval of the
conduction-to-valence bandgap CV . In a previous examination
of the QD IBSC [8], a bandgap of 1.95 eV is considered to be
necessary for candidate barrier materials. This consideration is
derived from detailed balance calculations [23] where the solar
irradiance is concentrated at 46 000, which is the physical
limit of geometrical concentration (i.e., the reciprocal of the
geometrical dilution factor [1]). The bandgap of 1.95 eV is associated with the apex of the efficiency curve found there. Here,
the authors consider a more practical geometrical concentration
factor of 1000. In addition, unlike previous calculations [23],
when obtaining the maximum-power-detailed balance efficiencies, the authors include the following as parameters in the
search space: the energetic width of intermediate states [49] and
the spectral extent of the absorption overlap [50]. Similarly to

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 3, MARCH 2008

Fig. 2. Maximum power isoefficiency contour plot of an IBSC under a geometric solar concentration of 1000 (shown here are the 50% and 55% contours)
as a function of the energetic gaps IV and CI (see Fig. 1). Contours obtained
using the rapid and precise computation method of [51].

previous calculations [23], all calculations here are performed

in the limit where electronic transitions are coupled to radiative
interactions with the absorber. To what extent is this actually the
case may be determined by future experimental investigation.
Fig. 2 shows the upper theoretical maximum power isoefficiency contour plot of an IBSC under a geometric concentration
of 1000. Allowing that the energetic widths of intermediate
states corresponding to these efficiencies are negligiblewhich
is the case in the high-efficiency regime (50%)inspection
of Fig. 2 reveals that the conduction-to-valence bandgap must
be between 1.25 and 3.13 eV. Before moving on to the next
design rule, the authors point out an advantage of computing a
two-dimensional isoefficiency contour plot as compared to the
one-dimensional plot previously presented in the literature [23].
This advantage stems from the fact that the complete set of ordered pairs (IV , CI ) that result in efficiencies greater than 50%
is visible. The upshot is a larger search space with which to
apply the design rules, and hence, a greater likelihood of obtain
suitable material systems.
D. Conduction Band Discontinuity Accommodates States
Associated With High Efficiencies
The fourth design rule specifies the minimum value of the
conduction band discontinuity CB , which defines the confining potential (see Fig. 1). This rule demands that the confining
potentials depth accommodates positioning the intermediate
states to enable high-efficiency solar energy conversion. The
minimum quantitative value for the conduction band discontinuity depends upon the conduction-to-valence bandgap. These
values are obtainable from the data comprising the isoefficiency
contour plot (see Fig. 2).
E. Direct Bandgaps
The fifth design rule demands that both the confined material and barrier material must be direct bandgap materials. For

LEVY AND HONSBERG: NANOSTRUCTURED ABSORBERS FOR MULTIPLE TRANSITION SOLAR CELLS

high-efficiency approaches, especially where the height of the

absorbing medium may be limited by existing fabrication techniques, it is essential to have the largest absorptivity possible. For
this reason, indirect bandgap materials, which require phonon
exchange for light absorption, are excluded because their absorption coefficients are lower than those of direct bandgap materials. Further, the use of an indirect bandgap barrier requires
thicker devices to significantly trap light and a longer duration of
epitaxial growth, which may increase the opportunity for defect
formation.

709

TABLE I
MATERIAL SYSTEMS WITH WHICH TO IMPLEMENT NANOSTRUCTURED
ABSORBERS FOR MULTIPLE TRANSITION SOLAR CELLS

F. Confined Material or Barrier Material is a Binary

Semiconductor
The sixth design rule simplifies the technological complexity
of the multistack fabrication process. It is indubitably simpler
to epitaxially fabricate a binary semiconductor than a ternary or
a quaternary semiconductor.
G. Substrate is Commercially Available
The seventh design rule stipulates that the substrate must be
commercially available. In general, the crystalline quality of
devices grown by molecular beam epitaxy is much better if the
substrate is commercially available. In addition, the experimental repeatability of measurements is more likely if the substrate
is subject to industrial process control. At the time of writing,
the following materials are commercially available: InP, GaAs,
and InAs.
The design rules offer quantitative criteria to identify material
system suitable for multiple transition solar cells implemented
with nanostructures. These rules may be applied to a set of
sufficiently characterized materials, as is done next.
III. APPLICATION TO THE IIIV SEMICONDUCTORS AND THEIR
TERNARY ALLOYS
The design rules are applied to the technologically important
IIIV zinc blend and wurtzite compound semiconductors and
their ternary alloys. Specifically, the following compounds and
their ternary alloys are considered for any of the three components (confined/barrier/substrate) of the material system: GaAs,
GaSb, GaP, GaN, AlAs, AlSb, AlP, AlN, InAs, InSb, InP, and
InN. The following parameters for all these binaries and their
ternary alloys are required to identify material systems that conform to the design rules: alc , the lattice constant; Eg , the band in
the valley; EgX , the bandgap in the X valley; EgL , the bandgap
in the L valley; VBO, the valence band offset; and, when available, (), (), (X), (X), (L), and (L), the Varshni
parameters of the , X, and L bandgaps, respectively. The review of Vurgaftman et al. is the sole source of information used
by the authors in modeling these material properties [45].
The design rules are applied to the technologically important IIIV semiconductors at two distinct temperatures. These
temperatures are 300 K, the terrestrial operating temperature
of a solar cell, and 4 K, a temperature at which a significant amount of research and development of these types of
material systems is likely to be performed [24], [25]. At

both temperatures, only the following material systems are

identified: InP0.85 Sb0.15 /GaAs/InP, InAs0.40 P0.60 /GaAs/InP,
InAs0.40 P0.60 /GaAs/InP, InAs/GaAs0.88 Sb0.12 /InP, and In/
GaAs0.70 P0.30 /GaAs. Table I gives the conduction-to-valence
bandgaps and conduction band discontinuities of these material systems at both 300 K and 4 K. It needs be noted that,
at 300 K, the band alignments of both the InAs/GaAsSb/InP
and InP/GaAs/GaAs material systems may yield, in principle, a
conversion efficiency between 48% and 50%. They are included
here to provide material systems whose quantum-confined regions are composed of a binary semiconductor.
Negligible valence band discontinuities are experimentally
inferred for two of these four systems: InAsP/GaAs [52],
[53] and InAs/GaAsSb [54], [55]. As regards the QD IBSC,
InAs/GaAsSb structures are the most promising. There are two
reasons for this. First, antimonide acts as a surfactant [56], [57],
which inhibits the generation of interfacial defects, and thus,
leads to superior optical activity [58]. Second, it is indubitably
less complex to constitute binary QD within a ternary barrier
than vice versa. The following concludes these findings.
IV. CONCLUSION
Four candidate material systems are identified for further development of multiple transition solar cells implemented with
nanostructures. The materials systems provide heterojunctions
at the transition between type I and type II heterojunctions, resulting from their negligible valence band discontinuities, and
offer the possibility for strain compensation during the growth
of multilayer structures. Resultant from several advantageous
properties, the InAs/GaAsSb is particularly suited to further investigation of a QD IBSC. A key tool used to identify these
material systems is a maximum power isoefficiency contour
plot of an IBSC under a geometric concentration of 1000.
This contour is presented here for the first time.
ACKNOWLEDGMENT
M.Y. Levy is gracious to the people of the Instituto de
Energa Solar, Universidad Politecnica de Madrid, for the scientific and interpersonal exchanges during the summers of 2006
and 2007, and Dr. A. Norman for his trenchant remark regarding
the Varshni parameters.

710

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 3, MARCH 2008

REFERENCES
[1] W. Shockley and H. J. Queisser, Efficiency of p-n junction solar cells,
J. Appl. Phys., vol. 32, pp. 510519, 1961.
[2] M. Wolf, Limitations and possibilities for improvement of photovoltaic
solar energy converters. Part I: Considerations for Earths surface operation, in Proc. Inst. Radio Eng., vol. 48, no. 7, pp. 12461263,
1960.
[3] P. Landsberg, H. Nussbaumer, and G. Willeke, Bandband impact ionization and solar cell efficiency, J. Appl. Phys., vol. 74, no. 2, pp. 14511452,
1993.
[4] R. T. Ross and A. J. Nozik, Efficiency of hot-carrier solar energy converters, J. Appl. Phys., vol. 53, no. 5, pp. 38133818, 1982.
[5] R. Schaller and V. Klimov, High efficiency carrier multiplication in PbSe
nanocrystals: Implications for solar energy conversion, Phys. Rev. Lett.,
vol. 92, no. 18, pp. 186601-1186601-4, 2004.
[6] M. C. Beard, K. P. Knutsen, P. Yu, J. M. Luther, Q. Song, W. K. Metzger,
R. J. Ellingson, and A. J. Nozik, Multiple exciton generation in colloidal
silicon nanocrystals, Nano Lett., vol. 7, no. 8, pp. 25062512, 2007.
[7] S. P. Bremner, R. Corkish, and C. B. Honsberg, Detailed balance efficiency limits with quasi-Fermi level variations, IEEE Trans. Electron
Devices, vol. 46, no. 10, pp. 19321939, Oct. 1999.
[8] A. Mart, L. Cuadra, and A. Luque, Quantum dot intermediate band solar
cell, in Conf. Rec. 28th IEEE Photovoltaic Spec. Conf., Anchorage, AK,
2000, pp. 940943.
[9] A. Luque, A. Mart, and L. Cuadra, Thermodynamic consistency of subbandgap absorbing solar cell proposals, IEEE Trans. Electron Devices,
vol. 48, no. 9, pp. 21182124, Sep. 2001.
[10] J. Nelson, J. Barnes, N. Ekins-Daukes, B. Kluftinger, E. Tsui, K. Barnham,
C. T. Foxon, T. Cheng, and J. S. Roberts, Observation of suppressed
radiative recombination in single quantum well p-i-n photodiodes, J.
Appl. Phys., vol. 82, no. 12, pp. 62406246, 1997.
[11] A. Luque, A. Mart, N. Lopez, E. Antolin, E. Canovas, C. Stanley,
C. Farmer, L. J. Caballero, L. Cuadra, and J. L. Balenzategui, Experimental analysis of the quasi-Fermi level split in quantum dot intermediate-band
solar cells, Appl. Phys. Lett., vol. 87, no. 8, pp. 083505-1083505-3,
2005.
[12] A. Mart, L. Cuadra, and A. Luque, Partial filling of a quantum dot
intermediate band for solar cells, IEEE Trans. Electron Devices, vol. 48,
no. 10, pp. 23942399, Oct. 2001.
[13] A. Mart, L. Cuadra, and A. Luque, Intermediate band solar cells, in
Next Generations Photovoltaics: High Efficiency Through Full Spectrum
Utilization, A. Mart and A. Luque, Eds. Bristol, U.K.: Inst. of Physics
Publishing, 2004, ch. 7, pp. 150160.
[14] E. F. Schubert, Cambridge studies in semiconductor physics and microelectronic engineering1, in Doping in IIIV Semiconductors,
H. Ahmed, M. Pepper, and A. Broers, Eds. Cambridge, U.K.: Cambridge
Univ. Press, 1993, ch. 1, pp. 4144.
[15] O. Madelung, Introduction to Solid-State Theory. Heidelberg, Germany:
Springer-Verlag, 1978, pp. 287, 382, 435447.
[16] O. L. Lazarenkova and A. A. Balandin, Miniband formation in a quantum
dot crystal, J. Appl. Phys., vol. 89, pp. 55095515, 2001.
[17] A. Mart, L. Cuadra, and A. Luque, Design constraints of the quantumdot intermediate band solar cell, Physica E, vol. 14, no. 1/2, pp. 150157,
2002.
[18] M. Y. Levy, A. Mart, C. Honsberg, and A. Luque, Calculation of the band
properties of quantum dot intermediate band solar cells with hydrogenic
impurities, in Proc. 19th Eur. Photovoltaic Energy Conf., Paris, France,
2004, pp. 332335.
[19] A. Luque, A. Mart, C. Stanley, N. Lopez, L. Cuadra, D. Zhou, J. L.
Pearson, and A. McKee, General equivalent circuit for intermediate band
devices: Potentials, currents and electroluminescence, J. Appl. Phys.,
vol. 96, no. 1, pp. 903909, 2004.
[20] A. Luque, A. Mart, N. Lopez, E. Antolin, E. Canovas, C. Stanley,
C. Farmer, and P. Diaz, Operation of the intermediate band solar cell
under nonideal space charge region conditions and half filling of the intermediate band, J. Appl. Phys., vol. 99, no. 9, pp. 94503-194503-9,
2006.
[21] A. Mart, E. Antolin, C. Stanley, C. Farmer, N. Lopez, P. P. Daz,
E. Canovas, P. Linares, and A. Luque, Production of photocurrent due
to intermediate-to-conduction-band transitions: A demonstration of a key
operating principle of the intermediate-band solar cell, Phys. Rev. Lett.,
vol. 97, no. 24, pp. 247701-1247701-4, 2006.
[22] A. G. Norman, M. C. Hanna, P. Dippo, D. H. Levi, R. C. Reedy, J. S. Ward,
and M. M. Al-Jassim, InGaAs/GaAs QD superlattices: MOVPE growth,
structural and optical characterization, and application in intermediate-

[23]
[24]

[25]

[26]

[27]
[28]

[29]

[30]

[31]
[32]

[33]

[34]
[35]
[36]

[37]
[38]
[39]
[40]
[41]
[42]
[43]

band solar cells, in Conf. Rec. 31st IEEE Photovoltaic Spec. Conf., Orlando, FL, 2005, pp. 4348.
A. Luque and A. Mart, Increasing the efficiency of ideal solar cells
by photon induced transitions at intermediate levels, Phys. Rev. Lett.,
vol. 78, pp. 50145017, 1997.
N. Ekins-Daukes, C. Honsberg, and M. Yamaguchi, Signature of intermediate band materials from luminescence measurements, in Conf. Rec.
31st IEEE Photovoltaic Spec. Conf. (IEEE Cat. No. 05CH37608), Lake
Buena Vista, FL, 2005, pp. 4954.
M. Y. Levy, N. Ekins-Daukes, and C. Honsberg, Photouminescent signature of absorbing medium with deep levels based upon absorption coefficicents derived from parabolic band approximation, presented at the
22nd Eur. Photovoltaic Energy Conf., Milan, Italy, 2007.
B. Kochman, A. D. Stiff-Roberts, S. Chakrabarti, J. D. Phillips, S. Krishna,
J. Singh, and P. Bhattacharya, Absorption, carrier lifetime, and gain
in InAs-GaAs quantum-dot infrared photodetectors, IEEE J. Quantum
Electron., vol. 39, no. 3, pp. 459467, Mar. 2003.
A. Sibille, Miniband transport, in Semiconductor Superlattices: Growth
and Electronic Properties, H. T. Grahn, Ed. Singapore: World Scientific,
1995, ch. 2, p. 37.
M. Levy, C. Honsberg, A. Mart, and A. Luque, Quantum dot intermediate band solar cell material systems with negligible valence band
offsets, in Conf. Rec. 31st IEEE Photovoltaic Spec. Conf. (IEEE Cat. No.
05CH37608), Lake Buena Vista, FL, 2005, pp. 9093.
Y. Sugiyama, Y. Nakata, K. Imamura, S. Muto, and N. Yokoyama,
Stacked InAs self-assembled quantum dots on (001) GaAs grown by
molecular beam epitaxy, Jpn. J. Appl. Phys. 1, Reg. Papers, Short Notes,
vol. 35, no. 2B, pp. 13201324, 1996.
K. Shiramine, Y. Horisaki, D. Suzuki, S. Itoh, Y. Ebiko, S. Muto, Y. Nakata,
and N. Yokoyama, TEM observation of threading dislocations in InAs
self-assembled quantum dot structure, J. Crystal Growth, vol. 205, no. 4,
pp. 461466, 1999.
C. H. Roh, Y. J. Park, K. M. Kim, Y. M. Park, F. K. Kim, and K. B. Shim,
Defect generation in multi-stacked InAs quantum dot/GaAs structures,
J. Crystal Growth, vol. 226, no. 1, pp. 17, 2001.
K. Stewart, M. Buda, J. Wong Leung, L. Fu, C. Jagadish, A. Stiff-Roberts,
and P. Bhattacharya, Influence of rapid thermal annealing on a 30 stack
InAs/GaAs quantum dot infrared photodetector, J. Appl. Phys., vol. 94,
no. 8, pp. 52835289, 2003.
M. Gutierrez, M. Hopkinson, H. Y. Liu, J. S. Ng, M. Herrera, D. Gonzalez,
R. Garcia, and R. Beanland, Strain interactions and defect formation
in stacked InGaAs quantum dot and dot-in-well structures, Phys.
E: Low-Dimensional Syst. Nanostruct., vol. 26, no. 14, pp. 245251,
2005.
K. Akahane, N. Ohtani, Y. Okada, and M. Kawabe, Fabrication of ultrahigh density InAs-stacked quantum dots by strain-controlled growth on
InP(311)B substrate, J. Crystal Growth, vol. 245, pp. 3136, 2002.
P. Lever, H. H. Tan, and C. Jagadish, InGaAs quantum dots grown with
GaP strain compensation layers, J. Appl. Phys., vol. 95, no. 10, pp. 5710
5714, 2004.
Z. H. Zhang, C. F. Xu, K. C. Hsieh, and K. Y. Cheng, Defect-free 50-layer
strain-balanced InAs quantum dots grown on InAlGaAs/InP for infrared
photodetector applications, J. Crystal Growth, vol. 278, no. 14, pp. 61
66, 2005.
K. Akahane, N. Yamamoto, and N. Ohtani, Long-wavelength light emission from InAs quantum dots covered by GaAsSb grown on GaAs substrates, Physica E, vol. 21, pp. 295299, 2004.
Z. H. Zhang, K. Y. Cheng, C. F. Xu, and K. C. Hsieh, Defect-free 100layer strain-balanced InAs quantum dot structure grown on InP substrate,
Appl. Phys. Lett., vol. 89, no. 6, pp. 63115-163115-3, 2006.
A. S. Brown and M. A. Green, Intermediate band solar cell with many
bands: Ideal performance, J. Appl. Phys., vol. 94, pp. 61506158,
2003.
T. Trupke, M. A. Green, and P. Wurfel, Improving solar cell efficiencies
by up-conversion of sub-band-gap light, J. Appl. Phys., vol. 92, pp. 4117
4122, 2002.
T. Trupke, M. A. Green, and P. Wurfel, Improving solar cell efficiencies of high-energy photons, J. Appl. Phys., vol. 92, pp. 16681674,
2002.
U. Laheld, F. Pedersen, and P. Hemmer, Excitons in type-II quantum dots:
Finite offsets, Phys. Rev. B, Condens. Matter, vol. 52, no. 4, pp. 2697
2703, 1995.
C.-K. Sun, A. Wang, J. Bowers, B. Brar, H.-R. Blank, H. Kroemer,
and M. Pilkuhn, Optical investigations of the dynamic behavior of
GaSb/GaAs quantum dots, Appl. Phys. Lett., vol. 68, no. 11, pp. 1543
1545, 1996.

LEVY AND HONSBERG: NANOSTRUCTURED ABSORBERS FOR MULTIPLE TRANSITION SOLAR CELLS

[44] H. Munekata, T. Smith III, F. Fang, L. Esaki, and L. Chang, Electrons and
holes in InAs-Ga(Al)Sb(As) quantum wells, J. Phys. Colloq., vol. 48,
no. C-5, pp. 151154, 1987.
[45] I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, Band parameters for
IIIV compound semiconductors and their alloys, J. Appl. Phys., vol. 89,
no. 11, pp. 58155875, 2001.
[46] N. Ekins-Daukes, K. Barnham, J. Connolly, J. Roberts, J. Clark, G. Hill,
and M. Mazzer, Strain-balanced GaAsP/InGaAs quantum well solar
cells, Appl. Phys. Lett., vol. 75, no. 26, pp. 41954197, 1999.
[47] N. Ekins-Daukes, K. Kawaguchi, and J. Zhang, Strain-balanced criteria
for multiple quantum well structures and its signature in X-ray rocking
curves, Crystal Growth Des., vol. 2, no. 4, pp. 287292, 2002.
[48] N. Nuntawong, S. Birudavolu, C. Hains, S. Huang, H. Xu, and D. Huffaker,
Effect of strain-compensation in stacked 1.3 m InAs/GaAs quantum dot
active regions grown by metalorganic chemical vapor deposition, Appl.
Phys. Lett., vol. 85, no. 15, pp. 30503052, 2004.
[49] A. S. Brown, M. A. Green, and R. P. Corkish, Limiting efficiency for a
multi-band solar cell containing three and four bands, Physica E, vol. 14,
pp. 121125, 2002.
[50] L. Cuadra, A. Mart, and A. Luque, Influence of the overlap between
the absorption coefficients on the efficiency of the intermediate band solar
cell, IEEE Trans. Electron Devices, vol. 51, no. 6, pp. 10021007, Jun.
2004.
[51] M. Y. Levy and C. Honsberg, Rapid and precise calculations of energy
and particle flux for detailed-balance photovoltaic applications, SolidState Electron., vol. 50, pp. 14001405, 2006.
[52] E. Ribeiro, R. Maltez, W. Carvalho, D. Ugarte, and G. Medeiros-Ribeiro,
Optical and structural properties of InAsP ternary self-assembled quantum dots embedded in GaAs, Appl. Phys. Lett., vol. 81, no. 16, pp. 2953
2955, 2002.
[53] R. L. Maltez, E. Ribeiro, W. Carvalho, Jr., D. Ugarte, and G. MedeirosRibeiro, Controlling alloy composition of InAsP self-assembled quantum
dots embedded in GaAs, J. Appl. Phys., vol. 94, no. 5, pp. 30513056,
2003.
[54] H. Y. Liu, M. J. Steer, T. J. Badcock, D. J. Mowbray, M. S. Skolnick,
P. Navaretti, K. M. Groom, M. Hopkinson, and R. A. Hogg, Longwavelength light emission and lasing from InAs/GaAs quantum dots covered by a GaAsSb strain-reducing layer, Appl. Phys. Lett., vol. 86,
pp. 143108-1143108-3, 2005.
[55] H. Y. Liu, M. J. Steer, T. J. Badcock, D. J. Mowbray, M. S. Skolnick,
F. Suarez, J. Ng, M. Hopkinson, and J. P. R. David, Room-temperature
1.6 m light emission from InAs/GaAs quantum dots with a thin GaAsSb
cap layer, J. Appl. Phys., vol. 99, pp. 46104-146104-3, 2006.
[56] M. Copel, M. C. Reuter, E. Kaxiras, and R. M. Tromp, Surfactants in
epitaxial growth, Phys. Rev. Lett., vol. 63, no. 6, pp. 632635, 1989.

711

[57] M. Larsson, W. Ni, K. Joelsson, and G. Hansson, Growth of high quality

Ge films on Si(lll) using Sb as surfactant, Appl. Phys. Lett., vol. 65,
no. 11, pp. 14091411, 1994.
[58] H. Shimizu and S. Saravanan, Buffer-material dependence of InAs quantum dots on GaAs substrate, Appl. Phys. Lett., vol. 88, no. 4, pp. 41119141119-3, 2006.
Michael Y. Levy was born in New York in 1975.
He received the dual B.S. degrees in electrical engineering and physics from Portland State University,
Portland, OR, in 2001, and the M.S. degree in electrical engineering in 2004 from Georgia Institute of
Technology, Atlanta, where he is currently working
toward the Ph.D. degree on photovoltaic absorbing
media with an intermediate band of electronic states.
Currently, he is an Interim Researcher at the University of Delaware, Newark. His current research
interests include energy policy, building integrated
photovoltaics, and high-efficiency solar energy conversion.
Mr. Levy received a U.S. National Science Foundation International Research Fellowship Program Postdoctoral Award. In 2008, he will commence
research at the Instituto de Energa Solar, Universidad Politecnica de Madrid.

Christiana Honsberg (S87M91SM03) received the Ph.D. degree from the University of
Delaware, Newark, in 1992.
From 1993 to 2000, she was a faculty member at
the University of New South Wales, Sydney, Australia, where she worked on high-efficiency commercial silicon solar cells, advanced concept photovoltaics, and developing the worlds first undergraduate degree program in photovoltaic engineering. In
2001, she joined the School of Electrical and Computer Engineering, Georgia Institute of Technology,
where she was involved in ultrahigh-efficiency solar cells capable of reaching the thermodynamic limits of conversion of solar radiation. In 2004, she
joined the University of Delaware, where she was engaged in developing the
university program on high-performance photovoltaic devices, encompassing
high-efficiency silicon, new materials for tandem solar cells, and new concepts in photovoltaic energy conversion. Dr. Honsberg was awarded a National
Science Foundation Integrative Graduate Education and Research Traineeship
(NSF-IGERT) in 2006 to develop a graduate education and research program in
solar hydrogen.

The Development and Application of the Solar Lights

in Wuhan city, China
Xia Chen

Jinyu Wen

Department of Power System and its Automation

Huazhong University of Science and Technology
Wuhan, China

Department of Power System and its Automation

Huazhong University of Science and Technology
Wuhan, China

AbstractTraditional
illumination
lights
are
energyconsumption and not economical in the long term. While the
solar lights are safe, easy to maintenance, independent of
conventional energy, no pollution to environment. With the cost
of the solar cells decreasing, the solar lights have been applied
widely. Meanwhile photovoltaics are one of the fastest growing
solar energy technologies. Photovoltaic devices, commonly called
solar cells or modules, use semiconductor material to directly
convert sunlight into electricity. In this paper, firstly the
development of solar light in Wuhan city is introduced. A
comprehensive analysis on the principle and the structure of the
solar light is conducted, and the mathematical model of solar
cells is proposed. Through tests on the solar cells, the I-V and PV characteristic curves are obtained which verify the theoretical
curves.
Keywords- solar lights; Photovoltaic cell; I-V curve; energy

I.

INTRODUCTION

Shortage of energy is dangerously affecting the economic

growth of the developing countries. Solar power technology
has been primarily dominated by use of solar cells. Arrays of
photovoltaic cells convert light from sun to electricity by using
semiconductor material. Although the cost of solar cells is still
high, the use of photovoltaic cells has expanded dramatically in
recent years for the severe energy crisis [1].
The conversion of solar energy into power is being
recognized as a promising alternative to meeting future energy
demands and significant world-wide efforts are under way to
develop solar energy application that will be competitive with
other energy resources.

possesses lots of advantages, such as its security, easy to

maintenance, independent of conventional energy, no pollution
to environment, furthermore, it can be installed depending on
the practical requirement and neednt be connected with the
grid, facilitating the installation procedure and avoiding
destroying the grass or road surface[3]. Therefore, it can be
applied on places such as plaza, campus, streets, highway and
so on.
The solar lights are being increasingly used in Wuhan city,
thanks to the governments support. They are scattering in
every part of this city,
Due to the free pollution and security of the solar lights, the
Wuhan botanical garden introduces about 2000 solar lamps to
protect the plants from heat generated by conventional lamps,
which may disorganize the growth law of plants.
In the park of river beach, Wuhan, about 200 solar lights
are implemented. The technologies of wind and solar energy
are combined to provide electricity for lights, which achieves a
100% energy saving rate. Beside, on major roads of East Lake
Development Zone, a large scale of solar lights has been a
beautiful spectacle attracting the citizens attention.
To welcome to Beijing Olympic Games, more than 4900
solar lamps including lights, landscape lamps and grass lamps
developed by the photovoltaic institutions of Wuhan city are
installed in Beijing. These solar lamps illuminate the whole
Beijng and illustrate the concept of Green Olympic Games.
II.

SYSTEM PRINCIPLE ANALYSIS

While solar energy is abundant enough to reduce reliance

on nonrenewable energy sources and capable of meeting
energy
demands
without
significant
environmental
degradation, the high cost of this technology constrains its
widespread use. But the large-scale conversion of solar energy
to generate power is also restricted to favorable locations with
abundant sunshine [2]. And even in these locations, attenuation
and interruption of sun irradiance by unfavorable weather and
by the diurnal cycle leads to requirements for energy storage to
provide a dependable power supply. Consequently, solar
energy remains to be developed on this issue.

A. Design principle
The general design principle of the solar light system is as
follows [4]:

The solar light system, which is developed and

implemented by the Rixin technology company, has been
widely installed in Wuhan city, Hubei province of China under
the auspice of Wuhan government. The solar light system

B. System structure
The solar light illumination system, in essence, is a system
of converting light from sun to electricity, which is called as
silicon photovoltaic module. Silicon panel receives the sunlight

On the premise of satisfying the power demand of the load,

ensure minimizing the capacity of the battery module and the
power of the solar cell module. Due to the fact that during the
design of the photovoltaic generation system, each decisionmaking will result in a significant influence on the cost of the
system. The investment will double owing to the inadequate
choice.

978-1-4244-2487-0/09/$25.00 2009 IEEE

and converts it to electricity, which will be stored in the

battery. The photographic switch will control the electricity
output, realizing the light on/off depending on the day and
night alternating [5].
The solar light system as Fig.1 shown comprises the solar
cells (photovoltaic module), control box, and DC energy saving
lamp, where the control box includes battery, controller and
various input and output ports.

The photovoltaic conversion efficiency is given by:

I U
P
= m m = max
Pm
Pin

(3)

Where I m and U m represent the maximum current and

voltage, respectively, Pmax is the maximum output power
when the solar cell receives the sun irradiation, and Pin is the
input power of the solar cell.
III.

CIRCUIT MODEL OF THE SOLAR CELL

The circuit model of the solar cell is shown as the Fig.2.

The equivalent shunt resistance Rsh existing in the internal
material is usually large, while the equivalent series resistance
Rs in the external material is relatively small [8]. The
relationship between voltage and current of the output load
RL can be illustrated as:

U +I R
q
I RL = I s I so exp
(U RL + I RL Rs ) 1 RL RL s (4)
Rsh
AkT

Figure1. The solar streetlight system

The power of the photovoltaic module provided by Rixin

Company is 10W, responsible for collecting solar energy. The
controller possesses high quality and reliability, which is
designed with a microcontroller for automatic lighting control
functions. Meanwhile, there are two types of DC output
voltage level, 12V and 24V. The controller protects the battery
from overcharge and deep discharge, furthermore, limits the
battery voltage to a maximum level in the case of surplus
charge and disconnects the load to prevent deep discharge in
the case of deficiency in charge. Besides, the controller is in
charge of monitoring the charge state of battery to ensure a full
utilization of the energy. The battery located in the controller
box is free of maintenance.
C. Battery efficiency and photovoltaic conversion efficiency
As for the battery, it is a critical factor for determinating the
working life of the whole illumination system.
The output efficiency of the battery is also called the charge
efficiency [7]. The battery used in the system should be
reversible; in fact, the practical battery cant be a desirable
energy-storage device. There must be a certain energy loss
during its operation, which can be defined by the capacity
output efficiency and the energy output efficiency. The
formulations below are used to calculate these two quantities.

C =

Cdis
100%
Cch

(1)

Q =

Qdis
100%
Qch

(2)

Where, Cdis represents the output electric quantity for

discharge; Cch represents the input electric quantity for
charge; Qdis represents the output energy for discharge; Qch
represents the input energy for charge;

Where q , k represents the electron charge and the Boltzmann

constant, respectively. A is the ideal coefficient of the solar
panel. T is the temperature of the solar panel. I so represents
the reverse current in the case of saturation.

Figure2. The circuit diagram of the solar cell

The technology parameters of the series of MBF125 of

the silicon solar cells module are shown in Table 1. Aimed to
this type, a comprehensive test is performed to obtain the
current-voltage (I-V) and power-voltage (P-V) characteristic
curves of solar cells module. I-V curve shows the behavior
of solar cells under all load conditions from open circuit to
short circuit. Through connecting the module to a variable
load, varying the resistance from zero to infinity to acquire
the I-V curve [9].
TableI.

PARAMETER LIST OF SILICON SOLAR CELLS

Type

Parameters

Electricity
Performance

Wp
Peak Power Output (W)
Vmp
Maximum Power Voltage (V)
Imp
Maximum Power Current (A)
Isc
Short Circuit CurrentA
Voc
Open Circuit Voltage(V)
Insulation Resistance( M )
Maximum System Voltage

MBF125
125
17.5
7.15
7.87
21.5
50
800 V

DC

In Fig.3, the static I-V characteristic curve indicates the

relationship between the current and the voltage transmitted in

the solar cell module obtained at constant sun irradiance and
temperature. The curve is determined by three points: short
circuit current I sc , open circuit voltage Voc , current and
voltage at maximum power point and I mpp . Voltage Vmpp at
maximum power point depends on sun irradiance G and cell
temperature T. The open circuit or short circuit operation of
the solar cell will not damage itself; in fact, it is this feature
that realizes the charge and discharge control of the system. In
practice, according to the practical weather condition and
temperature, the operating point of the solar cell will be
changed.

G
(10)
1 + Vmpp (T Ts )
Gs
Where parameters I scs , Vocs , I mpps and Vmpps are defined at
Vmpp ( G , T ) = Vmpps

standard conditions(G=1000/m2 and T=25 C).

The typical P-V curve of solar cell is also illustrated in
Fig.2. It is obvious that the peak power is located at the point
where dP / dV = 0 , based on P = V I , another expression is
obtained:

dI I
+ =0
(11)
dV V
As for the model in Fig.1, the value of dI / dV at the peak
power point can be calculated. By comparing the dI / dV with
I mpps / Vmpps , the values of A and Rs in the model which are
critical can be obtained by computer software. The desirable
estimation of A and Rs is achieved when the absolute value of
dI

dV

V =Vmpp

I mpp is minimized.

Vmpp

IV.

CONCLUSION

The solar light is a kind of novel and energy-saving

illumination system, possessing considerable advantages
compared with the conventional one. However, its high price
has been a bottleneck restricting its wide application.
Therefore, how to reduce the cost of the solar light is a main
concern.
Figure 3. The characteristic curve of solar cell

Where FF is the fill factor.

FF = ( I mpp Vmpp ) / ( I sc Voc )

(5)

In literature [10], the model of solar cells usually involves

five or six parameters. To obtain a simplified one-diode model
containing only four parameters, rewrite the expression in
[11], the voltage is given by:

I I
V = C2 Voc log 1 + sc
(6)

I sc C1

Where: C1 = 1 ( I mpp / I sc ) exp ( Vmpp / ( C1 Voc ) )

C2 =

Vmpp / Voc 1

ACKNOWLEDGMENT

log (1 I mpp / I sc )

Coefficients C1 and C2 depends on parameters I sc , Voc

I mpp and Vmpp are functions of sun irradiance G and
temperature T:
G
(7)
I sc ( G, T ) = I scs 1 + I sc (T Ts )
Gs

Voc (T ) = Vocs + Voc (T Ts )

I mpp ( G , T ) = I mpps

On the other hand, the power generated by utilizing the

solar energy absorbed by a given area of solar panel can be
made full use if the two technologies, photovoltaic cells to
generate electric power directly and solar thermal panels to
generate heat power. The combination of these two
technologies offers distinct advantages in cost savings in
production and installation. However, the drawback of this
solution lies in the efficiency of both photovoltaic cells and
solar thermal panels reduced as compared to the individual
application. Thus, the future work will further develop the kind
of system which will contribute to solve the problem of energy
shortage.

G
1 + I mpp (T Ts )
Gs

(8)
(9)

The work has been supported by Rixin Technology

Company. The author would like to thank Jinyu Wen for the
instruction on the paper, also appreciate Lei Yang from Rixin
Technology Company for the contribution to this paper.
REFERENCES
[1] Jong,M.J.m:System Studies on Combined PV/Thermal panels, Elftes
Symposium Thermische Solarenergie, 9-11, May, 2001, Kloster Banz,
Staffelstein,Germany
[2] A. Filion, Renewable energy options-overview and tend, Proceedings
of the 28th Annual National Conference of the Solar Energy Society of
Canada, Kingston, August 2003.
[3] Noboyuki Kasa,Takahiko Iida, Liang Chen, Flyback Inverter
Controlled by Sensorless Current MPPT for Photovoltaic Power

[4]

[5]
[6]
[7]
[8]
[9]

[10]

[11]

System, IEEE Transaction on Industrial Electronics, USA, No, 4

August 2005, Vol.52, pp.1145-1152
O.Wasynczuk, Modelling and dynamic performance of a linecommuntated photovoltaic inverter system IEEE Transactions on
Energy Conversion,Vol.4 No.3,September 1989.
Benner J P, Kazmerski L. Photovoltaics gaining greater visibility [J].
IEEE Spectrum, 1999,29(9):34-42
Ramakumar R, Bigger J E. Photovoltaic systems[J]. Proceedings of the
IEEE,1993, 81(3): 365-377.
Zhao Xiuchun, Zhang Xi, Li Peifang. A new style controller for
photovoltaic lighting system[J]. Energy Engineering, 2004(5): 28-32
Texas Instruments Corp. Improved charging methods for Lead-Acid
batteries using the UC3906[R]. Texas, USA: TI corp.1999
Liu, S, R.A.Dougal, R.A: Dynamic multiphysics model for solar
array,IEEE Transaction on Energy Conversion, Vol.17, No.2, June
2002, pp.193-200.
Gow, J.A; Manning,C.D.: Development of a photovoltaic array model
for use in power-electronics simulation studies, IEE Proc. Electr. Power
Appl.,Vol. 146, No.2, March 1999. pp. 193-200.
Xiao, W.; Dunford, W.G.; Capel, A: A novel modeling method for
photovoltaic cells, IEEE Power Electronics Specialists Conference,
Aachen, Germany, 2004.

Design, Characterization and Fabrication of Silicon Solar Cells for >50%

Efficient 6-junction Tandem Solar Cells
Ngwe Soe Zin, Andrew Blakers, Evan Franklin, Vernie Everett

The Australian National University, Canberra ACT 0200, Australia

http://solar.anu.edu.au/
Email: soe.zin@anu.edu.au
Abstract- A major objective for photovoltaic conversion

is to develop high efficiency solar cells. Many

approaches are under investigation - MUltiple Junction
Solar Cell, Multiple Spectrum Solar Cell, Multiple
Absorption Path Solar Cell, Multiple Energy Solar Cell,
and Multiple Temperature Solar Cells [1]. The Multiple
Junction Solar Cell approach based on a six-junction
tandem solar cell has been adopted to achieve
conversion efficiency of greater than 50% in the VHESC
program sponsored by DARPA [2]. In six-junction
tandem solar cells, individual solar cells are stacked on
one another and each solar cell absorbs the best-matched
slice of the solar spectrum. Silicon is one of the cells in
the tandem stacks, and absorbs photon energy of 1.42 1.1 eV. The role of the silicon cell is to convert 7% of
the light incident on the tandem stack into electricity.
Other cells in the stack contribute the balance of the
electricity. Key design parameters for the silicon cells
are that it should have dimensions of 2.5 x 8 mm 2 and it
needs to transfer light with energy of less than 1.1ev to
the underlying solar cells.
In this paper, discussion is made of the design of the
silicon cell. Minority carrier recombination at surfaces
and in the volume, internal quantum efficiency,
resistance losses, free carrier parasitic absorption, optical
reflection, light trapping, and light absorption must be
traded off against each other. PCID modeling is used to
analyze the various parameters and produce estimates of
short circuit current, fill factor and open-circuit voltage
of the cell [3]. In addition, characterization of solar cell
by photoconductance measurement to analyze carrier
recombination and emitter saturation current as well as
to predict the open-circuit voltage of solar cell [4, 5] is
presented. Discussion of cell fabrication process
followed by I-V testing is presented. Completed solar
cells were tested in ANU using an in-house fabricated
current-voltage flash tester [6] under AM1.5D.
I. CELL SRUCTURE AND DESIGN

The external dimensions of the silicon solar cell have to

be 2.5 mm 2 in width and 8mm in length in order to fit
the cell package. Cells were fabricated using 450llm
thick <100> p-type float-zone 10cm wafers. The cells
have an active n-type emitter region on both front and
back surfaces of the cell with the dimension of 6.5 x
1.9 mm 2 Metal contacts are made to both surfaces of

978-1-4244-1641-7/08/$25.00 2008 IEEE

the cell. The n-contact to the external world is on the

sunward side and the p-contact is on the back of cell.
Metal contacts are designed with a spacing of 1.9mm in
the lateral direction and 6.5 mm in the lengthwise
direction as shown in the figure 1. Multiple cells are
processed simultaneously on each wafer until the
metallization step is completed, at which time they are
diced out of the wafer to form individual solar cells.

:r""

.;l~
.... : I'

.L~,~~~------Y

Fig.l Top view and 3D view of Silicon solar cell

1.1 Illumination
Sunlight with power of 2W/cm' (20 suns) is incident
on the top of the package. GaAs and other high bandgap
cells will absorb much of this light. Light of energy
<1.42eV (875nm) will be transferred to the Si cell. Light
of energy <1.1 eV must be transferred to underlying cells.
Therefore, BSR and texturing cannot be used. A
polished cell will have to be thick, 0.5-2mm, to have a
reasonable efficiency of conversion in the weakabsorption wavelength range 875-1100nm.
1.2 AR Coating
The cells will be surrounded by an optically thick
n=1.4 medium. Based on modeling calculations, oxide
and nitride coating thickness of 30nm and 90nm
respectively gives rise to a loss of 1% and 3% in air and
under encapsulation, and is reasonably close to optimum
for both.
1.3 Contact Recombination
Contact recombination will be suppressed with small
contacts (~1 %) and/or heavy doping beneath the contacts.
1.4 Surface Recombination
For a surface area of 50 mm 2 in total, a reasonably
good surface passivation will be grown by using a
thermal oxide.
1.5 Edge Recombination
Edge recombination in small cells is a formidable
problem. Since the cells are small, they need to be

processed in a host wafer, for practical reasons. However,

it is highly desirable to cut the cell mostly out of the host
wafer before the final oxidation to allow oxide
passivation to minimize recombination at the cut/scribed
edge.
1.6IQE
To achieve IQE above 99%, the diffusion length will
be kept above 3-4 times the cell thickness. Bifacial cells
can be roughly twice as thick for the same IQE because
they have a junction on both surfaces.

should be around 630mV. Internal quantum efficiency is

also predicted using PC1D to achieve more than 90% as
shown in Fig.3.
Internal Quantum Efficiency Curve (PC1D)
100
80

~ 60

w
Q

40
20

1.7 Resistive Losses

For a 50f..lm, 100f..lm and 500f..lm thick polished cell,
65%, 74% and 87% of the 875-1100nm light will be
absorbed in the top half of the cell respectively. For a
100f..lm thick textured cell, 64% of the light will be
absorbed in the top half of the cell. Therefore current
sharing in a cell with a junction on both surfaces is a
significant but not overly important factor in reducing
emitter resistive losses. Additionally, the electron current
in the rear phosphorous diffused layer of a bifacial cell
will be small, and so will resistive losses.

o
200

......
400

......--..

600

1000

800
Wavelength

....

1200

1400

(nm)

Fig.3 Internal Quantum Efficiency modeled by PC1D

III. CHARACTERISATION AND FABRICATION
Before the fabrication of the cells, characterization of
material performance was conducted for minority carrier
life time and emitter saturation current (Joe) at
equivalent I sun intensity after each process step.
Characterization process is shown in table 1.

1.8 Lateral Diffusive Losses

Electron and hole currents can be extracted from the
two edges of the cell or the two ends or both. The sheet
resistance of a 500 micron thick, I ohm-cm wafer is 20
ohms/sq, which is much lower than the phosphorus
diffusion sheet resistance. Therefore electron resistive
losses are of more concern than hole resistive losses.
The best arrangement is to have n-contacts at both edges
and both ends of the active area. These must be on top
because most of the current is generated close to the top
surface. The p-contacts can be on the bottom, on both
edges and optionally on both ends as well. Based on
calculation, estimated resistive loss will be around 5%.

Table 1 Sequence of Characterization Process

II. MODELING BY PCID

PCID modeling is used to analyze the short circuit
current, open circuit voltage, fill factor and internal
quantum efficiency under 1 sun illumination condition.

Figures 4 illustrate that the adequate life time of around

560f..ls was maintained after all processes. From the
graph it can be seen that life time has dropped sharply
after heavy phosphorous diffusion and subsequent
annealing is just enough to maintain the life time.

Current-Voltage Curve (PC1D)

'(]4

0.2

'."'"

0.2

0.4

Stt'p
Diffusion

Oxidation
LPCVD
Diffusion
Oxidation
Diffusion
FGA 1
FGA2

Dt'tail
Emitter formation
Grow thin passivation oxide

Deposit AR coating
n++ formation
To fonn diffilsion barrier
p+ fonnation
Forming gas annealing to promote
hydrogenation
Forming gas annealing to promote
hydrogenation

Life-time of Sample after each process step

..

0.6

08

.
Voc(mV)

Fig.2 Current-Voltage curve by PC1D modeling

Modeling in fig. 2 estimates that short circuit current
should be around 4.2mA while open circuit voltage

978-1-4244-1641-7/08/$25.00 2008 IEEE

Process

step

FigA Characterization of Minority Carrier Life Time

after each Process Step
Values of surface recombination were kept low
throughout the process except after the heavy

phosphorous diffusion. Value of emitter saturation

current was maintained at around 25 fAl cm 2 after the
last processing step. Figure 5 illustrates the
characterization of Joe.
Joe of Sample after each process step
250.0

:c- 200.0

150.0
.. 100.0

.s:

50.0

0.0

1-+- 1.006-151

L__~__~__~__~__~::~~~~25.6
Fig. 6 Completed Si solar cells before dicing
Process

Step

Fig. 5 Characterization of Emitter Saturation Current

Cells were fabricated using standard silicon process
techniques, as shown in the table 2. Completed cells
were diced out from wafer to form individual cell.
Figure 6 shows the completed cell before dicing.
Table 2 Fabrication of silicon solar cells
StE'P

DE'taii

Series resistance is caused by inadequate ohmic contact

formed between the metal and diffusion. Table 3
illustrates the values of open circuit voltage, short circuit
current and fill factor of the cells tested without having
spectral filtering (eg using a GaAs cell).

iUIl

... tlIoQ~1-

11111

1.lID

Ul!

D..UI

n.sm o.~

11m

Si Etch
RCA Clean

HF:HNO etch
To clean organic/inorganic contaminant

Oxidation
Photolithography
Laser Scribe

Thick oxide for diffusion barrier

To make opening for diffusion
Create slots through the wafer

Fig. 7 Result of current-voltage for cells tested with I-V

flash tester

Diffusion
Oxidation

Emitter formation
Grow thin passivation oxide

Table 3 Voc, Isc and FF of Cells tested

LPCVD
RIE

Deposit AR coating
Make opening for n++ diffilsion

Diffusion

n++ fonnation

RIE

To make opening for p+ diffusion

Diffusion

p+ fonnatioll

Metallization

Deposit metal CrlPd/Ag

Sintering

To foml good ohmic contact

Dicing

To cut cells in individual

Testing

I-V testing

IV. TESTING
After dicing, individual cells were wired with electrodes
on both sides of n and p regions. Cells were tested with
current voltage flash tester under 1 sun illumination
condition. Of many cells, the best performing cells are
selected for further characterisation. Cells tested are
observed to have high shunt resistance but significant
series resistance, as shown in figure 7.

978-1-4244-1641-7/08/$25.00

2008 IEEE

oQ.~

............................'
.".

.~"

.~"

Voe
(mV)

Ce116wfr2
Ce117wfr2
Ce118wfr2
Ce119wfr2

612
619
619
610

Ise
(mA)

4.25
4.10
4.34
4.00

FF
9%)

75.9
27.4
66.2
54.9

Erf
(%)

16.5
5.79
14.8
11.2

V. CONCLUSIONS

Cells have been completed that demonstrate the absence

of shunts and a good fill factor. However, the open
circuit voltage and current are low compared with the
results expected from modeling. There were known
problems with the processing which will be fixed in
subsequent batches, which is likely to lead to sharply
improved efficiency.
REFERENCES

[1] CB Honsberg, AM Barnett, D Kirkpatrick,

"Nanostructured Solar Cells For High Efficiency
Photovoltaics," Photovoltaic Energy Conversion,
IEEE 4th World Conference, May 2006, Hawaii,
U.S.A

[2] Allen Barnett, Christiana Honsberg, Douglas

Kirkpatrick, Sarah Kurtz, Duncan Moore, David
Salzman, Richard Schwartz, Jeffrey Gray, Stuart
Bowden, Keith Goossen, Michael Haney, Dan
Aiken, Mark Wanlass, Keith Emery, "50% Efficient
Solar Cell Architectures and Designs," Photovoltaic
Energy Conversion, IEEE 4th World Conference,
May 2006, Hawaii, U.S.A
[3] D. A. Clugston and P. A. Basore, "PCID ver. 5: 32
bit solar cell modeling on personal computers," 26th
Photovoltaic Specialist Conference, September 1997,
California, U.S.A
[4] R. A. Sinton and A. Cuevas, "Contactless
determination of current-voltage characteristics and
minority-carrier lifetimes in semiconductors from
quasi-steady-state photoconductance data," Applied
Physics Letter, Volume 69, Issue 17, 1996, Pages
2510-2512.
[5] Cuevas and R. Sinton, "Prediction of the opencircuit voltage of solar cells from the steady-state
photoconductance," Progress in Photovoltaics
Research and Applications, Volume 5, Issue 2, 1997,
Pages 79 ~ 90.
[6] William M Keogh, Andrew W. Blakers, Andres
Cuevas, "Constant voltage I-V curve flash tester for
solar cells," Solar Energy Materials and Solar Cells,
Volume 81, Number 2, 2004, Pages 183-196.

978-1-4244-1641-7/08/$25.00 2008 IEEE

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET

Using Patent Data to Analyze the Development of the ext Generation of Solar Cells
Fang-Mei Tseng, Yi-Wei Chu, Ya-Ni Peng
Yuan Ze University, International Business Dept., Tao-Yuan, Taiwan, R.O.C
Abstract--With the shortage of raw materials for the
production of crystalline silicon solar cells, the next generation
of solar cells has reached the perfect stage for development.
While in the past researchers have laid emphasis on the
development of crystalline silicon solar cells, the application of
the next generation of crystalline silicon solar cells is different.
Among the thin film solar cells, the most widely-recognized is
the a-Si thin film solar cell with the greatest potential to be
developed and many manufacturers have already invested in
R&D. In this study, we employ the patent portfolio proposed by
Ernst to embark on the analysis of technology development
regarding the a-Si thin film solar cell and find that the major
technology field has reached the mature stage in the technology
life cycle; moreover, four patent strategic clusters are obtained.
The results regarding the company level and technology level of
the clusters are integrated with the profile data and the
development focus to realize the patent performance, technology
capacity and R&D background concerning the technological
fields of the a-Si thin film solar cell and to propose patent
strategies to companies.

I. INTRODUCTION
In recent years, the energy industries throughout the world
have been affected by the dramatic rise in the price of fossil
fuel energy, the short supply of energy, and the crisis brought
about by environmental pollution. This has resulted in the
promotion of a significant change in the industrial framework.
With the sharp increase in world population and economic
growth, the need for energy has been stimulated, and this in
turn has given rise to a sustained increase in the international
price of oil. In August 2006, the international crude oil price
increased to US$78.65 a barrel, and in April 2008 the price
broke through the US$100 a barrel landmark, a feat never
accomplished before. However, as the fossil fuel resources
will eventually become depleted, in the long-run a rise in
prices will surely be an inevitable trend.
In addition, humans have overused the traditional fossil
fuels and this has resulted in the increased density of carbon
dioxide in the atmosphere, which has worsened the global
warming problem, and caused serious damage to the natural
environment. Therefore, the Kyoto Protocol that was signed
by countries worldwide officially went into effect on
February 16th, 2005. The Kyoto Protocol set standards to
control greenhouse gas emissions by its members.
Of all the green energies, solar energy has experienced the
most vigorous growth, and its energy consumption continues
to rise. Due to the rapid growth of the PV market, the demand
for solar cells has increased significantly. The upstream
silicon, found in crystalline silicon solar cells, started to be in
short supply in 2004 and its prime cost increased. The next
generation of solar cells consists of various types, including

the second generation thin film types (amorphous (a-Si), cSi, etc.), compound types, and the third generation organic
types (dye-sensitized solar cell (DSSC), etc.). At present,
within the Si thin film solar cells, the a-Si thin film solar cell
is the main product of interest and has the most
developmental potential in terms of market share. Currently,
many enterprises have already entered the field of a-Si thin
film solar cells and are intensely competing with each other,
as obtaining the patent for this will be the major objective of
their R&D results. Furthermore, through patent analysis,
firms can acquire valuable information on the technological
trends and their competitors respective positions, identify
important technological developments, or engage in patent
portfolio management[1,3,16]. Moreover, further decisions in
regard to the direction of the company patent portfolio will
greatly benefit the corporate strategic planning in the future
[8].
This study uses a-Si thin film solar cells as an example to
modify the patent portfolio analysis proposed by Ernst [8],
through which to further understand the R&D of a-Si thin
film solar cells in the following ways:
1. Using patent information to analyze the developmental
trend of the a-Si thin film solar cell technology and
preparing a report analyzing the patent portfolios among
the main competitive corporations, thereby providing a
valuable source of reference for domestic enterprises and
academia;
2. Using the CHI Research patent indicators to categorize
corporations according to their technical competency in
order to understand their patent technology, and using the
methods and concepts of the patent portfolios as sources
of reference to list which corporations are the technology
leaders and the potential competitors, and which ones are
falling behind;
3. Exploring the patent strategy, technical competency, and
R&D focus on the a-Si thin film solar cell field, which can
be used as a source of reference to draft a patent strategy
and plan R&D directions in the future.
II. LITERATURE REVIEW
A. Patent Analysis
Patents can safeguard the inventions of enterprises and
bring about economic benefits. Through patent data, we also
can see the industry tendency as well as the competitive
power of enterprises or countries [4,19]. Patent analysis is a
method used to transfer these data to useful patent
information. It is a good tool for planning R&D and
managing intellectual property, and it is often used to analyze
competition in relation to techniques and trends [7,12]. The

3064

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET
main process of patent analysis is to use statistical analysis,
multivariate analysis or other quantitative models to analyze
and interpret each field of patents (such as the application
date, assignee name, assignee country, international
classification,etc.) to obtain patent information.
We consider summarized four aspects of patent analysis
from previous papers.
1. Discussing R&D inputs and
outputs: Trajtenberg [18] discussed the relationship between
patents and profit or R&D expenditures; Griliches [11]
pointed out that patent data can be an alternative indicator of
invention-related inputs and outputs when R&D data is
lacking. 2. Measuring a companys technical abilities: Jaffe
[15] used patent analysis to characterize the technological
position of U.S. manufacturing firms; Choung et al. [6]
investigated the evolution of the technological capabilities of
Korean semiconductor firms. 3. Planning a companys
strategies: Ernst [8] used the patent data of mechanical
engineering companies to evaluate the companies
technologicaly ability and demonstrated the application of
patent portfolios for strategic R&D planning purposes. 4.
Discussing market value and financial performance.
However, the above researchers elaborated mostly on the
application of patent information or used statistical methods
to carry on perform an empirical analysis of the companies
patent information. They did not further investigate the
technicalque position or patent strategy of companies. For
this reason, we aim at to remedy this deficiency to by
providing a thorough discussion. We revise the patent
portfolio analysis proposed by Ernst to measure the
technologicaly ability of the sample companies in the nextgeneration solar cell industry and Chen use the patent
portfolio analysis to assess technological ability of solar cell
vendors [5,8,9]. We also use multivariate analysis to develop
a set of suitable patent indicators, determine the patents
position, and induct select groups to discuss the patent
characteristics and the specialty of each group. Through these
analyses, we want seek to provide the more objective
information of regarding the firms R&D strategy.
B. Patent Indicators
Many research organizations and scholars develop
different patent indicators by themselves. The following is an
introduction to different patent indicators, which we use to
design a set of suitable and easy-to-estimate indicators to help
us to carry out the patent portfolio analysis, which objectively
evaluating the firms technological ability and R&D strategies.
CHI patent indicators are brought up by CHI Research
(now is ipIQ) to analyze corporate patent and technological
strength. CHI patent indicators include three standard patent
indicators, four citation indicators and two science linkage
indicators. Narin [17] showed how various technology
indicators (including NP, CII, TCT) can be used to
characterize the position of the EU in 13 technologies
compared to the North American region and to Japan, and be
used by corporate management in competitive analysis.
Huang et al. [14] thought that those indicators were also a

very good criterion to evaluate the science and technology

competitiveness of a particular country. Hsu [13] analyzed
the patent portfolios of important companies in the Magnetic
Random Access Memory (MRAM) industry, and. He used
those indicators to evaluate those companies. He also
employed factor analysis and cluster analysis to divide and
confirm the patent activities and quality to position of firms.
C. Patent Portfolio Analysis
Although Brockhoff [2] proposed the technology level,
Ernst [8] proposed the company level, and combined these
two levels to analyze the patent portfolio. By analyzing a
companys patent portfolio, we can acquire relevant
information regarding a competitors R&D strategies that will
help us to assess the competitive potential of technologies.
There are two dimensions, namely, patent activity and
patent quality, at the company level [8] Patent activity
measures the level of R&D activities, whereas patent quality
measures the impact of these activities. Patent applications
are the indicator of patent activity. Patent quality is measured
by calculating an index of patent indicators. Patents granted,
valid patents, patent applications in major foreign countries
and patent citations have frequently been identified as
indicators of patent quality [8]. The patent portfolio on the
technology level also has two dimensions, namely, the
relative patent position and the technologys attractiveness.
The relative patent position is derived from the number of
patent applications by the firm relative to the number of
patent applications of its most active competitors. The
attractiveness of each technological field is assessed by using
the growth rates of patent applications[10].
III. RESEARCH METHODOLOGY
We adopt the methodological structure of the patent
portfolio analysis proposed by Ernst [8], and assess the
performance regarding the patents of the a-Si thin film solar
cells of the leading 20 companies and research centers in
terms of R&D capacity based on 10 patent indicators at the
company level. For the technology level, we employ the
United States Patent Classification (UPC) to discover the
mainstream technological field of the a-Si thin film solar cell
and utilize the technology life cycle to estimate the
development status of the technology. We also use 4
indicators to discuss the research focus and technical
competency.
A. The Company Level
Ernst [8] utilized patent applications to measure patent
activity and used the share of granted patents, the share of
valid patents, the share of US patents, and the citation ratio as
the patent quality index. However, utilizing only patent
applications to measure patent activity is somehow
insufficient; in addition, information on the corporations
shares of granted patents, valid patents, and US patents are
not easy to obtain, thus giving rise to difficulties in measuring

3065

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET
patent quality. As a result, we adopt the 9 patent indicators
drawn up by CHI Research, in addition to another type of
patent citation indicator. The simplicity involved in using
these 10 patent indicators for measurement facilitates the
execution of the patent portfolio analysis methodology.
We adopt factor analysis to reveal the 10 patent indicators
potential characteristics, condensing them into a few factors
and using these extracted factors to measure the corporations
overall patent performance instead of using the traditional
approach of analyzing only the patent quality and quantity
performances. In the next step, we employ a two-step cluster
analysis to categorize the corporations patent competency for
confirming the patent position of each corporation and look
further into the factor differences amongst them. This helps
us to understand the characteristics of each category and to
acquire crucial information. The following list describes the
10 indexes used in this study:
1. Number of Patents (NP): A count of a companys issued
patents in the a-Si thin film solar cell field. By tracking
the number of patents, we can observe the trend of
companies technological development;
2. Patent Growth Percent in Area (PGPA): The change in the
number of patents from one time period to another,
expressed as a percentage. A higher growth rate means
that the technologies are more emphasized;
3. Percent of Company Patents in Area (PCPA): The number
of patents in the a-Si thin film solar cell field divided by
the total number of patents for that company, expressed as
a percentage. This is used to measure the R&D situation
of a company and how the company emphasizes the
technology;
4. Cites Per Patent (or Citation Index, CI): A count of the
citations received by a companys patents from
subsequent patents divided by the total number of patents
for that company. It is used to evaluate the technological
impact of patents;
5. Current Impact Index (CII): The number of times a
companys previous five years of patents are cited in the
current year, relative to all patents in the U.S. patent
system. It can measure the influence of a company in the
most recent five years;
6. Technology Strength (TS): The number of patents
multiplied by the current impact index. It can measure the
scale of the impact of a company on a technological field;
7. Technology Cycle Time (TCT): This is the median age in
years of the U.S. patent references cited on the front page
of the companys patents and is used to measure the speed
of innovation or how fast the technology is turning over;
8. Science Linkage (SL): The average number of science
references cited on the front page of the companys
patents. It mainly evaluates the relationship between the
patenting technique and the scientific research of a
company;
9. Science Strength (SS): The number of patents multiplied
by the science linkage. It can measure the scientific
development level of a company;

10. Number of citations (NC): A count of the citations

received by a companys patents from subsequent patents.
It is also used to evaluate the technological impact of
patents.
B. The Technical Level
Before going into the technological level of patent
portfolio analysis, it is essential to confirm the technology
field that is included in the a-Si thin film solar cell technology.
All the current classifications of technology adopt the IPC
established by the World Intellectual Property Organization
and the UPC.The main difference between the IPC and UPC
is that the latter updates monthly and reviews the previously
approved patents. Therefore, the UPC can respond quickly
and appropriately classify the rapidly-changing technology.
Moreover, the principles of the two are different as well. The
IPC classifies patents based on their claims, whereas the UPC
differentiates among patents based on the patent information
and places emphasis on the function of the manufacturing
process and on intrinsic characteristics. Using the UPC
classification as a patent analyzing method can give us a
better command of the patent technical content [20]; therefore,
we use the UPC method as a standard for classifying a-Si thin
film solar cell technologies.
After deciding on related technical classifications that
would be included in the a-Si thin film solar cells, we draw a
patent portfolio graph of the technology level according to
the analytical methodology of Ernst[8]. The vertical axis
represents the technologys attractiveness (Relative Growth
Rate, RGR or Relative Development Growth Rate, RDGR)
and the horizontal axis is the Relative Patent Position (RPP);
the calculation method is as:
1. RGR= average growth of patent numbers per
technological field between 1997 and 2006/ average
growth of patent numbers in all technological fields
between 1997 and 2006;
2. RDGR= average growth of patent applications between
2002 and 2006/ average growth of patent
applications between 1997 and 2001;
3. RPP= number of a firms patents in the technological field
/ number of patents in the technological field from
the most active competitor.
We use the technology level to discuss the development
status of the current a-Si thin film solar cell mainstream
technology as well as the relative technical strengths and
weaknesses of corporations in each field to understand their
degree of investment. The technology life cycle is also
applied to estimate the technological field development status
and to offer future technical strategy suggestions to the
companies.

3066

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET
IV. ANALYSIS AND DISCUSSION
A. Retrieval and Selection of Patent Data
USPTO is retrieved via the Patent Guider developed by
Learning Tech Corp. in this study. Moreover, the key words
of the patent related to the a-Si thin film solar cell as the
fundamental of data retrieval are collated by a literature
review and expert interviews. As a result, the key words are
(photovoltaic or solar cell) AND thin film AND
(amorphous silicon or a-Si or amorphous Si). As the
conditions for retrieval are decided, the cases applied,
approved and announced during 1/1/1976 to 11/27/2007 are
inquired of based on the fields of the abstract and title in the
USPTO. In the end, 806 patents and 263 assignees are
obtained.The following steps are processed for the further
realization of the assignees and the efficiency of the patent
analysis and for narrowing the extent of patent retrieval to
match the demands of this study: unification in accordance
with companies and corporate M&A, and deletion of the
personal patents and the patents with the citation number 0.
Through manual reading and confirmation article by article,
618 patents and 67 assignees are obtained. Then we list the
leading 20 companies with comprehensive R&D capacities in
the Appendix A by Patent Guider software.
B. Patent Portfolio Analysis
1. The Company Level
To inquire regarding the patent performance of the above
list of 20 companies and research centers, 9 patent indicators
proposed by CHI Research are employed with another patent
citation indicator for the factor analysis of the patent
indicators. Two patents of Teijin Limited are for the same
year, so it is out of the question to calculate the patent growth;
accordingly, the indicator of patent growth is removed.
According to the Kaiser method, the components with
eigenvalues larger than 1 will be reserved for factors
numbering not more than 6 and accounting for 70% or more
of the cumulative percentage of variance. The KMO and
Bartle test are used with the individual MSA variables, all of
which exceed 0.5. The MSA for 2 indicators, the technology
cycle time and science linkage, are both less than 0.5, and so
they are deleted; subsequently, the factor analysis proceeds
for the other 7 indicators following the same steps. Although
3 factors are extracted, the third factor contains only one
variable, namely, the citation index, and therefore CI is
isolated. The factor analysis proceeds with the 6 remaining
patent indicators. In the end, 2 factors are extracted; The
cumulative percentage of variance is 91.796%, indicating that
the factors extracted are able to represent the raw-data
structure. Factor 1 is referred to as Patent Activity. Factor 2 is

referred to as Recent Impact. In addition, an independent

variable is the citation index; this citation index measures the
quality of the patents, and so is referred to as Patent Quality.
With the new variables, the performance regarding the
original variables of the samples is standardized, and the
standardized values are used in the cluster analysis that
follows. In the first phase, the companies are clustered with
regard to the performance of the 3 factors through Wards
Method of hierarchical cluster analysis, with the similarity
among the 3 factors being measured using the squared
Euclidean distance. The number of clusters is judged by the
increment of the coefficient of agglomeration. The increasing
ratio of the coefficient of agglomeration is the highest as the
clusters are altered from 4 to 3, and so there are 4 clusters.
In the second phase, the final cluster center based on
Wards Method is the starting point of K-Means for clustering.
Through the repeated iteration and classification, the
company number of each cluster is 1, 1, 7 and 11 in order. We
continue our research as the characteristics of the clusters are
realized by variance analysis. As for the test of the
homogeneity of the variances of the 3 factors, the p-values of
Patent Activity, Current Impact and Patent Quality are
respectively 0.022, 0.241 and 0.085. Patent Activity has a pvalue of less than 0.05; that is, the population variance of
each cluster is different. As for the individual test of the 3
factors, the p-value of Patent Activity, Recent Impact and
Patent Quality is 0.000, which means that differences exist
among the 4 clusters regarding the 3 factors. Accordingly, the
difference is compared by the Games-Howell test and named
cluster A is referred to as the Leading Technology Cluster.
Cluster B is referred to as the Emerging Potential Cluster.
Cluster C is referred to as the Technology Quality
Orientation Cluster. Cluster D is referred to as the
Technology Lag Cluster. The names of the clusters are
collated in Appendix A .
2. The Technology Level
We adopt UPC as the classification standard concerning
the technologies of the a-Si thin film solar cell. UPC 136, 257
and 438 have more numbers of patents; as a consequence, we
define the preceding 3 technology categories as the major
technological fields of the a-Si thin film solar cell in the
appendix B. A patent portfolio figure adopts the technology
attractiveness (i.e., RGR and RDGR) as the indicator of the
x-axis and the relative patent position (RPP) as the indicator
of the y-axis and the importance of a TF as the size of
technology strength (i.e., the area occupied in the patent
portfolio figure). With the three preceding patent indicators in
accordance with the major technological fields and subjects,
we draw patent portfolio figures as Figures 1 and 2.

3067

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET

RPP
TF1

TF3

TF2

RDGR
Figure 1: The Patent Portfolio for the Technology Level (RGR)

RPP

TF1

TF3

TF2

RDGR
Figure 2: The Patent Portfolio for the Technology Level (RDGR)

(1) TF1 (Batteries: thermoelectric and photoelectric)

Among all the technological fields, the patent number for
TF1 is the highest, but RGR and RDGR are negative and
lower than the other 2 technological fields, indicating that
TF1 is experiencing negative growth. Based on speculation,
TF1 may be mature or may encounter a technology
bottleneck. The companies investing in TF1 should use the
patent map to search for another cutting-edge technological
field or else improve the original technology.
(2) TF2 (Active solid-state devices)
RGR of TF2 is higher than that of TF1 and lower than that

of TF3, but RDGR of TF2 is higher than that of both TF1 and
TF3. Compared to the other 2 technological fields, RDGR of
TF2 is relatively higher, meaning that perhaps TF2 is worth
being developed by the companies merely investing their
research resources in TF1 or TF3.
(3) TF3 (Semiconductor device manufacturing: process)
Among all the technological fields, RGR of TF3 is the
highest, but RDGR of TF3 is lower than that of TF2,
indicating that development is worthwhile despite there being
less potential to grow than in the case of TF2. The companies
investing in TF3 may stop investing excessive R&D

3068

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET
resources and search for another technological field in
addition at the same time when developing TF3.
With the cumulative patent number and time of the three
TFs, we use Peral curve to draw a TLC figure as Figure 3.
Figure 3 demonstrates that the development trend of the 3
major fields matches S-curve pattern in the mature period.
That is, the progress of technology slows down in the mature
period. As companies are familiar with the technology
applied to products, the rate of patent licensing will decline.
In addition, the competition among companies is intense. In
other words, R&D staff members in this industry have
accumulated enough experience and knowledge regarding the
technology, so that the commercialized technology is
frequently improved. Since most customers have bought
products developed using such technology, the related
knowledge has been made known and learned publicly.
Therefore, companies should formulate patent strategies such
as authorization, avoidance of designs with patent protection
and improved development, proper patent strategies are to be
chosen for the beneficial usage of patents.

Figure 3: Technology Life Cycle of Three Major TFs

C. Integration of Patent Portfolio Analytical Results

The analytical results regarding the company level and
technology level of the clusters are integrated with the profile
data and development focus to realize the patent performance,
technology capacity and R&D background concerning the
technological fields of the a-Si thin film solar cell and to
propose patent strategies to companies. The performance
integration is shown in Appendix B.
1. Leading Technology Cluster
There is only one company in this cluster, Canon. As a
camera manufacturer, it has greatly diversified its business.
Canon has 253 patents related to the a-Si thin film solar cell
with the most output in terms of patents. Its achievements in
R&D are located evenly in TF1, TF2 and TF 3 and it is a
leader in all 3 fields. Although the solar cell is not the
operating focus of Canon, R&D regarding such technology is
heavily emphasized. The excellent performance of Canon
regarding patent activity and patent quality demonstrates the
ownership of the technology and fundamental patents related
to a-Si thin film solar cell and the firms status as a
Technology Leader. Canon may apply for patent licensing
via technology expansive R&D for the future development of

solar cells and as the technology involved in the chips

changes.
2. Emerging Potential Cluster
There is only one company in this cluster, Evergreen Solar.
It develops, manufactures and sells solar power products.
Evergreen Solar owns only 5 patents related to the a-Si thin
film solar cell with the main R&D investment taking place in
TF1 and TF3 and without patent output in TF2. Meanwhile,
the relative patent position in the 3 technological fields is low.
Despite the low patent output, Evergreen Solar has had a
huge recent impact as an emerging potential company. In
addition, Evergreen Solar is involved in 1st-generation and
2nd-generation solar cells in spite of its small scale.
Evergreen Solar has a large-scale technology investment but
low technological capacity in TF1 and TF3, indicating its low
R&D efficiency. As a consequence, it is necessary to lower
the R&D proportion for each of the 2 fields. In addition,
Evergreen Solar should search for another technological field
that is worth cutting into to make up for the deficiency in
growth potential in both TF1 and TF3.
3. Technology Quality Orientation Cluster
There are 7 companies in this cluster, including Energy
Conversion Devices and Teijin Limited. A brief introduction
to Energy Conversion Devices (ECD) is provided below.
ECD is a large-scale corporation. United Solar, a whollyowned subsidiary of ECD, has become the world leader in
amorphous thin-film photovoltaic technology and related
commercial applications. ECD (United Solar is involved)
with 40 patents in relation to the a-Si thin film solar cell is the
company ranked second. It enjoys a secondary status in terms
of the relative patent position and the highest status in terms
of the technologys importance in the three major
technological fields, indicating that the R&D core matches
the major technology. The patent strategy is to stress the
patent quality, and so it is one of the firms in the Technology
Quality Orientation Cluster, which showing that it stresses
not only the expansion of capacity but also the patent output
quality of patent output.
With the its large-scale investment and high
technologically capacity in the three major technological
fields simultaneously, ECD must enhance its the R&D
capacity constantly and embark on technology improvement
in the future to maintain its leading status and to keep its
competitive advantages while consolidating position in
relation to the a-Si thin film solar cell.
4. Technology Lag Cluster
There are 11 companies in this cluster, including Sharp,
Pacific Solar, etc. They are referred to as the technology
lag cluster given the fact that there is no breakthrough
technology arising from the a-Si thin film solar cell field and
some of the key indicators have fallen far behind. Sharp is
the leading manufacturer of solar cells and modules. To
remain in top position, it has entered the a-Si thin film solar

3069

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET
cell field from the silicon cell. When both of them are
compared, the conversion efficiency of the a-Si thin film
solar cell is relatively low and the module efficiency is about
6% to 10%. Sharp changed the tandem of its a-Si thin film
solar cell from two tiers to three in order to improve the
conversion efficiency from the original 8.6% to 10%. With 19
patents related to the a-Si thin film solar cell, Sharp has put
the most R&D effort into TF1 and TF2 and has maintained a
competitive patent position in the three mainstreams. Sharps
R&D strategy, however, focuses more on Patent Activity
rather than Patent Quality. Therefore, its technology is
relatively less competitive in relation to the a-Si thin film
solar cell. Although Sharp is ranked third among the a-Si thin
film solar cell manufacturers in terms of market share, it is
not as dominant as in the patent. The indicator expressed by
the Percent of Company Patents in Area has shown that Sharp
is mostly devoted to the silicon cell instead of the thin film
solar cell. We suggest that only if there is a key thin film solar
cell patent can Sharp remain in a leading position.
As for Kaneka, the citation index ranks it among the
bottom four, which has shown that its technology and market
are significantly different. The main product produced by
Kaneka is the tandem on the glass of a single a-Si solar cell
and module, where the module efficiency can reach up to 8%.
Haneka actively develops Hybrid (a-Si/Poly-Si, tandem on
glass) in order to improve product efficiency. Consequently,
the citation index and current impact indicator of Kaneka are
both quite low, which results in a technology lag cluster.
V. CONCLUSIONS
The research targets of this study are mostly large-scale
resource-integrated enterprises, such as Sharp, Canon, Fuji
Electric and Mitsubishi, which have diversified their
activities into the semiconductor or TFT-LCD fields. The
technologies, materials, equipment and manpower of these
sectors are similar to those of the a-Si thin film solar cell
industry. Besides, silicon is a defective item and is a waste
material in semiconductor manufacturing. Therefore, the
inferior silicon has become the material used to produce the
solar cell. For these two reasons, for those firms to enter the
solar cell field may be regarded as an extension based on
their original technologies. Moreover, half the firms have
developed the first and second generations of solar cells. On
the one hand, the cost structure has made it hard for silicon
cell manufacturers to reduce costs so that they search for
another development. On the other hand, some of the
equipment and technologies are common to both generations,
such as PECVD, and hence many silicon cell manufacturers
will enter the thin film field to engage in parallel
development. The key findings of this study are as follows:
First, the numbers of patents can indicate the situation and
trend for a particular field. As for the solar cell industry,
manufacturers have turned to the a-Si thin film solar cell due
to the shortage of silicon. Besides, the a-Si thin film solar cell
technology and its patents have grown rapidly in recent years

for the reason that the technology can reduce costs. The
conversion efficiency, however, is less effective than that of
the silicon cell so that not all competitors have the same
manufacturing process. We find out that Canon is in the
leading technology cluster, which has dominant performance.
Evergreen Solar is in the emerging potential cluster and has
kept growing. The remaining companies need to put more
effort into key patents even though they currently have high
market share. They may lose their competitive advantage if
their technologies fall behind, if their conversion efficiency
remains inferior, or if they have to rely on technology turnkey services.
Second, there is one circumstance where firms may
engage in R&D but without manufacturing. Take Canon for
example. Canon has not regarded the solar cell as its main
product. This is also the case with the firms that provide
technology turn-key services. Ulvac is an early developer for
thin film solar equipment and is the major supplier to wellknown Japanese thin film solar cell manufacturers, such as
Sharp and Sanyo. Given that fact that Ulvac has provided
mature equipment and the thin film solar industry is growing
rapidly, Ulvac has started to provide turn-key services. This
will lead to further growth as R&D firms file patents and then
transfer technology to others. Firms may then choose to
specialize in either R&D or manufacturing instead of both.
Third, Campell [3] pointed out that the patent indicator is
an important tool when exploring technological development.
However, the messages revealed by patents are always less
evident than those related to technological development,
because the period of the patent review is too long. It usually
takes more than one year when patents are issued. Besides,
some firms do not file patents or file the wrong ones for
strategic purposes to confuse competitors. All of the aspects
will affect the patents quality. Nevertheless, the technology
for the a-Si thin film has already reached the mature stage.
Through R&D it is possible to have a breakthrough to
improve the products conversion efficiency by up to 10% to
12%. Therefore, every firm needs to put more effort into
R&D.
REFERENCES
[1]

[2]
[3]
[4]

[5]
[6]

[7]

3070

Ashton, W. Bradford and Sen, Rajat K.; Using patent information in

technology business planning, Research Technology Management, pp:
42-46, 1988.
Brockhoff, K.; Instruments for patent data analyses in business firms,
Technovation, Vol.12(1), pp: 41-58, 1992.
Campball, R. S.; Patent Trends as a Technological Forecasting Tool,
World Patent Information, Vol.3, pp.137-143, 1983.
Chakrabarti A. K.; Technology transfers and knowledge interactions
among defence firms in the USA: an analysis of patent citations,
International Journal of Technology Management, Vol.9, Nos 5/6/7,
pp:757-770, 1994.
Chen. T. W.; Use Patent Portfolio Analysis to Assess Technological
Ability of Solar Cell Vendors, Aletheia University, Taipei, Taiwan, 2005.
Choung J. Y., Hwang H. R., Choi J. H., Rim M. H.; Transition of
Latecomer Firms From Technology Generators: Korean Semiconductor
Firms, World Development, Vol.28(5), pp: 969-982, 2000.
Chung, Y. J.; Using Patent Data To Explore The Research and

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET

[8]

[9]
[10]
[11]
[12]

[13]

[14]

Development Strategy of the companies in Memory Card Industry,

master dissertation, National Yunlin University of Science and
Technology, Yunlin, Taiwan, R.O.C.,2004.
Ernst, H.; Patent Portfolios for Strategic R&D Planning, Journal of
Engineering and Technology Management, Vol.15(4), pp: 279-308,
1998.
Ernst, H.; Patent Information for Strategic Technology Management,
World Patent Information, Vol.25(3), pp: 233-242, 2003.
Fendt, H.; Strategische Patentanalyse , Blick in die Zukunft,
Writschaftswoche, Vol. 29, pp.40-45, 1983.
Griliches,Z.; Patent statistic as economic indicators: A survey , Journal
of Economic Literature, Vol.28, pp: 1661-1707, 1990.
Hall, B. H. and Z. Griliches and J. A. Hausman; Patents and R&D is
there a lag ?, International Economic Review, Vol.27(2), pp: 265-283,
1986.
Hsu, S. S.; A Modified Method for Patent Portfolio Analysis: A Study
on Magnetic Random Access Memory Technology, Aletheia University,
Taipei, Taiwan, R.O.C.,2005.
Huang, M. H. and D. Z. Chen and H. W. Chang,; National Science and
Technology Competitiveness: An Aspect from Patentometrics, Bulletin

of the Library Association of China, Vol.70, pp: 18-30, 2003.

[15] Jaffe A.; Technological Opportunity and Spillovers of R&D: Evidence
from Firms Patent, Profits and Market Value, The American Economic
Review, Vol.76(5), pp.984-1001,1986.
[16] Mogee, M. E.; Using Patent Data For Technology Analysis And
Planning, Research Technology Management, Vol.34( 4), pp: 43-49,
1991.
[17] Narin, F.; Patents as indicators for the evaluation of industrial research
output, Scientometrics, Vol.34(3), pp: 489-496, 1995.
[18] Trajtenberg, M., A Penny for your quotes: Patent Citations and the
Value of Innovations, RAD Journal of Economics, Vol.21(1), pp:
172-187,1990.
[19] Tsai, C. L.; Technology analysis for front-end industry using patent
roadmap and technology roadmapa case study based on CT-FED,
master dissertation, Feng Chia University, Taichung, Taiwan,
R.O.C.,2006.
[20] Wu, C. T.; The Research of Industry Patent Application and Patent
StrategyComputer Heat Dissipation Industry, National Sun Yat-sen
University, Kaohsiung, Taiwan, 2005.

APPENDIX A: THE R&D CAPACITY OF THE LEADING 20 COMPANIES

Ranking
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
18
17
20
19

Patent
number
253
40
33
36
32
10
19
19
9
14
12
9
4
9
7
5
2
3
6
4

Company (R&D capacity)

Canon Kabushiki Kaisha (826.9)
Energy Conversion Devices, Inc. (243.6)
BP Solar International Ltd. (120.2)
Sanyo Electric Co., Ltd. (85.8)
Kaneka Corporation(70.9)
The University of Delaware(55.4)
Fuji Electric Co., Ltd.(46.9)
Sharp Kabushiki Kaisha(48.8)
General Electric Company(45.7)
Semiconductor Energy Laboratory Co., Ltd. (37.2)
Siemens Solar Industries, L.P.(36.8)
Exxon Research and Engineering Company(32.2)
Agency of Industrial Science & Technology(23.0)
Mitsubishi Company(17.8)
TDK Corporation(16.9)
Evergreen Solar, Inc. (14.2)
Teijin Limited(13.6)
International Business Machines Corporation(13.4)
Midwest Research Institute(8.1)
Pacific Solar Pty. Limited(7.1)

3071

Self citation
457
26
32
10
5
0
5
4
3
4
0
2
0
0
0
6
0
0
1
1

Other
citation
80
123
37
24
20
31
14
16
23
12
16
14
13
5
8
2
8
7
0
1

Cluster
A
C
C
D
D
C
D
D
C
D
D
D
A
D
D
B
C
C
D
D

PICMET 2009 Proceedings, August 2-6, Portland, Oregon USA 2009 PICMET
APPENDIX B: THE PERFORMANCE INTEGRATION OF THE PATENT PORTFOLIO
Cluster
Leading
Technology
Cluster
Emerging
Potential
Cluster
Technology
Quality
Orientation
Cluster

Company
Canon

Patent Position
TF1
TF2
H
H

TF3
H

Importance
TF1
TF2
H
H

Group

Generation

TF3
H

1,2

H
H
H
L
L

H
H
H
L
L

O
O
O
-

2
1,2
1,2
1,2
1,2

L
H
H
H
L
L
H

L
L
H
H
L
L
H

O
O
O
O
O
O
X

1,2
2
1,2
2
2
1,2
1,2

L
L

L
L

O
O

2
2

L
L
H
H

L
L
H
H

O
O
X
-

1,2
2
1,2
1,2,

Evergreen Solar

Energy Conversion Devices

H
H
H
H
BP Solar
H
H
H
H
The University of Delaware
M
M
H
H
General Electric
M
L
L
L
Agency of Industrial Science &
L
L
L
L
Technology
International Business Machines
L
L
L
L
Teijin Limited
L
L
L
L
Technology
Sanyo Electric
H
H
H
H
Lag Cluster
Kaneka
H
H
H
L
Fuji Electric
H
H
H
L
Sharp
H
H
H
L
Semiconductor Energy
H
H
H
H
Laboratory
Siemens Solar
H
M
L
L
Exxon Research and
M
L
L
L
Engineering Company
Mitsubishi
M
M
H
L
TDK
L
L
L
L
Pacific Solar
L
L
L
H
Midwest Research Institute
L
L
L
H
(Note: H: higher than the median of the item, M: equal to the median, L: lower than the median)

3072

ANALYSIS OF DEFECTS AND IMPURITIES IN NEW (In)GaAsN MATERIALS

FOR CONCENTRATOR MULTI-JUNCTION SOLAR CELLS
Masafumi Yamaguchi, Yoshio Ohshita, Nobuaki Kojima1, Hidetoshi Suzuki and Boussairi Bouzazi
Toyota Technological Institute, 2-12-1 Hisakata, Tempaku, Nagoya 468-8511, Japan
ABSTRACT
Target of the Japanese Innovative Photovoltaics R&D Program started from fiscal year 2008 is to develop
high efficiency solar cells with conversion efficiency of more than 40% and low electricity cost of less than 7
JPY/kWh until 2050. In this program, concentrator III-V compound multi-junction (MJ) tandem solar cells
have great potential for high efficiencies of over 50% and low cost of less than 30JPY/W. InGaAsN material is
good candidate for 4-6 junction solar cells. In our preliminary stage of our project, 2% and 11.3% efficiencies
have been obtained with GaAsN p-n and InGaAsN p-i-n single-junction solar cells. In order to increase
efficiency of single and 4-junction solar cells, it is very important to understand and reduce defects and
impurities in the grown films. In this paper, characterization of the CBE (Chemical Beam Epitaxy)-grown
GaAsN films by Hall effect and DLTS (Deep Level Transient Spectroscopy) measurements is presented.
1. Introduction
In previous Japanese Photovoltaics R&D Program,
Takamoto et al1) has reported
In0.48Ga0.52P/In0.01Ga0.99As/Ge triple junction solar
cells with 31.7% under 1-sun AM1.5G and 39.2%
under 200-suns. New Japanese Innovative
Photovoltaics R&D Program has been started from
fiscal year 2008 and the target in this program is to

develop high efficiency solar cells with conversion

efficiency of more than 40% and low electricity cost
of less than 7 JPY/kWh until 2050. In order to
realize next generation MJ solar cells with super
high efficiencies of over 45%, the 4-junction solar
cell using an InGaAsN junction has been proposed
to satisfy such a demand2) .

Fig. 1. Our R&D scenario for realizing super high- efficiency of 45% and low cost of less than 7 Japanese
Yen/kWh with concentrator MJ solar cells.

978-1-4244-2950-9/09/$25.00 2009 IEEE

002332

Figure 1 shows our R&D scenario for realizing

super high- efficiency of 45% and low cost of less
than 7 Japanese Yen/kWh with concentrator MJ
solar cells. We are studying new materials such as
InGaAsN and InGaN. In this paper, characterization
of the CBE-grown CBE (Chemical Beam Epitaxy)grown films by Hall effect and DLTS (Deep Level
Transient Spectroscopy) measurements is presented.

2. Chemical beam epitaxy (CBE) growth of

(In)GaAsN
The quaternary InxGa1-xAs1-yNy has been studied
for 1.3-1.55m laser diode applications, and has
been found3) to have a 1.0eV band gap suitable for
high-efficient MJ solar cells2). Several techniques
such as metal organic vapor phase epitaxy
(MOVPE), molecular beam epitaxy (MBE) have
been widely used to grow InGaAsN thin films.
However, the optical and electrical properties of
them are drastically deteriorated with the
incorporation of nitrogen (N). The anticipated
efficiencies for the application of MJ solar cells have
not been achieved due to the short minority-carrier
diffusion length caused by the generation of Nrelated recombination centers. Another issue is the
carbon incorporation and N compositional
fluctuation in InGaAsN films grown by MOVPE at
high temperature. Hence, the lower growth
temperature is required. With MBE technique, no
carbon contamination is observed, but the crystal
quality is deteriorated by the damage of N2 plasma
as a nitrogen source. To resolve such problems, a
new technique substituted for a conventional method
in growth of InGaAsN is required, so we have
proposed a chemical beam epitaxy (CBE)
technique4), in which thin films are grown in low
pressure and temperature, and the N compound
sources are used. CBE is the only technique that
incorporates both the beam nature of MBE yet uses
all vapor sources. All sources are supplied to the
surface of substrate using metal organic gases and
decompose on the growing surface under low
pressure (~10-2Pa). In this study, we have focused
especially on the N incorporation in GaAsN thin
films grown by CBE as a function of growth
temperature and substrate orientation. We have also
proposed a new flow-rate modulation CBE (FMCBE) method in order to increase N incorporation
and to reduce C and H incorporation in films.
3. Crystalline and electrical properties of GaAsN
films
Figure 2 shows XRD-FWHM (full width at half
maximum) of GaAsN (004) rocking curves as a
function of N composition. XRD-FWHM values
decrease with decreasing temperature at range of

978-1-4244-2950-9/09/$25.00 2009 IEEE

400480oC, while the N composition increases. At

lower temperature region (<360 oC), XRD-FWHM
values increase rapidly due to the increasing of
lattice-mismatch with increasing N incorporation.
The minimum XRD-FWHM of 35.4 arcsec with N
composition 0.9% was obtained at 380 oC
temperature. This value is close with that of GaAs
substrate. All XRD- FWHM values obtained in this
study were lower than those obtained in GaAsN
films grown by other MOCVD (DMHy), MBE (N2),
and CBE (N2) techniques. That means that good
quality GaAsN fillms have been obtained by CBE
compared to the other growth methods.

Fig. 2. XRD-FWHM of GaAsN (004) rocking

curves as a function of N composition in comparison
with those of films grown by MOCVD and MBE.

Fig. 3. Electron mobility of GaAsN films obtained

in this study at various N concentrations (%) in
comparison with those of films grown by MOCVD
and MBE.
Figure 3 shows the electron mobility of GaAsN
films grown by CBE at various N concentrations

002333

(%). The solid line was calculated theoretically by

assuming the alloy scattering of N in GaAsN. We
have successfully obtained the higher electron
mobility (2000cm2/Vs) in GaAsN films, even
though the N composition is over 1%. For hole
mobility in GaAsN films, we have obtained
200cm2/Vs, which is similar with that of p-GaAsN
films (N 1%) reported until now. High electron
mobility obtained in this study implies that the
compositional fluctuation of N can be controlled by
low temperature CBE growth. However, the
concentrations of impurities (H, C) are still over
1018cm-3 at low growth temperature. Further study
for lowering impurity concentration is needed.

4. GaAsN single-junction solar cells and DLTS

analysis of defects in GaAsN films

have been used to solve problems for low minoritycarrier lifetime of 1-10 ps in the films. Figure 5
shows efficiency of homo junction GaAsN solar
cells as a function of minority carrier lifetime. It is
very important to understand origins of defects in
(In)GaAsN films and to reduce defects density in
order to increase minority-carrier lifetime and to
realize high efficiency (In)GaAsN single-junction
and 4-junction solar cells.
In this study, majority and minority traps in ptype GaAsN films grown by CBE were measured by
the DLTS method and a one electron trap with high
capture cross section was observed by using double
carrier pulse DLTS technique.
Although 6 electrons traps EC1 to EC6 at 0.228,
0.266, 0.292, 0.303, 0.349, and 0.660 eV form the
conduction band minimum Ec, respectively, were
detected, densities of 2 traps EC1 and EC2 are found
to be dependent on N concentration in GaAsN film
as shown in Fig. 6. Because a decrease in EC1 peak
height using a double carrier pulse DLTS technique
was observed with the injection of majority carrier,
the electron trap EC1 is thought to act as a
recombination center.

Fig. 4. I-V curve of a p-GaAs/i-n InGaAsN heterojunction cell with intrinsic layer wideth of 600nm.

Fig. 6. DLTS spectra of GaAsN film grown by CBE

as a function of N concentration.

Furthermore, its origin will be discussed. Figure

7 shows energy levels of possible defects in
GaAsN6). The possible atomic structures of these
defects are thought to be (N-As)As and/or (N-N)As
on single As sites.

Fig. 5. Efficiency of homo junction GaAsN solar

cells as a function of minority carrier lifetime.
Although 11.27% efficiency has been obtained
with an InGaAsN p-i-n single-junction cell5) in
previous study as shown in Fig. 4, insertion of
intrinsic layer and GaAs/InGaAsN hetero structure

978-1-4244-2950-9/09/$25.00 2009 IEEE

002334

3) M. Kondow, T. Kitatani, M. C. Larson, K.

Nakahara, K. Uomi and H. Inoue, J. Crystal
Growth, 188, 255 (1998).
4) M. Yamaguchi, T. Warabisako and H. Sugiwura,
J. Crystal Growth 136 29 (1994).
5) N. Miyashita, Y. Shimizu, N. Kobayashi,
M.Yamaguchi and Y. Okada, Proceedings 4th
World Conference on Photovoltaic Energy
Conversion, (2006) p. 869.
6) S.B. Zhang and S-H. Wei, Phys. Rev. Lett. 86,
1789 (2001).

Fig. 7 Energy levels of possible defects in GaAsN6).

Summary
Outline of the Japanese Innovative Photovoltaics
R&D Program was presented, and epitaxial grown
of GaAsN films by using CBE and characterization
of the films by Hall effect and DLTS measurements
were also reported. We have demonstrated high
quality CBE-grown GaAsN films with the
minimum XRD-FWHM of 35.4 arcsec and the
higher electron mobility (2000cm2/Vs) compared
to those with GaAsN films grown by the other
methods. CBE is thought to offer a new means for
growing InGaAsN films that are necessary for use
as the third junction in super high efficiency 4junction AlInGaP/GaAs/InGaAsN/Ge solar cells.
However, one of major problems in (In)GaAsN is
lower minority carrier lifetime because only 2% and
11.3% efficiencies have been obtained with GaAsN
p-n and InGaAsN p-i-n single-junction cells. We
have characterized defects in CBE-grown GaAsN
by using DLTS measurements. The electron trap
EC1 in CBE-grown p-type GaAsN is found to act as
a recombination center. Possible origin of defect
level EC1 is also discussed.

Acknowledgments
This work is partially supported by the New
Energy and Industrial Technology Development
Organization (NEDO) under the Ministry of
Economy, Trade and Industry (METI).
References
1) T. Takamoto, M. Kaneiwa, M. Imaizumi and M.
Yamaguchi, Progress in Photovoltaics 13, 495
(2005).
2) S. R. Kurtz, D. Myers and J.M. Olson,
Proceedings of the 26th IEEE Photovoltaic
Specialists Conference, (1997) p. 875.

978-1-4244-2950-9/09/$25.00 2009 IEEE

002335

INVESTIGATION OF OPTICAL ABSORPTION IN THIN-FILM SIIGE SOLAR CELLS

Fariborz Jahanshah1, M. Khiza~ , D. Modisette2, Ron Manginell3,
Nowshad Amin1, K. Sopian1, and Saleem H. Zaidi2
1JKEES and SERI, National University of Malaysia, Malaysia
2Gratings, Incorporated, Albuquerque , New Mexico
3Sandia National Laboratories

ABSTRACT

Thin-film, high efficiency crystalline Si solar cells promise

significant cost savings in contrast with conventional
substrate solar cells. These savings are achieved through
reduced Si usage assuming highly (- 20-30%) efficient
thin film Si solar cells can be manufactured. A necessary
condition for high efficiency solar cells in thin-film
configurations is complete optical absorption of sunlight.
However, due to its indirect bandgap, Si has inherently
weak absorption in near IR region so thicker films are
required to accommodate longer path lengths. Even
incorporation of scattering mechanisms (geometrical,
diffractive, and physical) is not enough to achieve
complete absorption in Si thin-films .
We propose thin-film crystalline Si/Ge solar cell
configuration in which the top surface comprises of thin Si
film and the bottom surface of thin Ge film. This
heterojunction solar cell combines the best attributes of
both materials with strong Si absorption in the visible
followed by absorption in Ge through most of the near IR
region. The absorption spectrum ranges from 300 nm to
1600 nm in this solar cell with only two thin film layers.
This approach has not been investigated extensively in
past due to technical challenges associated with growth of
high-quality Ge layers on Si. We have demonstrated
growth of very high quality SixGe1-x and Ge films on
nanostructured Si substrates, thus, identifying a promising
pathway in addressing the fundamental problem of
pseudomorphic growth of Ge on Si, a necessary condition
for high efficiency heterojunction Si/Ge thin-film solar cells.
Preliminary optical measurements on Si/Ge films exhibit
strong absorption dependence on growth quality.
1. INTRODUCTION

The high cost of Si wafers negatively impacts the ongoing

transitioning of electrical generation from carbon-based
fuels to sunlight-based renewable energy resources [1).
Reduced crystalline Si usage in a thin (- 10 IJm) film solar
cell configuration has the potential to significantly reduce
energy conversion costs assuming efficiencies in - 20-30
% range [2). However, silicon due to its indirect band gap
has inherently weak absorption in near IR range [3). This
weak absorption fundamentally limits the efficiency of all
Si solar cells in thin films due to incomplete optical
absorption, and in thick films due to bulk recombination
losses. Surface texturing mechanisms have been

978-1-4244-2950-9/09/$25.00 2009 IEEE

effectively employed to reduce reflection as well as

enhance near IR absorption [4-6); the lack of absorption
near and above the band gap reduces performance.
Although Si absorbs effectively in the solar spectrum (Fig.
1, left), a substantial fraction of the solar spectra is not
utilized (Fig. 1, center); particularly in thin films .
The optical absorption (Fig. 1, center) of other
semiconductor substrates such as Ge and Ino.s3Gao.47As
extends over much of the available solar spectrum, and
may be used to absorb sunlight more effectively. However,
these are expensive and toxic materials with limited
quantities in earth's crust. Here, we propose an alternate
approach by based on Si/Ge thin-film solar cell
configuration (Fig. 1, right) in order to effectively use most
of the available sunlight. In contrast with multi-junction
solar cell technology based on multi-layer growth of
compound semiconductor films on Ge substrates [7], this
approach has not been extensively investigated
presumably due to the following two reasons:
Since there is substantially higher power in the short
wavelength spectra, therefore, it is better to grow multilayer structures that selectively absorb short wavelengths
leaving long wavelength absorption to the Ge substrate,
and
Technical challenges associated with Si/Ge material
system.
However, as evident in Fig. 1 (left), the light intensity
decreases more slowly at longer wavelengths than at
shorter wavelength with peak power at - 0.6 IJm.
Therefore in a thin-film Si-Ge configuration (Fig. 1, right), it
is theoretically possible to absorb all photons in the - 0.40.7-lJm range with Si thickness in less than 10 IJm. The
longer wavelengths from - 0.7-1.6-lJm range are absorbed
.'"",~..........,.....!""",,!,"""'I Top contact
I

Sithinfilm
I

I
I

Ge thm film

Botton contact
Figure 1. One-sun AM 1.5 solar spectra (left), spectral
absorption coefficients for Si, Ge, and Ino.s3Gao.47As
substrates (center), and proposed solar cell configuration
(right).

000608

in Ge films with thickness < 20 11m (Fig. 1, center). It is

therefore possible to substantially enhance solar cell
efficiency in the configuration of Fig. 1 (right). Finally, the
quality of heteroepitaxial growth of Ge on Si has been
significantly improved by using the nanostructures.
2. HETEROEPITAXIAL GROWTH OF SiXGe1.X AND GE
LAYERS ON PATTERNED Si SUBSTRATES

Development of Ino.s3Gao.47As thin-films on Si substrates is

extremely challenging due to large (- 9 %) lattice and
thermal expansion mismatches [8]; it is not addressed
here. The growth of high quality Ge films on Si by
chemical vapor deposition (CVD) and molecular beam
epitaxy (MBE) is less challenging because of smaller (- 4
%) lattice mismatch [9]. However, a 4% lattice mismatch
and nearly 50% thermal expansion mismatch between Ge
and Si creates three major challenges for growing high
quality Ge films on Si substrates: (a) High surface
roughness due to pronounced <110> crosshatch pattern
(Fig. 2 (left) which creates additional difficulties in
lithographic patterning [10], (b) High dislocation densities
(Fig. 2, center) responsible for enhanced recombination
and dark currents, and (c) High density of microcracks
(Fig. 2 right) [11], and wafer bowing [12]. Wafer bowing
and microcracks are particularly undesirable due to wafer
breakage and electrical shorts. A traditional approach of

example of defect free 100 % Ge film of - 10-J.1m

thickness grown by CVD on nanostructured Si substrate.
Figure 4 (center and right) also plots the reciprocal space
maps (RSMs) of x-ray diffraction (e-2e) of Ge film on the
planar and nanostructured structures. The full width at half
maximum (FWHM) of the Ge layer of the nanostructured
structures (Fig. 4 (center was significantly narrow relative
to that of the un-patterned structure (Fig. 4 (right, which
suggest that the quality of the nanostructured Ge was
superior that of the planar Ge. The FWHM of the
reciprocal space maps of Ge epilayers on nanostructured
and planar Si was 93 and 373 arc sec respectively. The
RLP (Reciprocal lattice point) of the substrate is located at
1
k.Ll001]= 0.731 A- along the k.l[001] axis. The RLP of the
1
uniform Ge layer is at k.l[001]= 0.704 A along the k.l[001]
axis. It was observed that the Ge layers were relaxed as
expected. Furthermore, a crystallographic tilt of the Ge
layer with respect to the substrate was observed. This can
be recognized in the (004) diffraction as a

Figure 3. Some of the structures investigated for SiGe and

Ge growth on Si substrates: 20-J.1m square and 10-J.1m
deep mesa structures (left), one-dimensional (1D), 1-J.1m
period, 6-J.1m deep pattern (center), and 1-J.1m period, 4J.1m deep two-dimensional post pattern (right).
Figure 2. Heteroepitaxial Ge and SiGe growth on Si
substrates: cross-hatch pattern due to misfit dislocations
(left), transmission electron microscope (TEM) view of
threading dislocations (center), and top scanning electron
microscope (SEM) view of microcracks in the Ge layer
(right).
reducing lattice and thermal mismatches is to form a
"virtual" substrate by growing a graded composition of the
desired heteroepitaxial film on a defect-free Si substrate
[13]. A constant composition layer of the desired lattice
parameter can then be grown on this buffer layer. Graded
SiGe films have been demonstrated with very low (2x106
ern") defect densities [14]. Recently, single relaxed Ge
deposition on Si obtained superior quality epilayers
compared to graded multi- relaxed SiGe buffer layers [15].
Mathews et at. first proposed that limiting the lateral
dimensions of the sample prior to growth could reduce the
dislocation density [16]. We have extended this approach
to CVD growth of SiGe and Ge epilayers on patterned Si
substrates [17-22], and have recently demonstrated high
quality Ge layers, with dislocation density consistently
lower than 5x10 s cm-2 , when grown on nanostructured Si
substrates [18]. Figure 3 shows cross-sectional SEM
views of three of the structures. Figure 4 (left) shows an

978-1-4244-2950-9/09/$25.00 2009 IEEE

Figure 4. Epitaxially grown 10-J.1m thick Si films on

nanostructured Si substrates: TEM cross-sectional view
(left), RSMs of x-ray diffraction of Ge films on planar
(center), and nanostructured surfaces (right).
deviation of the diffraction maxima from the kll[001]=0 line
(vertical dashed lines in Fig. 4). The full width at half
maximum (FWHM) of the peak determines the equality of
the layer in the bulk. The X-ray RSM shows that the
FWHM is narrowed in the nanostructured surface, which
means that the mosaic tilts or the defect density is
reduced. The full width half maximum peaks of the
reciprocal space maps of Ge epilayers on nanostructured
Si show 93 arc sec; the best reported elsewhere for Ge
films on Si.
This extensive investigation of epitaxial growth as a
function of Si feature dimensions has demonstrated that
defect density is a sensitive function of Si seed layer as

000609

shown in Fig. 5. This work indicates that defect density

reduction less than 105 cm-2 is achievable with Si seed
structure widths less than 0.02 urn. This work has helped
identify the role of Si nanostructures in reducing wafer
bowing and density of microcracks.
l e+ lO , --

- --

---,

'E l e+9
u

le+8

.~ le+7

c: le+6

Experimental
OJrve fitting (our data)

l e+5
Vi le+4

le+3

0.01

0.1

10

100

1000

10000

Patterned Width (J.lITl)

Figure 5. Defect density variation as a function of Si

pattern width, taken from reference.

4. OPTICAL TRANSMISSION MEASUREMENTS

In order to evaluate the thin-film Si/Ge solar cell concept,
we have developed a simple experimental setup aimed at
measurement of optical absorption of thin (20-30 urn)
Si/Ge films. A simplified diagram of the spectral optical
transmission
measurement
system
used
for
measurements described here is described in Fig. 7. A
computer-controlled filter wheel with 12 slots was used for
transmission measurements at selective wavelengths. An,
unbiased InGaAs photodetector detector operating in 900.1600 nm spectral range was used for optical
measurements. The signal from the detector was hooked
up to a lock-in amplifier; an optical chopper was used as
reference. All measurements were normalized to detector
response in air; the light source used was a halogen lamp.
Thin Si/Ge films were attached to IR-transparent dielectric
substrates for optical transmission measurements as a
function of Si thickness of Si; the Ge film thickness was
kept constant at - 10 urn.

--------1

3. Modeling

Photod et ecto r _ _

We have carried out preliminary modeling of thin-film

Si/Ge solar cells with front and back surface contacts.
Figure 6 shows some preliminary results using PCID5 [23]
and SILVACO [24] softwares. Although individual cell
simulations are reasonably accurate, the combined cells
L1V data is not accurate. This is likely due to incorrect
device configuration since these layers must be currentmatched with the requirement of a tunnel junction between
the two similar to that used in III-V multi-junction
compound semiconductor solar cells. We are continuing
this work by adopting an approach proposed by Wada et
aI., Si/Ge solar cells [25].

Foe us in g len

Glas 5 SU b strate

Gefilm

C o ll im at in glen 5

C h op per
C o lo r fi lt er w h e e l

Colle ct in g lens

Xe n on A r c li ght

Figure 7. Experimental setup for optical transmission

measurements of Si-Ge thin film layers.

(a)

(c)

(b)

(d)

Figure 6. Simulations of to-prn thick Si (a) and (b) to-urn

thick Ge solar cells using PCID5, and 40-l..lm thick Si on
to-urn thick Ge (c and d) solar cell using SILVACO.

978-1-4244-2950-9/09/$25.00 2009 IEEE

Typical room temperature spectral response from Ge films

grown on planar and grating surfaces have been plotted in
Fig. 8 in 1050-1650-nm range. The transmission data
exhibits a noticeable increase in the transmittance through
Ge films grown on planar surfaces in contrast with films
grown on a grating surface. The lower transmittance
through gratings side of the sample may be attributed to
enhanced absorption in the higher quality Ge film on thin
grating surface; although the transmission trend for both
samples is similar. The reflectivity of the front surface Ge
layer grown on planar (rough surface) and grating (smooth
surface) has not been accounted for in these
measurements and may be playing a role here. Finally, for
these measurements, some absorption takes place in the
underlying Si film.
Figure 9 shows optical transmission measurements fro Ge
films only. Contrary to spectral response results obtained
from Si/Ge samples with thick Si film, a measurable
increase in the optical transmittance from Ge film grown
on grating surface has been observed. This higher
transmittance appears to be due to higher-quality

000610

6. REFERENCES
____ 1 . 4 . - - - - - - - - - - - - - - - - - - - - ,
;J
--- Reference

< 1 2 -___

Si/Ge without gratings

----.t.---- Si/Ge with gratings

~ 1.0

--

- -0 - -0 - -0 - -0 - -0 - -0

'5 0.8
'"
=
~

l-

Eo-<
"0

0'6 ~

0.4

. V~.==:::-~.~

.t:::l 02

';

5
lo

0-0

0.0
1050

A_

1200

1350

1500

1650

1800

A(nm)
Figure 8. Optical transmission measurements from Si/Ge
films in 1-1.6-llm wavelength range.

____ 1 . 4 , - - - - - - - - - - - - - - - - - - ,
;J
--- Reference

<
__
___ 1.2

Si/Ge without gratings

----.t.---- Si/Ge with grat ings

~ 1.0

.s

'5
'"
g

- -0 - -0 - -0 - - 0 - -0 - - 0

0.8

0.6

i:: =::::.--\ _=
5

l-

O.O-t-~-..-~-r-~--r-~-,--~-i

1050

1200

1350

1500

1650

1800

A(nm)
Figure 9. Optical transmission measurements from Ge
films in t -t .s-um wavelength range.
smoother Ge film that has no defects and no scattering
(Fig. 4, left) in contrast with highly rough (Fig. 2, left) Ge
films that enhance scattering and thus lead to reduced
normal-incident transmission.
5. SUMMARY AND FUTURE WORK

[1] A. A. Munzer, K. T. Holdermann, R. E. Schlosser, and

S. Sterk, IEEE Trans. Elect. Dev. 46,2055 (1999).
[2] C. Hebling, S. W. Glunz, J. 0 Schumacher, J.
Knobloch, 14th EUPVSEC, 1997.
[3] M. A. Green and M. J. Keevers, Prog. In Photovoltaics :
Research and Applications , 3, 189 (1995).
[4] P. Campbell and M. A. Green, J. Appl. Phys. 62, 1
(1987).
[5] W. H. Southwell, Opt. Lett. 8, 584 (1983)
[6] Saleem H. Zaidi, D. S. Ruby, K. Dezetter, and J. M.
Gee, 29th IEEE PVSC, 142 (2002).
[7] US Patent 7309832 - Multi-junction Solar Cell Device
(2007).
[8] M. Yamaguchi, J. Mater. Res. 6, 376 (1991).
[9] M. Yamaguchi, J. Mater. Res. 6, 376 (1991).
[10] M. A. Lutz, R. M. Feenstra, F. K. Legoues. P. M.
Mooney, and J. O. Chu, Appl. Phys. Lett. 66, 724 (1995).
[11] A. Ackaert, L. Buydens, D. Lootens, P. Van Daele,
and P. Demeester, Appl. Phys. Lett. 55, 2187 (1989).
[12] S. Sakai, Appl. Phys. Lett. 51,1069 (1987).
[13] O. Wada and J. Crow, Integrated Optoelectronics ,
Academic Press (1995).
[14] M. T. Currie, et aI., Appl. Phys. Lett. 72,1718 (1998).
[15] O. Nur, et aI., Semicond. Sci. Technol., 15, L25
(2000).
[16] J. W. Mathews, S. Mader, and T. B. Light, J. Appl.
Phys. 41,3800 (1970).
[17] Ganesh Vanamu, A. K. Datye, R. L. Dawson and S.
H. Zaidi, Applied Physics Letters, 2006.
[18] Ganesh Vanamu, A. K. Datye and S. H. Zaidi, Applied
Physics Letters, 80, (2006).
[19] Ganesh Vanamu, A. K. Datye and S. H. Zaidi, J. Vac.
Sci. & Technol. 823, 1622-1629 (2005).
[20] Ganesh Vanamu, A. K. Datye and S. H. Zaidi, Journal
of Crystal Growth 280, 66-74, (2005).
[21] Ganesh Vanamu, J. Robbins, T. A. Khraishi, A. K.
Datye and S. H. Zaidi, J. Electronic Materials 34, 5,
(2005).
[22] Saleem H. Zaidi, US Patent, 6, 835, 246 82, issued
on Dec. 28, 2004.
[23] www.pv .unsw.edu.au.
[24] www.silvaco.com.
[25] Wada et aI., US Patent 6743974 (2004).

7. ACKNOWLEDGEMENTS
Partial funding for this work was provided by WPAFB and
the University Kebangsaan Malaysia.

A preliminary analysis of thin-film Si/Ge solar cell

configuration has been carried out. Solar cells in this
configuration can potentially absorb all incident sunlight
(AM 1.5 G). While significant materials growth issues
exist, we have demonstrated that defect-free Ge films can
be grown on nanostructured Si surfaces. Preliminary
Optical transmission measurements have indicated higher
transmission through smoother films grown on grating
surfaces. A more detailed modeling, structural, and
transmission analysis is in progress.

978-1-4244-2950-9/09/$25.00 2009 IEEE

000611

Calculating the Efficiency of Dye Sensitized

Solar Cells with Industry Based
Silicon Solar Cells
Komal Magsi, Emilia Macdonald, and Advisor Professor Leon Shterengas

Abstract Our senior design project deal with two types of

solar cell technologies: industry standard Si-based and the

emerging one based on novel dye sensitized materials. Dye
sensitized solar cells will be fabricated in a laboratory setting.
FTIR and Raman microscopy will be used to study the surface
make-up of the novel solar cells. A thin film flexible
photovoltaic cell will be characterized. The efficiency of the
dye sensitized solar cells will be compared to that of thin film
Si-based solar cells. We will design the experiments and
perform measurements using LabView controlled data
acquisition. The prospects to improve the efficiency of the
novel solar cell technology will be discussed.
Index Terms - Amorphous semiconductors, Dyes, Lasers,
Photovoltaic cells

I. INTRODUCTION

n our project we will be testing the different efficiencies of

the new technology of dye sensitized solar cells (DSSC) and
the more common amorphous silicon solar cell. A brief
description of the technology of the solar cells we will be
testing and important concepts to the project is included
below.
Si-Based Solar (amorphous silicon):
Amorphous is characterized by an indefinite shape with noncrystal material. Due to irregular atomic arrangements,
reciprocal action between photons and silicon atoms occurs
more frequently in amorphous silicon. Irregular atomic
arrangement allows more light to be absorbed. Ultra thin
amorphous silicon film is used. Unbreakable structure allows
for rough weather durability. Roll to roll manufacturing
processing.
Silicon is a four-fold conditioned atom that is normally
tetrahedrally bonded to our four neighboring silicon atoms. In
crystalline silicon this tetrahedral structure is continued over a
large range, forming a well-formed lattice (crystal).
In amorphous silicon this long range order is not present and
the atoms form a continuous random network. Not all the
atoms within amorphous silicon are four-fold coordinated.

978-1-4244-5550-8/10/$26.00 2010 IEEE

Due to the disordered nature of the material some atoms have

a dangling bond. These dangling bonds are defects in the
continuous random network, which cause anomalous electrical
behavior.
Dye Sensitized Solar Cells:
The inventors of DSSC are Michael Grtzel and Brian
O'Regan at the cole Polytechnique Fdrale de Lausanne in
1991. The basic structure of DSSC is photosensitive layer of
ultrathin, nano-sized semiconductor crystals over a thin layer
of titanium dioxide. Photons hit photosensitive layer and
allow electrons to flow. Electrons form on the Ti02 layer create
a current. Liquid Electrolyte was needed to carry electrons
from one layer to next. Dye is made of amorphous organic
material. Dye is coated onto the Ti02. Dye absorbs light and
attracts excited electrons which generate charge.
Energy conversion efficiency:
This is the ratio between the useful output of an energy
conversion machine and the input, in energy terms. The useful
output may be electric power, mechanical work, or heat.
Energy conversion efficiency is not defined uniquely, but
instead depends on the usefulness of the output.
Power:
The watt (symbol: W) is a derived unit of power in the
International System of Units (SI). It measures rate of energy
conversion. One watt is equivalent to 1 joule (J) of energy per
second.
II. SYSTEM DESIGN
A. Design Constraints
Solar Cell Performance:
A solar cell is a pn junction device with no voltage directly
applied across the junction. The solar cell will convert photon
power into electrical power and deliver this power to a load.
These devices have long been used for the power supply of
satellites and space vehicles, and also as the power supply to
some calculators.

2
The photocurrent IL produces a voltage drop across the
resistive load, which forward biases the pn junction. The
forward-bias voltage produces a forward bias current IF as
indicated in the figure. The net pn junction current, in the
reverse-bias direction is:
I = IL IF = IL IS [exp

-1]

response of both solar cells peaks around the same wave

length of 400 to 500nm.
(a)

(2.0)

As the diode becomes forward biased, the magnitude of the

electric field in the space charge region decreases, but does not
go to zero or change direction. The photocurrent is always in
the reverse-bias direction. There are two limiting cases of
interest. The short-circuit condition occurs when R=0 so that
V=0. The current in this case is referred to as the short-circuit
current or: I = Isc = IL
The second limiting case is the open circuit condition and
occurs when R .

(b)

The net current is zero and the voltage produced is the opencircuit voltage. The photocurrent is just balanced by the
forward-biased junction current so we have:
I = 0 = IL Is [exp

-1], and Voc = Vt ln (1 +

) (2.1)

Conversion Efficiency of Solar Concentration:

The conversion efficiency of a solar cell is defined as the ratio
of output electrical power to incident optical power. For the
maximum power output we write:
=

x 100% =

x 100%

(2.2)

The maximum possible current and the maximum possible

voltage in the solar cell are Isc and Voc respectively. The ratio
ImVm/IscVoc is called the fill factor and is a measure of the
realizable power from a solar cell. Typically the fill factor is .7
and .8.

Figure 2.1: (a) The Spectral response of silicon (b) The

Spectral Response of DSSC
Now it is important to look at the solar spectrum to see if the
peak absorption wavelength of the cells is in the wavelength
of visible light. Figure 2.2 shows the solar spectrum where it
is easy to see that the wavelength of 400nm to 500nm fall
around the peak. The peak of the solar spectrum is where the
most solar energy is produced so it is a perfect wavelength for
the solar cells to operate at.

The conventional pn junction solar cell has a single

semiconductor bandgap energy. When the cell is exposed to
the solar spectrum, a photon with energy less than Eg will have
no effect on the electrical output power of the solar cell. A
photon with energy greater than Eg will contribute to the solar
cell output power, but the fraction of photon energy that is
greater than Eg will eventually only be dissipated as heat.
Technical:
One of the technical constraints of the project is the spectral
response of silicon and DSSCs. When testing for the
efficiency of both solar cells it is important that the
wavelength of the light shined on both cells hit them around
the peak of their absorption. Figure 2.1 shows the spectral
response of both silicon and DSSC. Both the spectral

Figure 2.2: The Solar Spectrum

3
Financial:
There are no financial constraints due to the fact that the
project will be very inexpensive. The only purchase that will
be needed is a laser diode to shine light on the solar cells and
the amorphous silicon cells. Figure 2.3 shows two laser
diodes found on ebay.com, the left ones starting price was
$9.99 and the right ones starting price was $14.99. The
amorphous silicon cells are also readily available and
inexpensive. The rest of the equipment although very
expensive will be provided by Professor Leon Shterengas in
the lab. Proving that cost for this project will be minimal.

(b)

Figure 2.4 (a) Spectrometer (b) Inside of a spectrometer

(a)

Figure 2.3: 405 nm, HD-DVD Blue-Ray, Blue, Laser diodes

B. Design Considered:
For our project, there were many designs available. Based
upon the design constraints, inexpensive equipment, and
environmental feasibility we chose to implement the design
described later in the report. We have included a number of
pictures of our lab equipment below for further understanding.
The spectrometer shown in figure 2.4 will be used to measure
the spectral response of the solar cells

(b)
(a)

Figure 2.5 (a) Laser hat (b) Solar cells in parallel and series

4
The laser hat shown in figure 2.5 (a) will be used to hold the
laser diode as well as control the lasers voltage and
temperature.
Figure 2.5 (b) is an example of solar cells used for testing.
The multimeter shown in figure 2.6 will be used to measure
voltage and current for analysis. The multimeter can be
controlled by the computer using LABview through the
connector shown in figure 2.6 (a).
(a)
Figure 2.7 (a) Connector (b) Laser Diode Controller

(b)

Figure 2.8

Figure 2.6 (a) Multimeter (b) Back of multimeter 488 GPIB

bus

Figure 2.7 (b) is a laser diode controller which we will use to

control the laser diode
(a)

Figure 2.9
Figure 2.8 and 2.9 Examples of LabView which will be used
to control data in our lab to our discretion.
We decided to use blue laser diodes to shine light onto the
solar cells due to the fact that they are reasonably price and
easy to control
C. Final Design

(b)

The final design will consist of 2 apparatuss. The first

apparatus will have a silicon based solar cell under a laser
diode controlled by lab view parameters. The second

5
apparatus will have a dye sensitized solar cell under a laser
diode controlled by lab view as well. Due to the spectral
response of the dye and amorphous silicon, we have chosen a
laser diode that falls within the range of 400-500nm.
We will most likely be using a laser diode with the parameters
of 405nm (Figure 2.3). The amorphous silicon cell is readily
available to us in Professor Leons lab. The DSSC will be
made in a lab that we have access to on campus so this does
not pose any great issues.
Building the dye sensitized solar cell and silicon based solar
cell:
Figure 2.10 Circuit used to measure I-V curve on oscilloscope
The dye sensitized solar cell will be made in a lab. The silicon
based solar cell will be made available to us without any
design from our part.
In order to measure the efficiency of the solar cells we will be
looking at the I-V curves, an example shown in figure 2.13 of
both cells under a blue laser diode. We will measure the IV
curve using two different methods. LAB VIEW will be used
when necessary and applicable.
The first method is by using an oscilloscope to get a solid
curve. First we will build the circuit show in figure 2.10. R1
should be 1K, the frequency of the sin wave should be 1kHz
and the voltage across the secondary winding of the
transformer should be in the range of 5V to 10V. The resistor
R1 not only serves as a means of limiting the current but can
also be used to monitor current because the voltage across the
resistor corresponds to the current across the solar cell. Due
the fact that the solar cell and the resistor R1 are in series the
current across both of them is equal. Equation 2.4 shows how
the voltage across the resistor corresponds to the current
across the solar cell in mA.
(2.4)
A transformer is used to provide DC isolation of the circuit
ground from the signal generator ground. DC isolation is
important because we need to simultaneously display the
voltage of the solar cell and the voltage of R1 on the
oscilloscope. A floating signal source is created by
grounding the AC signal source and the transformer. This
floating signal source allows us to place a ground were the
solar cell the resistor meet. The voltages of the solar cell and
the resistor are then measured from the points shown on figure
2.10 and the ground between them. The top connection in
figure 2.10 is connected to the X-input of the oscilloscope
(obtaining the solar cells voltage signal) and the bottom
connection in figure 2.10 is connected to the Y-input of the
oscilloscope (obtaining the resistors voltage signal). Display
both of the voltages on the oscilloscope reverse the voltage
signal from the resistor (Y) to obtain set the oscilloscope to the
X-Y regime and the I-V characteristic curve should be
displayed for the solar cell.

The second method of obtaining the I-V curve is by

performing point-to-point measurements. First we will build
the circuit show in figure 2.11 Voltage and current
measurements will be performed with the FLUKE 45 DMM.
Next we will set up the DMM to simultaneously perform two
measurements (voltage and current of the diode). For the
DMM to do that it must be connected to the circuit as shown
in the figure 2.12 and select the voltage to be displayed in the
main display and the current in the second display if the
DMM. Then measure the current at different voltages of the
solar cell.

Figure 2.11 Circuit used to measure I-V curve using point-topoint measurements

Figure 2.12 How to connect the above circuit to a DMM

Both of the methods for measuring the I-V curve will be used
for the silicon solar cells and the DDSC as well as different
amounts of power provided by the laser diodes. For each cell
a graph can be made containing both the curve from the
oscilloscope and in the point-to-point measurements. Once we
have completed these graphs we will use them compare the
efficiency of silicon solar cells and the DSSC.
Characterization:

6
III. SYSTEM IMPLEMENTATION AND TESTING
IV Measurements:
There are a few ways the IV curve will be measured for this
experiment.
How will measurements be made?
Primarily, plot the simulated diode I-V characteristics in linear
scale, i.e. I versus V. Identify the voltage range in which this
dependence is linear. From the slope of the linear dependence
of I on V in this voltage range find the diode series resistance.
Secondly, plot the same simulated diode I-V it in semi-log
scale, i.e. ln(I) versus V. Identify the voltage range in which
the dependence of ln(I) on V is linear. Determine the diode
non-ideality factor from the slope of the I-V characteristic in
semi-log scale.

We are currently in the beginning stages of our testing. We

have begun by shining light onto amorphous silicon solar cells
and testing the output power of the solar cells. We have found
that blue lasers maybe not be the right light to shine onto our
solar cells and are currently testing other sources of light.
IV. DISCUSSION
A. MULTIDISCIPLINARY ISSUES
Solar cells can easily be applied to a non-EE area. This
technology is applicable in materials science, nanotechnology,
vehicles, home improvement, commercialization, marketing,
green energy, and much more. Though our project focuses
mainly on the technical aspect of these cells, there is much
more that goes into them. The actual material of these cells is
an extremely important consideration. Material science in
many universities nation-wide and abroad are opening up
special programs to further study organic materials and dyes
that may increase the efficiency of these cells.
Nanotechnology is also another applicable field for solar cell
technology. The study of minute particles for the transfer
electrons is exactly what TiO2 is in the dye sensitized solar
cells.
Some vehicles use solar panels for auxiliary power, such as for
air conditioning, to keep the interior cool, thus reducing fuel
consumption. In 1975, the first practical solar boat was
constructed in England. By 1995, passenger boats
incorporating PV panels began appearing and are now used
extensively. In 1996, Kenichi Horie made the first solar
powered crossing of the Pacific Ocean, and
the sun21 catamaran made the first solar powered crossing of
the Atlantic Ocean in the winter of 20062007. There are
plans to circumnavigate the globe in 2010.

Figure 2.13 An expected IV curve

Figure 2.14
The final design was chosen based on the constraints of the
silicon and dye sensitized solar cells constraints. As explain
earlier in the paper the laser diode used to shine light on the
solar cells must be in the wavelength of the peak of the solar
spectrum of both solar cells. The peak for both solar cells is
around 400nm -500nm. The laser diodes shown in figure 2.3
are 405nm which fits into that constraint.

In 1974, the unmanned AstroFlight Sunrise plane made the

first solar flight. On 29 April 1979, the Solar Riser made the
first flight in a solar powered, fully controlled, man carrying
flying machine, reaching an altitude of 40 feet (12 m). In
1980, the Gossamer Penguin made the first piloted flights
powered solely by photovoltaics. This was quickly followed
by the Solar Challenger which crossed the English Channel in
July 1981. In 1990 Eric Raymond in 21 hops flew from
California to North Carolina using solar power. Developments
then turned back to unmanned aerial vehicles (UAV) with
the Pathfinder (1997) and subsequent designs, culminating in
the Helios which set the altitude record for a non-rocketpropelled aircraft at 29,524 meters (96,860 ft) in 2001.
The Zephyr, developed by BAE Systems, is the latest in a line
of record-breaking solar aircraft, making a 54-hour flight in
2007, and month-long flights are envisioned by 2010.
B. PROFESSIONAL AND ETHICAL ISSUES
Nuclear and coal power plants used today have an associated
health risk for those who live near them. These methods of
producing energy have several negative attributes worth

7
noting. Specifically, mining and nuclear waste are both
hazardous to people in terms of direct physical harm as well as
the direct physical harm of the environment in which these
facilities exist. Not only are the people who are near the
plants negatively affected, but the long-term effects of acid
rain and global climate change hurt ecosystems that we
depend on for our survival.
Our current energy use is unethical because it harms life
around the globe. Outputs from coal plants cause toxic
chemicals to accumulate in the bodies of plants and animals.
Habitat destruction causes extinction and displacement of
species. For example, hydropower has severely threatened the
lives of salmon in the Pacific Northwest--species have become
regionally extinct in many parts. Solar power provides a
natural way to sustain life in all parts of the world without
causing harm to any species.
C. IMPACT ON SOCIETY
Given the current energy crisis our study of DSSC will only
impact society and environment in a positive way. Providing
a less expensive method of collecting solar energy will
positively impact society and the environment because it will
aid to decrease its dependence on fossil fuels. In parts of the
world that are deprived of any means to electricity, solar cells
can perhaps help if a way is found to make them less
expensive. Refrigeration and cooling, two necessary
ingredients for a comfortable life are being denied to humans
in hundreds of cities abroad.
Solar energy is better to our environment than any known
traditional forms of energy especially against fossil fuel and
coal. It has many uses such as for electricity production or for
heating of water through using photovoltaic. Solar energy can
be used to power heated swimming pools; it can be used to
power cars or attic fans, calculators and other appliances. Due
to high prices of the electric bill using conventional sources,
the demand of solar energy is increasing. The worldwide
demand for solar energy is greater than its supply.
V. SUMMARY AND CONCLUSIONS
Our senior design project deal with two types of solar cell
technologies: industry standard Si-based and the emerging one
based on novel dye sensitized materials. Dye sensitized solar
cells will be fabricated in a laboratory setting. FTIR and
Raman microscopy will be used to study the surface make-up
of the novel solar cells. A thin film flexible photovoltaic cell
will be characterized.
The efficiency of the dye sensitized solar cells will be
compared to that of thin film Si-based solar cells. We will
design the experiments and perform measurements using
LabView controlled data acquisition. The prospects to
improve the efficiency of the novel solar cell technology will
be discussed.
ACKNOWLEDGMENT
The authors would like to sincerely thank their advisor,

Leon Shterengas, for his continuous help, support and patience

through this endeavor. Additional thanks to the advice
received by a panel of mentors provided by Professor Wendy
Tang.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]

Basics of Laser Diodes, Tektronix, Inc., 2002.

D. Russell D. (2004, April). The Diode Laser. Optics & Photonics News
[Online]. Available: http://www.opticsinfobase.org/
S. Leon. Measurement of DC Voltages and Currents. ESE 211
Electronics Lab A, Lab 2. Department of Electrical Engineering. State
University of New York at Stony Brook, NY, 2009.
S. Leon. Semiconductor Diodes. ESE 211 Electronics Lab A, Lab 8.
Department of Electrical Engineering. State University of New York at
Stony Brook, NY, 2009.
D. Carsten. Notes on Disordered Matter. Published 2008
Available: http// blog.disorderedmatter.eu. March 5th, 2008.
N.M. Thantsha. Opto-electronic analysis of silicon solar cells by LBIC
investigations and currentvoltage characterization. December 2009,
Pages 4445-4448. Physics B: Condensed Matter.
Available: http://wintowtinting.com.au/ Sunscreen Window Tinting &
Graphics
M. Alberto. Photoregulation of Anthocyanin Synthesis. Plant Physiol.
1984 June; 75(2): 447453.
N. Donald A. An Introduction To Semiconductor Devices. McGrawHill 2006.

K. Magsi is currently studying Electrical Engineering

with a minor in Materials Science at Stony Brook University. Her
current research includes the comparison of Ph and Ramaan based
spectral management for Si cell application. She is being advised under
the honorable Professor Charles Fortmann who received his PhD in solar
technology from Stanford University. Previous research she has done
includes the study of dye sensitized solar cells and the comparison of
different dyes such as chlorophyll and anthocyanine. She has worked
under the Dean of Materials Science, Professor Gary Halada during the
summer of 2009. She is also being trained to study X-Ray topography
with regards to stress strain defects with Professor Dudleys group.
Komal received first place at the Advanced Energy Conference in
November of 2009 where she presented research she had done on
Efficiency Increase for Dye Sensitized Solar Cells. In her free time she
serves as editor-in-chief of the Stony Brook MSA Minaret for which she
has received a Student Media Services Award in 2009. She hopes to
further her studies in the engineering and materials science field with a
specialization in sustainability and renewable energy.

E. Macdonald was born in Toronto, Ontario, Canada.

She is currently studying for her bachelors degree in
electrical engineering major at SUNY Stonybrook,
New York, America, planning to graduate May 2010.

2010 International Conference on Design & Technology of Integrated Systems in Nanoscale Era

Optical properties of BxInyGa1-x-yAs/GaAs

grown by Metal Organic Chemical Vapor
deposition for Solar cell
R. Hamila, F. Saidi, H. Maaref
Laboratoire de micro-optolectroniques et nanostructures, Facult des Sciences de Monastir, Avenue de
lEnvironnement 5000 Monastir, Tunisia

Ph. Rodriguez, L. Auvray, and Y. Monteil

Laboratoire Multimatriaux et Interfaces. Universit Claude Bernard Lyon 143, Boulevard du 11 Novembre
1918, France
E-mail address: radhiahamila@yahoo.fr (R.Hamila)
Abstract:

semiconductor alloy that is substantially latticematched to the silicon substrate. The polarity of the
active p-n junction cell is either p-on-n or n-on-p.

The present invention further includes a method for

substantially lattice matching single-crystal III-V
semiconductor layers by including boron in the
chemical structure of active cells layers in multijunction solar. Solar photovoltaic devices, i.e., solar
cells, are devices capable of converting solar radiation
into usable electrical energy. The energy conversion
occurs as the photovoltaic-effect which occurs in a cell
composed of a p-type semiconductor layer adjacent to
an n-type semiconductor layer, here after referred to
as p-n junction cell. Solar radiation impinging on a
solar cell and absorbed by active region of
semiconductor
material
generates
electricity.
Therefore, a quaternary material III-V semiconductor
BInGaAs has been tested for the application in solar
cells [1]. Single layer has been grown lattice matched
on GaAs using Metal Organic Chemical Vapor
deposition (MOCVD). Optical study has been
achieved of B0.0108In0.36Ga0.062As quantum well. At
room temperature (300 K) PL study has shown an
asymmetric PL band is around 1.19 eV of the emission
energies.
Based in these experimental results, we
have suggested that the band gap energies of BInGaAs
QW could be adequate for active cells layers in multijunction solar.

Emitter layer
Base layer

Tunnel junction layer

Emitter layer

Bottom active
substrate

Figure.1. Design of double-junction solar cell based on BInGaAs

as absorber material [5]

The present invention further includes a method for

substantially lattice matching single-crystal III-V
semiconductor layers with silicon substrate by
including boron and/or nitrogen in the chemical
structure of these layers. BGaAs and BInGaAs
alloys [2,3] are candidates for applications in multijunction solar cells and detectors. BxGa1-xAs with
boron concentrations up to 4% is obtained by
metalorganic
chemical
vapour
deposition
(MOCVD) [2] or molecular-beam epitaxy (MBE).
Different solar cell structures have been grown

Keywords: BInGaAs, DDX, MOCVD, PL, Solar cell

I.

INTRODUCTION:

A multi-junction device further comprises one or

two subsequent active cells each having a base layer
and an emitter layer Fig1. with inter connection
tunnel junctions between, each active cell. At least
one layer that forms each of the top and middle
active cells is composed of single-crystal III-V

978-1-4244-6340-4/10/$26.00 2010 IEEE

Top active

-1-

2010 International Conference on Design & Technology of Integrated Systems in Nanoscale Era

under varying growth conditions on differently

oriented GaAs substrates ((001), 6 off (001)
towards (111)A, and 2 off (001) towards (
[4]. BxInyGa1-x-yAs structures can grow lattice
matched to GaAs. In this paper, we have
investigated HRXRD and Photoluminescence (PL)
of BxGa1-x-yInyAs QW structure grown on GaAs by
Metal Organic Chemical Vapor Deposition
(MOCVD), with high indium composition up to yIn=
36%. We have focused our study on the influence of
boron incorporation into InGaAs, its effects on the
PL emission. At room temperature (300 K) PL band
is around 1.19 eV. Therefore, BInGaAs single
crystal semiconducteur can absorb a different part of
the wide energy distribution of photon in sunlight.
BInGaAs may be the active cells in the multijunction solar cell device.
II.

RESULTS AND DISCUSSIONS

III.

In order to estimate indium solid composition

of (B)InGaAs, samples were investigated by
HRXRD. Fig 2.a. shows symmetric diffraction /2
scans using (004) plane for the BInGaAs epilayer.
We have assumption that the boron composition
incorporation is the same into BGaAs one of about
1.08%. Using the asymmetric (115) plan diffraction
reciprocal space, we can determine the indium
composition. From the angular difference
between the BInGaAs and GaAs peaks we calculate
lattice mismatch witch is related to y as follow:


EXPERIMENTS DETAILS

 

(1)



(2)

Where:

The growth of BInGaAs quantum well has been

performed by atmospheric-pressure MOVPE in a Tshape horizontal reactor ant growth temperature is
580C. The QW was grown on (001) GaAs
substrates misoriented 1 towards [110] direction.
Diborane
(B2H6),
triethylgallium
(TEG),
trimethylindium (TMI) and arsine (AsH3) were used
as boron, gallium, indium and arsenic sources,
respectively. Hydrogen was used as carrier gas.
AsH3 flow rate was kept constant, leading, for
BInGaAs layers, to V/III ratios ranging between 120
and 460. The boron composition was varied by
varying V/III flux ratio [6]. HRXRD measurements
for (004) plane reflection was performed
systematically with the copper target (CuK1=
1.54056 ) radiation from a Discover D8 (40 Kv 55
mA) high power X ray generator. The lattice

(3)

(4)

And theory of elasticity: P 0.5-1.0 this is related to

the Poison ratio and the relaxation degree, B is the
GaAs Bragg angle for (004) reflection and x is the
boron composition.

10

GaAs

BInGaAs epilayer

10

Intensity (a u)

10

mismatch of the epilayer was evaluated and

the indium composition was determined using

Vegards law from separation angle between the
epilayer and substrate (400) diffraction peaks. The
Poisson
ratio
was
estimated
to
0.31.
Photoluminescence measurements were carried out
between 10 and 300K while keeping the samples in
a closed-cycle helium circulation cryostat. The
excitation wavelength used is the 514,5 nm line of
the cw Ar+ laser. The emission was dispersed by a
high-resolution spectrometer and detected by a
thermoelectrically cooled InGaAs photodetector
with a built-in amplifier.

BInGaAs

10

10

10

10

10

10

31,5

32,0

32,5

33,0

/2

33,5

34,0

34,5

Figure.2.a. High resolution X-ray diffraction /2 spectra of

BxInyGa1-x-yAs/GaAs epilayer.

For BInGaAs QW grown with same condition to

BInGaAs epilayer Fig 2.b shows BxInyGa1-x-yAs
structure grow lattice matched to GaAs, that we
observe an asymmetric peak at the left. We have
attributed to the combination of GaAs and

-2-

35,0

2010 International Conference on Design & Technology of Integrated Systems in Nanoscale Era

compostion respectively. Quantum well emission is

observed at 1.19 eV (1.04 m). A performance of
solar cell based silicon increase when absorbed
photons with Si increase. A contribution of our
active layer based BInGaAs is to absorb part of the
range of visible to emit in the vicinity of 1.19 eV.
These photons are subsequently absorbed by Si to
create enough current. We noticed that boron
incorporation strongly decreases PL intensity, as
already observed with nitrogen incorporation [7].
The full- width at half-maximum (FWHM) of the
PL band is 35 meV. This suggests the performance
of optical quality of structure.

BInGaAs. It should be noted that all layers are

almost completely relaxed. Thus the crystal lattice
parameter and crystal symmetry match that of the
underlying substrate. This lattice matching condition
results in epitaxial layers with minimal strain, few
defects and thus superior electrical properties.

10

BInGaAs QW

10

intensity (a u)

10

10

Irrespective of the application, and as with any

energy generation system, efforts have been ongoing
to increase the output and efficiency of multijunction solar cells. In terms of output multiple cells
or layers having different energy band gaps have
been stacked so that each cell or layer can absorb a
different part of the wide energy distribution of
photon in sunlight. The stacked arrangement has
been provided in monolithic structure on a single
substrate or on multiple substrates. Thus the
BInGaAs QW offer the new band gap energy. In a
multiple cell device, semiconductor materials are
typically lattice matched to form multiple p-n or (np) junctions. The p-n or (n-p) junctions can be of the
homo-junction or heterojunction type. When solar
energy is received at a junction, minority carriers
(i.e. electrons and holes) are generated in the
conduction and valence bands of the semiconductor
materials adjacent the junction. A voltage is thereby
created across the junction and a current can be
utilized from. As the solar cell energy passes to the
next junction, which can be optimized to a low
energy range, converted into a useful current.

10

10

10

-1

10

30

31

32

33

34

35

36

/2
Figure.2.b. High resolution X-ray diffraction /2 spectra of
BxInyGa1-x-yAs/GaAs quantum well.

At room-temperature (300 K) PL studies

evidenced QW emission wavelength ranging
between 0.94 and 1.04 m, depending on the well
composition and thickness.

BInGaAs QW
T=300 K
2
Pexe=80W/cm

PL Intensity (a u)

Epl = 1,19 eV

IV.

1.1

1.2

1.3

CONCLUSION:

In conclusion we have studied optical properties

of B0.0108In0.36Ga0.062As/GaAs quantum well
structure grown by MOVPE at 580C in lattice
matched to GaAs. At room temperature (300 K) PL
band is around 1.19 eV of the emission energies.
Solar radiation impinging on a solar cell and
absorbed by active region of BInGaAs
semiconductor material generates electricity. In
terms of output multiple cells or layers having
different energy band gaps have been stacked so that
each cell or layer can absorb a different part of the
wide energy distribution of photon in sunlight.

1.4

Energy (eV)

Figure.3. room temperature photoluminescence (300K) spectra of

B0.0108In0.36Ga0.062As /GaAs quantum well

Fig 3. shows the PL spectrum for the sample

with x=1.08% and y= 36.32% of boron and indium

-3-

2010 International Conference on Design & Technology of Integrated Systems in Nanoscale Era

REFERENCES
[1] G. Leibiger, C. Krahmer, J. Bauer, H. Hernberger, V.
Gottschalch, J. Crystal Growth 272, 732-738 (2004)
[2] V. Gottschalch, G. Leibiger, G. Bendorf, J. Crystal Growth
248, 468 (2003)
[3] R. Asomoza, V. A. Elyukhin, and R. Pena-Sierra Appl. Phys.
Lett 78, 2494 (2001)
[4] J .F. Geisz, D.J.Friedman, J.M.Olson, S.R.Reedy,
A.B.Swartzlander, B.M. Keyes, A.G. Norman, Appl. Phys. Lett
76, 1443 (2000)
[5] Tom MarkvartLuis and Castafier , Solar Cells:
Materials,Manufacture and Operation, Elsevier Ltd, ISBN13:978-1-85617-457-1 (2006)
[6] H. Dumont, D. Rutzinger, C. Vincent, J. Dazord, Y. Monteil,
F. Alexander, and J.L.Gontner, Appl. Phys. Lett 82, 1830 (2003)
[7] L. Auvray, H. Dumont, J. Dazord, Y. Monteil, J,Bouix, C.
Bru Chevallier, J. Crystal Growth 221, 475 (2007)

-4-

EFFICIENT BULK HETEROJUNCTION SOLAR CELLS INCORPORATING CARBON

NANOTUBES AND WITH ELECTRON SELECTIVE INTERLAYERS
Golap Kalita, Koichi Wakita and Masayoshi Umeno
Chubu University, Department of Electronics and Information Engineering
1200 Matsumoto cho, Kasugai-shi, Aichi 487-8501, Japan
E-mail: golapkalita@yahoo.co.in
ABSTRACT
Here, we demonstrate incorporation of carbon
nanotubes (CNTs) in the active layer of donor-acceptor
based solar cells and thereby significantly enhanced
device performance. In this work, bulk heterojunction solar
cells were fabricated with P3HT and PCBM incorporating
chemically modified CNTs. Enhanced device performance
was achieved for P3HT:PCBM solar cells incorporating a
certain amount of modified CNTs in the active layer.
P3HT:PCBM bulk heterojunction solar cells were
fabricated with a thin interlayer of TiOx between the active
layer and top electrode as a electron selective layer. The
TiOx layer also acts as a barrier for shorting and shunting
in the device, caused by the presence of metallic CNTs. In
the fabricated device, CNTs provide efficient charge
transportation path and the TiOx layer acts as an electron
selective layer. The enhanced device performance with
introduction of CNTs is attributed to better charge
transportation. Work function of MWNTs is in the range of
4.5~5.1 eV, which is close to valance band of polymer; it
signifies possible hole transportation through MWNTs in
the active layer. As well as, we have studied affect of
MoO3 and C60 as a hole and electron selective interlayer
respectively for P3HT:PCBM bulk heterojunction solar
cells. Fabricated device with MoO3 layer compare to
PEDOT:PSS device shows better fill factor and small
improve in open circuit voltage thought the short circuit
current density was less. The initial degradation of
P3HT:PCBM solar cells with MoO3 and C60 interlayer at
the respective electrode is reduced that of the device with
PEDOT:PSS interlayer.
INTRODUCTION
The discovery of ultra fast photo induced charge
transfer at donor-acceptor interface has brought lot of
interest in organic photovoltaic devices based on
conjugated polymer and fullerene (C60) derivatives. [1]
Organic solar cells fabricated with either spin-coating or
printing offer an alternative to the conventional inorganic
solar cells with such potential advantages as low-cost,
ease of manufacturing, non-toxicity and lightness. In a
bulk heterojunction solar cell, photogenerated exciton
dissociation occurs at polymer (donor) C60 (acceptor)
interface and free carriers (electron and hole) are
transported through the percolation path to the respective
electrodes. [2,3] Conjugated polymer poly (3hexylthiophene) (P3HT) has been most widely used as
electron donor along with [6,6] phenyl C61 butyric acid

978-1-4244-5892-9/10/$26.00 2010 IEEE

methyl ester (PCBM), as electron acceptor for fabrication

of bulk heterojunction solar cells. In the last few years,
significant enhancement in device performance of
P3HT:PCBM bulk heterojunction solar cells has been
achieved and demonstrated. [4-6] Power conversion
efficiency of such type of bulk heterojunction solar cells
need further enhancement for commercial application.
One of the main factors limiting the conversion efficiency
of organic solar cells is due to poor carrier mobility in the
active layer and charge collection in the electrodes. [7] We
have previously demonstrated that with incorporation of
CNTs in organic solar cells, there can be improved charge
transportation and thereby improved device performance
[8,9]. Incorporation of CNTs in P3HT:PCBM bulk
heterojunction solar cells were reported and enhanced
device performance were observed. On the other hand, it
has been reported that electron and hole selective
interlayer in bulk heterojunction solar cells can enhance
device performance. The electronic properties of TiOx
layer suggest that it can be suitable electron selective
layer in between the active layer and the cathode.
Similarly, thermally evaporated molybdenum trioxide
(MoO3) were studied as alternative of PEDOT:PSS as hole
selective interlayer. It has been observed that ITO coated
with acidic PEDOT:PSS layer can degrade the
P3HT:PCBM active layer and thereby deterioration in solar
cell performance. [10-12]
In this paper, we report fabrication of P3HT:PCBM bulk
heterojunction solar cells with incorporation of modified
MWNTs. The solar cells were fabricated with a thin
interlayer of TiOx between the active layer and cathode as
electron selective and hole blocking layer. Similarly device
were fabricated with a MoO3 as hole collecting layer
instead of PEDOT:PSS. Initial stability test of the
fabricated solar cell with MoO3 and C60 interlayer in
comparison to PEDOT:PPS device are presented.
EXPERIMENTAL
For fabrication of bulk heterojunction solar cells, Indium
tin oxide (ITO) coated glass substrates with a resistivity of
10 cm were used. ITO substrates were spin-coated at
2000 rpm with a thin layer of poly (3,4
ethylenedioxythiophene)
poly
(styrenesulfonate)
(PEDOT:PSS). Thickness of the PEDOT:PSS films were
around 10 nm and kept constant for all devices. Device
fabricated with MoO3 interlayer was thermally evaporated
and film thickness was kept constant around 10 nm.
Photoactive organic absorber P3HT:PCBM solution in
dicholorobenzene was spin-coated at 1500 rpm on top of

000090
1

Surface of pristine MWNTs were chemically modified

through oxygen plasma treatment for incorporation in
conjugated polymer. With oxygen plasma treatment,
hydroxyl and carboxyl functional groups were grafted to
the surface of MWNTs. Chemically modified MWNTs with
oxygen plasma treatment were used in these studies.
(a)

15
2

RESULTS AND DISCUSSION

functionalized with oxygen plasma treatment shows

homogeneous dispersion in a conjugated polymer with
sonication for several hours. MWNTs (0.1 wt %) were
mixed with P3HT solution in dicholorobenzene and finally
PCBM was added to the solution such that P3HT:PCBM
concentration is 1:1. The thickness of the P3HT:PCBM
(1:1) film is around 90 nm as measured from the cross
section. In figure 1(a) TEM image of modified MWNTs is
presented. From TEM image it was observed that the
multilayer structure of CNTs remains intact, without any
structural destruction with oxygen plasma treatment.
Induced defect appear in the side wall of MWNTs with
grafting of oxygen species with helps in interecting with
organic solvents or polymers. Figure 1(b) presents TEM
image of a P3HT:PCBM composite film incorporating
modified MWNTs. In the P3HT:PCBM composite film,
functionalized MWNTs were well dispersed through out
the film. The TEM shows some prolonged nanotubes only
in the edge of the film as the MWNTs were embedded with
the polymer composite. Again, it was observed that there
was no pin hole formation on the top of the active layer
with MWNTs incorporation.

Current Density (m A/cm )

the PEDOT or MoO3 film. For CNTs incorporated device,

modified CNTs were added to P3HT solution and then the
PCBM. Thermal annealing was carried out in inert
atmosphere at 150 C for 15 min. Device fabricated with
C60 as electron selective layer was thermally evaporated
on the P3HT:PCBM active layer. Device with TiOx layer
were fabricated by spin coating from Ti(OC3H7)4 solution in
ethanol and hydrolyzed in air. Al electrodes were thermally
-3
evaporated at a pressure better than 510 Pa.
Transmission electron microscopic measurement of
P3HT: PCBM composite films were carried out by using a
JEOL-2100F field emission transmission electron
microscope operated at 200 V. Current voltage (IV)
characteristics of the fabricated solar cells were measured
at room temperature (25 C) using JASCO SS-200 W
solar simulator in the dark and under Air Mass (AM) 1.5
simulated solar radiation.

ITO/PEDOT/MWNTs+P3HT:PCBM/TiOx/Al
ITO/PEDOT/MWNTs+P3HT:PCBM/TiO2/Al
ITO/PEDOT/P3HT:PCBM/TiOx/Al
ITO/PEDOT/P3HT:PCBM/Al

10

(c)
(b)

ITO/PEDOT/P3HT:PCBM/TiO2/Al

(a)

ITO/PEDOT/P3HT:PCBM/Al

0
-5
-10
-15
-1

-0.5

0.5

Voltage (V)
(b)

Figure 1 Transmission electron micrograph of (a)

functionalized MWNTs sample (b) P3HT:PCBM composite
film incorporating functionalized MWNTs.
Effective application of MWNTs along with the polymer
matrix strongly depends on the homogeneous
dispersibility. In our previous studies
MWNTs

978-1-4244-5892-9/10/$26.00 2010 IEEE

Figure 2 I-V characteristics curve of the fabricated devices

with the structure (a) ITO/PEDOT:PSS/P3HT:PCBM/Al
(b) ITO/PEDOT:PSS/P3HT:PCBM/TiOx/Al
(c) ITO/PEDOT:PSS/MWNTs+P3HT:PCBM/TiOx/Al under
2
illumination (AM 1.5 100 mW/cm ).
Solar cells were fabricated with and without MWNTs
incorporation inserting a TiOx interlayer and device
performances were compared. Figure 2 presents I-V
characteristics under 1.5 AM solar radiation of the
fabricated device with and without MWNTs incorporation
and TiOx interlayer. Fabricated device with normal
structure ITO/PEDOT:PSS/P3HT:PCBM/Al shows Jsc, Voc,
2
FF and as 6.49 mA/cm , 0.51 V, 36.5 % and 1.21 %,
respectively. Similarly, device fabricated inserting the
electron selective layer TiOx in between the P3HT:PCBM
and Al layer shows much better device performance.
Fabricated
device
with
the
structure
ITO/PEDOT:PSS/P3HT:PCBM/TiOx/Al shows Jsc, Voc, FF

000091
1

P3HT:PCBM active layer in bulk heterojunction solar cells.

P3HT:PCBM bulk heterojunction solar cells were
fabricated on a 5 nm thick thermally evaporated MoO3
layer. Fabricated device with MoO3 as hole collecting
electrode shows better device performance than that of
the device fabricated with PEDOT:PSS. The I-V
characteristic with MoO3 interlayer in comparison to
PEDOT:PSS device is shown in figure 4. Fabricated
device with MoO3 layer having the structure
ITO/MoO3/P3HT:PCBM/Al shows Jsc, Voc, FF and as
2
5.93 mA/cm , 0.548 V, 51 % and 1.64 %, respectively. The
enhancement of device performance is due to better Voc
and FF thought Jsc was reduced slightly with MoO3 as hole
collecting interlayer.

10
2

Current density (mA/cm )

and as 7.8 mA/cm , 0.514 V, 45 % and 1.8 %,

respectively. Device performance was enhanced with the
TiOx interlayer due to significant enhancement in Jsc and a
small improvement in FF. In the fabricated device, the
TiOx layer act as hole blocking and electron selective layer
which reduce the possibility of recombination of the
generated exciton. With the interfacial layer electron
selectivity of the top electrode increase significantly. High
current in the forward direction and low leakage current
were observed with the TiOx layer. Functionalized MWNTs
were incorporated in the fabricated device in presence of
the separating interlayer of TiOx. Device fabricated with
the structure ITO/PEDOT:PSS/MWNTs+P3HT:PCBM/TiOx
2
/Al shows Jsc, Voc, FF and as 10.52 mA/cm , 0.50 V,
42 % and 2.2 %, respectively. In the device with
incorporation of MWNTs there is significant improvement
in Jsc and consequently conversion efficiency. The
enhanced device performance with introduction of CNTs is
attributed to better hole transportation and large surface
area for exciton dissociation in the active layer.

8
6

Al/PCBM:P3HT/PEDOT:PSS/ITO(a)
Al/C60/PCBM:P3HT/MoO3/ITO(c)

4
2

Al/PCBM:P3HT/MoO3/ITO (b)

0
-2
-4
-6
-8
-0.5

0.5

Voltage (V)
Figure 3 Energy band diagram of the fabricated device
with incorporation of MWNTs and TiOx layer as electron
selective layer presenting possible charge transportation
path in the device.

Figure 4 I-V characteristics curve of the fabricated devices

with the structure (a) ITO/PEDOT:PSS/P3HT:PCBM/Al
(b) ITO/MoO3/P3HT:PCBM/Al
(c) ITO/MoO3/P3HT:PCBM/C60/Al under illumination (AM
2
1.5 100 mW/cm ).

Possible
charge
transportation
with
MWNTs
incorporation and TiOx as electron selective layer is
explained in terms of energy band diagram as shown in
figure 3. The Work function of MWNTs is in the range of
4.5~5.1 eV, which is close to HOMO level of P3HT; it
signifies possible hole transportation through MWNTs. It is
expected that surface chemical groups, due to
functionalization increase the work function of the MWNTs.
With increase work function of MWNTs it will be much
closer to HOMO level of P3HT and provide efficient hole
transportation. Whereas TiOx layer is expected to work as
an efficient hole blocking layer due to deep HOMO level
and suitable LUMO level close to Fermi energy of Al as
shown in the energy band diagram.
In a different set of experiments, organic solar cells were
fabricated with MoO3 as a hole selective interlayer at the
anode instead of PEDOT:PSS. It has been previously
demonstrated that PEDOT:PSS is aqueous and naturally
acidic, which would cause interface instability in organic
solar cells. Hence, PEDOT:PSS facilitate degradation of

Fabricated solar cells with MoO3 as hole collecting

interlayer was further modified with a thin C60 interlayer in
between PCBM:P3HT and Al electrode. Fabricated device
with the structure ITO/MoO3/P3HT:PCBM/C60/Al shows Jsc,
2
Voc, FF and as 6.06 mA/cm , 0.55 V, 50 % and 1.7 %,
respectively. The device performance is almost similar to
that of the device with only MoO3 interlayer. There is only
small improvement in Jsc, this may be due to better
electron collection with C60 interlayer at the cathode. Here,
we selected C60 as interlayer; since it can provide direct
contact to the PCBM domain in the active layer and the
cathode, while preventing contact of the polymer layer to
the cathode. The thin C60 layer provides electron
selectivity at the interface and thereby reducing
recombination photo generated charges. The energy band
diagram of the fabricated device with MoO3 as hole
selective and C60 as electron selective interlayer is
presented in figure 5. The band energy of MoO3 and C60
materials shows possible hole and electron collection in
the respective interface of the fabricated devices.

978-1-4244-5892-9/10/$26.00 2010 IEEE

000092
1

P3HT:PCBM interface. Similarly, solar cells fabricated with

the thin C60 layer in electron collecting interlayer can play
an important role in stabilizing device performance. At the
cathode, there can be charge accumulation in the
interface between the P3HT:PCBM layer and Al electrode
and which lead to degradation of the active layer. From
the viewpoint of the degradation induced by carrier
accumulation, this indicates that the C60 layer can
smoothly extract electrons from the photoactive layer to
the Al cathode without hindering cell performance.
CONCLUSIONS

Figure 5 Energy band diagram of the fabricated device

with MoO3 as hole selective and C60 as electron selective
interlayer.

Conversion efficiency(%)

Fabricated solar cells with PEDOT:PSS and MoO3 as

hole collecting interlayer performance was studied in air to
observe the initial degradation. The degradation process
shows strong influence of cathode and anode
configuration. Figure 6 presents device performance as a
function of time under similar solar irradiation of AM 1.5
2
100 mW/cm and same device area.
1.6
1.4
1.2
1
0.8
0.6
(Al/PCBM:P3HT/PEDOT:PSS/ITO)

0.4

(Al/PCBM:P3HT/MoO3/ITO)

0.2

In conclusion, we have demonstrate P3HT:PCBM bulk

heterojunction solar cells fabrication with chemically
modified MWNTs incorporation. Improved device
performance was achieved for P3HT:PCBM solar cells
incorporating a certain amount of modified MWNTs in the
active layer. P3HT:PCBM bulk heterojunction solar cells
were fabricated with TiOx interlayer as a electron selective
layer. Fabricated device with modified structure shows Jsc,
2
Voc, FF and as 10.52 mA/cm , 0.50 V, 42 % and 2.2 %,
respectively. The TiOx layer also acts as a barrier for
shorting and shunting in the device, caused by the
presence of metallic CNTs. As well as, we have studied
affect of MoO3 and C60 as a hole and electron selective
interlayer respectively for P3HT:PCBM bulk heterojunction
solar cells. Fabricated device with MoO3 layer compare to
PEDOT:PSS device shows better fill factor and small
improve in open circuit voltage though the short circuit
current density was less. P3HT:CPBM solar cells with
MoO3 and C60 interlayer at respective electrode shows
reduced initial degradation that of the device with
PEDOT:PSS, thereby improving the lifetime of the solar
cells was increased.
ACKNOWLEDGEMENT

(Al/C60/PCBM:P3HT/PEDOT:PSS/ITO)

0
0

10

12

Time (h)
Figure 6 Device performance as a function of time with
device configuration (a) ITO/PEDOT:PSS/P3HT:PCBM/Al
(b) ITO/MoO3/P3HT:PCBM/Al
(c) ITO/MoO3/P3HT:PCBM/C60/Al under illumination (AM
2
1.5 100 mW/cm ).
From the stability measurement, it shows that with MoO3
as hole collecting layer instead of PEDOT:PSS layer is
much more stable. It has been demonstrated that in an
organic solar cells the hygroscopic PEDOT:PSS layer can
adsorption water in air, thereby affecting the active layer.
Measurements of the charge mobility and hole injection
after air exposure showed that the degradation increases
the resistance of the PEDOT:PSS/blend layer interface.[]
On the other hand, It has been observed that PEDOT:PSS
layer accelerate oxidation of the aluminum electrode at the
aluminum/photovoltaic layer interface. Whereas, device
fabricated with MoO3 interlayer can prevent water
adsorption in air and providing a stable interface with the

978-1-4244-5892-9/10/$26.00 2010 IEEE

The Author is grateful to Japan Society for Promotion of

Science (JSPS) for financial support and fellowship and
their collaborators.
REFERENCES
[1] S. Sariciftci, L. Smilowitz, A. J. Heeger and F. Wudl,
Photoinduced Electron Transfer from a Conducting
Polymer to Buckminsterfullerene, Science 258, 1992,
1474-1476.
[2] J. M. Halls, C.A. Walsh, N.C. Greenham, E.A.
Maeseflia, R.H. Friend, S.C. Oratti and A.B. Holmes,
Efficient photodiodes from interpenetrating polymer
networks, Nature 376, 1995, 498.
[3] M. Granstrom, K. Petrisch, A.C. Arias, A. Lux, M.R.
Andersoon and R.H. Friend, Laminated fabrication of
polymeric photovoltaic diodes, Nature 395, 1998 257260.
[4] S.E. Shaheen, C.J. Brabec, N.S. Sariciftci, F. Padinger,

000093
1

T. Fromherz and J.C. Hummelen, 2.5% efficient organic

plastic solar cells Appl. Phys. Lett. 78, 2001, 841.
[5] C.J. Brabec, J.C. Hummelen and N.S. Sariciftci,
Plastic Solar Cells, Adv. Funct. Mater. 11, 2001, 15-26.
[6] G. Kalita, M. Masahiro, W. Koichi, M. Umeno,
Nanostructured morphology of P3HT:PCBM bulk
heterojunction solar cells, Solid State Electronics 54,
2010, 447-451.
[7] G. Kalita, H.R. Aryal, S. Adhikari, R. Afre, T. Soga, M.
Sharon and M. Umeno Fullerene (C60) decoration in
oxygen plasma treated multiwalled carbon nanotubes for
photovoltaic application, Appl. Phys. Lett. 92, 2008,
063508.
[8] G. Kalita, S. Adhikari H.R. Aryal, M. Umeno, R. Afre, T.
Soga, M. Sharon, Cutting carbon nanotubes for solar cell
application, Appl. Phys. Lett. 92, 2008, 123508.

978-1-4244-5892-9/10/$26.00 2010 IEEE

[9] G. Kalita, H.R. Aryal, S. Adhikari, K. Wakita, M. Umeno,

Poly (3-octylthiophene)/fullerene heterojunction solar cell
incorporating carbon nanotubes J. Nanosci. Nanotech. 10,
2010, 3844-3848.
[10] K. Kawano, R. Pacios, D. Poplavskyy, J. Nelson D.
D.C. Bradley and J. R. Durrant, Solar Energy Materials
and Solar Cells 90, 2008 3520-3530
[11] Y. Kanai, T. Matsushima and H. Murata,
Improvement of stability for organic solar cells by using
molybdenum trioxide buffer layer, Thin Solid Films
518, 2009, 537-540.
[12] K. Kawano and C. Adachi, Reduced initial
degradation of bulk heterojunction organic solar cells by
incorporation of stacked fullerene and lithium fluoride
interlayers Appl. Phys. Lett. 96, 2010, 053307.

000094
1

DIAGNOSTIC STRUCTURES FOR EPITAXIAL THIN SILICON SOLAR CELLS

C. Paola Murcia1, Ruiying Hao1, Tim Creazzo1, Kevin Shreve1, Anthony Lochtefeld2, Michael Curtin2, Allen Barnett1.
1
Solar Power Program University of Delaware, Newark, DE, United States
2
AmberWave Systems, Salem, NH, United States
ABSTRACT
The value of a high performance thin silicon solar cell is
based on high open circuit voltage (Voc) which is highly
dependent upon surface and interface recombination. A
microelectronic approach with the series and parallel
fabrication of different device structures is presented. This
approach includes the fabrication of planar solar cells
based on different solar cell designs that maximize Voc.
Subsequently, using the same solar cell design, more
advanced epitaxial growth test structures were fabricated
on the same substrate to understand recombination losses
and Voc. In high performance thin silicon solar cells, Voc
is highly sensitive to surface recombination. Achieving a
good surface and interface passivation in epitaxial silicon
solar cells is a challenge, especially for the absorber back
surface. In this work, we used test structures embedded
in the solar cells to independently optimize Voc and the
back surface of a thin epitaxial absorber as well as other
silicon interfaces. Most epitaxial thin silicon solar cells are
fabricated with in-situ processes that lead to difficulties in
the independent optimization of such surfaces.
To
contribute to the design space in thin silicon solar cells, we
tested basic device designs and test structures that lead to
high Voc. The design of basic structures is important to
measure parameters that limit voltage. This is a diagnostic
tool for the fabrication of high performance thin silicon
solar cells. This work reports device results that exhibit a
path to high open circuit voltage, determined by device
design, interface recombination, and bulk lifetime. With
independent optimization of parameters that limit Voc and
therefore performance, this work contributes to the
understanding of thin silicon solar cells.
BACKGROUND
Thin silicon solar cells based on epitaxially grown
absorbers have recently reported efficiencies up to 16.9%
[1]. Fabrication of thin crystalline silicon solar cells based
on epitaxial absorbers is cumbersome and critical
parameters cannot be easily measured.
Different
approaches to a high performance thin silicon solar cell
have been reported before [1-6].
Epitaxial wafer
equivalent designs aim for a low cost solution based on
epitaxially grown active layers deposited in an in-situ
process [3]. Most of the recent literature on crystalline
thin-film silicon solar cells reports device results and
explains the fabricated solar cell structures. Previous
analyses of thin silicon solar cell designs show that for
high performance high open circuit voltage and good light
trapping are needed. Open circuit voltage precedes
optical confinement in importance to achieve high
efficiency [4, 7].
Current research presents the

978-1-4244-5892-9/10/$26.00 2010 IEEE

opportunity to analyze thin silicon solar cell structures from

first principles and design structures that allow
optimization of solar cell parameters that are important for
high performance.
(a)

(b)

Figure 1. Evolution of test structures for a high

performance thin silicon solar cell. (a)Thin silicon
solar cell using a SiO2 to passivate silicon-silicon
interfaces (b) Planar thin silicon solar cell
TEST STRUCTURE DESIGN
The high performance of a thin silicon solar cell is based
upon high open circuit voltage, which, in turn, is highly
dependent on surface recombination. We present: (1) a
microelectronic approach with the parallel fabrication of
planar devices that test advanced concepts to further
increase open circuit voltage, (2) design test structures
that can be used to resolve a path to high Voc, (3) test
structure results exhibiting studied voltage and surface
recombination characteristics that show a path to high
performance.
In our thin silicon solar cell design (Figure 1), a low cost
conductive substrate is used to grow a high performance
solar cell. The first step is to optimize the absorber bulk
lifetime during the epitaxy. A second step requires good
electrical properties (low series resistance) and good
optical properties (high reflectivity) at the back interface of
the solar cell. In this work we focus only on surface or
interface recombination. To systematically improve the
cell in Figure 1a, we start with a planar solar cell (Figure
1b) and then transfer to a lateral design (Figure 1a) by the
gradual implementation of a patterned dielectric (silicon
dioxide for this work) to passivate Si-Si interfaces and
study the path towards higher Voc.

002204

In previously reported wafer equivalent designs of planar

thin silicon solar cells (Figure 1b), the back surface
interface is composed of porous Si (PSi). PSi has been
chosen because of its good optical properties as a Bragg
reflector along with its compatibility with the epitaxial
growth of silicon layers [4,5]. This back interface has not
been quantitatively characterized yet in terms of
recombination losses.

Figure 2. Top view of active area morphology: lateral

overgrowth geometries for grid designs. The red
corresponds to the junction areas. The green area
corresponds to the total front surface area. AV/AL is a
measurement of the lateral vs. vertical growth and
varies according to the geometry or the absorber
thickness.
Microelectronic Approach
The ability to test structures and devices fabricated on the
same substrate and exposed to the same processes is
important. For this purpose, we present a mask design
that includes planar solar cells and test structures with
silicon-silicon and silicon-dielectric interfaces. All the
geometrical designs that test for morphology of the
epitaxial growth will be fabricated, tested and analyzed
using a single substrate. This facilitates the reduction of
error sources caused by processing differences from run
to run or differences from wafer to wafer.
Previous work has explored the use of dielectrics as part
of electrical interfaces by using silicon on insulator (SOI)
crystalline substrates [6,8]. Such designs aim to reduce
recombination and dark saturation current in order to
increase open circuit voltage.
Among some of the
potential benefits: first, dielectrics can increase the back
surface reflectance.
Second, they can serve as
passivation layers for the junction surfaces. Including a
dielectric intermediate layer opens a new design space for
test structures to analyze the effects of lateral overgrowth
and silicon / dielectric interfaces. By growing epitaxial thin
silicon solar cells using all these different geometries, we
can simultaneously test surface recombination, dielectric
passivation, epitaxial growth morphology and optical
properties embedded in these designs. All the previously
mentioned benefits will be measured by means of the Voc
of the solar cell.

978-1-4244-5892-9/10/$26.00 2010 IEEE

Layout Design
The active (or junction) area designs proposed below are
based on combinations of grid shapes. The goal is to
grow different absorber thicknesses to vary the area of the
silicon/dielectric interface (Figure 2). On the other hand,
different geometries can exhibit a dependence relationship
on Voc. Higher quality growths are possible for specific
dielectrics and the crystal defect density can be optimized.
Other characteristics of the epitaxial growth are:
coalescence and substrate crystal orientation vs. design
orientation.
Achieving coalescence increases the crystal defect
density which can then be reduced with a clever via
design. In general, the geometrical design of the active
area is a tool to determine contributions and losses from
the epitaxial absorber back surface regarding open circuit
voltage.
With our microelectronic approach, these
advanced grid designs are compared to their planar
equivalents.
SOLAR CELL FABRICATION
The microelectronic approach starts with the growth of a
dry oxide on the conductive Silicon substrate. The
substrate is (100) p-type crystalline silicon of
approximately 0.01ohm-cm bulk resistivity.
The dry
oxidation is done at 900C in a tube furnace. After the
oxide
growth,
active
areas
are
defined
by
photolithography.
The solar cell is grown by high
temperature atmospheric pressure chemical vapor
deposition (APCVD) with a high growth rate.

Figure 3. Scanning electron microscope (SEM) image

of a cross section of a grid test structure.
The absorber is preceded by a thin p+ Si growth.
Subsequently, the n-type Si (10um) and a thin n+ Si layer
for decreased contact resistance are grown. As front
surface passivation, a thin dry oxide is thermally grown on
the top surface of some samples. This thin dry oxide is
grown at 900C with high purity oxygen. Aluminum is e-

002205

beam evaporated as the back contact. At this stage, Voc

is measured using a Suns-Voc setup which uses a
separate monitor solar cell to measure the current (or light
intensity) in order to extrapolate a Voc measurement.

Figure 5. Experimental results of Voc dependence on

front surface area vs. back Si-Si interface area ratio.

Figure 4. Scanning electron microscope (SEM) image

of a top view of a grid test structure.
Other thermal treatments used include Rapid Thermal
Processing (RTP) in a forming gas atmosphere and
furnace annealing in a nitrogen and oxygen atmosphere.
Suns-Voc is measured after thermal processing. The grid
test structures are equivalent in area to the planar solar
cells, 25mm2. Data reported here corresponds to an
absorber thickness of 10um as previously stated. Other
processing steps involving testing procedures include
testing after etching vias through the passivation oxide
and partial dicing to eliminate probe resistance losses and
cross-talk.
EXPERIMENTAL RESULTS
Surface morphology analysis of the epitaxial solar cell was
done with scanning electron microscopy (SEM) (Figures 3
and 4). On average, the lateral growth equals the vertical
growth for most of the test structures. SEM images show
the propagation of faults through the lateral growth (Figure
3). The silicon growth is as expected and lateral growing
planes show a larger defect density. The following plots
show a summary of voltage data plotted as a function of
front surface and back surface areas. Planar solar cells
Suns-Voc measurements are above 550mV. Test patterns
Suns-Voc values are below 500mV. In order to locate the
losses, we use the voltage results from the test patterns.
There is a voltage relationship between the Si-oxide
interface area and Si-Si area. Voc data shows that
voltage is inversely proportional to front surface (active
area) vs junction surface (junction area) ratio. This is
shown in Figure 5. All data points are from cells on the
same substrate.

978-1-4244-5892-9/10/$26.00 2010 IEEE

We define a laterally grown active area as the total area

on the test pattern that grows on the oxide mask. The
vertically grown active area is the total junction area or
area where vertical growth occurs. This indicates that
structures with less laterally grown front surface have
higher voltage.
To determine sources of voltage loss, only the front
surface area was analyzed (Figure 6). Following the
arrows in the plot, we can read that when the line width is
kept constant and the spacing changes, the Voc does not
change. This is an indicator that the density of lateral
growth on corner points in the grid design does not affect
the voltage in these structures. On the other hand,
holding the spacing constant and increasing the line width
causes the L/V (lateral/vertical) area ratio to be larger, and
the voltage increases. This corroborates the result of
Figure 5.
A third example of the effect of the laterally grown area on
voltage loss is shown in Figure 7. Here Voc vs. total
perimeter of the active area present in the grid designs is
plotted. As the lateral area approaches the vertical area
(line width and spacing are closer) or the perimeter
decreases, the voltage increases. Back surface
recombination is appears to affect our test structures to
the point that the expected gain in Voc due to reduced
junction area (and recombination current) is cancelled out.
In general, these test structures present a lower than
expected open circuit voltage. The Voc is limited by (1)
bulk lifetime, as seen from the planar solar cell results,
and (2) by interface recombination as seen from the test
structures. Bulk lifetime is limited by the epitaxial growth
and processing conditions of our devices.

002206

understand the performance of advanced reflectors and

other electrical interfaces used in back surfaces. In
recently explored epitaxial wafer equivalent designs, all
interfaces within the active cell and between the solar cell
and the conductive substrate are silicon and silicon
dioxide. This research presented the design, fabrication
and analysis of test device structures that can be
optimized to increase open circuit voltage for thin silicon
solar cells. Improved understanding
of these
thin
crystalline silicon solar cell structures can lead to a path
for maximizing open circuit voltage.
ACKNOWLEDGEMENTS

Figure 6. Experimental results of Voc dependence on

light generating area (front surface).
Areas used to probe and measure voltage were also
considered as sources of voltage loss in this analysis. All
planar solar cells and test structures had the same contact
pads and the area of the pad is larger than the total
junction area. However, all solar cells had the same
testing pad, suggesting the variation of Voc in the test
patterns as seen in Figures 5, 6, 7 is attributed to losses in
the pattern itself and not the contact pad. In addition, all
planar solar cells had larger Voc than the test patterns.
Both planar and test structures were grown orthogonal to
the (100) direction and 45 degrees off-axis. There is
negligible voltage dependence on the pattern orientation.

This research was, in part, funded by the U.S.

Government Defense Advanced Research Projects
Agency under Agreement No.: HR0011-07-9-0005. The
views, opinions, and/or findings contained in this
article/presentation are those of the authors and should
not be interpreted as representing the official views or
policies, either expressed or implied, of the Defense
Advanced Research Projects Agency or the Department of
Defense.
Approved for Public Release, Distribution Unlimited.
REFERENCES
[1] K. Van Neiuwenhuysen et al. Epitaxial thin film silicon
solar cells with efficiencies up to 16.9% by combining
advanced light trapping methods and CVD emitters, 24th
EU PVSC. (2009)
[2] J. Van Hoeymissen et al. Thin-film epitaxial solar cells
on low cost substrates: closing the efficiency gap with bulk
Si cells using advanced photonic structures and emitters,
23rd EU PVSC. (2008)
[3] E. Schmich et al. n-type emitter epitaxy for crystalline
silicon thin-film solar cells. Prog. Photovolt. Res. Appl.
V16, P159.(2008)
[4] F. Duerinckx, et al. Optical path length enhancement
for >13% screenprinted thin film silicon solar cells. 21st
EU PVSEC (2006)
[5] F. Haase et al. Back contact monocrystalline thin-film
silicon solar cells from the porous silicon process. 24th
EUPVSC (2009)
[6] C.P.Murcia, et al. High performance thin silicon solar
cells grown on silicon, 24th EUPVSC (2009).

Figure 7. Voltage dependence on active area perimeter

CONCLUSIONS

[7] A. Barnett, C. Honsberg, et al. High Performance Thin

Silicon Solar Cells, EU PVSEC. (2007)
[8] R. Hao, et al., Thin silicon solar cells using epitaxial
lateral overgrowth structure, 34th IEEE PVSEC (2009).

Thin silicon solar cells with epitaxial absorbers include

new interfaces that should be tested in basic designs to

978-1-4244-5892-9/10/$26.00 2010 IEEE

002207

DIAGNOSTIC STRUCTURES FOR EPITAXIAL THIN SILICON SOLAR CELLS

C. Paola Murcia1, Ruiying Hao1, Tim Creazzo1, Kevin Shreve1, Anthony Lochtefeld2, Michael Curtin2, Allen Barnett1.
1
Solar Power Program University of Delaware, Newark, DE, United States
2
AmberWave Systems, Salem, NH, United States
ABSTRACT
The value of a high performance thin silicon solar cell is
based on high open circuit voltage (Voc) which is highly
dependent upon surface and interface recombination. A
microelectronic approach with the series and parallel
fabrication of different device structures is presented. This
approach includes the fabrication of planar solar cells
based on different solar cell designs that maximize Voc.
Subsequently, using the same solar cell design, more
advanced epitaxial growth test structures were fabricated
on the same substrate to understand recombination losses
and Voc. In high performance thin silicon solar cells, Voc
is highly sensitive to surface recombination. Achieving a
good surface and interface passivation in epitaxial silicon
solar cells is a challenge, especially for the absorber back
surface. In this work, we used test structures embedded
in the solar cells to independently optimize Voc and the
back surface of a thin epitaxial absorber as well as other
silicon interfaces. Most epitaxial thin silicon solar cells are
fabricated with in-situ processes that lead to difficulties in
the independent optimization of such surfaces.
To
contribute to the design space in thin silicon solar cells, we
tested basic device designs and test structures that lead to
high Voc. The design of basic structures is important to
measure parameters that limit voltage. This is a diagnostic
tool for the fabrication of high performance thin silicon
solar cells. This work reports device results that exhibit a
path to high open circuit voltage, determined by device
design, interface recombination, and bulk lifetime. With
independent optimization of parameters that limit Voc and
therefore performance, this work contributes to the
understanding of thin silicon solar cells.
BACKGROUND
Thin silicon solar cells based on epitaxially grown
absorbers have recently reported efficiencies up to 16.9%
[1]. Fabrication of thin crystalline silicon solar cells based
on epitaxial absorbers is cumbersome and critical
parameters cannot be easily measured.
Different
approaches to a high performance thin silicon solar cell
have been reported before [1-6].
Epitaxial wafer
equivalent designs aim for a low cost solution based on
epitaxially grown active layers deposited in an in-situ
process [3]. Most of the recent literature on crystalline
thin-film silicon solar cells reports device results and
explains the fabricated solar cell structures. Previous
analyses of thin silicon solar cell designs show that for
high performance high open circuit voltage and good light
trapping are needed. Open circuit voltage precedes
optical confinement in importance to achieve high
efficiency [4, 7].
Current research presents the

978-1-4244-5892-9/10/$26.00 2010 IEEE

opportunity to analyze thin silicon solar cell structures from

first principles and design structures that allow
optimization of solar cell parameters that are important for
high performance.
(a)

(b)

Figure 1. Evolution of test structures for a high

performance thin silicon solar cell. (a)Thin silicon
solar cell using a SiO2 to passivate silicon-silicon
interfaces (b) Planar thin silicon solar cell
TEST STRUCTURE DESIGN
The high performance of a thin silicon solar cell is based
upon high open circuit voltage, which, in turn, is highly
dependent on surface recombination. We present: (1) a
microelectronic approach with the parallel fabrication of
planar devices that test advanced concepts to further
increase open circuit voltage, (2) design test structures
that can be used to resolve a path to high Voc, (3) test
structure results exhibiting studied voltage and surface
recombination characteristics that show a path to high
performance.
In our thin silicon solar cell design (Figure 1), a low cost
conductive substrate is used to grow a high performance
solar cell. The first step is to optimize the absorber bulk
lifetime during the epitaxy. A second step requires good
electrical properties (low series resistance) and good
optical properties (high reflectivity) at the back interface of
the solar cell. In this work we focus only on surface or
interface recombination. To systematically improve the
cell in Figure 1a, we start with a planar solar cell (Figure
1b) and then transfer to a lateral design (Figure 1a) by the
gradual implementation of a patterned dielectric (silicon
dioxide for this work) to passivate Si-Si interfaces and
study the path towards higher Voc.

002204

In previously reported wafer equivalent designs of planar

thin silicon solar cells (Figure 1b), the back surface
interface is composed of porous Si (PSi). PSi has been
chosen because of its good optical properties as a Bragg
reflector along with its compatibility with the epitaxial
growth of silicon layers [4,5]. This back interface has not
been quantitatively characterized yet in terms of
recombination losses.

Figure 2. Top view of active area morphology: lateral

overgrowth geometries for grid designs. The red
corresponds to the junction areas. The green area
corresponds to the total front surface area. AV/AL is a
measurement of the lateral vs. vertical growth and
varies according to the geometry or the absorber
thickness.
Microelectronic Approach
The ability to test structures and devices fabricated on the
same substrate and exposed to the same processes is
important. For this purpose, we present a mask design
that includes planar solar cells and test structures with
silicon-silicon and silicon-dielectric interfaces. All the
geometrical designs that test for morphology of the
epitaxial growth will be fabricated, tested and analyzed
using a single substrate. This facilitates the reduction of
error sources caused by processing differences from run
to run or differences from wafer to wafer.
Previous work has explored the use of dielectrics as part
of electrical interfaces by using silicon on insulator (SOI)
crystalline substrates [6,8]. Such designs aim to reduce
recombination and dark saturation current in order to
increase open circuit voltage.
Among some of the
potential benefits: first, dielectrics can increase the back
surface reflectance.
Second, they can serve as
passivation layers for the junction surfaces. Including a
dielectric intermediate layer opens a new design space for
test structures to analyze the effects of lateral overgrowth
and silicon / dielectric interfaces. By growing epitaxial thin
silicon solar cells using all these different geometries, we
can simultaneously test surface recombination, dielectric
passivation, epitaxial growth morphology and optical
properties embedded in these designs. All the previously
mentioned benefits will be measured by means of the Voc
of the solar cell.

978-1-4244-5892-9/10/$26.00 2010 IEEE

Layout Design
The active (or junction) area designs proposed below are
based on combinations of grid shapes. The goal is to
grow different absorber thicknesses to vary the area of the
silicon/dielectric interface (Figure 2). On the other hand,
different geometries can exhibit a dependence relationship
on Voc. Higher quality growths are possible for specific
dielectrics and the crystal defect density can be optimized.
Other characteristics of the epitaxial growth are:
coalescence and substrate crystal orientation vs. design
orientation.
Achieving coalescence increases the crystal defect
density which can then be reduced with a clever via
design. In general, the geometrical design of the active
area is a tool to determine contributions and losses from
the epitaxial absorber back surface regarding open circuit
voltage.
With our microelectronic approach, these
advanced grid designs are compared to their planar
equivalents.
SOLAR CELL FABRICATION
The microelectronic approach starts with the growth of a
dry oxide on the conductive Silicon substrate. The
substrate is (100) p-type crystalline silicon of
approximately 0.01ohm-cm bulk resistivity.
The dry
oxidation is done at 900C in a tube furnace. After the
oxide
growth,
active
areas
are
defined
by
photolithography.
The solar cell is grown by high
temperature atmospheric pressure chemical vapor
deposition (APCVD) with a high growth rate.

Figure 3. Scanning electron microscope (SEM) image

of a cross section of a grid test structure.
The absorber is preceded by a thin p+ Si growth.
Subsequently, the n-type Si (10um) and a thin n+ Si layer
for decreased contact resistance are grown. As front
surface passivation, a thin dry oxide is thermally grown on
the top surface of some samples. This thin dry oxide is
grown at 900C with high purity oxygen. Aluminum is e-

002205

beam evaporated as the back contact. At this stage, Voc

is measured using a Suns-Voc setup which uses a
separate monitor solar cell to measure the current (or light
intensity) in order to extrapolate a Voc measurement.

Figure 5. Experimental results of Voc dependence on

front surface area vs. back Si-Si interface area ratio.

Figure 4. Scanning electron microscope (SEM) image

of a top view of a grid test structure.
Other thermal treatments used include Rapid Thermal
Processing (RTP) in a forming gas atmosphere and
furnace annealing in a nitrogen and oxygen atmosphere.
Suns-Voc is measured after thermal processing. The grid
test structures are equivalent in area to the planar solar
cells, 25mm2. Data reported here corresponds to an
absorber thickness of 10um as previously stated. Other
processing steps involving testing procedures include
testing after etching vias through the passivation oxide
and partial dicing to eliminate probe resistance losses and
cross-talk.
EXPERIMENTAL RESULTS
Surface morphology analysis of the epitaxial solar cell was
done with scanning electron microscopy (SEM) (Figures 3
and 4). On average, the lateral growth equals the vertical
growth for most of the test structures. SEM images show
the propagation of faults through the lateral growth (Figure
3). The silicon growth is as expected and lateral growing
planes show a larger defect density. The following plots
show a summary of voltage data plotted as a function of
front surface and back surface areas. Planar solar cells
Suns-Voc measurements are above 550mV. Test patterns
Suns-Voc values are below 500mV. In order to locate the
losses, we use the voltage results from the test patterns.
There is a voltage relationship between the Si-oxide
interface area and Si-Si area. Voc data shows that
voltage is inversely proportional to front surface (active
area) vs junction surface (junction area) ratio. This is
shown in Figure 5. All data points are from cells on the
same substrate.

978-1-4244-5892-9/10/$26.00 2010 IEEE

We define a laterally grown active area as the total area

on the test pattern that grows on the oxide mask. The
vertically grown active area is the total junction area or
area where vertical growth occurs. This indicates that
structures with less laterally grown front surface have
higher voltage.
To determine sources of voltage loss, only the front
surface area was analyzed (Figure 6). Following the
arrows in the plot, we can read that when the line width is
kept constant and the spacing changes, the Voc does not
change. This is an indicator that the density of lateral
growth on corner points in the grid design does not affect
the voltage in these structures. On the other hand,
holding the spacing constant and increasing the line width
causes the L/V (lateral/vertical) area ratio to be larger, and
the voltage increases. This corroborates the result of
Figure 5.
A third example of the effect of the laterally grown area on
voltage loss is shown in Figure 7. Here Voc vs. total
perimeter of the active area present in the grid designs is
plotted. As the lateral area approaches the vertical area
(line width and spacing are closer) or the perimeter
decreases, the voltage increases. Back surface
recombination is appears to affect our test structures to
the point that the expected gain in Voc due to reduced
junction area (and recombination current) is cancelled out.
In general, these test structures present a lower than
expected open circuit voltage. The Voc is limited by (1)
bulk lifetime, as seen from the planar solar cell results,
and (2) by interface recombination as seen from the test
structures. Bulk lifetime is limited by the epitaxial growth
and processing conditions of our devices.

002206

understand the performance of advanced reflectors and

other electrical interfaces used in back surfaces. In
recently explored epitaxial wafer equivalent designs, all
interfaces within the active cell and between the solar cell
and the conductive substrate are silicon and silicon
dioxide. This research presented the design, fabrication
and analysis of test device structures that can be
optimized to increase open circuit voltage for thin silicon
solar cells. Improved understanding
of these
thin
crystalline silicon solar cell structures can lead to a path
for maximizing open circuit voltage.
ACKNOWLEDGEMENTS

Figure 6. Experimental results of Voc dependence on

light generating area (front surface).
Areas used to probe and measure voltage were also
considered as sources of voltage loss in this analysis. All
planar solar cells and test structures had the same contact
pads and the area of the pad is larger than the total
junction area. However, all solar cells had the same
testing pad, suggesting the variation of Voc in the test
patterns as seen in Figures 5, 6, 7 is attributed to losses in
the pattern itself and not the contact pad. In addition, all
planar solar cells had larger Voc than the test patterns.
Both planar and test structures were grown orthogonal to
the (100) direction and 45 degrees off-axis. There is
negligible voltage dependence on the pattern orientation.

This research was, in part, funded by the U.S.

Government Defense Advanced Research Projects
Agency under Agreement No.: HR0011-07-9-0005. The
views, opinions, and/or findings contained in this
article/presentation are those of the authors and should
not be interpreted as representing the official views or
policies, either expressed or implied, of the Defense
Advanced Research Projects Agency or the Department of
Defense.
Approved for Public Release, Distribution Unlimited.
REFERENCES
[1] K. Van Neiuwenhuysen et al. Epitaxial thin film silicon
solar cells with efficiencies up to 16.9% by combining
advanced light trapping methods and CVD emitters, 24th
EU PVSC. (2009)
[2] J. Van Hoeymissen et al. Thin-film epitaxial solar cells
on low cost substrates: closing the efficiency gap with bulk
Si cells using advanced photonic structures and emitters,
23rd EU PVSC. (2008)
[3] E. Schmich et al. n-type emitter epitaxy for crystalline
silicon thin-film solar cells. Prog. Photovolt. Res. Appl.
V16, P159.(2008)
[4] F. Duerinckx, et al. Optical path length enhancement
for >13% screenprinted thin film silicon solar cells. 21st
EU PVSEC (2006)
[5] F. Haase et al. Back contact monocrystalline thin-film
silicon solar cells from the porous silicon process. 24th
EUPVSC (2009)
[6] C.P.Murcia, et al. High performance thin silicon solar
cells grown on silicon, 24th EUPVSC (2009).

Figure 7. Voltage dependence on active area perimeter

CONCLUSIONS

[7] A. Barnett, C. Honsberg, et al. High Performance Thin

Silicon Solar Cells, EU PVSEC. (2007)
[8] R. Hao, et al., Thin silicon solar cells using epitaxial
lateral overgrowth structure, 34th IEEE PVSEC (2009).

Thin silicon solar cells with epitaxial absorbers include

new interfaces that should be tested in basic designs to

978-1-4244-5892-9/10/$26.00 2010 IEEE

002207

2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

A leakage Current Limited Robust SRAM for Subthreshold/Nearthreshold

Applications
Yang Tongmu
Nanjing Branch, Shanghai Baosight Software Co.,Ltd.
College of Software Engineering, Southeast University
Nanjing, China
yangtongmu@baosight.com

AbstractIn this paper, a leakage current limited SRAM

bitcell operating in subthreshold/nearthreshold region is
demonstrated in IBM 0.13 m CMOS process. Proposed bitcell
exhibits wide hysteresis effect, making the design less
vulnerable to process variation. Its hold margin is 30.2% and
18.9% greater than conventional 6T SRAM and referenced
SRAM (at 400 mV). At the same times, the SRAM trip point
voltage changes according to bitline voltages. Its read margin is
45% and 9% greater than conventional 6T SRAM and
referenced SRAM (at 400 mV). Dynamic leakage cut off
transistor is utilized to reduce leakage current without
inducing dynamic energy penalty. Thus, compared to the
referenced 4kb SRAM array, proposed array optimum-energy
supply voltage is scaled from 400 mV to 310 mV

moment. Moreover, this technology has the disadvantage

that Static Noise Margin (SNM) can be prone to converge to
zero. Leakage energy becomes a critical part in the
subthreshold/nearthreshold SRAM design because it takes
more and more portion, especially at low system frequency.
So how to reduce it without dynamic energy penalty
becomes more important than ever.
This paper presents a leakage current limited robust
SRAM which can reduce leakage current without inducing
any dynamic energy penalty. The rest part of the paper is
organized as follows. Proposed robust SRAM bitcell and its
operations are described in section II. In last section,
proposed technique is validated, and the simulated results are
presented

Keywords- subthreshold/nearthreshold SRAM, low leakage

current, static noise margin, optimum-energy supply voltage

I.

12T SUBTHRESHOLD BITCELL

INTRODUCTION

Subthreshold/nearthreshold logic circuits are becoming

increasingly popular in ultra-low-power applications [1]-[4].
However, when the device feature size reaches the nanoscale
range, it is extremely challenge to design robust CMOS logic
for subthreshold/nearthreshold systems because of the
reduced voltage margin and the increased device variations.
The major difficulties with the conventional 6T SRAM (6T)
design in subthreshold/nearthreshold operation include: 1)
degraded immunity of the process variation; 2) poor
writability; 3) increased leakage energy; 4) limited number
of bitcells per bitline and reduced bitline sensing margin [5].
Various methods are adopted to tackle these problems.
Primarily, reference [2]-[4] decouple the SRAM bitcell
nodes from the bitline and hence retain the value during read
operations. Unfortunately, these single read methods tend to
favor readability in subthreshold/nearthreshold operation at
the expense of degraded writability. Reference [3] and [6]
use higher wordline voltage to increase the drive current of
the write access transistors. This improves the write margin
at the cost of additional power routing, extra voltage
generation circuitry, and increased power consumption.
Virtual supply rails have also been used in previous work to
improve bitcell writability [2]. In the design, the collapsed
virtual supply rails make it easier for the accessed bitcell to
flip its value. The collapsed supply voltage also deteriorates
the stability of other SRAM bitcells, which are connected to
the same bitline or wordline but not to be written at the

978-1-4244-7237-6/$26.00

II.

2010 IEEE

Figure 1.

Total energy and optimal Vdd of referencd 4kb SRAM array.

As voltage scales, the dynamic energy reduces. However,

because of the increased delay, the leakage energy increases.
Therefore, whether the total optimum Energy-per-Operation
(EPO) increases or decreases is uncertain. In fact, there is an
optimum-energy supply voltage point, operating at which
offers the best EPO. This optimum energy operating supply
voltage (Vopt) can also be obtained experimentally. Several
circuits are analyzed, and it is found that Vopt is normally in
weak subthreshold or nearthreshold region. Vopt of an
example circuit (4kb SRAM array of [1]) is obtained by a
simulation, which is indicated in Fig. 1.
It can be noted that the Vopt of 4kb referenced SRAM
(ST) array is 400mV. Further reducing Vdd below Vopt does
not yield any additional energy benefits. In fact, Vopt is

V9-26

2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

determined by many parameters. As for SRAM array with

extremely low switching probability and high leakage
current, it is sure to obtain a higher Vopt than other digital
blocks. Switching probability is determined by the SRAM
array architecture. Therefore, reducing leakage current can
reduce total energy consumption from: 1) the reduced
leakage energy itself, 2) the reduced energy benefit get by
reduced Vopt.
A. Circuit Design
In comparison with the abovethreshold designs,
subthreshold/nearthreshold static noise margin (SNM)
sensitivity to sizing changes is reduced. In this paper,
structural change of bitcell instead of the sizing change is
considered to enhance the immunity with the processvoltage-temperature (PVT) fluctuations.

(a)

(b)

Figure 2. (a) Proposed SRAM bitcell and (b) its transfer characteristics.

The circuit schematic of proposed bitcell is illustrated in

Fig. 2(a). Transistors P1-N3-N5 and P2-N4-N6 form two
coupling inverters. N9 and N10 are the access transistors.
The positive feedback from N7/N8 adaptively changes the
switching threshold of the inverter corresponding to the
direction of input transition. The NMOS transistors N1 and
N2, respectively gated by the signals Q and Q , are
introduced to the bitcell to minimize the leakage power
dissipation during standby mode.
The transfer characteristics of proposed bitcell, 6T and
the ST bitcell are shown in Fig. 2(b) respectively. The curves
of the wide dual threshold action (hysteresis effect) of
proposed design, indicates that proposed design has higher
stability (noise immunity) than 6T and ST bitcell.
Considering that PMOS drive-strength is weaker than that of
NMOS, the trip point of bitcell need to be just larger than
Vdd/2 during transferring from state 0 to state 1, and it
should be significantly lower than Vdd/2 during transferring
from state 1 to state 0 to detect and store the correct
signal from bitline BL and bitline BL . For proposed design
at Vdd of 400mV, the two trip points are 227 mV and 104
mV, respectively.

gate overdrive, large load capacitance and process variation.

The break of positive feedback and reduced pull-down
strength (serially connected NMOS) helps proposed SRAM
bitcell to achieve better writability.
Initially consider the condition that Q = 0 and Q =
1. In order to write a 0 to Q , bitline BL and bitline
BL are forced to be 1 and 0, respectively. The wordline
WL becomes high, which discharges the node Q and turns
on the transistor P1. This makes the source terminal of the
transistor N1 electrically connected to Q . Thus, N1 is turned
off because its Vgs is 0. And this condition helps to break
the feedback of latch. During the transition that node Q is
changing from state 0 to state 1, the feedback transistor
(N7) can increase the value of Q35 . All these conditions
make write access easier.
The value that node Q can be charged or discharged is
closely related to the length of the write period. Fortunately,
it is demonstrated that the data correctness is not affected
even if the node is charged to a value smaller than a full high
signal value or discharged to a value larger than a full low
signal value.

C. Standby Mode
In the standby mode, the supply voltage of SRAM bitcell
is cut off by the NMOS transistor N1/N2 to reduce the
leakage power. However, the supply voltage cannot be
reduced arbitrarily as bitcells will not be able to hold the
contents of the bitcells. So the size of N1/N2 transistors
should be set carefully. The comparison of hold SNM at low
supply voltages is shown in Fig. 3. As SRAM bitcells of [2],
[3] and [4] adopt the conventional inverter pair based
architecture, their transfer characteristics are therefore
similar with 6T in hold mode.
It can be clearly observed from the Fig. 3 proposed
bitcell exhibits superior hold margin. In IBM 0.13 m
process technology, the hold SNM of 6T and ST bitcells is
166 mV and 193 mV, while proposed bitcell exhibits 238
mV hold SNM at 400 mV, which is significantly better. Its
hold margin is 30.2% and 18.9% greater than 6T and ST (at
400 mV).
Since the cut-off transistors N1 and N2 neither introduce
extra load on the wordline nor require an auxiliary circuit to
switch the bitcell into a stand-by mode, N1 and N2 do not
have any influence on dynamic power dissipation and access
time. Proposed SRAM bitcell reduces leakage power without
increasing dynamic power dissipation in the standby mode.

B. Write Operation
Maintaining a sufficient write margin is challenging in
subthreshold/nearthreshold SRAM design due to the small

V9-27

2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

(a)

may charge Q to increase to a positive value; Q thus makes

a transition from Vdd to a lower voltage. No matter what the
memory structures are, a read failure (RF) event is occurred
when the memory stored value 0 is increased to positive
value higher than the trip point (VTRIP). Hence, the readfailure probability (PRF) is given by:
PRF = P[VREAD > VTRIP]
(1)
So, PRF can be reduced by increasing inverter switching
threshold. With the higher threshold voltage during state 0
to state 1, proposed design has 45% and 9% improvement
in the read SNM, compare to 6T and ST, shown in Fig. 4.

(b)

III.

(c)
Figure 3.

Hold margin of (a) 6T, (b) ST, and (c) proposed design (at 400
mV).

D. Read Operation
Read failure is the most critical problem in realizing
subthreshold memory as it imposes the smallest margin in
comparison with the hold and write margins.


EXPERIMENTAL RESULTS

Leakage current limited SRAM arrays (25616) is

validated in IBM 0.13 um, 8-metal CMOS process
technology. The threshold voltages of nMOS and pMOS
transistors are 355 mV and -325 mV, respectively. The
subthreshold slope factor is 82 mV/dec[7]. The channel
length of access transistors in proposed bitcell is set to 3X
of the minimum value (0.12 um) to utilize the Reverse Short
Channel Effect (RSCE) [8]. Thus write margin is improved.
The leakage current of leakage current limited 4 kb
SRAM array is 0.4 A for a supply voltage of 400 mV at
27. And the leakage current of proposed design is 64% of
that of ST (at 400 mV). It can be seen from Fig. 5, Vopt of
proposed SRAM array is scaled from 400 mV to 310 mV,
because of the reduced leakage current. Supply voltage of
proposed SRAM array can be scaled down with the rest of
the circuits when operating in the low-power mode.

9P9





5HDG610
P9






GFP9
9P9





(a)

(b)

Figure 5. Total energy of proposed SRAM array and referenced SRAM

array versus supply voltage.

IV.
(c)
Figure 4. The read margin of (a) 6T, (b) ST, and (c) proposed design (at
400 mV).

Consider the condition that Q = 0 and Q = 1.

During the read operation, the values stored in the nodes are
transferred to the bitlines. However, the bitline capacitance

CONCLUSION

In this paper, a leakage current limited SRAM bitcell

operating in subthreshold/nearthreshold is demonstrated in
IBM 0.13 m CMOS process. The unique characteristic of
proposed design is to have both stability and low leakage
characteristic in a bitcell. Its hold margin is 30.2% and
18.9% greater than conventional 6T SRAM and referenced
SRAM (at 400 mV). At the same time, its read margin is

V9-28

2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

45% and 9% greater than conventional 6T SRAM and

referenced SRAM (at 400 mV). Dynamic leakage cut off
transistor is utilized to reduce leakage current without
inducing any dynamic energy penalty. A leakage current
reduction of 36% is achieved as compared to referenced
SRAM design (at 400 mV 4kb). With proposed design, the
SRAM array optimum-energy supply voltage point of is
scaled from 400 mV to 310 mV. Thus, supply voltage of the
SRAM array can be scaled down with the rest of the circuits
when operating in the low-power mode.

[3]

[4]

[5]

[6]

REFERENCES
[1]

[2]

J.P. Kulkarni, K. Kim, K. Roy, A 160 mV Robust Schmitt Trigger

Based Subthreshold SRAM, IEEE Journal of Solid-State Circuits,
vol. 42, no. 10, pp. 2303 2313, 2007.
B. H. Calhoun and A. Chandrakasan, A 256 kb subthreshold SRAM
using 65 nm CMOS, in proceedings of IEEE International SolidState Circuits Conference, pp. 628629, 2006.

[7]
[8]

V9-29

N. Verma and A.P. Chandrakasan, A 256 kb 65nm 8T Subthreshold

SRAM Employing Sense-Amplifier Redundancy, IEEE Journal of
Solid-State Circuits, vol. 43, no. 1, pp.141 149, 2008.
T.H. Kim, J. Liu, J. Keane, and C.H. Kim, A High-Density
Subthreshold SRAM with Data-Independent Bitline Leakage and
Virtual Ground Replica Scheme, in proceedings of IEEE
International Solid-State Circuits Conference, pp.330 606, 2007.
T.H. Kim, J. Liu, and C.H. Kim, An 8T Subthreshold SRAM Cell
Utilizing Reverse Short Channel Effect for Write Margin and Read
Performance Improvement in proceedings of Custom Integrated
Circuits Conference, pp.241 244, 2007.
I.J. Chang, J.J. Kim, S.P. Park, and K. Roy, A 32kb 10T
Subthreshold SRAM Array with Bit-Interleaving and Differential
Read Scheme in 90nm CMOS, in proceedings of IEEE International
Solid-State Circuits Conference, pp. 388 622, 2008.
IBM CMOS 8RF (CMOS8RF) design Manual, Available from:
www.ibm.com
T.H. Kim, J. Liu, J. Keane, and C.H. Kim, A 0.2 V, 480 kb
Subthreshold SRAM With 1 k Cells Per Bitline for Ultra-LowVoltage Computing, IEEE Journal of Solid-State Circuits, vol.
43, no. 2, pp.518 529, 2008.

2011 29th IEEE VLSI Test Symposium

Calibrated high-efficiency testing and modelling methodologies for

concentrated multi-junction solar cells
Jeffrey F. Wheeldon, SUNLab, School of Information Technology and Engineering, University
of Ottawa, Ottawa, ON, Canada
Abstract
Multi-junction solar cells (MJSC) under concentrated sunlight are leading the way forward to record
efficiencies (>40%, under AM1.5D solar spectrums) and are ideally suited for utility-scale energy
production. MJSC are able to achieve high efficiencies by partitioning the solar spectrum using multiple
n-p junctions composed of III-V and IV group semiconductors. To reduce the cost associated with their
more complex epitaxial growth processes, MJSC are coupled with concentrator optics (typical
concentrations of 500-1000 suns). The University of Ottawa's SUNLab focuses its research on improving
the efficiency of concentrated high-efficiency MJSC systems. This research takes place in three crucial
areas: laboratory characterization of MJSC, numerical modelling of thermal, electrical and optical
properties of cells and systems, and outdoor testing of concentrated photovoltaic systems. Laboratory
testing of MJSC are performed under controlled temperature and spectral irradiance conditions, which
allow accurate efficiency comparisons of different solar cell designs. Flash and continuous solar
simulators illuminate the solar cells with intensities of up to 1000 suns. Flash solar simulators are the
standard method of characterizing solar cells under concentration, whereas testing under continuous
concentrated solar radiation provides critical additional information regarding thermal reliability (see Fig.
1(a)). MJSC are composed of subcells connected electrically in series, making their performance sensitive
to changes in the solar spectrum. Spectral sensitivity is quantified by performing quantum efficiency
measurements that determine a subcell's ability to extract an electron for a given incident photon. For
each experimental measurement, numerical modelling assists in determining the limiting elements within
a design. These numerical simulations cover the critical areas, such as the concentrating optics with
complex raying tracing, current-voltage characteristics of the MJSC using semiconductor device
simulators, and finite element method thermal models of the MJSC on carrier. Outdoor testing of a solar
system measures the efficiency under real-world conditions. The University of Ottawa's outdoor test site
has three solar tracking stations that measure the power output of a photovoltaic system while monitoring
the irradiance and spectrum of the sun, which determine overall system efficiencies (see Fig. 1(b)).

(b)

(a)

Figure 1: (a) Laboratory characterization of a MJSC under continuous concentrated solar radiation. (b)
University of Ottawa's solar tracker station, in which the performance of concentrated photovoltaic systems are
measured over long periods of time under real-world conditions.

978-1-61284-656-9/11/$26.00 2011 IEEE

209

Dynamics of Innovation in Solar Cell Industry: Divergence of Solar Cell Technologies

Yukihiko Nakata
Ritsumeikan Asia Pacific University
1-1, Jumonjibaru, Beppu, Oita 874-8577 Japan
nakata@apu.ac.jp
Abstract
The dynamics of innovation in solar cell industry was
researched based on the case studies of global solar cell
industries. What is the dynamics of innovations in the solar
cell industries? Japanese 1st group including Sharp, Kyocera,
Sanyo and Mitsubishi is mainly doing the business by crystal
silicon (c-Si) and amorphous Si (a-Si) solar cell. The
industrial cluster of Solar Island Kyusyu is focusing on thin
film solar cells. In USA, SunPower focuses on c-Si solar cell,
First Solar of 1st place is producing CdTe solar cell and
Nanosolar, Solyndra are developing CIGS. China including
Suntech mainly produced c-Si solar cell. Q-cells in Germany
expand from cell process to solar cell module and CIGS solar
cell. As the results of the case studies, the diverse
technologies of solar cell are taken for business by global
solar cell industry. Therefore, in case of the solar cell
industry, the innovation dynamics theory of Utterback can not
apply to the innovation. What are the conditions to accept the
diverse technologies such as the solar cell industries? To
answer the question, the process day of the solar cell
compared with LCD and semiconductor device. The main
process days for poly-Si and a-Si solar cell are less than 1
day. This means that solar cell technologies are simple
because they require shot process days to reduce the cost. The
conditions of diverse technologies in business are extracted
from the research results. First condition is simple technology,
which means short process day. Second condition is the
drastically increasing market. Third condition is high
motivation to enter the industry. Under the above conditions,
many companies want to enter the business even if their
technology level. In other words, many companies enter the
business using their adaptable technologies. As the results, the
diverse technologies are accepted for business.
I. Introduction
For environmental sustainability to be attained, the steady
increase in global greenhouse gas (GHG) emissions must be
reversed early in this century. One way to reduce GHG
emissions is to use sustainable energies, such as solar, wind,
hydro, biomass, and geothermal energies.
Japan implemented several incentive programs since it
began researching and developing solar cells in 1955 and first
applied them to a lighthouse. The "Sunshine Project" was
launched in July 1974 following the first oil crisis in 1973 to
promote R&D of solar cells. Japanese companies are
researching and developing many kind solar cell technologies.
As a result, Japanese companies became world leaders in the
production of solar cells [1]. But, its share rapidly fell down
recently [2], [3].
The U.S. share fell continuously from 40% of 1997 to 7%
of 2007. However, it increased to 13% of 2009 [2], [3]. Many
978-1-61284-952-2/11/$26.00 2011 IEEE

solar cell venture companies, such as SunPower, First Solar,

Nanosolar and Solyndra etc, entered using diverse solar cell
technologies in Silicon Valley and other area.
Several Chinese companies, such as SunTech, entered to
solar cells business by using many available technologies and
drastically increased its production share [2], [3].
Many European countries have introduced a feed-in tariff
(FiT) system, in which electricity produced by using
sustainable energy sources is bought by the electric utilities at
above market prices. The number of solar cell companies,
such as Q-Cells, emerged and increased its production share
[2], [3].
Utterback described about the dynamics of innovation as
follows [4]. Products move from high variety, to dominant
design, to incremental innovation on standardized. But,
according to the above cases, many solar cell companies
accepted the diverse technologies for business.
What is the dynamics of innovations in the solar cell
industries? What are the conditions to accept the diverse
technologies such as the solar cell industries?
However, there has been a lack of research on the
dynamics of innovation in solar cell industry.
In this paper, the dynamics of innovation in solar cell
industry was researched using the case studies of global solar
cell industries.
II. Literature Review
Utterback described about the dynamics of innovation as
follows [4]. The model consists that the rate of major
innovation for both products and follows a general pattern
over time, and that product and process innovation share an
important relationship. The rate of product innovation in an
industry is highest in fluid phase during its formative years.
The rate of major product innovation slows down and the rate
of major process innovation speeds up in transition phase
and may emerge dominant design. Then, the rate of major
innovation dwindles in specific phase.
Also he described the patterns in the emergence of radical
innovations [4]. It is researched that how a radical
technological innovation can emerge and successfully invade
the established technology.
The solar cell business begun from 1950s and is growing
up rapidly. Many companies entered to solar cell business by
using diverse technologies. In this paper, the applicability of
this theory to the solar cell industry was verified.
III. Methodology
In this paper, the dynamics of innovation in global solar
cell industries was researched.
The methodology is shown in Fig. 1. The technologies of
global solar cell industries are researched by the case study

1042

IEEE Int'l Technology Management Conference

method based on the theory of the dynamics of innovation [4]

and The S-curve [5], [6]. The case studies were conducted by
interview, attending conferences and exhibitions etc. at Japan
and Stanford University.1 The cases included Sharp in Japan,
SunPower, First Solar, Nanosolar and Solyndra in US,
Suntech in China and Q-Cells in Germany. Then, the
dynamics of innovation in solar cell industry was analyzed
Dynamics of Innovation
In Solar Cell Industry

Divergence of Technologies

Case Studies
Japan
Sharp

USA
SunPower
First Solar
Nanosolar
Solyndra

China
Suntech

Theories
Dynamics of
Innovation
The S-curve

high point of about 50% in 2004 [1]. But, it fell to 14 % in

2009 [3]. The U.S. share fell continuously from 40% of 1997
to 7% of 2007. However, it increased to 13% of 2009
including overseas production [3]. Germany increased the
share by introducing feed-in tariff (FiT) system, in which
electricity produced by using sustainable energy sources is
bought by the electric utilities at above market prices. Then
Germany kept its share of 14 % in 2009 [3]. China drastically
increased its share from 3% in 2004 to 36% in 2009 [3].
The paradigm in solar cell competition has been shifted
from 2004. In this explosive situation, the dynamics of in
solar cell industry was researched

Germany
Q-Cells

Background

Paradigm Shift
in Solar Cell Competition

Fig. 1. Methodology to analyze dynamics of innovation in

solar cell industry.
IV. Paradigm Shift in Solar Cell Competition
The competitive climate in solar cell industry was
researched to analyze the dynamics of innovation.
A. Solar Cell Production Amount
The solar cell production amount worldwide is drastically
increasing as shown in Fig. 2 [2], [3].

Fig. 2. Solar cell production amount worldwide

(Author made from [2], [3])
B. National Share of Solar Cell Production
Figure 3 plots the changes in national share of solar cell
production from 1997 to 2009 [3]. Japans share reached a
1 The author researched and developed the solar cell
technologies at Sharp Corporation for 20 years and researched
USA solar cell industry at Stanford University as Visiting
Professor, from October 2009 to March 2010. This paper was
written based on these experiences and research studies.

Fig. 3. National share of solar cell production

(Author made from [2], [3])
V. Case Studies of Global Solar Cell Industries
A. Japanese Solar Cell Industry
1) Sharp Corporation
Sharp had the highest production share of solar cells in the
world until 2006. Initially, single crystal silicon (c-Si) wafers
and poly silicon (poly-Si) wafers were purchased, and solar
cells and modules were produced. Later on, Sharp began to
expand to poly-Si wafers and poly-Si ingots for higher
vertical integration. However, Sharp had trouble maintaining
a stable supply of poly-Si raw material. Consequently, Sharp
dropped from 1st in 2007 and 4th in 2009 in terms of
production [3]. It recently expanded the business to poly-Si
raw material.
Sharp built a 21st century industrial complex for
stacked type amorphous silicon (a-Si) solar cells on a 1,000
MW/year scale with an LCD production line for the 10th
generation [7]. Sharp presented and demonstrated the utilityscale stacked type a-Si solar cells [8], [9]. Also, Sharp started
the operations at its a-Si solar cells plant on March 29, 2010
in Sakai City, Osaka Prefecture, Japan [10].
B. US Solar Cell Industry
1) SunPower
SunPower has been developing world record solar
technology since the 1970s. SunPower is one of the global
leaders in developing high-efficiency solar cells [11].
SunPowers solar innovation began nearly four decades
ago, when Richard Swanson was pondering ways to deal with
the oil crisis of the early 1970s. Swanson, who by then was a
professor of electrical engineering at Stanford, found
SunPower with the help of funds.

Richard Swanson mentioned I was working on his

doctorate in engineering at Stanford University. And, I joined
faculty of Stanford University and researched back surface
point-contact solar cell. I got the conversion efficiency of
28% at 500 suns and 23% of large area solar cell. Then, I
found SunPower. In 1991, I resigned from the professor of
Stanford University to devote full time to SunPower
Corporation. This is difficult decision. But, this is a good idea
to put the high-efficiency solar cell to commercial scale.
Then, I decided to devote.2
In 2008, SunPower signed an agreement with PG&E to
build the worlds largest 250 MW photovoltaic (PV)
power plant, set to begin energy delivery in 2010. Florida
Power & Light Co. also contracted to provide 25 MW of
electricity in 2009, and another 10 MW by the end of 2010
[12].
SunPower announced that, under a strategic supply
agreement, Toshiba will order 32 megawatts of highefficiency solar panels from SunPower in 2010 [13].
2) First Solar
First Solar was formed in 1999 and launched production
of commercial products in 2002 [14]. First Solar claims as
follows. Cadmium Telluride (CdTe) is a suitable
semiconductor, which enables it to convert solar energy into
electricity more efficiently [15]. First Solar claim the
manufacturing process transforms a piece of glass into a
complete solar module is less than 2.5 hours [14]. It
drastically expands the production amount at 1.5 times a year
and got 1st position in 2009 [3]. The production capacity
reached to 1228 MW in 2009 [3].
First Solar is the first PV maker to produce modules <
$1/watt ($0.85/watt in 3Q 2009). Conversion efficiencies of
the modules reached 11.1% in the 4th quarter of 2009 and the
module manufacturing cost was US$0.84 per watt. Production
cost per watt was lowered to US$0.80 per watt [15].
There are two weaknesses in First Solar. First, there is the
limitation of material of Cd and Te. Second, the toxicity of
Cd. Japan has stopped R&D of solar cell with Cd more than
10 years ago.
Benny Buller said First Solar is recycling solar modules
that enable substantially all components of the module,
including the glass and the encapsulated semiconductor
material, to be treated and processed into new modules or
other products. By implementing collection and recycling
programs, First Solar is preventing the modules from entering
the waste stream of local communities. 3
But, actual implementation in future is remains the most
crucial task to continue this business.
3) Nanosolar
Nanosolar has unique technology to produce cupper
indium gallium diselenide (CIGS) thin film solar cell using

2 The author interviewed Dr. Richard Swanson at headquarter

of SunPower in December 4th, 2009
3 The author interviewed Benny Buller, Director, Device
Improvement of First Solar in December 1st, 2009 at Hotel
Nikko San Francisco.

roll-to-roll low-cost nanoparticle ink printing process [16].

Nanoparticle ink is made by nanoparticles of 20nm in size. It
is coated onto a metal foil using coating/printing techniques
that work in normal atmosphere.
Nanosolar established manufacturing facility in San Jose,
California, as its U.S. site and a facility in the Berlin capital
region as its European site in 2006.
Martin Roscheisen said CIGS nanoparticle ink print on
aluminum foil to make solar cell. It has promising potential.4
4) Solyndra
Solyndra was founded by Christian Gronet in 2005 and
establish headquarters in Fremont, California in 2006 and
begin the production in 2007 [17].
Solyndra designs and manufactures cylindrical modules
incorporating CIGS thin-film technology. The panels employ
cylindrical modules which capture sunlight across a 360degree photovoltaic surface capable of converting direct,
diffuse and reflected sunlight into electricity. Solyndra claims
light weight and simple installation means fast projects and
low cost. The light weight panels install without penetrations
or array grounding, making this the easiest and fastest-toinstall rooftop solar system yet.
C. Chinese Solar Cell Industry
China became the 1st leading source of solar cells,
surpassing Japan and Germany, from 2008 [3]. Many Chinese
companies producing them are listed on overseas stock
markets [3]. Suntech Power, for example, is the worlds 2nd
leading manufacturer in 2009 [3]. Chinese companies focus
on c-Si solar cell. China has a large potential market for solar
cells, and autonomous businesses may emerge, from ones
producing poly-Si raw materials to ones marketing solar cell
systems.
1) Suntech Power
Zhengrong Shi took PhD in electrical engineering from
the University of New South Wales in Australia in
photovoltaic technology. He returned to China in 2001 to set
up his solar power company - Suntech Power found [18].
Suntech Power mainly focus on c-Si solar cell and rapidly
grown up to 2nd production share in 2009 [3]. Headquarters
is located at Wuxi, Jiangsu province, China. It has four
production sites at Wuxi, Luoyang, Qinghai and Shanghai and
8000 global employees.
Shi presented Suntech Power strategy [19]. Suntech
Power mainly focuses on c-Si solar cell and has researched
and developed solar cell with high conversion efficiencies of
19% of single c-Si and 17% of poly-Si, which is named
Pluto technology. Suntech Power is routinely utilizing the
Pluto technology to produce solar cell.
D. Germany Solar Cell Industry
1) Q-Cells
Q-Cells of Germany was established in 1999, and started
the production of the solar cell from 2001 [20]. The Europe
solar cell market was quickly expanded by the incentive
policies of the European governments. Q-Cells was growing
rapidly and took the 1st place of the production share of the
The author interviewed Martin Roscheisen, former CEO of
Nanoslar, at headquarter of Nanoslar in March 12, 2010.
4

Technology performance parameter

solar cell in 2007 and 2008 [3]. Q-Cells focused on the c-Si
and poly-Si solar cell process by the automation to reduce the
cost. However, Q-Cells got into red in 2008 and fell out of
first place by First Solar. Therefore, Q-Cells restructured the
solar cell business and expand from c-Si and poly-Si cell
process to solar cell module and CIGS solar cell [3].
E. Summary of Global Solar Cell Industries
Based on the above case studies of global solar cell
industry, the diverse technologies of solar cell are classified
according to the S-curve [6], [7], as shown in Fig. 4. c-Si and
poly-Si solar cells are mature technologies. A-Si thin film
solar is near maturity. Cupper gallium diselenide (CGS),
CIGS and CdTe solar cells are in growth stage. The
embryonic technologies are organic, spherical Si, quantum
dot and many embryonic solar cells, which many venture
companies are researching.
New
Invention
Period

Technology
Improvement
Period

Maturity
poly-Si
a-Si TF SC

Growth
CGS, CIGS

Aging

c-Si

MatureTechnology
Period

CdTe SC

Embryonic

Quantum dot

Spherical Si SC

Organic SC

Embryonic SC

Time

Fig. 4. Divergence of Solar Cell Technologies

on the S-curve (Author made)
Table 1. Technologies of global solar cell industries.
(Author made)
New Invention Period
Embryonic Stage
Quantum dot Organic SC
Spherical Si

Japan

USA

Technology Improvement Period Mature-technology Period

Growth Stage
Materiality Stage
CGS, CIGS a-Si TF-SC poly-Si SC
c-Si SC

CdTe SC

1st Group
(Sharp, Kyocera,
Sanyo,
Mit bi hi)
Solar Island
Kyusyu
SolarPower

Nanosolar
Solyndra
First Solar
Venture

China

Suntech
Other Compalies

Germany Q-cells
Production Level

R&D Level

Based on this categorizing, the technology progress level

was analyzed according to the theory of the S-curve as sown
in Table 1.
Japanese 1st group including Sharp, Kyocera, Sanyo and
Mitsubishi is mainly doing the business by c-Si and a-Si solar

cell. Industrial cluster of Solar Island Kyusyu, including

Fuji Electric Systems, Honda Soltec, Mitsubishi Heavy
Industries and Solar Frontier, is focusing on thin film solar
cells.
In USA, SunPower focuses on c-Si solar cell, First Solar of
1st place is producing CdTe solar cell and Nanosolar,
Solyndra are developing CIGS. In addition, many venture
companies are challenging many emerging technologies using
many materials and technologies.
China including Suntech mainly produced c-Si solar cell.
Q-cells restructured the solar cell business and expand from
cell process to solar cell module and CIGS solar cell [3].
As the result, the divers technologies are took for business
by global solar cell industry.
VI. Discussion
A. Dynamics of Innovations in Solar Cell Industries
From the summary of global solar cell industries, what is
the dynamics of innovations in solar cell industries?
Utterback described that the rate of product innovation in
an industry is highest, which means high variety products put
in the market in fluid phase [4].
However, solar cell business began from 1950s and run
beyond 50 years and some solar cell technologies, such as cSi and poly-Si solar cells, have moved into specific phase.
However, many companies are using the diverse technologies
in solar cell business. This means that the diverse
technologies in fluid phase, transition phase and specific
phase are simultaneously accepted for solar cell business.
Also, he described the patterns in the emergence of radical
innovations [4]. It is researched that how a radical
technological innovation can emerge and successfully invade
the established technology. He said that at the time an
invading technology first appears, the established technology
offers better performance or cost than dose the challenger,
which is still unperfected.
In case of flat television, LCD television had the share of
81% and plasma display pale television (PDP-TV) had 10%
in 2007 [21]. Two technologies are accepted the market. PDPTV is mainly producing by Panasonics depending on the
progress of research and development and investment.
However, only two technologies are surviving in flat
television and these numbers of technologies are deferent
from them of solar cell industry.
Christensen described about new read-write head
technologies [22]. The ferrite heads improved between 1976
and 1989, and then it was substituted with the thin-film heads.
Also, the thin-film heads was substituted with magnetoresistive heads. This means two technologies were accepted at
the same time.
In the solar cell industries, however, many technologies are
competing in the real business.
Therefore, in case of the solar cell industry, the innovation
dynamics theory of Utterback can not apply to the innovation.
B. Process days of Solar Cell compared with LCD and
Semiconductor
What are the conditions to accept the diverse technologies
in business such as the solar cell industries?
To answer the question, the process day and unit area
value of the solar cell compared with Liquid Crystal Display

(LCD) and semiconductor device, as shown in Fig. 5.

Most important requirement for solar cell is to reduce the
cost per generate electricity to compete with the other energy
sources, such as thermal, hydro and nuclear power generation.
Therefore, short process day and low unit area value of the
solar cell are required to reduce the cost.
The main process days for poly-Si and a-Si solar cell are
less than 1 day. But, the process day for an LCD is 4-5 days
and that for a semiconductor is 3060 days depending on the
products.
For this reason, unit area value is 6 cent /cm2 of poly-Si
solar cell and 5 cent/cm2 of an a-Si solar cell, comparing to
LCD of 7 cent/cm2 and 1200 cent/cm2 of semiconductor.
Especially, unit area value of LCD is rapidly reducing
because the LCD television market drastically increasing.
This means that diverse technologies of solar cell are
simple because they required shot process days and low unit
area value to reduce the cost.
1 0000

Unit Area Value (cent/cm2)

Semiconductor
1000

100

poly-Si
Solr Cell
10

a-Si
Solar Cell

LCD

1
1

10

100

Process Days (Day

Fig. 5. Process days and unit area value of Sola Cell, LCD
and semiconductor devices. (Author made)
C. Conditions of divers technologies
Utterback researched the dynamics of innovation from the
case studies of plate glass, assemble products, film
photography, ice etc [4]. However, as described above, in
case of the solar cell industry, the innovation dynamics theory
of Utterback can not apply to the innovation.
What are the conditions to accept the diverse technologies
such as the solar cell industries?
Therefore, the conditions of diverse technologies in
business are extracted from the results of the paradigm shift
of solar cell competition, case studies of global solar cell
industries.
First condition of diverse technologies is simple
technology, which means short process day, because it is easy
to enter the business by low input resources and time. In other
word, simple process accepts the diverged technologies to
enter the business. In the contrast, complex process needs to
integrate many technologies, not one technology, in case of
long process day, such as semiconductor. The industries with
complex processes not accept diverse technologies to simplify

the total processes and reduce the research and development

effort.
Second condition is the drastically increasing market to
enhance the entrance. As described above, solar cell market
drastically increasing for environmental sustainability
Third condition is high motivation to enter the industry. As
explained above, the solar cell markets are rapidly increasing
and many companies want to enter the solar cell business
even if their technology level.
Under the above conditions, many companies want to enter
the business even if their technology level. In other words,
many companies enter the business using their adaptable
technologies. As the results, the diverse technologies are
accepted for the business.
VIII. Conclusions
The dynamics of innovation in solar cell industry was
researched based on the case studies of global solar cell
industries.
What is the dynamics of innovations in the solar cell
industries?
Japanese 1st group including Sharp, Kyocera, Sanyo and
Mitsubishi is mainly doing the business by c-Si and a-Si solar
cell. The industrial cluster of Solar Island Kyusyu is
focusing on thin film solar cells. In USA, SunPower focuses
on c-Si solar cell, First Solar of 1st place is producing CdTe
solar cell and Nanosolar, Solyndra are developing CIGS. In
addition, many venture companies are challenging many
emerging technologies using many materials and
technologies. China including Suntech mainly produced c-Si
solar cell. Q-cells in Germany restructured the solar cell
business and expand from c-Si and poly-Si cell process to
solar cell module and CIGS solar cell
As the results of the case studies of the global solar cell
industries, the diverse technologies of solar cell are taken for
business by the global solar cell industry.
Therefore, in case of the solar cell industry, the innovation
dynamics theory of Utterback can not apply to the innovation.
What are the conditions to accept the diverse technologies
in business such as the solar cell industries?
To answer the question, the process day of the solar cell
compared with LCD and semiconductor device. The main
process days for poly-Si and a-Si solar cell are less than 1
day. This means that solar cell technologies are simple
because they required shot process days and low unit area
value to reduce the cost.
The conditions of diverse technologies in business are
extracted from the research results. First condition is simple
technology, which means short process day, because it is easy
to enter the business by low input resources and time. Second
condition is the drastically increasing market. Third condition
is high motivation to enter the industry.
Under the above conditions, many companies want to
enter the business even if their technology level. In other
words, many companies enter the business using their
adaptable technologies. As the results, the diverse
technologies are accepted for business.

Acknowledgments
I express my appreciation for Professor Richard B. Dasher
of U.S.-Asia Technology Management Center of Stanford
University to give me the opportunity to research at Stanford
University. I thank Japan Society for the Promotion of
Science for the financial support.
References
1. New Energy and Industrial Technology Development
Organization (NEDO), "Why did Japan achieve the
world's No. 1 share in solar cells? (Japanese),"
Nedobooks, 2007.
2. PV News, May and April 2008.
3. Outlook of Solar Cell Industry 2011, Sangyo Times
2010.
4. James M. Utterback, Mastering the Dynamics of
Innovation, Harvard Business School Press 1994.
5. J.C. Fisher and R. H. Pry Simple Substitution Model of
Technical Change. Technological Forecasting and
Social Change, vol. 3, pp. 75 88.
6. Tarek Khalil, Chapter 5 Technology Life Cycles,
Management of Technology, McGraw-Hill Higher
Education, (2000), pp80-94.
7. Press Announcement of Sharp Corporation, Sharp to
Develop Manufacturing Complex for the 21st Century,
http://sharp-world.com/corporate/report/plant/index.html,
31 July 2007.
8. Paul Wormser, Thin-Film Companies Heading for
Terawatt-hour Module Manufacturing: Micromorph,
presented at PHOTONs PV Thin Film Conference,
February 2, 2010, San Francisco.
9. Sharp Corporation, Utility-Scale Thin Film
Photovoltaics, exhibited at Solar Power International 09
Exhibition, Oct 27 29, 2009, Anaheim, California, USA.
10. Sharp Corporation, Operation Starts at Solar Cell Plant
in GREEN FRONT SAKAI, press-released in March 29,
2010.
11. SunPower, [Online]. Available:
http://us.sunpowercorp.com/business/
12. SunPower, PG&E Signs 800 MW Photovoltaic Solar
Power Agreements, [Online]. Available:
http://investors.sunpowercorp.com/releasedetail.cfm?Rel
easeID=328221

13. SunPower, 32-Megawatt Solar Power Supply

Agreement With Toshiba, [Online]. Available:
http://investors.sunpowercorp.com/releasedetail.cfm?Rel
easeID=447820
14. First Solar, [Online]. Available:
http://www.firstsolar.com/
15. PV Tech First Solar takes 2009 revenue past US$2
billion, published at 18 February 2010, [Online].
Available: www.pv-tech.org, http://www.pvtech.org/lib/printable/8181/
16. Nanosolar, (2011, January 9.) [Online]. Available:
http://www.nanosolar.com/
17. Solyndra (2011, January 9.) [Online]. Available:
http://www.solyndra.com/
18. Suntech Power Holdings Co., Ltd.2008 Corporate
Report, [Online]. Available:
http://phx.corporateir.net/External.File?item=UGFyZW50SUQ9MzQzMDg
4fENoaWxkSUQ9MzMwMjA2fFR5cGU9MQ==&t=1
19. Zhengrong Shi, Suntech, CEO Roundtable Discussion:
Developing the Utility Market for Solar Electricity,
presented at Solar Power International 09 Conference,
Oct 27 29, 2009, Anaheim, California, USA.
20. Q-Cells (2011, January 9.) History [Online]. Available:
http://www.q-cells.com/en/company/
qcells_se/history_and_facts/index.html
21. Hisakazu Torii , The 14th DisplaySearch Japan Forum
(Jan 30-31, 2008)
22. Clayton M. Christensen, The Innovators Dilemma,
Boston, Massachusetts, Harvard Business School Press,
1997.

Dynamics of Innovation in Solar Cell Industry: Divergence of Solar Cell Technologies

Yukihiko Nakata
Ritsumeikan Asia Pacific University
1-1, Jumonjibaru, Beppu, Oita 874-8577 Japan
nakata@apu.ac.jp
Abstract
The dynamics of innovation in solar cell industry was
researched based on the case studies of global solar cell
industries. What is the dynamics of innovations in the solar
cell industries? Japanese 1st group including Sharp, Kyocera,
Sanyo and Mitsubishi is mainly doing the business by crystal
silicon (c-Si) and amorphous Si (a-Si) solar cell. The
industrial cluster of Solar Island Kyusyu is focusing on thin
film solar cells. In USA, SunPower focuses on c-Si solar cell,
First Solar of 1st place is producing CdTe solar cell and
Nanosolar, Solyndra are developing CIGS. China including
Suntech mainly produced c-Si solar cell. Q-cells in Germany
expand from cell process to solar cell module and CIGS solar
cell. As the results of the case studies, the diverse
technologies of solar cell are taken for business by global
solar cell industry. Therefore, in case of the solar cell
industry, the innovation dynamics theory of Utterback can not
apply to the innovation. What are the conditions to accept the
diverse technologies such as the solar cell industries? To
answer the question, the process day of the solar cell
compared with LCD and semiconductor device. The main
process days for poly-Si and a-Si solar cell are less than 1
day. This means that solar cell technologies are simple
because they require shot process days to reduce the cost. The
conditions of diverse technologies in business are extracted
from the research results. First condition is simple technology,
which means short process day. Second condition is the
drastically increasing market. Third condition is high
motivation to enter the industry. Under the above conditions,
many companies want to enter the business even if their
technology level. In other words, many companies enter the
business using their adaptable technologies. As the results, the
diverse technologies are accepted for business.
I. Introduction
For environmental sustainability to be attained, the steady
increase in global greenhouse gas (GHG) emissions must be
reversed early in this century. One way to reduce GHG
emissions is to use sustainable energies, such as solar, wind,
hydro, biomass, and geothermal energies.
Japan implemented several incentive programs since it
began researching and developing solar cells in 1955 and first
applied them to a lighthouse. The "Sunshine Project" was
launched in July 1974 following the first oil crisis in 1973 to
promote R&D of solar cells. Japanese companies are
researching and developing many kind solar cell technologies.
As a result, Japanese companies became world leaders in the
production of solar cells [1]. But, its share rapidly fell down
recently [2], [3].
The U.S. share fell continuously from 40% of 1997 to 7%
of 2007. However, it increased to 13% of 2009 [2], [3]. Many
978-1-61284-952-2/11/$26.00 2011 IEEE

solar cell venture companies, such as SunPower, First Solar,

Nanosolar and Solyndra etc, entered using diverse solar cell
technologies in Silicon Valley and other area.
Several Chinese companies, such as SunTech, entered to
solar cells business by using many available technologies and
drastically increased its production share [2], [3].
Many European countries have introduced a feed-in tariff
(FiT) system, in which electricity produced by using
sustainable energy sources is bought by the electric utilities at
above market prices. The number of solar cell companies,
such as Q-Cells, emerged and increased its production share
[2], [3].
Utterback described about the dynamics of innovation as
follows [4]. Products move from high variety, to dominant
design, to incremental innovation on standardized. But,
according to the above cases, many solar cell companies
accepted the diverse technologies for business.
What is the dynamics of innovations in the solar cell
industries? What are the conditions to accept the diverse
technologies such as the solar cell industries?
However, there has been a lack of research on the
dynamics of innovation in solar cell industry.
In this paper, the dynamics of innovation in solar cell
industry was researched using the case studies of global solar
cell industries.
II. Literature Review
Utterback described about the dynamics of innovation as
follows [4]. The model consists that the rate of major
innovation for both products and follows a general pattern
over time, and that product and process innovation share an
important relationship. The rate of product innovation in an
industry is highest in fluid phase during its formative years.
The rate of major product innovation slows down and the rate
of major process innovation speeds up in transition phase
and may emerge dominant design. Then, the rate of major
innovation dwindles in specific phase.
Also he described the patterns in the emergence of radical
innovations [4]. It is researched that how a radical
technological innovation can emerge and successfully invade
the established technology.
The solar cell business begun from 1950s and is growing
up rapidly. Many companies entered to solar cell business by
using diverse technologies. In this paper, the applicability of
this theory to the solar cell industry was verified.
III. Methodology
In this paper, the dynamics of innovation in global solar
cell industries was researched.
The methodology is shown in Fig. 1. The technologies of
global solar cell industries are researched by the case study

1042

IEEE Int'l Technology Management Conference

method based on the theory of the dynamics of innovation [4]

and The S-curve [5], [6]. The case studies were conducted by
interview, attending conferences and exhibitions etc. at Japan
and Stanford University.1 The cases included Sharp in Japan,
SunPower, First Solar, Nanosolar and Solyndra in US,
Suntech in China and Q-Cells in Germany. Then, the
dynamics of innovation in solar cell industry was analyzed
Dynamics of Innovation
In Solar Cell Industry

Divergence of Technologies

Case Studies
Japan
Sharp

USA
SunPower
First Solar
Nanosolar
Solyndra

China
Suntech

Theories
Dynamics of
Innovation
The S-curve

high point of about 50% in 2004 [1]. But, it fell to 14 % in

2009 [3]. The U.S. share fell continuously from 40% of 1997
to 7% of 2007. However, it increased to 13% of 2009
including overseas production [3]. Germany increased the
share by introducing feed-in tariff (FiT) system, in which
electricity produced by using sustainable energy sources is
bought by the electric utilities at above market prices. Then
Germany kept its share of 14 % in 2009 [3]. China drastically
increased its share from 3% in 2004 to 36% in 2009 [3].
The paradigm in solar cell competition has been shifted
from 2004. In this explosive situation, the dynamics of in
solar cell industry was researched

Germany
Q-Cells

Background

Paradigm Shift
in Solar Cell Competition

Fig. 1. Methodology to analyze dynamics of innovation in

solar cell industry.
IV. Paradigm Shift in Solar Cell Competition
The competitive climate in solar cell industry was
researched to analyze the dynamics of innovation.
A. Solar Cell Production Amount
The solar cell production amount worldwide is drastically
increasing as shown in Fig. 2 [2], [3].

Fig. 2. Solar cell production amount worldwide

(Author made from [2], [3])
B. National Share of Solar Cell Production
Figure 3 plots the changes in national share of solar cell
production from 1997 to 2009 [3]. Japans share reached a
1 The author researched and developed the solar cell
technologies at Sharp Corporation for 20 years and researched
USA solar cell industry at Stanford University as Visiting
Professor, from October 2009 to March 2010. This paper was
written based on these experiences and research studies.

Fig. 3. National share of solar cell production

(Author made from [2], [3])
V. Case Studies of Global Solar Cell Industries
A. Japanese Solar Cell Industry
1) Sharp Corporation
Sharp had the highest production share of solar cells in the
world until 2006. Initially, single crystal silicon (c-Si) wafers
and poly silicon (poly-Si) wafers were purchased, and solar
cells and modules were produced. Later on, Sharp began to
expand to poly-Si wafers and poly-Si ingots for higher
vertical integration. However, Sharp had trouble maintaining
a stable supply of poly-Si raw material. Consequently, Sharp
dropped from 1st in 2007 and 4th in 2009 in terms of
production [3]. It recently expanded the business to poly-Si
raw material.
Sharp built a 21st century industrial complex for
stacked type amorphous silicon (a-Si) solar cells on a 1,000
MW/year scale with an LCD production line for the 10th
generation [7]. Sharp presented and demonstrated the utilityscale stacked type a-Si solar cells [8], [9]. Also, Sharp started
the operations at its a-Si solar cells plant on March 29, 2010
in Sakai City, Osaka Prefecture, Japan [10].
B. US Solar Cell Industry
1) SunPower
SunPower has been developing world record solar
technology since the 1970s. SunPower is one of the global
leaders in developing high-efficiency solar cells [11].
SunPowers solar innovation began nearly four decades
ago, when Richard Swanson was pondering ways to deal with
the oil crisis of the early 1970s. Swanson, who by then was a
professor of electrical engineering at Stanford, found
SunPower with the help of funds.

Richard Swanson mentioned I was working on his

doctorate in engineering at Stanford University. And, I joined
faculty of Stanford University and researched back surface
point-contact solar cell. I got the conversion efficiency of
28% at 500 suns and 23% of large area solar cell. Then, I
found SunPower. In 1991, I resigned from the professor of
Stanford University to devote full time to SunPower
Corporation. This is difficult decision. But, this is a good idea
to put the high-efficiency solar cell to commercial scale.
Then, I decided to devote.2
In 2008, SunPower signed an agreement with PG&E to
build the worlds largest 250 MW photovoltaic (PV)
power plant, set to begin energy delivery in 2010. Florida
Power & Light Co. also contracted to provide 25 MW of
electricity in 2009, and another 10 MW by the end of 2010
[12].
SunPower announced that, under a strategic supply
agreement, Toshiba will order 32 megawatts of highefficiency solar panels from SunPower in 2010 [13].
2) First Solar
First Solar was formed in 1999 and launched production
of commercial products in 2002 [14]. First Solar claims as
follows. Cadmium Telluride (CdTe) is a suitable
semiconductor, which enables it to convert solar energy into
electricity more efficiently [15]. First Solar claim the
manufacturing process transforms a piece of glass into a
complete solar module is less than 2.5 hours [14]. It
drastically expands the production amount at 1.5 times a year
and got 1st position in 2009 [3]. The production capacity
reached to 1228 MW in 2009 [3].
First Solar is the first PV maker to produce modules <
$1/watt ($0.85/watt in 3Q 2009). Conversion efficiencies of
the modules reached 11.1% in the 4th quarter of 2009 and the
module manufacturing cost was US$0.84 per watt. Production
cost per watt was lowered to US$0.80 per watt [15].
There are two weaknesses in First Solar. First, there is the
limitation of material of Cd and Te. Second, the toxicity of
Cd. Japan has stopped R&D of solar cell with Cd more than
10 years ago.
Benny Buller said First Solar is recycling solar modules
that enable substantially all components of the module,
including the glass and the encapsulated semiconductor
material, to be treated and processed into new modules or
other products. By implementing collection and recycling
programs, First Solar is preventing the modules from entering
the waste stream of local communities. 3
But, actual implementation in future is remains the most
crucial task to continue this business.
3) Nanosolar
Nanosolar has unique technology to produce cupper
indium gallium diselenide (CIGS) thin film solar cell using

2 The author interviewed Dr. Richard Swanson at headquarter

of SunPower in December 4th, 2009
3 The author interviewed Benny Buller, Director, Device
Improvement of First Solar in December 1st, 2009 at Hotel
Nikko San Francisco.

roll-to-roll low-cost nanoparticle ink printing process [16].

Nanoparticle ink is made by nanoparticles of 20nm in size. It
is coated onto a metal foil using coating/printing techniques
that work in normal atmosphere.
Nanosolar established manufacturing facility in San Jose,
California, as its U.S. site and a facility in the Berlin capital
region as its European site in 2006.
Martin Roscheisen said CIGS nanoparticle ink print on
aluminum foil to make solar cell. It has promising potential.4
4) Solyndra
Solyndra was founded by Christian Gronet in 2005 and
establish headquarters in Fremont, California in 2006 and
begin the production in 2007 [17].
Solyndra designs and manufactures cylindrical modules
incorporating CIGS thin-film technology. The panels employ
cylindrical modules which capture sunlight across a 360degree photovoltaic surface capable of converting direct,
diffuse and reflected sunlight into electricity. Solyndra claims
light weight and simple installation means fast projects and
low cost. The light weight panels install without penetrations
or array grounding, making this the easiest and fastest-toinstall rooftop solar system yet.
C. Chinese Solar Cell Industry
China became the 1st leading source of solar cells,
surpassing Japan and Germany, from 2008 [3]. Many Chinese
companies producing them are listed on overseas stock
markets [3]. Suntech Power, for example, is the worlds 2nd
leading manufacturer in 2009 [3]. Chinese companies focus
on c-Si solar cell. China has a large potential market for solar
cells, and autonomous businesses may emerge, from ones
producing poly-Si raw materials to ones marketing solar cell
systems.
1) Suntech Power
Zhengrong Shi took PhD in electrical engineering from
the University of New South Wales in Australia in
photovoltaic technology. He returned to China in 2001 to set
up his solar power company - Suntech Power found [18].
Suntech Power mainly focus on c-Si solar cell and rapidly
grown up to 2nd production share in 2009 [3]. Headquarters
is located at Wuxi, Jiangsu province, China. It has four
production sites at Wuxi, Luoyang, Qinghai and Shanghai and
8000 global employees.
Shi presented Suntech Power strategy [19]. Suntech
Power mainly focuses on c-Si solar cell and has researched
and developed solar cell with high conversion efficiencies of
19% of single c-Si and 17% of poly-Si, which is named
Pluto technology. Suntech Power is routinely utilizing the
Pluto technology to produce solar cell.
D. Germany Solar Cell Industry
1) Q-Cells
Q-Cells of Germany was established in 1999, and started
the production of the solar cell from 2001 [20]. The Europe
solar cell market was quickly expanded by the incentive
policies of the European governments. Q-Cells was growing
rapidly and took the 1st place of the production share of the
The author interviewed Martin Roscheisen, former CEO of
Nanoslar, at headquarter of Nanoslar in March 12, 2010.
4

Technology performance parameter

solar cell in 2007 and 2008 [3]. Q-Cells focused on the c-Si
and poly-Si solar cell process by the automation to reduce the
cost. However, Q-Cells got into red in 2008 and fell out of
first place by First Solar. Therefore, Q-Cells restructured the
solar cell business and expand from c-Si and poly-Si cell
process to solar cell module and CIGS solar cell [3].
E. Summary of Global Solar Cell Industries
Based on the above case studies of global solar cell
industry, the diverse technologies of solar cell are classified
according to the S-curve [6], [7], as shown in Fig. 4. c-Si and
poly-Si solar cells are mature technologies. A-Si thin film
solar is near maturity. Cupper gallium diselenide (CGS),
CIGS and CdTe solar cells are in growth stage. The
embryonic technologies are organic, spherical Si, quantum
dot and many embryonic solar cells, which many venture
companies are researching.
New
Invention
Period

Technology
Improvement
Period

Maturity
poly-Si
a-Si TF SC

Growth
CGS, CIGS

Aging

c-Si

MatureTechnology
Period

CdTe SC

Embryonic

Quantum dot

Spherical Si SC

Organic SC

Embryonic SC

Time

Fig. 4. Divergence of Solar Cell Technologies

on the S-curve (Author made)
Table 1. Technologies of global solar cell industries.
(Author made)
New Invention Period
Embryonic Stage
Quantum dot Organic SC
Spherical Si

Japan

USA

Technology Improvement Period Mature-technology Period

Growth Stage
Materiality Stage
CGS, CIGS a-Si TF-SC poly-Si SC
c-Si SC

CdTe SC

1st Group
(Sharp, Kyocera,
Sanyo,
Mit bi hi)
Solar Island
Kyusyu
SolarPower

Nanosolar
Solyndra
First Solar
Venture

China

Suntech
Other Compalies

Germany Q-cells
Production Level

R&D Level

Based on this categorizing, the technology progress level

was analyzed according to the theory of the S-curve as sown
in Table 1.
Japanese 1st group including Sharp, Kyocera, Sanyo and
Mitsubishi is mainly doing the business by c-Si and a-Si solar

cell. Industrial cluster of Solar Island Kyusyu, including

Fuji Electric Systems, Honda Soltec, Mitsubishi Heavy
Industries and Solar Frontier, is focusing on thin film solar
cells.
In USA, SunPower focuses on c-Si solar cell, First Solar of
1st place is producing CdTe solar cell and Nanosolar,
Solyndra are developing CIGS. In addition, many venture
companies are challenging many emerging technologies using
many materials and technologies.
China including Suntech mainly produced c-Si solar cell.
Q-cells restructured the solar cell business and expand from
cell process to solar cell module and CIGS solar cell [3].
As the result, the divers technologies are took for business
by global solar cell industry.
VI. Discussion
A. Dynamics of Innovations in Solar Cell Industries
From the summary of global solar cell industries, what is
the dynamics of innovations in solar cell industries?
Utterback described that the rate of product innovation in
an industry is highest, which means high variety products put
in the market in fluid phase [4].
However, solar cell business began from 1950s and run
beyond 50 years and some solar cell technologies, such as cSi and poly-Si solar cells, have moved into specific phase.
However, many companies are using the diverse technologies
in solar cell business. This means that the diverse
technologies in fluid phase, transition phase and specific
phase are simultaneously accepted for solar cell business.
Also, he described the patterns in the emergence of radical
innovations [4]. It is researched that how a radical
technological innovation can emerge and successfully invade
the established technology. He said that at the time an
invading technology first appears, the established technology
offers better performance or cost than dose the challenger,
which is still unperfected.
In case of flat television, LCD television had the share of
81% and plasma display pale television (PDP-TV) had 10%
in 2007 [21]. Two technologies are accepted the market. PDPTV is mainly producing by Panasonics depending on the
progress of research and development and investment.
However, only two technologies are surviving in flat
television and these numbers of technologies are deferent
from them of solar cell industry.
Christensen described about new read-write head
technologies [22]. The ferrite heads improved between 1976
and 1989, and then it was substituted with the thin-film heads.
Also, the thin-film heads was substituted with magnetoresistive heads. This means two technologies were accepted at
the same time.
In the solar cell industries, however, many technologies are
competing in the real business.
Therefore, in case of the solar cell industry, the innovation
dynamics theory of Utterback can not apply to the innovation.
B. Process days of Solar Cell compared with LCD and
Semiconductor
What are the conditions to accept the diverse technologies
in business such as the solar cell industries?
To answer the question, the process day and unit area
value of the solar cell compared with Liquid Crystal Display

(LCD) and semiconductor device, as shown in Fig. 5.

Most important requirement for solar cell is to reduce the
cost per generate electricity to compete with the other energy
sources, such as thermal, hydro and nuclear power generation.
Therefore, short process day and low unit area value of the
solar cell are required to reduce the cost.
The main process days for poly-Si and a-Si solar cell are
less than 1 day. But, the process day for an LCD is 4-5 days
and that for a semiconductor is 3060 days depending on the
products.
For this reason, unit area value is 6 cent /cm2 of poly-Si
solar cell and 5 cent/cm2 of an a-Si solar cell, comparing to
LCD of 7 cent/cm2 and 1200 cent/cm2 of semiconductor.
Especially, unit area value of LCD is rapidly reducing
because the LCD television market drastically increasing.
This means that diverse technologies of solar cell are
simple because they required shot process days and low unit
area value to reduce the cost.
1 0000

Unit Area Value (cent/cm2)

Semiconductor
1000

100

poly-Si
Solr Cell
10

a-Si
Solar Cell

LCD

1
1

10

100

Process Days (Day

Fig. 5. Process days and unit area value of Sola Cell, LCD
and semiconductor devices. (Author made)
C. Conditions of divers technologies
Utterback researched the dynamics of innovation from the
case studies of plate glass, assemble products, film
photography, ice etc [4]. However, as described above, in
case of the solar cell industry, the innovation dynamics theory
of Utterback can not apply to the innovation.
What are the conditions to accept the diverse technologies
such as the solar cell industries?
Therefore, the conditions of diverse technologies in
business are extracted from the results of the paradigm shift
of solar cell competition, case studies of global solar cell
industries.
First condition of diverse technologies is simple
technology, which means short process day, because it is easy
to enter the business by low input resources and time. In other
word, simple process accepts the diverged technologies to
enter the business. In the contrast, complex process needs to
integrate many technologies, not one technology, in case of
long process day, such as semiconductor. The industries with
complex processes not accept diverse technologies to simplify

the total processes and reduce the research and development

effort.
Second condition is the drastically increasing market to
enhance the entrance. As described above, solar cell market
drastically increasing for environmental sustainability
Third condition is high motivation to enter the industry. As
explained above, the solar cell markets are rapidly increasing
and many companies want to enter the solar cell business
even if their technology level.
Under the above conditions, many companies want to enter
the business even if their technology level. In other words,
many companies enter the business using their adaptable
technologies. As the results, the diverse technologies are
accepted for the business.
VIII. Conclusions
The dynamics of innovation in solar cell industry was
researched based on the case studies of global solar cell
industries.
What is the dynamics of innovations in the solar cell
industries?
Japanese 1st group including Sharp, Kyocera, Sanyo and
Mitsubishi is mainly doing the business by c-Si and a-Si solar
cell. The industrial cluster of Solar Island Kyusyu is
focusing on thin film solar cells. In USA, SunPower focuses
on c-Si solar cell, First Solar of 1st place is producing CdTe
solar cell and Nanosolar, Solyndra are developing CIGS. In
addition, many venture companies are challenging many
emerging technologies using many materials and
technologies. China including Suntech mainly produced c-Si
solar cell. Q-cells in Germany restructured the solar cell
business and expand from c-Si and poly-Si cell process to
solar cell module and CIGS solar cell
As the results of the case studies of the global solar cell
industries, the diverse technologies of solar cell are taken for
business by the global solar cell industry.
Therefore, in case of the solar cell industry, the innovation
dynamics theory of Utterback can not apply to the innovation.
What are the conditions to accept the diverse technologies
in business such as the solar cell industries?
To answer the question, the process day of the solar cell
compared with LCD and semiconductor device. The main
process days for poly-Si and a-Si solar cell are less than 1
day. This means that solar cell technologies are simple
because they required shot process days and low unit area
value to reduce the cost.
The conditions of diverse technologies in business are
extracted from the research results. First condition is simple
technology, which means short process day, because it is easy
to enter the business by low input resources and time. Second
condition is the drastically increasing market. Third condition
is high motivation to enter the industry.
Under the above conditions, many companies want to
enter the business even if their technology level. In other
words, many companies enter the business using their
adaptable technologies. As the results, the diverse
technologies are accepted for business.

Acknowledgments
I express my appreciation for Professor Richard B. Dasher
of U.S.-Asia Technology Management Center of Stanford
University to give me the opportunity to research at Stanford
University. I thank Japan Society for the Promotion of
Science for the financial support.
References
1. New Energy and Industrial Technology Development
Organization (NEDO), "Why did Japan achieve the
world's No. 1 share in solar cells? (Japanese),"
Nedobooks, 2007.
2. PV News, May and April 2008.
3. Outlook of Solar Cell Industry 2011, Sangyo Times
2010.
4. James M. Utterback, Mastering the Dynamics of
Innovation, Harvard Business School Press 1994.
5. J.C. Fisher and R. H. Pry Simple Substitution Model of
Technical Change. Technological Forecasting and
Social Change, vol. 3, pp. 75 88.
6. Tarek Khalil, Chapter 5 Technology Life Cycles,
Management of Technology, McGraw-Hill Higher
Education, (2000), pp80-94.
7. Press Announcement of Sharp Corporation, Sharp to
Develop Manufacturing Complex for the 21st Century,
http://sharp-world.com/corporate/report/plant/index.html,
31 July 2007.
8. Paul Wormser, Thin-Film Companies Heading for
Terawatt-hour Module Manufacturing: Micromorph,
presented at PHOTONs PV Thin Film Conference,
February 2, 2010, San Francisco.
9. Sharp Corporation, Utility-Scale Thin Film
Photovoltaics, exhibited at Solar Power International 09
Exhibition, Oct 27 29, 2009, Anaheim, California, USA.
10. Sharp Corporation, Operation Starts at Solar Cell Plant
in GREEN FRONT SAKAI, press-released in March 29,
2010.
11. SunPower, [Online]. Available:
http://us.sunpowercorp.com/business/
12. SunPower, PG&E Signs 800 MW Photovoltaic Solar
Power Agreements, [Online]. Available:
http://investors.sunpowercorp.com/releasedetail.cfm?Rel
easeID=328221

13. SunPower, 32-Megawatt Solar Power Supply

Agreement With Toshiba, [Online]. Available:
http://investors.sunpowercorp.com/releasedetail.cfm?Rel
easeID=447820
14. First Solar, [Online]. Available:
http://www.firstsolar.com/
15. PV Tech First Solar takes 2009 revenue past US$2
billion, published at 18 February 2010, [Online].
Available: www.pv-tech.org, http://www.pvtech.org/lib/printable/8181/
16. Nanosolar, (2011, January 9.) [Online]. Available:
http://www.nanosolar.com/
17. Solyndra (2011, January 9.) [Online]. Available:
http://www.solyndra.com/
18. Suntech Power Holdings Co., Ltd.2008 Corporate
Report, [Online]. Available:
http://phx.corporateir.net/External.File?item=UGFyZW50SUQ9MzQzMDg
4fENoaWxkSUQ9MzMwMjA2fFR5cGU9MQ==&t=1
19. Zhengrong Shi, Suntech, CEO Roundtable Discussion:
Developing the Utility Market for Solar Electricity,
presented at Solar Power International 09 Conference,
Oct 27 29, 2009, Anaheim, California, USA.
20. Q-Cells (2011, January 9.) History [Online]. Available:
http://www.q-cells.com/en/company/
qcells_se/history_and_facts/index.html
21. Hisakazu Torii , The 14th DisplaySearch Japan Forum
(Jan 30-31, 2008)
22. Clayton M. Christensen, The Innovators Dilemma,
Boston, Massachusetts, Harvard Business School Press,
1997.

Solar PV Technology Value Chain in Respect of New Silicon Feedstock

Materials: A context of India and its ambitious National Solar Mission
Hasita Kaja
(Under-graduate student ECE branch)

D.T. Barki

Nalla Narasimha Reddy Engineering College

Noble Energy Technologies Ltd. (NEST)

Koremula village, near Ghatkesar mandal,

Hyderabad.

A-9, Aero View Towers, Shamlal Buildings,

Begumpet, Hyderabad 500 016

Email: hasitakaja@rocketmail.com

Email: dtbarki@usa.net,
dtbarki@solarnest.net ; Web:
www.solarnest.net

Abstract:
Achieving energy security through sustainable
technologies is the most challenging objective that
needs to be fulfilled within the context of rising
energy demand in a country like India. Although
nuclear energy is expected to meet a part of Indias
energy demand by 2030, this would still be non
substantial and more importantly an unsafe way of
producing power as demonstrated by a series of
negative events that have marred its reputation as
a sustainable technology, particularly in the wake
of recent nuclear disaster at the Fukushima power
plant in Japan. And with increasing pressure on
every country to switch over from fossil-fuel based
energy to clean and renewable energies, India has
rightly announced its ambitious National Solar
Mission (NSM) to install 20 GW of solar power by
2022 in Nov 2009. This paper discusses how India
can achieve this target by adopting various solar
PV technologies and processes. Until not so far
back in time, poly-silicon (Si) production primarily
catered to the semiconductor markets; however,
the world now realizes that much silicon is
required for the fast growing solar industry as well.
The only Si production process that has dominated
the world market has been the Siemens process,
which is highly capital and energy intensive.
Further this process is not carbon-neutral, since it
uses coke and woodchips as reducing agents,
causing considerable deforestation. Scientists and
technologists across the world realize that spending

so much money and energy to produce silicon can

be avoided and have thereby begun to research on
alternative technologies to produce silicon for the
solar industry. As a result several new and
disruptive technologies have emerged such as solar
grade silicon (SoG) also known as
UMG
(Upgraded Metallurgical Grade) silicon, which
involve substantially lesser resources in terms of
capital & energy and are therefore carbon-neutral
at the end of the day. Solar grade silicon of purity
99.9999% still dominates the PV industry across
the world. In this paper, we give an account of
UMG silicon methodology with the view that it has
the potential to be a disruptive technology for
existing high-purity solar grade silicon producers
at the same time a revolution for the PV industry.
For instance UMG-Si uses 1/3rd the cost, 1/15th the
energy and 1/10th capex (capital expenditure) of
Siemens process. For many years to come Silicon
(Si) solar cells will contribute significantly towards
production of mono and multi crystalline PV
modules. If India has to achieve the target, over
and above set by the NSM then it must begin to
focus on producing Si by adopting new processes
that have been aforementioned. By doing so, India
could emerge as a true world leader in solar PV,
marching alongside countries like Germany who
are determinedly working towards meeting a
quarter of their energy requirements through solar
technologies by 2050.

I.

INTRODUCTION

There is a pressing need to accelerate the development of

advanced clean energy technologies in order to address the
global challenges of energy security, climate change and
sustainable development. This challenge was acknowledged
by the ministers from G8 countries, China, India, South
Korea, Japan, Italy, France, Germany and USA.
The energy policy of India is largely defined by the country's
burgeoning energy deficit and increased focus on developing
alternative sources of energy, particularly nuclear, solar and
wind energy.
About 70% of India's energy generation capacity is from
fossil fuels, with coal accounting for 40% of India's total

energy consumption followed by crude oil and natural gas at

24% and 6% respectively. India is largely dependent on fossil
fuel imports to meet its energy demands by 2030, India's
dependence on energy imports is expected to exceed 53% of
the country's total energy consumption.
Why Solar:
Due to rapid economic expansion, India has one of the
world's fastest growing energy markets and is expected to be
the second-largest contributor to the increase in global energy
demand by 2035, accounting for 18% of the rise in global
energy consumption. Given India's growing energy demands
and limited domestic fossil fuel reserves, the country has
ambitious plans to expand its renewable and nuclear power
industries. India envisages to increase the contribution of
nuclear power to overall electricity generation capacity from
4.2% to 9% within 25 years. The country has five nuclear

reactors under construction (third highest in the world) and

plans to construct 18 additional nuclear reactors (second
highest in the world) by 2025.
The production of nuclear electricity is very expensive.
Approximately 21000 crore rupees are needed to produce 1
megawatt. If we use water for production we need only 6000
crore rupees to produce 1 megawatt (reference from Google:
nuclear power hazards). And moreover, nuclear energy has
many ill effects on the living organisms. As we saw in Japan,
earthquakes and tsunami destroyed the Fukushima town and
badly affected the Japanese economy and man power. After
this Japan has closed its three reactors which were producing
nuclear electricity. But radiation of nuclear rays has not
stopped yet. The speed of radiation has crossed the normal
living conditions mark. Poison of the radiation has mixed in
the water, milk, fruits, vegetables and other food products.
But, the Indian government is very eager to produce nuclear
electricity at any cost. The dangers from nuclear radiation
are very hazardous. Scientists have not found any method to
nullify the dangers caused by nuclear ash. There are other
natural sources which can be utilized for our energy purpose.
We can use solar energy and wind energy, resources which
are abundant and importantly benign in nature towards
human beings and the environment in general.

II.

INDIAS NATIONAL SOLAR

MISSION

Silicon is the second most abundant element in the earths

crust, in the form of oxide as sand or quartz and as silicate (in
the form of organic clay and mica). Silicon is mainly used in
electronic industry and solar industry. The semi-conductor
industry suffered a strong dip in the manufacture of poly
silicon. Up to 1998 the solar PV sector was dependant on offspec silicon it obtained from semi-conductor industry. But
since the production of silicon from quartz, the total market
got quintupled. Despite thinner solar cells and other
efficiency gains the corresponding silicon consumption has
quadrupled, according to the latest solar study.
It is only because several manufacturers of poly-silicon are
now able to produce slightly less pure materials for photo
voltaic industry, that annual growth rates of 34%and 54% are
possible. However, the continuing good availability of solar
grade silicon meant, that the efforts of the photovoltaic
industry to establish a feedstock supply independent of
semiconductor industry were pursued with insufficient
intensity. Its strange that India with a massive plan
generating 20 GW of solar power within a decade, has no
company domestically producing silicon. About one ton of
polysilicon is required to produce 10 MW of solar cells. Its
learnt that there are about 6-7 solar companies (Reliance,
TataBP, BHEL/BEL, Lanco Solar and so on) which are
toying with plans of setting up of silicon manufacturing
facilities under Semiconductor policy. The purpose of this
paper is to discuss and recommend alternate methods of
silicon production.

The Jawaharlal Nehru National Solar Mission (JNNSM) was

launched a few days before the UN Climate Change Summit
in Copenhagen last year. On March 5th 2010, the Ministry of
New and Renewable Energy (MNRE) published the Draft
Guidelines for Selection of New Grid Connected Solar
Projects under Phase 1 of the Jawaharlal Nehru National
Solar Mission. The mission will be carried out in three phases
and aims to do the following:

To create a policy framework for deployment of

20,000 MW by 2022;

To add 1,000 MW of grid solar power by 2013

and another 3,000 MW by 2017.

The target for 2017 may be highly based on the availability

of international finance and technology transfer.

Figure 1: High purity quartz lumps ready for shipping to a

solar grade silicon mfg co., Photo Courtesy: NEST,
Hyderabad

THE AGE-OLD SIEMENS

III.
PROCESS: A poly-Si manufacturing process that we can
bid good-bye to!
The only available, largely used technology for producing
silicon has been the Siemens process. Silicon is available as
silica in the earths crust. It is extracted and purified for
processing. PV modules require high purity quartz to extract
silicon from it. In order to crack open the oxygen bonds in
silica the carbothermic reduction process is used. This
requires high amount of energy and other raw materials like
charcoal, coal and coke.
Figure 2: High purity quartz lumps (1.5 to 5)
Photo Courtesy: NEST, Hyderabad
The raw material, Silicon

Conventional Chemical routes based on electronic grade

silicon (Siemens Method)
In the chemical route, metallurgical grade silicon is reacted
with HCl and transformed into the gaseous phase. The gas

phase is distilled and the by-product is removed before the

gas (often Trichloro Silane, TCS) is decomposed, on
filaments in a bell jar CVD reactor. Conventional plants for
the production of electronic grade silicon produce silicon via
distilled (chloro-) silane-gasses.
The plants require large investments for relatively modest
production capacities. They also consume huge amounts of
electrical energy. Several companies (Hemlock Wacker,
Solar Grade Silicon LLC (2300 MT/year), Tokuyama, Joint
Solar Silicon) have all increased their capacities. The
processes are still based on (chloro-) silanes but in several
cases they have begun using new deposition technology,
often fluidized beds.
This Siemens process has many shortcomings that cannot be
overlooked and hence cannot be used for future production.
In order to produce a single (1) metric ton (MT) of silicon
metal, raw material inputs of 2.8MT quartz, 1.4MT of coal,
and 2.4MT of wood chips are required. The reduction
reaction is very energy intensive requiring about 13
megawatt hours of electricity to produce 1 ton of silicon and
contributing over one third or 36% to the silicon metal
production costs. In addition, about 1.4 MT of silicon metal
are required to produce 1 MT of polysilicon suitable for
crystalline silicon PV. In this process about 40% of the quartz
starting material is converted to silica fume. The impurities
Al and Ca can be removed to very low levels, but in practice
this is prevented by large heat losses. Besides high-energy
intensity, deforestation takes place due to the use of coal,
charcoal, wood and wood chips as additional feedstock to
quartz. Under these circumstances, one can get a doubt and
ask how clean the clean-tech option of solar energy is?
Quartz, being a primary source of silicon for solar cells,
cannot be sidelined, though it requires mining and hence a
contributor to the deforestation. But the good news for the
solar industry is that there are alternate methods available
which are almost carbon-neutral.
To overcome these disadvantages there are many new and
disruptive technologies such as solar grade silicon (SoG) also
known as UMG (Upgraded Metallurgical Grade) silicon,
which involve substantially lesser resources in terms of
capital and energy and are therefore carbon neutral at the end
of the day. Interestingly, several companies such as RSI
silicon-USA, Calisolar-Canada, Becancour Silcion-Canada
among many other companies have started production of
solar grade silicon on commercial scale. The silicon giants
such as Dow Corning-USA (PV 1011) and Elkem-Norway,
had been working constantly to perfect the solar grade silicon
routes. Let us take a look at these options in brief, as the
elaborate discussion of these alternate solar grade silicon
technologies is beyond the scope of this paper.

[Fig 2: Silicon grade silicon (from Becancour Silicon

Canada, shown along with the basic feedstock, quartz.
Courtesy: Photon International Magazine]

IV.

DIFFERENT ROUTES FOR SILICON

FEEDSTOCK TO SOLAR CELLS

The silicon value chain to feedstock for solar cells is shown

in figure 3. In principle, there are several different ways to
SoG-Si.
1. The traditional route to semiconductor silicon is shown on
the left hand side of Figure 3. Scrap, rejected, and non-prime
material from this production is the main supply route of
today.
2. Several companies (Wacker, ASIMI, etc) are making a big
effort to economize the chemical route by shifting from rod
decomposition of TCS to fluidized bed in the last step of the
poly-silicon production
3. The purification of metallurgical silicon has, so far, not
resulted in a commercially viable process for a feedstock to
solar cells. Several companies (Elkem, FerroAtlantica etc)
are working on this route to SoG-Si.
4. The last potential source of solar grade silicon might be
from the direct carbothermic reduction of quartz and carbon
as indicated on the right hand side of figure 3. The Solsilc
route and RSI Silicon are examples of this.

Fig 3: Alternative routes for solar grade silicon, (Source:

New Feedstock Materials, Aud Waernes, Ola Raaness, et al)

V.

DIRECT METALLURGICAL ROUTE

SOLSILC
The Direct Metallurgical route produces silicon from ultrahigh purity raw materials. Silicon production according to the
SOLSILC technology has resulted in silicon of very high
purity and has already been running at a pilot scale of 50-100
ton/yr. This carbothermic production route can, in principle,
reduce the cost of feedstock to below 15/kg. India should
consider working on this technology.
The idea behind the SOLSILC Technology was to develop a
new two-step high temperature plasma-process for solar
grade silicon (SoG-Si) production, based on carbothermic
reduction. The SOLSILC Technology relied on expertise of
the metallurgical production methods for Silicon and SiC,
and PV cell processing.
The work involves with the development of a low-purity Siprocess facility (say, 6N purity instead of 10N or 11N) in
order to control the production process before a clean
environment is established. In order to control the purity of
the silicon, the selection and supply of raw materials is
crucial. Specially selected Carbon Black and quartz are used
in the process.
The bottleneck, as of now, that all producers of SoG-Si based
on carbothermic reduction of quartz face, is to remove carbon
and metals impurities like iron, aluminium etc from the
silicon. The source of the impurities is mainly the quartz. In
order to comply with the objective of low cost, this has to be
done on large-scale in a cost effective and reliable manner. A
large number of experiments were carried out first in small
scale.
Kazakhstan carbothermic route
A research group at the Kazakh National Technical
University in Kazakhstan is also working on a carbothermic
reduction process for solar grade silicon. They are using very
clean and high purity quartz from deposits in Kazakhstan.

Silicon by electrolysis
Electrolysis has been the way to aluminium for more than
100 years. Silicon can also be produced by electrolysis. A
process was patented by Stubergh in 1994. The process is
based on the dissolution of quartz in a fluoride-containing
bath at 1000C, and decomposition of the quartz into silicon
and oxygen. Silicon precipitates at the electrode. Afterwards,
the silicon is crushed and cleaned with acid. In order to
obtain a clean product, the silicon is melted and thereafter
crystallized into ingots for subsequent sawing into wafers.
The slag that comes with the silicon from the electrolyte has
a high solubility for impurities and can be separated from the
silicon. Today, Stubergh has scaled up his tests to a larger
laboratory scale. The work is partly done in cooperation with
SINTEF and IFE. He has established a company Norwegian
Silicon
Refinery and has made plans for a full scale test facility.
SINTEF is also doing research on production of silicon by
electrolysis. RSI Silicon, USA, has also adopted similar
electrolytic process that consumes only 1/15th energy as
compared to Siemens process. And the process is almost
carbon-neutral.

SUMMARY AND
RECOMMENDATIONS

VI.

Fossil fuels are dwindling and exhausting. Nuclear energy

option is clearly losing its ground in light of recent
Fukushima disaster. Solar PV technology is the clear energy
option today. The solar PV industry has reached its
commercial phase, and its the best option for a meaningful
and sustainable development. Prudence has to be exercised in
selecting the alternate silicon routes to avoid far-reaching
damages to the environment and economy. In order to
succeed technologically, commercially with self-reliant
silicon routes, Indian Ministry should look at adopting solar
grade silicon route as it starts implementing the Solar
Missions goals. Economical and environmental goals are
also imminent. The carbon footprint of solar grade silicon is
1/14th that of silicon produced through Siemens route.

VII.

Metallurgical purification (Upgraded MG UMG)

Elkem-route for cleaning of metallurgical silicon
Elkem has been working on the development of a new
process for the production of SoG Si (or UMG) for a very
long time. Elkem is the worlds leading company in the
production of metallurgical silicon. In this method, the
chosen route to produce SoG-Si is to further enhance its
standard metallurgical products through slag refining. They
have developed a two step refining process. The first step is
slag refining of the silicon to remove certain elements. The
next step is a leaching process that will remove other
elements. The energy consumption of the silicon process is
20 25% of the traditional Polysilicon route. Elkem Solar
has shown that their SoG-Si may be used to produce cells of
an efficiency of 15 to 16%. Elkem will build a pilot-scale
plant in 2005.

REFERENCES

1.

J. Kraiem et al, High performance solar cells

made from 100% UMG Silicon obtained via the
photosil process.

2.

D.T.Barki, Path-breaking Direct to Silicon

(DTS) Approach to Obtain Crude Silicon from
High Purity Quartz, 22nd European Conference,
Milan, Italy, [2006]

3.

Sun & Wind Energy, 2/2006,

4.

Aud Waernes, Ola Raaness, Bart Geerlings et al,

New feedstock materials, ECN Netherlands &
Sintef Norway

5.

Edgar A Gunther, Making Silicon Metal for

Solar,

214

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 2, APRIL 2012

Wide Band Gap Gallium Phosphide Solar Cells

Xuesong Lu, Member, IEEE, Susan Huang, Martin B. Diaz, Nicole Kotulak, Ruiying Hao, Robert Opila,
and Allen Barnett, Life Fellow, IEEE

AbstractGallium phosphide (GaP), with its wide band gap of

2.26 eV, is a good candidate for the top junction solar cell in a
multijunction solar cell system. Here, we design, fabricate, characterize, and analyze GaP solar cells. Liquid phase epitaxy is used to
grow the semiconductor layers. Four generations of GaP solar cells
are developed and fabricated with each solar cell structure being
designed and improved based on the first principles analyses of the
predecessor solar cells. Quantum efficiency and currentvoltage
measurements are used to analyze the solar cell performance and
to develop predictive models. We create a GaP solar cell with an
efficiency of 2.42% under AM 1.5G one sun illumination.
Index TermsGallium phosphide (GaP), liquid phase epitaxy
(LPE), quantum efficiency (QE), wide band gap.

I. INTRODUCTION
ECAUSE of band gap limitations, a single-junction solar
cell cannot achieve efficiency higher than 30% under one
sun illumination. However, higher efficiency can be reached
with tandem solar cells [1]. Over the past three years, the efficiency reported of the two-terminal triple-junction solar cells
has increased systematically from 41.1% [2] to 41.6% [3] and
to 42.3% [4]. The highest reported efficiency so far for a twoterminal triple-junction solar cell system from Solar Junction
has the efficiency of 43.5 2.5% [5]. One way to further improve the efficiency of a multijunction solar cell system is to
increase the number of junctions beyond 3 [1], [6]. Nell et al.
have developed a spectral pn junction model for tandem solar
cell design [6]. In their paper, they demonstrated that theoretical
efficiencies are 28.3% for a single-junction solar cell and 39.1%,

Manuscript received August 16, 2011; revised November 15, 2011; accepted
December 18, 2011. Date of publication February 23, 2012; date of current
version March 16, 2012.
X. Lu was with the Department of Electrical and Computer Engineering, University of Delaware, Newark, DE 19716 USA. He is now with Applied Materials
Company, Santa Clara, CA 95054 USA (e-mail: luxuesong001@gmail.com).
S. Huang was with the Department of Materials Science and Engineering, University of Delaware, Newark, DE 19716 USA. She is now with the
IBM Thomas J. Watson Research Center, Yorktown Heights, NY 10598 USA
(e-mail: rongxue.huang@gmail.com).
M. B. Diaz and N. Kotulak are with the Department of Electrical and Computer Engineering, University of Delaware, Newark, DE 19716 USA (e-mail:
mdiaz@udel.edu; nkotulak@udel.edu).
R. Hao was with the Department of Electrical and Computer Engineering,
University of Delaware, Newark, DE 19716 USA. She is now with Crystal Solar
Company, Santa Clara, CA 95054 USA (e-mail: ryhao@udel.edu).
R. Opila is with the Department of Materials Science and Engineering, University of Delaware, Newark, DE 19716 USA (e-mail: opila@udel.edu).
A. Barnett was with the Department of Electrical and Computer Engineering,
University of Delaware, Newark, DE 19716 USA. He is now with the School
of Photovoltaic and Renewable Energy Engineering, University of New South
Wales, Sydney 2052, Australia (e-mail: allen.barnett@unsw.edu.au).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JPHOTOV.2011.2182180

Fig. 1.

Five-junction solar cell system with dichroic mirror [9].

44.5%, and 57.3% for two-, three-, and ten-junction solar cells
under AM 1.5G, respectively.
A five-junction solar cell system has been developed for a
separated electrical contact system with lateral spectrum splitting design, which could theoretically achieve over 50% efficiency [7][9]. The lateral spectrum splitting system design
is described in [9]. It provides several benefits, including increased theoretical conversion efficiency, reduced current mismatch losses, reduced lattice mismatch problems, increased flexibility in material choices, and device designs that can more
closely approach the ideal performance limits of existing solar
cell technology. In addition, the multijunction solar cell system
based on a silicon platform has the unique advantages of meeting both efficiency and cost targets. Fig. 1 shows the structure
of this five-junction solar cell system with optics [9]. Light first
passes through the wide band gap solar cell. Then, the dichroic
mirror is used to split the light spectrum.
A first principle calculation described elsewhere [10] shows
that this five-junction solar cell system can achieve the efficiency
of 50% when the top junction solar cell has a band gap between
2.1 and 2.7 eV (the band gaps of the bottom four solar cells
are shown in Fig. 1). This leaves a broad range for material
selection.
In this study, we choose GaP as the material for the top
junction solar cell. The advantages of this include proper band
gap (2.26 eV), high carrier mobility [11] allowing high collection probability, high crystal quality, and well-developed

2156-3381/$31.00 2012 IEEE

LU et al.: WIDE BAND GAP GALLIUM PHOSPHIDE SOLAR CELLS

technology [12]. GaP is also a good candidate for solar cells in

space applications because it is stable at high temperatures [13].
The efficiency of a five-junction solar cell system with a GaP
solar cell as the top junction is 3.6% absolute higher than the
four-junction solar cell system. In order to achieve this 3.6%
efficiency gain, the GaP solar cell must have three important
features. First, it should have a high Vo c , i.e., a characteristic
that would come from a solar cell device with a high bulk lifetime and a low surface recombination velocity. Second, it must
efficiently collect the minority carriers generated by absorbed
photons. Third, it must have good optical properties, including
good transparency for photons that are not absorbed. The last
feature requires the solar cell to have a physical thickness equal
to or less than its active region thickness [14], and it requires
the reflection loss from the surface to be minimized.
The GaP epitaxial layers were grown by liquid phase epitaxy
(LPE). The metal contacts were deposited using electron beam
evaporation. The solar cells were isolated using a dicing saw
resulting in a final cell area of 0.16 cm2 . JV and quantum
efficiency (QE) measurements were used to test the performance
of the solar cell devices. Recently, the highest efficiency for a
GaP solar cell was reported as 2.6% from [15]. That solar cell
had a short-circuit current density (Jsc ) of 1.81 mA/cm2 , an open
circuit voltage (Vo c ) of 1.48 V, and a fill factor (FF) of 77%. Their
epitaxial layers were grown by molecular-beam epitaxy. In this
study, our best GaP solar cell that is shown in Fig. 2 has the
efficiency of 2.42%. The Vo c , Jsc , and FF of this solar cell are
1.550 V, 1.97 mA/cm2 , and 79.4%, respectively. We note that
each of the individual parameters of the solar cell in this study
is higher than the parameters reported in the 2.6% solar cell cited
previously [15]. The current densities of each generation GaP
solar cell are measured outdoors under calibrated conditions for
accuracy.

II. QUANTUM EFFICIENCY

Quantum efficiency (QE) is the ratio of the number of minority carriers collected by the solar cell device to the number
of photons of a given energy incident on the device. The QE
involves external QE (EQE) and internal QE (IQE). The EQE
of a solar cell device includes the effect of reflection. The IQE
denotes the efficiency with which photons that are not reflected
can generate collectible minority carriers. To calculate IQE, we
divide the solar cell devices into three parts: the emitter region,

IQEe =

215

Fig. 2. Measured JV curve of the best GaP solar cell in this study under AM
1.5G one sun illumination (the fourth-generation GaP solar cell).

the base region, and the depletion region. Then, the total IQE of
the solar cell is the sum of the IQEs of the three regions [16]:
IQEtotal () = IQEe () + IQED () + IQEB ()

(1)

where IQEe , IQED , and IQEb , are the IQEs of the emitter,
depletion, and base regions, respectively. The IQE of the emitter
region is as in (2), shown at the bottom of the page, where is the
absorption coefficient of GaP; Le and De are the diffusion length
and the diffusivity of the minority carriers in the emitter region,
respectively; Sf is the front surface recombination velocity; and
We is the thickness of the emitter (without the depletion region).
The IQE of the depletion region is
IQED = exp(We )[1 exp(Wd )]

(3)

where Wd is the thickness of the depletion region. The IQE of

the base region is as in (4), shown at the bottom of the page,
where Lb and Db are the diffusion length and diffusivity of
the minority carriers in the base region, respectively; Sr is the
back surface recombination velocity; and Wb is the thickness of
the base region. In this study, the diffusion length in the base
region is much less than the base region thickness. Therefore,
the Sr value would not affect the solar cell performance. In the
simulation, the value of Sr was fixed at 1 105 cm/s.
The previous equations allow us to calculate the IQE curve
of a GaP solar cell. The values of some critical parameters for
the different cell regions, such as diffusion length and surface



(Sf Le /De + Le ) e W e [(Sf Le /De ) cosh We /Le + sinh(We /Le )]
Le
W e

exp
e
2 L2e 1
(Sf Le /De ) sinh(We /Le ) + cosh(We /Le )

(2)

IQEb =

Lb
exp[(We + WD )]
2 L2b 1


Lb (Sr Lb /Db ) [cosh W b/Lb exp(Wb )] + sinh(Wb /Lb ) + Lb exp(Wb )

(Sr Lb /Db ) sinh(Wb /Lb ) + cosh(Wb /Lb )

(4)

216

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 2, APRIL 2012

ting the JV curves using the two-diode equation indicates that

J01 , which represents the recombination from the bulk region,
is 830 A/cm2 and J02 , which represents the recombination
from the space charge region, is 315 A/cm2 . Since this high
J02 degrades the performance of this first-generation GaP solar
cell, we conclude that decreasing the recombination from the
space charge region will improve the performance of the solar
cell. Equation (5), which follows, is the two-diode equation that
is used in the simulation [17]:


q(V IRs)
1
J = J01 exp
n1 kT


q(V IRs)
(V JRsh)
1 +
JL .
+ J02 exp
n2 kT
Rsh
(5)
Fig. 3.

Structure of the first-generation GaP solar cell.

recombination velocity, are obtained by fitting the IQE model

to the measured IQE curve. These deduced parameters are then
used to calculate the dark saturation current and the Vo c of the
solar cell device. By comparing the calculated values with the
JV measurements, we can further verify the accuracy of these
parameters. In this study, the IQE curve was measured by combining the measurement results of the EQE and the reflection.
III. CHARACTERIZATION AND ANALYSIS
A. Characterization of the GaP Epitaxial Layer
Both n-type and p-type GaP epitaxial layers were grown on
GaP substrates in a horizontal LPE system. The detailed growth
procedure is described in [8] and [14].
Various test methods were used to characterize the epitaxial
layers so that the growth conditions could be optimized. SEM
and optical microscopes were used to view the surface morphology of the layer. The dopant concentration of the epitaxial
layer was measured by the Hall measurement. The thickness of
the epitaxial layer was measured using a profilometer and the
cross-section SEM. After optimization, we found that a highquality GaP epitaxial layer with proper thickness and dopant
concentration could be grown via LPE.
B. First-Generation GaP Solar Cell
After epitaxial layer growth was optimized, we fabricated
the GaP solar cell. PdSiPd was used as the Ohmic contact
material for n-type GaP, and PdZnPd was used as the Ohmic
contact material for p-type GaP. Fig. 3 shows the structure of
the first-generation solar cell, i.e., a GaP pn junction grown on
an n-type GaP substrate.
The n-type and p-type GaP layers used Te and Zn, respectively, as the dopant source. The JV testing for this device was
performed under dark condition and AM 1.5G illumination.
This solar cell does not have AR coating on top. Its Vo c , Jsc , and
FF are 1.32 V, 0.54 mA/cm2 , and 69%, respectively. Curve fit-

IQE curve fitting gives the diffusion lengths in the emitter and
base region of 0.55 and 0.04 m, respectively. The front surface
recombination velocity is approximately 1 6 cm/s.
The analysis of the JV and IQE curves indicates that the
first-generation GaP solar cell has three problems: high value of
J02 that indicates a high space charge region recombination rate;
a low-diffusion length in the n-type base region; and a high front
surface recombination velocity. Therefore, we developed a new
solar cell structure in which the open circuit voltage of the GaP
solar cell was improved by decreasing the high recombination
rate in the space charge region.
C. Second-Generation GaP Solar Cell
One way to decrease the space charge region recombination
rate is to use the relatively high-quality p-type epitaxial layer as
the lower doped base region. It is known that the width of the
space charge region in each side is determined by the dopant
concentrations. If one side of a pn junction is relatively lower
doped, the space charge region in that side will be relatively
wider. For the first-generation GaP solar cell, the n-type base
region has a much lower diffusion length than the p-type emitter
region, which can be seen from the IQE curve-fitting results of
the first-generation solar cell. Most of the space charge region
is located in the low-quality n-type base region because it was
relatively lower doped. Thus, the recombination rate in that
space charge region was high. However, if we use the higher
quality p-type region as the lower doped base, most of the space
charge region will be located in this high-quality p-type region
so that the recombination in the space charge region will be
greatly decreased. As a result, J02 will decrease, and Vo c of the
solar cell will increase.
Table I lists the calculated Vo c values for different J02 values
assuming that the other parameters remain unchanged. Based
on our simulation, if J02 decreases to 1 1016 A/cm2 , then
Vo c of the solar cell will increase to 1.50 V. However, the further
decrease of J02 will not increase Vo c significantly, because J01
will dominate the performance of the solar cell.
Fig. 4 shows the structure of the second-generation GaP solar
cell with a GaP pn junction grown on a p-type substrate. The
only change from the first generation to the second generation

LU et al.: WIDE BAND GAP GALLIUM PHOSPHIDE SOLAR CELLS

217

TABLE I
CALCULATED Vo c VALUES FOR DIFFERENT J0 2 VALUES

Fig. 6. Measured and simulated IQE curves of the second-generation GaP

solar cell.

Fig. 4.

Structure of the second-generation GaP solar cell.

Fig. 5.

JV curve of the second-generation GaP solar cell under one sun.

is the polarity of the solar cell. In our experiment, the p-type

substrate has a thickness of 450 m and a dopant concentration
of 1 1018 cm3 . The dopant sources for the p-type and n-type
GaP layers are Zn and Te, respectively.
Fig. 5 shows the JV curve of the second-generation GaP
solar cell under one sun illumination. Vo c , Jsc , and FF of this
solar cell are 1.535 V, 0.95 mA/cm2 , and 81.4%, respectively.
This Vo c value matches the simulated result quite well. The
efficiency of this solar cell is 1.18%. This solar cell does not
have an AR coating on top. According to the JV curve-fitting
results, the value of J01 is 3 1030 A/cm2 , and the value of
J02 decreases to 9 1017 A/cm2 . This result verifies that our
new structure decreases the space charge region recombination
rate.

The QE curve fitting of the second-generation GaP solar cell

is plotted in Fig. 6. The parameters used in this IQE simulation
are 0.11 and 2 m for the diffusion lengths in the emitter and
the base, respectively. The front surface recombination velocity
is found to be approximately 1 106 cm/s.
In order to further analyze this solar cells performance, we
performed more IQE simulations and changed one parameter
at a time to determine the influence of each parameter on the
overall performance of the solar cell.
Fig. 7 shows the simulated IQE curves for different parameters. The simulation results show several relationships. First,
increasing the diffusion length in the emitter region can increase
the IQE at a short-wavelength range. Second, the front surface
recombination rate will not affect solar cell performance if the
diffusion length in the emitter is low. Third, if the quality of the
emitter layer is improved, the reduced front surface recombination velocity can lead to a significant increase in the IQE at
short wavelengths. If the front surface recombination velocity
is decreased from 1 106 to 1 103 cm/s and the diffusion
length in the emitter is taken as 2 m, the IQE will improve from
20% to approximately 90% at short wavelengths (280450 nm).
Fourth, if the diffusion length in the base region is improved, the
IQE at longer wavelengths range (450550 nm) can be further
increased.
Based on our simulation results, the diffusion lengths in the
emitter and base regions, and the surface recombination velocity
are all very important in achieving high-performance GaP solar
cells. However, increasing the diffusion lengths is more important in improving the performance of our second-generation
GaP solar cell. Thus, our next third- and fourth-generation GaP
solar cells will focus on increasing the diffusion lengths in the
emitter and base regions based on the second-generation solar
cell structure, respectively.
D. Third- and Fourth-Generation GaP Solar Cells
In order to improve the diffusion length in the emitter, we
carefully went through the whole growth procedure. Based on
the analysis, we decided to change the dopant source for the

218

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 2, APRIL 2012

TABLE II
PARAMETERS OF THE FOURTH-GENERATION OF GAP SOLAR CELLS

Fig. 8. Measured IQE curve of the second-, third-, and fourth-generations

GaP solar cells.

Fig. 7. Simulated IQE curves with different parameters. (a) Changed Le and
Sf and (b) changed Lb . The unit for Le and Lb is micrometers and for Sf is
centimeters per second.

n-type GaP layer. In the previous generations of solar cells, the

dopant source was Te. Te has a relatively high vapor pressure
that makes it difficult to control the dopant concentration in the
n-type GaP growth layer. Moreover, according to the literature,
Te will stimulate the formation of nonradiative recombination
centers in the growth layer [18]. Thus, we changed the dopant
source to Sn.
Our third-generation GaP solar cell has a Si3 N4 /SiO2 doublelayer antireflection coating on top. The thickness of the Si3 N4
and the SiO2 layers are 50 and 20 nm, respectively. The performance of this solar cell shows an obvious improvement. The
JV curve parameters are listed in Table II. IQE curve fitting
gives a diffusion length in the n-type emitter of 0.19 m, i.e.,
a length that is much higher than that of the second-generation
solar cell. The diffusion length in the base and the front surface
recombination velocity remains the same as those in the secondgeneration solar cell. This indicates that Sn as the n-type dopant
source increases the diffusion length in the n-type growth layer.
Our fourth-generation GaP solar cells focused on the improvement of the diffusion length in the base region based on

the third-generation solar cell results. The main change to the

LPE growth process is the decrease in the growth rate of the
p-type base layer. In the third-generation solar cell, the base
layer was relatively thick (8 m); therefore, the growth rate of
the layer was 4 nm/s in order to shorten the time of one growth
cycle. For this new structure, we use a slower growth rate of
1.6 nm/s. With epitaxial growth, the atoms need sufficient time
to reach their lowest energy state and duplicate the crystal lattice structure; this slower growth rate leads to higher crystalline
quality.
The JV curve of the fourth-generation GaP solar cell is
shown in Fig. 2. The same AR coating that we used for the
third-generation cell was deposited on top of this solar cell. The
efficiency of this solar cell is 2.42% under AM 1.5G one sun
illumination. Table II lists the parameters of all the four generations of GaP solar cells. The values for diffusion lengths and
surface recombination velocity come from IQE curve fitting.
The consistent increase in efficiency demonstrates the subsequent improvement of each generation of a GaP solar cell.
Fig. 8 shows the measured IQE curves of the second-, third-,
and fourth-generation GaP solar cells.
Simulated IQE curves for GaP solar cells with the fourthgeneration structure and the indicated parameters are shown
in Fig. 9. They demonstrate the pathway to the theoretical
maximum efficiency of 12.58% under AM 1.5G one sun illumination. A solar cell with Le , Lb , and Sf of 5 m, 20 m
and 1103 cm/s, respectively, produces the best IQE curve and
achieves this 12.58% efficiency. The other curves in the figure

LU et al.: WIDE BAND GAP GALLIUM PHOSPHIDE SOLAR CELLS

219

IV. CONCLUSION
Four generations of GaP solar cells have been developed, fabricated, and analyzed. Their performance has been tested using
two primary methods: JV and QE. A predictive QE model has
been developed to further analyze the solar cells performance
and to improve the solar cell design. Continuous improvements
in design with each generation of solar cell led us to a GaP
solar cell with the best reported efficiency grown by LPE. This
GaP solar cell has the Vo c of 1.55 V, the Jsc of 1.97 mA/cm2 ,
and the FF of 79.4%. Its efficiency is 2.42%. Further increasing
the diffusion length in the emitter and the base and further decreasing the front surface recombination velocity will allow us
to substantially increase the efficiency.
Fig. 9. Simulated IQE curves of GaP solar cells based on the fourth-generation
structures with different parameters. The unit for Le and Lb is micrometers and
for Sf is centimeters per second.
TABLE III
CALCULATED VALUES OF Jsc , J0 , Vo c , FF, AND EFFICIENCY
FOR A FOURTH-GENERATION GAP SOLAR CELL WITH THE PARAMETERS
LISTED IN THE FIRST THREE COLUMNS

ACKNOWLEDGMENT
The authors would like to thank N. Faleev for his help with the
XRD measurements, J. Mutitu for his help in depositing the AR
coating, and J. Allen from the Institute of Energy Conversion
for his help with the QE measurements.
REFERENCES

demonstrate the contribution of each parameter (Le , Lb and Sf )

to the performance of the solar cell:


qV
1 ISC .
I = I0 exp
(6)
kT
Taking the parameters used in Fig. 9, Jsc , J0 , Vo c , FF, and efficiency were calculated for the GaP solar cells. We calculate Vo c
from the single-diode equation [19] for the solar cells. Table III
lists these results. Equation (6) is the single-diode equation that
is used in the simulation.
To achieve a GaP solar cell with 12.58% efficiency, the following issues must be addressed. In order to improve the quality
of the p-type base layer, a better quality p-type substrate with
fewer impurities needs to be found so that the diffusion length
of the epitaxial layer grown on top of it can be improved [20].
The n-type emitter growth conditions also need to be optimized
to achieve a longer diffusion length. A thin emitter layer would
further improve the collection probability of the minority carriers in this region. Therefore, optimum growth conditions need
to be developed to grow thin n-type GaP (less than 0.3 m)
via LPE. A proper surface passivation method, such as growing
AlGaP on top of the GaP solar cell device [21], would decrease the front surface recombination velocity. Moreover, optimizing the metallization procedure for the GaP solar cells
would improve the FF. Once the models that are described earlier are developed, a GaP solar cell with 12.58% efficiency can
be achieved.

[1] A. De Vos, Detailed balance limit of the efficiency of tandem solar cells,
J. Phys. D, Appl. Phys., vol. 13, pp. 839846, 1980.
[2] F. Dimroth et al., Metamorphic GaInP/GaInAs/Ge triple-junction solar
cell with >41% efficiency, in 34th IEEE Photovolt. Spec. Conf., Philadelphia, PA, 2009, pp. 19331937.
[3] R. King et al., Band-gap-engineered architectures for high-efficiency
multi-junction concentrator solar cells, in Proc. 24th Eur. Photovolt.
Solar Energy Conf. Exhib., Hamburg, Germany, 2009, pp. 5561.
[4] S. Wojtczuk et al., InGaP/GaAs/InGaAs concentrators using Bi-facial
epi-growth, in 35th IEEE Photovolt. Spec. Conf., Honolulu, HI, 2010,
pp. 12591264.
[5] (2011). [Online]. Available: http://cleantechnica.com/2011/04/19/solarjunction-breaks-concentrated-solar-world-record-with-43-5-efficiency/.
[6] M. Nell and A. Barnett, The spectral pn junction model for tandem
solar cell design, IEEE Trans. Electron Devices, vol. ED-34, no. 2,
pp. 257266, Feb. 1987.
[7] J. Gray et al., Efficiency of multijunction photovoltaics systems, in Proc.
33rd IEEE Photovolt. Spec. Conf., San Diego, CA, 2008, pp. 16.
[8] X. Lu et al., Gallium phosphide solar cells for multijunction systems, in
Proc. 34th IEEE Photovolt. Spec. Conf., Philadelphia, PA, 2009, pp. 1113
1116.
[9] A. Barnett et al., Very high efficiency solar cell modules, Prog. Photovolt.: Res. Appl.., vol. 17, pp. 7583, 2009.
[10] X. Lu et al., A high performance multi-junction solar cells fabricated by
liquid phase epitaxy, in Proc. 24th Eur. Photovolt. Solar Energy Conf.
Exhib., Hamburg, Germany, 2009, pp. 299304.
[11] I. Fritz et al., High temperature electron mobilities in LPE-grown gallium
phosphide, J. Electron. Mater., vol. 14, pp. 7384, 1985.
[12] I. Tan et al., Wafer bonding of 75 mm diameter GaP to AlGaInPGaP
light-emitting diode wafers, J. Electron. Mater., vol. 29, pp. 188194,
2000.
[13] S. Bittner et al., GaP solar cells with InGaP quantum well for high
temperature application, presented at the Proc. 37th IEEE Photovolt.
Spec. Conf., Seattle, WA, 2011.
[14] X. Lu et al., Design, fabrication, characterization and analysis of wide
band gap gallium Phosphide solar cells, in Proc. 25th Eur. Photovolt.
Solar Energy Conf. Exhib., Valencia, Spain, 2010, pp. 212217.
[15] C. Allen et al., Results of a gallium phosphide photovoltaic junction with
an AR coating under concentration of natural sunlight, Solar Energy
Mater. Solar Cells, vol. 95, pp. 26552658, 2011.
[16] W. Yang et al., Internal quantum efficiency for solar cells, Solar Energy,
vol. 82, pp. 106110, 2008.
[17] G. Araujo, E. Sanchez, and M. Martil, Determination of the twoexponential solar cell equation parameters from empirical data, Solar
Cells, vol. 5, p. 199, 1982.

220

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 2, APRIL 2012

[18] K. Kazmierski, The effect of tellurium doping and growing conditions

of n-type layer on quantum efficiency of red-emitting gallium phosphide
LEDs, Electron. Technol., vol. 13, pp. 99110, 1982.
[19] W. Shockley, M. Sparkes, and G. Teal, PN junction transistors, Phys.
Rev., vol. 83, pp. 151162, 1951.
[20] H. Alawadhi, R. Vogelgesang, A. Ramdas, T. Chin, and J. Woodall, Indirect transitions, free and impurity-bound excitons in gallium phosphide:
A revisit with modulation and photoluminescence spectroscopy, J. Appl.
Phys., vol. 82, pp. 43314337, 1997.
[21] T. Prutskij et al., GaPAlGaP heterojunction photosensors with photosensitivity maximum at 380420 nm, Solar Energy Mater. Solar Cells,
vol. 37, pp. 349355, 1995.

Xuesong Lu (M11) was born in Fuyang, China. He

received the B.S. and M.S. degrees in materials science from Sichuan University, China, in 2003 and
2006, respectively. He received the Ph.D. degree from
the Department of Electrical and Computer Engineering, University of Delaware, Newark, in 2011. During
his Ph.D. study, his research was mainly focused on
high band gap GaP solar cells grown via the liquid
phase epitaxy method.
He is currently a Senior Process Engineer with
Applied Materials Company, Santa Clara, CA, where
he is working on a project involving high-efficiency crystalline Si solar cells.

Susan Huang received the B.S. degree in physics

from McGill University, Montreal, QC, Canada, and
the M.S. degree in nanoscale science and engineering from the State University of New york at Albany,
where she worked on semiconductor-organic hybrid
solar cells. She received the Ph.D. degree in materials science and engineering from the University of
Delaware, Newark, where his research was focused
on epitaxial growth of GaP on silicon for solar cell
applications.
She is currently with the IBM Thomas J. Watson
Research Center, Yorktown Heights, NY, where she has been involved in the
investigation on novel structures for silicon-based photovoltaics. Her research
interest includes photovoltaics.

Martin B. Diaz was born in Midland, TX. He received the B.S.E.E degree from the University of
Nebraska-Lincoln in 2009. He received the Ph.D. degree in electrical engineering from the University of
Delaware, Newark, in 2011.
His research interest includes high-performance,
next-generation photovoltaic solar cells. In 2009, he
joined the Solar Power Program with the University
of Delaware. His current research is focused on liquid
phase epitaxy growth of IIIV solar cells on silicon.

Nicole. A. Kotulak received the B.A. degree in

physics from Notre Dame of Maryland University,
Baltimore, in 2009. She then joined the University of Delaware as an NSF-IGERT Fellow and received the M.S. degree in electrical and computer
engineering in 2011 for her work on the growth
of GaP on Si via liquid phase epitaxy for applications in very high efficiency photovoltaic devices. She is currently working toward the Ph.D.
degree with the University of Delaware, Newark,
where she is involved in research on high-efficiency
photovoltaics.

Ruiying Hao received the M.S. degree in materials science and engineering from Sichuan University,
China, in 2006, where she worked on CdTe solar cells
and received the Ph.D. degree in electrical and computer engineering from the University of Delaware,
Newark, in 2011 for research on single crystalline
thin Si solar cells based on epitaxial absorber.
She is currently with Crystal Solar Company,
Santa Clara, CA, working on Si solar cells. Her research interest includes photovoltaics.

Robert Opila received the Ph.D. degree in chemistry

from the University of Chicago, Chicago, IL, in 1982.
He was with Bell Laboratories, where he studied
the role of surfaces and interfaces in electronic materials and devices. He was named a Distinguished
Member of Technical Staff and promoted to Technical Supervisor. Since 2002, he has been with the
Department of Materials Science and Engineering,
the University of Delaware, Newark, where his research interest in the surfaces and interfaces of
electronic materials has grown to photovoltaics and
thermoelectrics.
Dr. Opila is a Fellow of the American Vacuum Society. He is currently an
Editor of the Journal of Applied Surface Science.

Allen Barnett (LF06) received the B.S. and M.S.

degrees from the University of Illinois, Urbana, and
the Ph.D. degree from Carnegie-Mellon University,
Pittsburgh, PA, all in electrical engineering.
He has been with the School of Photovoltaics
and Renewable Energy Engineering, University of
New South Wales, Sydney, Australia, as Professor of
Advanced Photovoltaics since September 2010. He
joined the University of Delaware (UD), Newark, in
1976 as the Director of the Institute of Energy Conversion and a Professor of electrical engineering. He
left UD in 1993 to devote his attention full time to AstroPower, Inc., which
became the fourth largest independent solar cell manufacturer in the world. He
returned to UD in 2003 and was the Executive Director of the Solar Power
Program; a Research Professor with the Department of Electrical and Computer
Engineering; and a Senior Policy Fellow with the Center for Energy and Environmental Policy. He has supervised 26 Ph.D. dissertations, including seven
Ph.D. and three M.S. degrees in 2011. He has authored or coauthored more than
280 publications. He holds 28 U.S. patents. He actively consults for government
agencies, institutional investors, and private companies.
He is the recipient of the IEEE William R. Cherry Award for outstanding
contributions to the advancement of photovoltaic science and technology and
the Karl W. Boer Solar Energy Medal of Merit. He is also the recipient of seven
R&D 100 Awards for new industrial products.

THREE TERMINAL SI-SI:GE SOLAR CELLS

1

Lu Wang , Yi Wang , Andrew Gerger , Anthony Lochtefeld , Kevin Shreve , Allen Barnett
Solar Power Program, Electrical and Computer Engineering, University of Delaware, Newark, DE 19716, USA
2
AmberWave, Inc, 13 Garabedian Drive, Salem, NH 03079, USA

ABSTRACT

The multi-junction system can be simplified into the

following structure after the replacement by the three
terminal solar cells.

A 3-terminal Si-Si:Ge tandem solar cell, which consists

of a top silicon solar cell and a bottom Si:Ge solar cell
sharing a common base, has been designed and
fabricated. This solar cell is designed to replace standalone Si solar cell and Si:Ge solar cell in a multi-junction
solar cell system in order to simplify the module,
improve its performance and reduce its cost. Initial
results show that both Si solar cells and Si:Ge solar
cells in this structure can achieve good performance.
The best Si solar cell shows a Voc 598.5mV, Jsc
2
22.9mA/cm and FF 77.3% under one sun, best Si:Ge
2
solar cell shows a Voc 189mV and Jsc 3.26mA/cm
below silicon filter. None of these solar cells have antireflection coatings.

Figure 2 Schematic of a multi-junction solar cell

system with a three terminal solar cell

INTRODUCTION
Si:Ge solar cell grown on Si can theoretically contribute
an absolute 5.7% efficiency below Si to a high
performance multi-junction solar cell system at 50 suns
as shown in Figure 1 [1 2]. In this system, the optics
needs to be designed carefully for the back side of
transparent Si solar cell to optimize the long wavelength
light transmission. Photon losses cannot be eliminated
in this way. In this paper a 3-terminal Si-Si:Ge tandem
solar cell is designed and fabricated by adding a Si solar
cell on Si substrate of Si:Ge solar cell. It can replace
stand-alone Si solar cell and Si:Ge solar cell in the
multi-junction solar cell system as shown in Figure 1 [2].

Single junction silicon solar cells have achieved 25%

efficiency, which is approaching realistic efficiency limits
[3]. To further improve solar cells efficiency, Si-Si:Ge
heterostructure solar cells have been developed to
enhance long wavelength light performance [4,5].
Encouraging results of Si:Ge solar cells grown on Si
substrates were reported by our group [1]. This makes
the combination of Si solar cells and Si:Ge solar cells
practical. However, the efficiency of a monolithic
stacked two terminal Si-Si:Ge solar cells is limited by
current mismatch effect. In this paper, a 3-terminal SiSi:Ge tandem solar cell consisting of a top silicon solar
cell and a bottom Si:Ge solar cell sharing the same
base, which can take advantage of both silicon solar
cells and Si:Ge solar cells and overcome the current
mismatch effect and optics losses between them, will be
reported.
THREE TERMINAL SOLAR CELLS DESIGN AND
EQUIVALENT CIRCUIT
The basic structure of designed 3-terminal Si-Si:Ge
tandem solar cell is illustrated in Figure 3. The top cell is
a silicon solar cell with a band gap of 1.12eV. Contacts
for base and emitter of Si solar cell are on the same
side. Al contact of Si solar cell is also shared by the
bottom Si:Ge solar cell with approximately 88% Ge
concentration. 5um P type base and 1um N type emitter
of Si:Ge solar cell was grown by Reduced pressure
chemical vapor deposition technology (RPCVD) at
AmberWave Inc. On top of emitter, a N type Si cap is
grown to passivate Si:Ge surface.

Figure 1 Schematic of a multi-junction solar cell

system [2]

1
978-1-4244-9965-6/11/$26.00 2011 IEEE

000286

contact were connected to the IV tester. While the Si:Ge

solar cell emitter was floating. Table 1 shows measured
data at one sun.
2

Si solar cell
Jsc (mA/cm ) Voc (mV)
FF
Outdoor 1 sun
22.9
598.5
77.3%
Table 1 Summary of a Si solar cell performance in
group one
When testing the Si:Ge solar cells IV curves, no good IV
curve was obtained, all curves behaved like that a Si PN
junction and a Si:Ge PN junction were connected in
series. The back side N type silicon diffused layer was
not etched away completely before P and N type Si:Ge
epitaxial layers were grown on it, so there were two PN
junctions on the back side. A second design was
developed to eliminate this problem based on an
epitaxial silicon emitter.

Figure 3 Schematic vertical cross section of a 3terminal Si-Si:Ge tandem solar cell
The current flow of the 3-terminal solar cells follows. All
generated holes will flow towards the common base,
and be collected by the base Al contact. Electrons
generated by silicon cell will flow towards the upper (Si)
emitter, and electrons generated in the Si:Ge cell will
flow towards the bottom cell (Si:Ge) emitter. The
equivalent circuit of ideal solar cells consists of two PN
junctions sharing common base (Figure 4).

Group Two Fabrication and Performance

These solar cells experienced no diffusion. Silicon solar
cell emitters with a depth 1 um and doping density
17
-3
5x10 cm were grown using RPCVD on a P type
silicon substrate. Next Si:Ge epitaxial layers were grown
on the other side of substrate, including a 5 um 86.3%
Ge P type Si:Ge base with doping density around
17
-3
10 cm , a 1 um 86.3% Ge n type Si:Ge layer with
17
-3
doping density 10 cm and a 200nm n+ Si cap by
AmberWave, Inc..
The rest fabrication steps are the same as in group one,
and these solar cells were tested in the same way as
previously described. And tested results are shown in
table 2 and table 3.

Figure 4 Equivalent circuits of an ideal three

terminal Si-Si:Ge solar cell

Si solar cell
Jsc (mA/cm )
Voc (mV)
FF
Outdoor 1 sun
17.5
583.3
81.8%
Table 2 Summary of a Si solar cell performance in
group two

EXPERIMENTALS AND RESULTS

To improve the performance of 3 terminal Si-Si:Ge solar
cells, three groups of solar cells were fabricated.. Their
fabrication process and test results are reported in the
following paragraphs.

Si:Ge solar cell Jsc (mA/cm ) Voc (mV)

FF
Outdoor 1 sun
1.92
180.0
No Data
16.7 suns
32.05
213.9
43.26%
Table 3 Summary of a Si:Ge solar cell performance
in group two

Group One Fabrication and Performance

The performance of the silicon solar cell is fair.

Compared to the silicon solar cell in group one, the
diffused emitter is better. The performance of Si:Ge was
a little bit surprise, because there were a lot of scratches
on the Si:Ge epitaxial layer after growth which replicates
any defect on the surface of silicon wafer. The emitter
growth of this Si:Ge solar cell is good.

In the first group, silicon wafers were POCl3 diffused to

form N type emitter (45/sq), then the back side
diffused region was etched so that epitaxial layers could
be grown on P type Si substrate. A 5 um P type Si:Ge
epitaxial layer as base with Ge concentration 91% and a
0.2 um N type silicon layer as emitter were grown
sequentially by AmberWave, Inc. using RPCVD. Silicon
cell and Si:Ge cell emitters were patterned via HNA (HF:
Nitric: Acetic) etching. Emitter contacts and base Al
contacts were deposited using E-beam evaporation
followed by RTP annealing.
Testing was performed outdoors. The Si solar cell and
Si:Ge solar cell were tested separately. To measure the
performance of the Si solar cell in a three terminal solar
cell structure, the Si solar cell faces towards the sun
directly, and its emitter contact and the Al common base

A third group was made to improve the performance of

Si:Ge solar cells, since great improvement of efficiency
of Si:Ge solar cells has taken place in our group[1].
Figure 5 are actual pictures of 3 terminal Si-Si:Ge solar
cells.

2
978-1-4244-9965-6/11/$26.00 2011 IEEE

000287

9.7 suns
79.08
282
44.9%
Table 5 Summary of group three Si:Ge solar cell
performance
2

Si solar cell
Jsc (mA/cm ) Voc (mV)
FF
Outdoor 1 sun
16.0
553.2
78.9%
Table 6 Summary of group three Si solar cell
performance
Figure 5 Actual pictures of Si solar cell (left) and
Si:Ge solar cell (right)

The performance of the Si:Ge solar cell is improved as

expected. While the silicon solar cell Voc is low due to
the rough surface of substrate on which N type silicon
emitter was grown.

Our Best Si:Ge Solar Cell Performance

The best mono-junction Si:Ge solar cell has been made
in our group and these best results will be integrated
and progress reported at the conference [1]. Figure 6
shows the IV curve under 30 suns, and table 4 shows
measured data under 30 suns [1].

Further improvement can be made by using a double

side polished substrate, so that both silicon epitaxial
layer and Si:Ge epitaxial layer can be grown on polished
surfaces. A diffused silicon emitter as in group one can
also be reconsidered for the silicon solar cells instead of
epitaxial layer. Additionally light trapping optics can be
included in an advanced structure.
SUMMARY
A Si solar cell and a Si:Ge solar cell can be integrated
into a three terminal Si-Si:Ge solar cell, which can be
applied in the multi-junction solar cell system to replace
the former independent two solar cells. This change can
both simplify the module, improving the performance
and reduce its cost. Better performance of 3 terminal SiSi:Ge solar cells will be achieved by improving structure
design, fabrication and application of light trapping
optics.

Figure 6 IV curves of 88%Ge Si:Ge solar cell below

Si filter at 30 suns [1]
Concentration

Ge
Jsc
Voc
FF
Eff
2
(%)
(mA/cm ) (mV) (%)
(%)
1 sun
88
3.41
263
68
0.61
102.3
30 suns
88
350
66
0.79
Table 4 Summery of the best Si:Ge solar cell
performance [1]

ACKNOWLEDGEMENT
This research was, in part, funded by the U.S.
Government Defense Advanced Research Projects
Agency under
Agreement
No.:
HR0011-07-90005. The views, opinions, and/or findings contained
in this article/presentation are those of the authors and
should not be interpreted as representing the official
views or policies, either expressed or implied, of the
Defense Advanced Research Projects Agency or the
Department of Defense.
The author would also like to thank China Scholarship
Council for providing stipends during his Ph.D.
education.

Group Three Fabrication and Performance

The 3 terminal Si-Si:Ge solar cells growth conditions,
dopant concentration had been updated, according to
the best single junction Si:Ge solar cell [1]. And Si:Ge
layer was grown on the polished side of substrate. But
silicon epitaxial layer was grown on the unpolished side
of substrate. The emitter of the silicon solar cell is 0.5
18
-3
um thick with a doping density 1x10 cm , the P type
15
-3
Si:Ge base is 5 um thick with doping density 5x10 cm
and 88% Ge, and N type Si:Ge emitter layer is 1 um
18
-3
thick with doping density 1x10 cm and 88% Ge.
Above there is a 200nm N type Si cap with a doping
18
-3
density 5x10 cm . The rest of the fabrication steps and
testing are the same as previously described.
Tested results are shown in table 5 and table 6.
Test Condition
One sun
Under Si filter

Jsc (mA/cm )
8.16
3.26

Voc (mV)
215
189

REFERENCES
[1] Yi Wang, et al. Design, Fabrication, Characterization
and Improvement of Si:Ge Solar Cell below Si Solar
Cell In A Multi-junction Solar Cell System, this
conference.
[2] Allen Barnett, et.al. Very High Efficiency Solar Cell
Modules, Prog. Photovolt: Res. Appl. 17 (2009) 7583.

FF
56.7%
No data

3
978-1-4244-9965-6/11/$26.00 2011 IEEE

000288

[3] Martin A. Green*, The Path to 25% Silicon Solar

Cell Efficiency: History of Silicon Cell Evolution,
Prog. Photovolt: Res. Appl. 2009; 17:183189.
[4] Khalid Said, et al. Design, Fabrication, and Analysis
of Crystalline Si-SiGe Heterostructure Thin-Film
Solar Cells, IEEE TRANSACTIONS ON
ELECTRON DEVICES, VOL. 46, NO. 10,
OCTOBER 1999.
[5] S.A. Healy and M.A. Green, Efficiency
enhancements in crystalline silicon solar cells by
alloying with germanium, Solar Energy Materials
and Solar Cells 28 (1992) 273-284 North-Holland.

4
978-1-4244-9965-6/11/$26.00 2011 IEEE

000289

FIELD PERFORMANCE EVALUATION AND MODELLING OF SPECTRALLY TUNED

QUANTUM-WELL SOLAR CELLS
Matthew Norton1 , Alison Dobbin2, Alexander Phinikarides1,
Tom Tibbits2, George E. Georghiou1 and Sylvain Chonavel3
1
Photovoltaic Technology, ECE, University of Cyprus, Kallipoleos 75, Nicosia 1678, Cyprus
2
QuantaSol Ltd, 14 Kingsmill Business Park, Chapel Mill Road, Kingston upon Thames, Surrey, KT1 3GZ, UK
3
Whitfield Solar Ltd., Webbs Court, 8 Holmes Rd., RG6 7BH Reading, UK

INTRODUCTION
ABSTRACT
Monolithic multi-junction solar cells are becoming prevalent in concentrator photovoltaic (CPV) systems due to
their high demonstrated conversion efficiencies. However, these devices often operate at sub-optimal levels
due to current mismatch losses arising between each
p-n junction as a result of natural variation in the solar
spectrum at the earths surface. The use of quantum
wells in solar cell design affords improved control over the
spectral response of the cells via band-gap engineering.
This makes it possible to tailor the spectral response of
a cell for optimum performance under a given annual
spectral resource. Establishing the optimum spectral
response for a given location is a major challenge of this
approach to cell design, and relies heavily on averaged
and modelled data based on combinations of satellite and
sparse ground measurements. In addition, there is only
limited field experience of such technology to date.
The purpose of this work is to investigate through
experiment the effect of incorporating multi-quantumwell (MQW) structures into photovoltaic cells to respond
to a specific range of annual irradiation spectra and
to compare the results obtained with those predicted
through modelling. A number of triple-junction cells of
different design have been placed side-by-side on an
accurate solar tracker, and current-voltage characteristics
of each taken at regular intervals over a few months.
The outputs are then compared to those predicted by
a model of cell performance that includes simulated
spectra, generated with the SMARTS program, specific
to that location and period of time.
Results obtained from outdoor testing indicate that the
cells designed to provide improved current matching
under the spectral conditions in which they have
been tested have performed more consistently than
conventional cell designs. This is evident when the
short-circuit current is normalised to a fixed direct
normal irradiation and temperature, and plotted against
atmospheric depth. Detailed analysis of the spectral
resource is expected to reveal additional information on
the performance of the cells.

978-1-4244-9965-6/11/$26.00 2011 IEEE

High-efficiency solar cells are finding promising terrestrial applications within concentrator photovoltaic (CPV)
systems that offer to deliver higher overall conversion
efficiencies than conventional flat panel photovoltaic
installations. Most high-efficiency cells are designed as
monolithic devices with two or three junctions in order
to exploit a broad range of the solar spectrum. Whilst
the solar spectrum is constant outside of the atmosphere,
terrestrial deployment of these cells can be complicated
by the varying content of the solar spectrum due to
atmospheric attenuation processes. The spectral content
of sunlight over a year varies with location on earth, with
the result that a similar cell will not necessarily deliver the
same energy yield from place to place, even if the total
annual irradiation levels are similar.
Thanks to advances in nanotechnology, quantum well
solar cells can now be designed and fabricated to
deliver optimum performance under the specific spectral
resource of a given location on earth. This is done
by careful design of materials with band-gaps suited
to the particular spectra available. Compared with
a conventional multi-junction cell design, a tuned cell
should then exhibit a more stable power output over a
range of atmospheric conditions and depths.
Cells have now been designed with a variety of different
spectral responses [1, 2]. The designs have been based
on the results of analyses of the spectral resource at
a number of locations. These analyses make use of
a combination of real and modelled data based on
satellite and ground-based measurements. Whilst this
approach offers convenience and avoids the cost of
long-term site measurements, it is also likely to introduce
a wide margin of error. Consequently, it is desirable to
obtain real performance measurements of a range of cell
designs under varying environmental conditions to both
determine the relative advantages of tuning cells, and to
correlate the results with the models used to design the
cells.

000534

APPROACH

1000

Experimental apparatus

900

To investigate advantages of using spectrally tuned cells

over conventional designs, a side-by-side comparison
of the output of a range of cells operating under
identical conditions is needed. Ideally, all cells will
be operating under identical light concentration levels
and intensity distribution, with identical spectra and
operating temperatures. To this end, a number of
individual lens-cell combinations have been fabricated
that achieve concentration levels of 310 suns. These
have been mounted side-by-side on an accurate solar
tracker in Nicosia, Cyprus and connected to a currentvoltage (I-V) characterisation system using 4-wire Kelvin
probes. A calibrated pyrheliometer and spectroradiometer collimator are mounted alongside these, allowing
the measurement of direct normal irradiance (DNI) and
spectrum together with the electrical characteristics of
each cell. In addition, ambient and cell temperatures
have been recorded as well as wind speed and direction,
humidity and pressure. I-V measurements of each cell
are performed every 5 minutes.

800

Data analysis
To more clearly examine the relationship between
atmospheric depth and cell output, the data are filtered
to remove days of poor atmospheric clarity. This is done
by using the DNI as an indicator of clearness, k , based
on a method described by Merten and Andreu [3, 4]. In
this approach, DNI is plotted against atmospheric depth,
and then a curve is fitted to describe the upper limit
of the recorded DNI at each atmospheric depth, as in
Figure 1. In effect, the fitted curve describes a very clear
atmosphere (k approaches 1). The clearness value of
each data point can then be calculated as a function of the
DNI, I , the top-of-atmosphere DNI, I0 , and atmospheric
depth, M , as in equations 1 and 2 below.

I(M ) =

I
I(M )

(1)

I0 e0.2M

(2)

Modelling
Hourly solar spectra for the test location in Nicosia,
Cyprus were calculated using the SMARTS model. Inputs
to the SMARTS model included Aerosol Optical Depth
data taken from the CUT-TEPAK NASA AErosol RObotic
NETwork site in Southern Cyprus, and precipitable
water calculated from on-site measurements of ambient

978-1-4244-9965-6/11/$26.00 2011 IEEE

Data points
Fitted curve

DNI [Wm2 ]

700
600
500
400
300
200
100
0

10

15

20

25

30

35

Atmospheric Depth [AM]

Figure 1 Scatter plot of direct normal irradiance

versus atmospheric depth.
temperature and humidity.
The DNI and Average
Photon Energy (APE) of the modelled spectra have been
compared with on-site measurements and a good match
between model and data was found [5, 6]. These spectra
were then filtered through the transmissivity function of
the CPV optics and used in a calculation of hourly power
output for various cell designs.
The total energy harvest predicted for each cell design
was compared and the design expected to produce
most energy was chosen to go on-sun in Cyprus. A
standard Quantasol Multi-Quantum Well (MQW) cell that
is current matched under standard test conditions (25 C,
ASTM AM1.5d G173-03 spectrum) and a conventional
triple-junction solar cell were put on-sun in the same
system / location. The cell and spectral models were
used to generate detailed data sets for the anticipated
cell performance, which included short-circuit current,
open circuit voltage, and fill factor as well as the
maximum power output. These data could then be
compared with the real measurement results to indicate
the accuracy of the predictions. Figure 2 shows the result
of a simulation performed to calculate the anticipated
current generation variation of a triple junction cell across
different atmospheric depths.
RESULTS AND DISCUSSION
The collected data have been analysed to look for trends
in the cell output over a range of atmospheric conditions.
Using the techniques described in this paper, the data
were filtered to examine only reasonably clear days, i.e.
where k was above 0.75. The depth of atmosphere was
calculated using the time of day and pressure readings

000535

1.1
Relative Isc

from the site, and the short-circuit current (Isc ) was

normalised to the current output of the cell at a DNI of
900 Wm-2 and AM2. By plotting the relative observed
short-circuit current against the depth of atmosphere, as
shown in Figure 3, it can be seen that the conventional
cell design displays a lower relative output at higher
atmospheric depths than the MQW cell.

0.9
0.8
0.7
1.5

2.5
3
3.5
4
Atmospheric Depth [AM]

4.5

2.5
3
3.5
4
Atmospheric Depth [AM]

4.5

1
0.9
1.1
0.8
0.7

1.5

2
2.5
3
3.5
4
Atmospheric Depth [AM]

4.5

Figure 2 Normalised current output of the simulation

performed for a triple junction cell over the outdoor
measurement period.
The behaviour of the MQW cell across the range of
atmospheric depths studied implies that it is limited by
a single junction. In contrast, the plot of the normalised
conventional cell current exhibits the characteristic knee
where the limiting junction changes due to the variation in
spectrum.
for each junction of the 3J stack ,
Modelled IscA Zs
indicate that the top cell will be predominantly responsible
for limiting the current over the range of modelled spectra.
This is expected from MQW cells that have a middle
junction with an increased spectral response range and
hence current output. However, there was a discrepancy
between the rate of current drop predicted by the model,
and the observed behaviour. It is not yet clear whether
this is an artefact introduced by the cell housing and
measurement accuracy, or the accuracy of the model
in predicting cell operating temperatures and spectral
content of light for the specific measurement period.
Further measurement over a longer period of time,
along with more detailed modelling of cell behaviour will
help identify the reason for discrepancies between the
modelled and measured cell performance.
An expected result of using MQW structures is an
increased performance in red-rich spectra that occur at
high atmospheric depths, and this was observed. The
witnessed fill-factors for the two cells, shown in Figure
4, agreed with the results presented in Figure 3. The
fill factor for the MQW cell remained almost constant,
decreasing slightly with increased atmospheric depth.
Conversely, the conventional cell displayed an increased

978-1-4244-9965-6/11/$26.00 2011 IEEE

Relative Isc

Normalised Isc [A]

1.1

1
0.9
0.8
0.7
1.5

Figure 3 Scatter plots of side-by-side current measurements of two different cell designs, filtered to
show clear sky behaviour. (Above: a conventional
cell output; below: a MQW cell output)
fill factor at higher atmospheric depths, consistent with a
current mismatch between junctions.
An examination of the power output of the two cells has
so far not been able to indicate a clear difference in
performance characteristics of the two designs. This is
because whilst the Isc of the conventional cell dropped
with increased atmospheric depth, the corresponding
increase in fill factor maintained the power output of the
cell for the range of spectral conditions tested so far.
CONCLUSIONS
The depth of atmosphere has been shown to be a
useful approximate indicator of spectral content when
measured data is filtered to reveal clear-sky conditions.
The particular analyses conducted show that a k value
calculated as described from the atmospheric depth and
DNI alone is a convenient parameter to use in filtering
data for clear sky conditions.
The results obtained so far indicate that there is a
more consistent output of the tuned quantum well solar
cells over a range of atmospheric depths. Since the
atmospheric depth directly affects the spectrum of the
light incident on the cells, this can be interpreted as an
indication that the cells are better suited to the range of
spectra that they have been exposed to. Further work

000536

Progress in Quantum Well Solar Cells, page 187.

CRC, 2010.

Fill Factor [a.u.]

1
0.95
0.85
0.8
0.75
0.7
1.5

Fill Factor [a.u.]

[3] J. Merten and J. Andreu. Clear separation of seasonal

effects on the performance of amorphous silicon solar
modules by outdoor I/V-measurements. Solar Energy
Materials & Solar Cells, 52:1125, 1998.

0.9

2.5
3
3.5
4
Atmospheric Depth [AM]

4.5

1
0.95

[5] Alison Dobbin, Matthew Norton, George E.

Georghiou, Matthew Lumb, and Tom N. D. Tibbits.
Energy harvest predictions for a spectrally tuned
multiple quantum well device utilising measured and
modelled solar spectra. In CPV-7 conference, Las
Vegas, USA, April 2011.

0.9
0.85
0.8
0.75
0.7
1.5

[4] T.R. Betts, R. Gottschalg, and D.G. Infield. Spectral

irradiance correction for pv system yield calculations.
In 19th European Photovoltaic and Solar Energy
Conference. WIP Munich, 2004.

2.5
3
3.5
4
Atmospheric Depth [AM]

4.5

[6] M. Norton, A. Dobbin, and G.E.Georghiou. Solar

spectrum prediction and comparison with measurements. In CPV-7 conference, Las Vegas, USA, April
2011.

Figure 4 Scatter plots of side-by-side normalised fill

factor variations of two different cell designs, filtered
to show clear sky behaviour. (Above: a conventional
cell; below: a MQW cell)
is needed to reconcile the predicted short-circuit current
variation across atmospheric depths with the observed
behaviour, and this will be done through a more detailed
analysis of the temperature and spectral model accuracy.
The tests described in this paper have indicated that the
conventional triple junction cells suffer more significant
current mismatch losses than those cells constructed with
MQW structures within them. The work has already
demonstrated an improvement in current output over a
range of spectral conditions, and further measurement is
underway to quantify the effect on annual power output.
REFERENCES
[1] M.P. Lumb, A.L. Dobbin, D.B. Bushnell, K.H. Lee, and
T.N.D. Tibbits. Comparing the energy yield of (IIIV) multi-junction cells with different numbers of subcells,. In 6th International Conference on Concentrating Photovoltaic Systems, CPV-6, Freiburg, Germany,
April 2010. American Institute of Physics.
[2] K.W.J. Barnham, I.M. Ballard, B.C. Browne, D.B.
Bushnell, J.P. Connolly, N. J. Ekins-Daukes,
M. Fuhrer, R. Ginige, G. Hill, A. Ioannides, et al.
Nanotechnology for Photovoltaics, chapter 5: Recent

978-1-4244-9965-6/11/$26.00 2011 IEEE

000537

Study on the effect of reverse current blocking layer for concentrator solar cells
1

Myoung-Gyun Suh* , Ji-Eun Chang , Dong-Ho Kim , Kyu-Sang Kim , and Yun-Gi Kim
Energy Lab, Samsung Advanced Institute of Technology, Yongin-si, Gyeonggi-do, Republic of Korea
2
R&D CAE Group, Samsung LED, Suwon-si, Republic of Korea

ABSTRACT
In high concentration photovoltaic (HCPV) systems, III-V
multijunction solar cells work under high concentration
2
condition at a small size, normally less than 1cm . In this
case, the metal coverage ratio and the current density
become larger than those in the case of flat-plate solar
cells. Due to the high current density, combined with the
sheet resistance of solar cells, most of currents reverse to
photocurrents flow vertically under front metal grids. This
non-uniformity is critical when the metal coverage ratio is
large as the case of concentrator solar cells.

In this paper, the effects of metal coverage ratio and nonuniform reverse current density (i.e. forward bias current
density) on the performance of InGaP/InGaAs/Ge triplejunction solar cells under high concentration were
analyzed in detail. Moreover, by introducing reverse
current blocking layer between metal busbar and AlInP
window layer, we have successfully reduced the unwanted
reverse current and achieved significant improvement in
cell efficiency under high concentration condition.

Introduction

(a)

Recently, the efficiency of III-V multijunction solar cell has

reached above 40% at high concentration conditions
[1,2,3]. To optimize solar cell performance under high
concentration, the effects of metal grid design, sheet
resistance and tunnel junction resistance with increasing
concentration have been studied [4,5]. Since concentrator
solar cells have much smaller size than ordinary flat-plate
solar cells, the effect of series resistance of metal grids
lessen, while perimeter recombination loss and metal
coverage ratio (front metal grid area / cell size) become
much more important [5].
Studies on the dark I-V characteristics of solar cells under
high concentration have shown that the current density
under metal grids is much higher than the current density
between metal grids [7]. This non-uniformity of current
density goes worse as the current density level increases.
This is understood to be the result of the sheet resistance
of window layer and emitter layer, and it can be reduced
by optimizing metal grid pitch and doping level of epitaxial
layers. However, we cannot avoid the sheet resistance
induced loss under even higher current density since it is
proportional to the square of current density.
In this study, we have analyzed the effect of non-uniform
reverse current density (i.e. forward bias current
density) on the performance of solar cells and we suggest
one way to solve this problem, reduction of effective metal
coverage without disturbing photocurrent collection of
metal grids. First, we evaluated the effect of metal busbar
size, which is the largest portion of metal coverage, on the
solar cell performance with varying concentration ratio.
Furthermore, to improve the solar cell efficiency under
high concentration, metal busbar is deposited on antireflection coating (ARC) layer, which also acts as reverse
current blocking layer (RCBL).

(b)

Figure 1. Epitaxial structure of InGaP/InGaAs/Ge

triple-junction solar cells used in this study: (a)
Schemetic image (b) SEM image

[*] Corresponding author : mg.suh@samsung.com

978-1-4244-9965-6/11/$26.00 2011 IEEE

000556

Experimental

Results and Discussion

Through Metal-Organic Vapor Phase Epitaxy (MOVPE)

method, weve grown InGaP/InGaAs/Ge triple-junction
solar cells on p-type Ge substrate (Figure 1). After the
epitaxial growth process, several steps of device
fabrication have followed; n-metal (Ti/Au/Ag) deposition,
Ge back surface lapping, GaAs contact layer etching, pmetal (Au/Ag) deposition, and anti-reflection coating
(ARC) layer deposition. Other steps, such as wafer
cleaning, photolithography and dicing, have done properly
in the fabrication process. All the deposition processes are
done by e-beam evaporator and MgF2/ZnS double layers
are used for ARC. For the fabrication of solar cells with
reverse current blocking layer (RCBL), the n-metal grid
fingers are formed as same, but the n-metal busbar is
deposited after the ARC process (Figure 2).
The solar cells used in this study have approximately
28.50.5% conversion efficiency at 1 sun (AM1.5G). The
entire cell size is 0.5cm x 0.6cm and the active area of
solar cells is 0.5cm x 0.5cm. Metal grid finger width and
pitch are fixed to 8 and 140 , respectively.
The cell efficiency under 1 sun condition was measured
using an ordinary class A solar simulator. To correct the
spectral difference from the standard spectrum, a triplejunction cell with the same epi-structure was used as a
reference cell. Moreover, the measured data were crosschecked with subcell currents calculated through the
convolution of measured External-Quantum-Efficiency
(EQE) and standard photon density spectrum.
The cell efficiency under high concentration condition was
measured using Spectrolabs T-HIPSS (Terrestrial-High
Intensity Pulsed Solar Simulator).

To analyze the effect of metal coverage ratio on

concentrator solar cells, we have measured dark I-V
curves with three different effective metal coverage ratios.
From Figure 3(a), we could see clearly different two
regimes; metal coverage independent regime at low
current level and metal coverage dependent regime at
high current level. This can be explained as following; at
low current level, the current density is uniform and the
forward bias current doesnt depend on the metal
coverage ratios. However, at high current level, the current
density under metal grids is much higher due to the
increased effect of sheet resistance. Hence, the forward
bias current depends on the metal coverage ratios.
(a)

0.7
0.6

0.010

0.4

0.008

Current (A)

Current (A)

0.5

effective metal coverage ratios

0.01
0.05
0.34

0.3

0.006

0.004

0.002

0.2

0.000
0.0

0.5

0.1
0.0
0.0

1.0

1.5

2.0

2.5

3.0

Voltage (V)

0.5

1.0

1.5

2.0

2.5

3.0

Voltage (V)
(b)

Ratio of Dark Currents

effective metal coverage ratios

0.01
0.05
0.34

0
2.5

2.6

2.7

2.8

2.9

3.0

Voltage (V)
Figure 2. Schematic device structure of solar cells w/
and w/o reverse current blocking layer (RCBL)

978-1-4244-9965-6/11/$26.00 2011 IEEE

Figure 3. (a) Dark I-V curve of solar cells with different

effective metal coverage ratios. (b) Ratio of dark
currents above 2.5V (normalized to the case of
effective metal coverage ratio 0.01).

000557

w/ RCBL
0.067
28.45 %
30.08 %
30.90 %
29.43 %
26.95 %

Metal Coverage Ratio

Cell Efficiency (1sun)
Cell Efficiency (38 suns)
Cell Efficiency (75 suns)
Cell Efficiency (299 suns)
Cell Efficiency (454 suns)

w/o RCBL
0.229
28.30 %
29.54 %
30.35 %
27.73 %
26.21 %

Table 1. Cell efficiency of solar cells w/ and w/o RCBL

(a)

Current Density (mA/cm )

15

10
w/ RCBL
w/o RCBL

Voc (V)
2
Jsc (mA/cm )
FF (%)
Efficiency (%)

0
0.0

0.5

w/ RCBL
2.383
13.73
86.95
28.45

1.0

w/o RCBL
2.379
13.70
86.83
28.30

1.5

2.0

2.5

Voltage (V)
(b)

2.0

454 suns (w/o RCBL)

299 suns
75 suns
38 suns
454 suns (w/ RCBL)
299 suns
75 suns
38 suns

1.8

Current (A)

1.6

454 suns

1.4
1.2
1.0

0.6

0.2
0.0
0.0

Conclusion

299 suns

0.8

0.4

In Figure 3(b), we can also see that the metal coverage

ratio dependence of dark current increases as the forward
bias increases. However, there seems to be no welldefined relationship.
After analyzing the dark I-V characteristics, we have
measured light I-V curves of solar cells with two metal
coverage ratios, 0.067 and 0.229. The two solar cells have
exactly the same cell size, active area and grid design, but
have different effective metal coverage ratios by
controlling the area of RCBL.
Figure 4(a) shows that the cell efficiency at 1 sun has
increased 0.15%, from 28.30% (w/o RCBL) to 28.45% (w/
RCBL), which is negligibly small. This is because, at 1 sun,
the current density is so low that the sheet resistance does
not significantly disturb current spreading and the current
density is relatively uniform. Hence, metal coverage ratio
is less important than the cell size and the recombination
current under 1 sun condition.
While, the difference in the cell efficiencies become larger
under concentration conditions (Table 1). At 75 suns, the
cell efficiency has increased significantly from 30.35%
(w/o RCBL) to 30.9% (w/ RCBL). At higher concentration
conditions, 299 suns and 454 suns, an unexpected
change of the slope in I-V curve is found (Figure 4), but
still a meaningful improvement is observed.
This can be understood that RCBL reduces the reverse
current (i.e. forward bias current), which mostly flows
vertically under metal grids due to high current density and
sheet resistance. Consequently, the maximum power point
shifts toward upper-right in the light I-V curve, which
means efficiency increase. Here, it is important not to
disturb photocurrent collection and not to increase series
resistance when we introduce RCBL.
Since the front metal grid design is not optimized for high
concentration condition, samples used in this experiment
show relatively small efficiency increase (< 3%) and peakefficiency concentration (< 100 suns). By optimizing grid
design and lowering sheet resistance, it is possible to
further increase the cell efficiency and the peak-efficiency
concentration ratio.

75 suns
38 suns

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Voltage (V)

Figure 4. Light I-V curve of solar cells w/ RCBL (red)

and w/o RCBL (black) ; (a) under 1sun (b) under
concentration

978-1-4244-9965-6/11/$26.00 2011 IEEE

We have studied the effect of metal coverage ratio on the

performance of concentrator solar cells. In the I-V curves,
the operating condition clearly divides into two regimes;
low current density regime where the total current is
independent of metal coverage ratio, and high current
density regime where the total current is dependent of
metal coverage ratio. This is due to the non-uniformity of
current density, which is induced from the effect of sheet
resistance at high current density.
Therefore, it is very important to reduce metal coverage
ratio for concentrator solar cells without disturbing
photocurrent collection and increasing series resistance.
By introducing reverse current blocking layer, the cell
efficiency has increased 0.55% at 75 suns, which is larger
than 0.15% at 1 sun. We believe that this study about the

000558

metal coverage ratio and the solar cell structure with

RCBL could be helpful for the development of high
efficiency concentrator solar cells.

ACKNOWLEDGEMENT
The authors gratefully acknowledge the technical support
by Jin-Soo Moon and Tak-Kyun Ro in the module
processing. The authors also thank to previous team
members, Joo-Sung Kim, Sang-Moon Lee, and Taek Kim.
REFERENCES
[1] R.R. King, et al., Band-Gap-Engineered Architectures
for High-Efficiency Multijunction Solar Cells, Proceedings
th
24 European Photovoltaic Solar Energy Conference and
st
th
Exhibition, Hamburg, Germany, 21 -25 September 2009.
[2] J.Geisz, D.Friedman, J.Ward, et al., 40.8% efficient
inverted triple-junction solar cell with two independently
metamorphic junctions, Appl. Phys. Lett. 93 (12),
p.123505, 2008
[3] W.Guter, J.Schne, S.Philipps, M.Steiner, G.Siefer,
A.Wekkeli, E.Welser, E.Oliva, A.Bett, and F.Dimroth,
Current-matched triple-junction solar cell reaching 41.1%
conversion efficiency under concentrated sunlight, Appl.
Phys. Lett. 94, p.223504, 2009
[4] K.Nishioka, T.Takamoto, et al., Evaluation of
InGaP/InGaAs/Ge
triple-junction
solar
cell
and
optimization of solar cells structure focusing on series
resistance for high-efficiency concentrator photovoltaic
systems, Solar Energy Materials and Solar Cells, 90 (9),
pp. 1308-1321.
[5] Omer Korech, Baruch Hirsch, Eugene A. Katz, and
Jeffrey M. Gordon, High-flux characterization of ultrasmall
multijunction concentrator solar cells, Appl. Phys. Lett. 91,
p.064101, 2007
[6] G.Duggan and I.M.Ballard, Evaluation of competing
triple junction concentrator cells by electroluminescence,
th
34 IEEE PVSC, pp. 655-659, 2009
[7] B.Galiana, C. Algora, and I. Rey-Stolle, Explanation
For the Dark I-V Curve of III-V Concentrator Solar Cells,
Prog. Photovolt: Res. Appl. 16, pp. 331-338, 2008

978-1-4244-9965-6/11/$26.00 2011 IEEE

000559

THIN CRYSTALLINE SILICON SOLAR CELLS WITH METALLIC BACK REFLECTOR

1

C. Paola Murcia , Ruiying Hao , Christopher Leitz , Anthony Lochtefeld , Allen Barnett
Electrical and Computer Engineering, University of Delaware, Newark, DE, United States
2
AmberWave Inc., Salem, NH, United States

ABSTRACT
Thin-film crystalline silicon (c-Si) solar cells have the
potential for very high efficiency through high open circuit
voltage (Voc) [1]. The best performance reported in thinfilm crystalline silicon solar cells with absorber thickness
below 20 microns is 16.9% efficiency for a solar cell grown
epitaxially on a crystalline silicon conductive substrate [2].
The efficiency potential of thin c-Si solar cells is above
20%. In previously reported thin c-Si solar cells, layers of
porous silicon have been included as a Bragg reflector in
the light trapping design or as a separation layer in the
mechanical design [3, 4]. The Bragg reflector uniformity
and reliability has been analyzed [5]. The challenges of
this Bragg reflector are uniformity and the maximum
achievable optical gain.
This work presents a systematic approach to the design,
fabrication, testing and analysis of thin c-Si solar cells.
Solar cell designs presented include a thin c-Si solar cell
with no light trapping structures and a thin c-Si solar cell
with an optical design comprised of a metallic back
reflector on the back surface and a chemical texture on
the front surface. These baseline designs enable us to
separately measure and localize voltage and current
losses. The metallic back reflector has low electrical
losses, good mechanical stability, and uniform optical
properties. To optimize the baseline designs we measure
open circuit voltage (Voc) and external quantum efficiency
(EQE). Voc shows high material quality in the absorber
layer. EQE shows an evidence of light trapping with a
current gain of 3.6%.
BACKGROUND
Despite the indirect band gap that leads to lower
absorption, silicon is the most used material for
commercial solar cells and modules. From material
properties, silicon has a low absorption for most of the
energy available in the solar spectrum. The relatively low
band gap makes silicon unsuitable for very high efficiency
solar cells especially when used in a single junction
structure.
Part of this is due to the limitations of
conventional thick solar cell designs. The very low energy
conversion in the infrared requires either optimal light
trapping structures or a material thickness of a few
hundred microns that effectively collects the generated
power.
The high performance potential of thin crystalline silicon
solar cells is based on the smaller active volume that can

978-1-4244-9965-6/11/$26.00 2011 IEEE

lead to reduced bulk recombination and hence, higher

voltage. Thin film solar cells have different requirements
because the collection of carriers is much closer to where
they are generated. Furthermore, the much smaller active
volume represents lower recombination losses within the
solar cell. In summary, thin solar cells have the potential
of higher performance because less material can achieve
the same conversion efficiency of a conventional thick
solar cell.

Figure 1. Ideal thin-crystalline silicon solar cell design.

MOTIVATION
The availability of semiconductor grade silicon and its
price has varied throughout the years and has an impact
on the prices of commercial solar panels. Numerous
techniques have been used to make thin crystalline silicon
solar cells including growing the solar cell on a handle.
Approaches to thin silicon solar cells have included the
epitaxial growth of the solar cell on a Czochralski. (CZ)
conductive substrate. Such efforts have optimized a solar
cell on a CZ substrate, and then introduced a lower quality
silicon substrate (metallurgical grade) as a growth
substrate [6]. Other approaches have also transferred the
cell to a mechanical handle [5]. The highest efficiency
achieved for a thin film crystalline silicon solar cell thinner
than 25 microns and grown on a CZ c-Si substrate is
16.9% [2]. This solar cell uses a Bragg reflector as part of
the light trapping design. No other alternatives for the
optical design of the back surface have been strongly
studied for thin crystalline silicon solar cells.
In this work we present the design, fabrication and testing
of solar cells to systematically include a light trapping
structure. High voltage is an indicator of a high quality
absorber material. Quantum efficiency measurements
demonstrate a gain in the infrared region, due to internal
reflection from the back metallic reflector and a chemical
texture on the front surface.

000661

THIN SILICON SOLAR CELL BASELINE

Design
The first baseline design is a thin solar cell grown on a
conductive silicon wafer (see Fig. 2). This structure allows
the development of processes to increase bulk lifetime
and reduce surface recombination. A second design,
utilizing a proprietary process from AmberWave Inc,
includes the same thin silicon solar cell with a metallic
back reflector (see Fig 3). With the first design, the bulk
lifetime and surface recombination can be optimized. The
second design is used to optimize the optical design.

design is to include the optical design for enhanced

absorption.
In previously reported thin crystalline silicon solar cells the
optical design includes a front surface texture and a back
Bragg reflector achieved with a multi layer stack of porous
silicon of different porosities [3]. Our approach uses a
chemical etch that combines hydrofluoric acid and
hydrogen peroxide to create a micro texture on the front
surface [7]. Scanning electron microscope images confirm
that black regions correspond to a more porous layer (see
Figure 4). The back surface of the solar cell is a
conductive, metallic reflector.
Metallic reflectors can be easily integrated into
large area fabrication of back reflectors in thin c-Si solar
cells. Previous approaches to metallic back reflectors and
state of the art high efficiency conventional silicon solar
cells have demonstrated that the Si-metal interface has
high recombination. The optical properties are not ideal
either.
To improve reflection and passivation, a
dielectric/metal has been used in thin c-Si [8]

Figure 2. Baseline solar cell: Thin silicon solar cell grown

on c-Si.
Fabrication
The epitaxial growth technique chosen is reduced
pressure chemical vapor deposition (RPCVD). RPCVD
allows the growth of single crystal silicon films at high
temperature. The growth process requires that the solar
cell is grown on a surface that is compatible with silicon for
epitaxial growth at temperatures above 1000C. Crystalline
silicon is the best candidate as a growth substrate for the
baseline solar cell.

Figure 3. Thin silicon solar cell with the optical design

(back metallic reflector, and chemical texture on the front)
on a crystalline silicon handle for support.

The solar cell design is comprised of a thin epitaxial

absorber grown on a conductive CZ silicon wafer (see Fig
2). Contact to the absorber on the front can be done via
thermal diffusion or epitaxial growth of a n+ Si layer. For
this work, we have used a phosphorous oxychloride
(POCl3) thermal diffusion to form a highly doped layer for
ohmic contact.
The front contact is deposited with
photolithography and lift-off. The contact to the n+ layer is
TiPdAg and the contact to the p+ layer is Al.
The goal of this first solar cell is to optimize the
photovoltaic cell. The key parameters for the solar cell are
bulk lifetime and surface recombination velocity on both
surfaces. All the key parameters are optimized throughout
the fabrication. The processing steps developed with the
baseline design are valuable for the fabrication of the
subsequent solar cell that includes the light trapping
structures. The next step towards a high performance

978-1-4244-9965-6/11/$26.00 2011 IEEE

Figure 4. SEM top view of the chemical texture (top) vs the

non textured front surface)

000662

TESTING AND ANALYSIS

Electrical J-V characteristics are measured with a solar
2
simulator that uses an ELH bulb calibrated to 1000W/m ).
A 36% current increase and a 2% voltage increase are
estimated due to an optimized anti-reflection coating. With
the baseline solar cell on c-Si we report 631mV and above
15% efficiency.
To test the optical enhancement due to the optical design
on the second solar cell design, quantum efficiency
measurements exhibit the energy conversion efficiency as
a function of depth. While voltage is an indicator of the
diode performance, by measuring the EQE we can
diagnose where the current losses are.
Voc

Jsc

FF

Eff.

mV

mA/cm2

S19

631.4

21.49

75.01

10.18

14.05

S27

620.5

22.04

81.32

11.12

15.35

S27

618.8

20.50

81.76

10.37

14.31

Cell

Eff. w/AR
%
Figure 5. EQE measurement of a baseline solar cell, and
a baseline solar cell with a metallic back reflector.
CONCLUSIONS

Table 1. Best solar cell parameters measured for baseline

solar cells (epi on c-Si). The last column extrapolates a
36% increase of current and a 2% increase in voltage due
to a double layer anti reflection coating. No light trapping
structures are present (no front texture, no back reflector)
on these devices.
The short circuit current for the thin solar cell on crystalline
2
silicon is above 20 mA/cm . With the 36% increase from a
double layer anti reflection coating, we can collect
2
between 27 and 29 mA/cm for a structure that has no
light trapping structures. This means that (1) The light
path inside the solar cell is perpendicular to the front
surface and (2) there is no reflection from the back
surface. High fill factors are achieved for most solar cells
with an optimal front contact geometry and passivation of
the solar cell perimeter.
To analyze where the current losses are located, EQE
measurements of the thin solar cell with no light trapping
(see Fig 2) and the thin solar cell with the optical design
(see Fig 3) are compared (see Fig 5). The better quantum
efficiency response in the infrared portion of the spectrum
indicates a gain that comes from the internal reflection
from the metallic layer on the back of the solar cell and the
front texture. The solar cells fabricated are 18 microns in
thickness and the optical absorption decays at 850nm
(see Fig 4). Any increased absorption in this region is due
to the back reflector. This result is very encouraging as an
alternative for the back surface optical design of high
performance thin silicon solar cells.

978-1-4244-9965-6/11/$26.00 2011 IEEE

A systematic approach to high performance thin silicon

solar cells has been presented. Two solar cells designs
were fabricated, measured and analyzed.
Critical
parameters that contribute to high efficiency in thin film
crystalline silicon solar cells are measured with the testing
of the baseline designs.
Efficiency above 15% is reported for a thin c-Si solar cell
with no diffractive light trapping structures. Light trapping
comprised of a front texture and the metallic back reflector
show a current increase in the infrared. The optimal
combination of a surface texture for increased optical path
length and metallic back reflectors can increase the
efficiency of thin crystalline silicon solar cells.
ACKNOWLEDGEMENT
This research was, in part, funded by the U.S.
Government Defense Advanced Research Projects
Agency under Agreement No.: HR0011-07-9-0005. The
views, opinions, and/or findings contained in this
article/presentation are those of the authors and should
not be interpreted as representing the official views or
policies, either expressed or implied, of the Defense
Advanced Research Projects Agency or the Department of
Defense.
REFERENCES
[1]
Barnett, A.M. and C.B. Honsberg. High
performance thin silicon solar cells. in 23rd European
Photovoltaic Solar Energy Conference. 2007.

000663

[2]
Nieuwenhuysen, K.V., et al. Epitaxial thin film
silicon solar cells with efficiencies up to 16.9% by
combining advanced light trapping methods and CVD
emitters. in 24th European Photovoltaic Solar Energy
Conference. 2009. Hamburg, Germany.
[3]
Duerinckx, F., et al., Simulation and
implementation of porous silicon reflector for epitaxial
silicon solar cells. Progress in Photovoltaics: Research
and Applications, 2008. 16: p. 399-407.
[4]
Haase, F., et al. Back contact monocrystalline
thin-film silicon solar cells from the porous silicon process.
in 34th IEEE Photovoltaic Specialists Conference. 2009.
[5]
Hoeymissen, J.V., et al., Thin-film epitaxial solar
cells on low cost Si substrates: closing the efficiency gap
with bulk Si cells using advanced photonic structures and
emitters, in 23rd European Photovoltaic Solar Energy
Conference2008: Valencia, Spain. p. 2037-2041.
[6]
Schmich, E., et al., n-Type emitter epitaxy for
crystalline silicon thin-film solar cells. Progress in
Photovoltaics: Research and Applications, 2008. 16: p.
159-170.
[7]
Tsujino, K. and M. Matsumura, Morphology of
nanoholes formed in silicon by wet etching in solutions
containing HF and H2O2 at different concentrations using
silver nanoparticles as catalysts. Electrochemica Acta,
2007. 53: p. 28-34.
[8]
Reuter, M., et al., 50um thin solar cells with
17.0% efficiency. Solar Energy Materials and Solar Cells,
2009. 93: p. 704-706.

978-1-4244-9965-6/11/$26.00 2011 IEEE

000664

THE PERFORMANCE OF THIN FILM SOLAR CELLS EMPLOYING PHOTOVOLTAIC

ZnSe/CdTe, CdS/CdTe and ZnTe/CdTe HETEROJUNCTIONS
1

T. Potlog , N. Spalatu , V. Fedorov , N. Maticiuc , C. Antoniuc , V. Botnariuc and J. Hiie , T. Raadik , V. Valdna
1
Physics Department, Moldova State University, A. Mateevici str. 60, MD 2009, Chisinau, Moldova
2
Department of Materials Science, Tallinn University of Technology, Ehitajate tee 5, 19086, Tallinn, Estonia

ABSTRACT
This paper focuses on the photovoltaic parameters of
ZnSe/CdTe, CdS/CdTe, and ZnTe/CdTe thin film
heterojunction solar cells. ZnSe/CdTe, CdS/CdTe, and
ZnTe/CdTe thin film heterojunction solar cells were
fabricated by Close Space Sublimation (CSS) on TCOcoated glass substrates. All types of solar cells were
fabricated in a superstrate configuration. The thickness of
ZnSe and ZnTe layers was varied in order to adjust the
solar cell performance. A similar cadmium chloride
solution for the treatment of a CdTe layer with an elevated
temperature air annealing of the completed devices before
the back contact deposition was applied to ZnSe/CdTe
and CdS/CdTe thin film heterojunctions solar cells with
exception of ZnTe/CdTe. All cells were characterized
through light and dark current density-voltage (J-V)
measurements
and
quantum
efficiency
(QE)
measurements. The saturation current, ideality factor and
photovoltaic parameters for all thin film heterojunction
solar cells are presented. The investigation at the room
2
temperature under illumination of 100 mW/cm through the
wide gap components of ZnSe/CdTe, CdS/CdTe, and
ZnTe/CdTe heterojunctions showed a value of conversion
efficiency () of solar energy to electric energy about
4.7%, 9.9%, and 1.3%, respectively. The incorporation of
Zn at the ZnSe and CdTe interface doubles the short
circuit current density and improves the performance of
ZnSe/CdTe thin film heterojunction solar cells.
INTRODUCTION
The primary aim of this research paper is to develop low
cost photovoltaic systems. Routes to decrease cost per
Watt are either to reduce the cost of the technology or to
increase its efficiency. The first of these is addressed by
the use of thin film technology for low cost deposition at
low temperatures and with low material usage. The
second is addressed by using multiple threshold devices
to push efficiencies beyond the Schockley-Queisser limit
for single band-gap PV cells [1]. Our studies will focus on
the development of new electronic devices for PV
applications. Materials of interest for PV systems include
II-VI compound thin film semiconductors, especially CdTe
and transparent conducting oxides required for the
fabrication of solar cells. The main goals of this paper are
to develop polycrystalline growth techniques for
ZnSe/CdTe and ZnTe/CdTe thin film heterojunction solar
cells on glass/TCO superstrate configurations and
compare them with 10% efficiency CdS/CdTe solar cells

978-1-4244-9965-6/11/$26.00 2011 IEEE

fabricated at the Moldova State University [2]. The high

short-circuit current density (Jsc) is obtained in CdS/CdTe
solar cells using thinner CdS films as window layer.
However, the photovoltaic conversion efficiency does not
exceed 16.5% [3]. The II-VI compounds could be used as
components in tandem cell systems consisting of two or
more junctions based on compounds with different band
gaps. Therefore, this paper describes the results of the
preparation and study of the photoelectrical properties of
ZnSe/CdTe, CdS/CdTe, and ZnTe/CdTe thin film
heterojunction solar cells.
SAMPLE PREPARATION
Thin layers ZnSe, CdTe, ZnTe and CdS were obtained
using the close-space sublimation (CSS) method on glass
2
substrates with an area of up to 2 cm . The important
process parameters of the CSS method are the
temperatures of the substrate and source. CdTe thin films
were deposited at a temperature of the source and
o
o
substrate of 340 C and 590 C, respectively. This
technological regime provides the best efficiency on
CdS/CdTe solar cells; therefore, this regime was used for
the preparation of the ZnSe/CdTe and ZnTe/CdTe thin film
heterojunction solar cells. The CdS/CdTe thin film
heterojunction solar cells were prepared onto SnO2/glass
substrates (produced by Solaronix) by the successive
growth of CdS and CdTe and denoted as C1, C2, and C3
in dependence on the type of the back contact used. The
ZnSe/CdTe was obtained by the same procedure and
noted as Z1, Z2, and Z3 in dependence on the thickness
of ZnSe. Also, the same procedure was used to prepare
ZnTe/CdTe thin film solar cells, which were denoted as
ZT1, ZT2, and ZT3 in dependence on the thickness of
ZnTe. Te/Ni was used as an ohmic back contact to CdTe
in the case of ZnSe/CdTe and ZnTe/CdTe thin film solar
cells and TCO for CdS, ZnSe, and ZnTe.
The thickness of CdTe in the preparation of all types of
thin film heterojunction solar cells was ~6 m. The
thickness of ZnSe and ZnTe was varied in order to adjust
the solar cell efficiency. Undoped ZnTe layers had a
6
8
resistivity of (10 -10 ) cm. Therefore, the ZnTe thin films
were grown from crystals doped with As. The resistivity
measured at room temperature for thin films obtained at
o
o
2
3
Tsub= 430 C and Ts= 660 C reached a value of (10 -10 )
cm. The hole concentration in ZnTe for this
16
-3
technological regime achieves (2.6-7.8) 10 cm at 300
2
K, and their mobility is 50-85 cm /Vs. It is well known that
the CdCl2 treatment is an important step in the formation
of a solar cell containing CdTe. The annealing of the

001365

ZnTe/CdTe heterojunction solar cells in a CdCl2 saturation

solution leads to the detachment of the structure from the
SnO2/glass substrate. Therefore, indium was used to
prepare ZnTe/CdTe structures. A thin layer of indium was
evaporated on the surface of the CdTe of ZnTe/CdTe
structure, and then the structure was annealed in vacuum
o
at 250 C. The resistivity of the CdTe:In reaches a value of
4
~10 cm. Undoped ZnSe films are high-resistive: =
8
10
(10 -10 ) cm. The deposition of Zn thin layer by thermal
o
evaporation on ZnSe and annealing at 470 C reduced the
5
resistivity till 10 cm. It was found that an increase in the
o
o
annealing temperature from 410 C to 470 C results in an
increase in the donor concentration by an order of
17
-3
magnitude to a value of 10 cm . The study of electrical
properties showed the contribution of noncontrolled donortype native defects to increasing conductivity of the ZnSe
samples. Thicknesses of about 0.12 m for CdS are
sufficient to avoid the formation of shunting paths. The
17
-3
electron concentration in CdS is 6.510 cm .

decreasing thickness (see Table 1), which indicates that,

in this case, the energy band gap of the ZnSe films
increases. This particular behavior of the ZnSe films
seems to be related with changes in the band structure
induced by changes in their thickness. To determine the
value of the band gap of the ZnSe and ZnTe films, the
2
(h) versus (h) plots are presented in Fig. 2. It is
evident that all graphs for all films have straight line
portions in the short wavelength region, which confirms
the direct band gap nature for all the ZnSe films. The (h)
2
versus (h) plots gave us a band edge of Eg = 2.670 eV
for ZnSe thin films with smaller thicknesses. The highest
value of Eg found for ZnSe films was 2.684 eV, which is
fairly close to the accepted value of 2.7 eV for the energy
band gap of bulk ZnSe. The values are found to be
comparable with those from earlier reports [4, 5].
14

2.5x10

ZnSe- 4 min
ZnSe- 8 min
ZnSe-11 min

14

2.0x10

Optical characterization of ZnSe and ZnTe thin films

The transmission spectra of as deposited ZnSe and ZnTe
thin films on glass substrates were studied in a
wavelength range of 200 to 1200 nm.

( h )2, (eV/cm)2

RESULTS AND DISCUSSION

14

1.5x10

14

1.0x10

13

5.0x10

0.0
2.5

100

2.6

2.7

2.8

2.9

3.0

h , eV

80

Figure 2. Dependence of (h) versus (h) for

ZnSe thin films.

T, %

60

Table 1 Thickness and band gap values for the

ZnSe and ZnTe thin films

40
glass/ZnSe_ 1
glass/ZnSe_ 2
glass/ZnSe_ 3
glass/ZnTe_ 1
glass/ZnTe_ 2
glass/ZnTe_ 3

20
0
300

600

900

1200

, nm

Figure 1. Effect of film thickness on

transmittance of ZnSe and ZnTe thin films.

the

The transmittance of the ZnSe and ZnTe films with

different thicknesses is presented in Fig. 1. The ZnSe and
ZnTe thin films were denoted as (1-3). The thickness of
the ZnSe buffer layers was determined using the following
relation [5]:

d=

12
2(1n2 2 n1 )

(1)

where n1 and n2 are the refractive indexes at two adjacent

maxima (or minima); 1 and 2 the respective wavelength
values. The transmittance of the ZnSe films shifts in the
transmission threshold to lower wavelengths with

978-1-4244-9965-6/11/$26.00 2011 IEEE

ZnSe

deposition time

d, m

Eg, eV

4 min

1.4

2.684

8 min

2.7

2.675

3
ZnTe
1
2
3

11 min
deposition time
4 min
8 min
11 min

4.1
d, m
1.8
3.9
5.2

2. 670
Eg, eV
2.247
2.245
2.242

The transmittance spectra of the as-deposited ZnTe thin

films do not exhibit interference maxima and minima.
Therefore, the thickness of ZnTe was estimated using the
procedure proposed by [6] to an accuracy of better than
1% with similar accuracies in the values of n. The
thickness corresponds to a value of (1) 1.8 m; (2) 3.9 m;
and (3) 5.2 m, respectively. The optical transmission of
these layers in the photosensitivity region is 40-60%.
Figure 1 shows that the transmittance value also
decreases with increasing film thickness. This behavior
agrees well with other investigations [7]. The energy band-

001366

gap estimated from Fig. 3 is about 2.25 eV and is very

close to the reported bulk value at room temperature for
ZnTe [8] and weakly varies with thicknesses.

Figure 4 shows for ZnSe/Zn/CdTe thin film solar cells that

n=f(T) exhibits two linear regions. It is well-known that, if
the current transport is dominated by a thermal process,
the current-voltage relationship takes the form

eU ,
I = I O exp

nkT

( h )2, (eV/cm)2

4x10

Glass/ZnTe-1
Glass/ZnTe-2
Glass/ZnTe-3

3x10

(3)

14

2x10

12

1x10

C1_n1
C1_n2
Z3_n2
Z3_n1
ZT2_n

8
2.25

2.30

2.35

2.40

h, eV

0
2.20

10

6
2

Figure 3. Dependence of (h) versus (h) for

ZnTe thin films.

4
2

ZnSe/CdTe thin film heterojunction solar cells

280

The study of the dark current-voltage characteristics of

several ZnSe/CdTe solar cells with different thicknesses of
ZnSe shows that the direct curves of the I-U
characteristic shift in the direction of the current abscissa,
which corresponds to an increase in the build-in voltage
from 0.46 to 0.52 V. A further increase in the thickness of
the ZnSe layer shows that the build-in voltage remains
constant. Thus, this thickness of ZnSe was taken as a
value of the penetration depth of the contact field in the
n/CdTe region of the heterostructure. The rectification
coefficient of the heterostructure with dZnSe= 1.4 m
calculated at Ud=Uind = 1 V is about 2 and increases to 60
with increasing thickness of the ZnSe film. A further
increase in the thickness of the ZnSe layer results in an
increase in the series resistance of the heterojunction,
which limits the direct current that leads to a decrease in
the rectification coefficient. The achievement of high
efficiency for solar cells requires the understanding of the
junction transport. In order to obtain information on the
transport current mechanisms of the ZnSe/CdTe devices,
we measured current-voltage characteristics in the dark in
the temperature interval of (290-393) K. In the case of our
structures, the logarithm of forward current versus voltage
at different temperatures was analyzed and the ideality
factor was estimated. Figure 4 illustrates the temperature
dependences of ideality factor for all types of the thin film
solar cells. For the ZnSe/Zn/CdTe curves measured in the
above temperature interval at applied biases U< UD, the
current is an exponential function of voltage with two
values of ideality factor. When U becomes greater than
the diffusion potential (UD), the I-V characteristics can be
approximated by the linear dependence:
U U D
(2),
I=
R

where UD is the diffusion potential equal to the build-in

voltage and R is the series resistance of the cell.

978-1-4244-9965-6/11/$26.00 2011 IEEE

300

320

340

360

380

400

T,K

Figure 4. Temperature dependence of the ideality

factor for all types of thin film solar cells.
while if tunneling dominates, the I-V relationship has the
form
(4)
I = I OT exp( AU )
These results can be qualitatively explained as a
tunneling-dominated current due to carriers from the
valence band of the p/type material or the conduction
band of the n/ZnSe tunneling into interface states where
they recombine. The saturation current I0 is found to vary
according to the relation:
E
(5)
I o ~ exp a ,
kT
where the E a is the activation energy of charge carriers
in the forward bias. This region corresponds to a
recombination-limited current. The resulting activation
energy is 0.34 eV. We propose to use a model devised by
Donnelly and Milnes [9], which combines tunneling to
interfaces states and recombination at these states. We
attribute the activation energy of the thermal activation of
holes below the hole quasi-Fermi level in CdTe to an
energy at which two conditions are simultaneously
satisfied: (1) the barrier is thin enough for appreciable
tunneling to occur and (2) the concentration of inteface
states and their effective recombination cross sections are
high enough for appreciable recombination to occur.
The photovoltaic characteristics of thin film heterojunction
solar cells with different thicknesses of the ZnSe window
layer were investigated at 300 K and under illumination of
2
100 mW/cm . The current-voltage characteristics of these
thin film heterojunction solar cells are illustrated in Fig.5.
It is evident that the incorporation of Zn at the ZnSe and
CdTe interface (samples Z3) approximately doubles the

001367

-1.0

-0.8

-0.6

-0.4

-0.2

0
0.0

0.2

0.4

0.6

0.8

1.0

U, V
-5

-10
Z3
Z2
Z1

The photoelectrical properties of the ZnTe/CdTe thin film

heterojunction solar cells were investigated under
illumination through the wide band-gap ZnTe layer. The
current-voltage characteristics of the ZnTe/CdTe thin film
heterojunction solar cells are shown in Fig. 7.

-15

-20

Norm. Quantum efficency (a.u)

J, mA/cm

short circuit current density and improves the performance

of ZnSe/Zn/CdTe thin film heterojunction solar cells.

1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2

Z3
Z1
Z2

1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2

ZnTe/CdTe thin film heterojunction solar cells

The dark current-voltage characteristics exhibit only one
slope in the temperature interval of (290-393) K. Figure 4
illustrates the temperature dependence of the ideality
factor of the ZnTe/CdTe thin film solar cell with highest
efficiency. The forward I-U branches in the entire
temperature range show the temperature dependence of
ideality factor. The value of the ideality factor in this
temperature range varies with temperature from 2.9 to 1.1.
Based on these results, we can assume that the current
flow mechanism is generation-recombination. This is also
confirmed by the temperature dependence of saturation
current. The barrier height determined from the
temperature dependence of saturation current in the hightemperature region was 0.69 eV.

978-1-4244-9965-6/11/$26.00 2011 IEEE

Figure 6. Quantum efficiency of ZnSe/CdTe thin

film heterojunction solar cells with different
thicknesses of ZnSe.
The photovoltaic parameters are listed in Table 2. The
highest value of efficiency is obtained for the ZnTe/CdTe
thin film solar cell with a thickness of ZnTe of ~5.2 m. It
is ~1.3%.

The external quantum efficiency (EQE) (Fig. 6) for these

cells shows that the shape of these characteristics
depends on ZnSe layer thickness. In the case of samples
Z1, almost all space charge is situated in CdTe, and EQE
is determined by the electron-hole generation in it. The
redistribution of electron-hole pair generation occurs with
increasing thickness of ZnSe (samples: Z2 without Zn
layer at the interface and Z3 with Zn layer at the interface).
It is obvious that the contacting materials make an equal
contribution to photocurrent if and only if the thickness of
the epitaxial layer is on the order of the penetration depth
of the contact field. If the thickness is less than the
penetration depth, the major contribution to the EQE is
obtained in the sensitivity region of the narrow-bandgap
material; for the layer thickness higher than the
penetration depth of the contact field, it is in the sensitivity
region of the wide-bandgap material.

h (eV)

J, mA/cm

Figure 5. Illuminated current-voltage characteristics of

ZnSe/CdTe thin film heterojunction solar cells with
different thicknesses of ZnSe: (Z1) 1.4 m, (Z2) 2.7 m,
and (Z3) 4.1 m.

ZT1
ZT2
ZT3

-1.0

6
4
2

-0.5

0
0.0

U, V
0.5

1.0

-2
-4
-6
-8

Figure 7. Illuminated current-voltage characteristics of

the ZnTe/CdTe thin film heterojunction solar cells with
different thicknesses of ZnTe: (ZT1) 1.8 m, (Z2) 3.9
m, and (Z3) 5.2 m.
The EQE of these thin film heterojunction solar cells is
given in Fig. 8. The most effective separation of charge
carriers occurs in ZT2. This can be judged from the form
EQE
spectra.
of

001368

Table 2 The photovoltaic and electric parameters of the ZnSe/CdTe, CdS/CdTe, and ZnTe/CdTe thin film
heterojunction solar cells at room temperature

Sample

Jsc,
2
mA/cm

Ucd, V

FF, %

, %

Rs, cm

Z3 with Zn at the
interface
Z2
Z1
ZT1
ZT2
ZT3
C1
C2
C3

18.48

0.71

35.67

4.68

10.70
4.58
2.90
4.94
3.37
23.27
17.30
22.96

0.75
0.72
0.63
0.69
0.43
0.79
0.78
0.70

41.36
30.80
27.39
37.04
28.90
54.00
55.10
31.60

3.33
1.01
0.50
1.27
0.42
9.90
7.40
5.20

Rsh,
2
cm

Jo,
2
mA/cm

21.00

128.13

1.28E-2

2.89

34.70
110.37
234.52
53.67
90.91
11.66
10.67
86.42

187.83
218.75
321.64
332.15
188.92
1054.54
487.80
101.38

1.60E-5
1.11E-5
1.30E-4
1.86E-6
2.78E-4
1.07E-4
5.86E-3
3.15E-2

2.46
2.81
5.69
2.94
8.62
12.93
7.73
10.90

illuminations. It is evident that the J-V curve for the cell C3

with Sb layer at the back exhibits a stronger rollover in the
light in forward bias, while the thin film heterojunctions
solar cells C2 with Cu and C1 with Te at the back
improved their J-V load characteristics.

J, mA/cm

Norm. Quantum efficency (a.u)

It appears that, for samples Z1 and Z2, during CdTe

layer deposition, a high resistivity layer or a solid solution
at the interface is formed because of the mutual diffusion
of component-materials atoms, which determines the
heterojunction EQE. An increase in the thickness of the
ZnTe layer shifted the main peak to shorter wavelengths.
The main contribution to EQE in the case of ZT3 thin film
solar cells comes from electron-hole pair generation in
ZnTe.

C1
C2
C3

1.0
ZT1

0.8

-1.0

-0.5

20

10

0
0.0

ZT2

0.5

1.0

U, V

ZT3

-10

0.6
-20

0.4
-30

0.2
0.0
1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2

h (eV)

Figure 8. EQE of ZnTe/CdTe thin film heterojunction

solar cells with different thicknesses of ZnTe: (ZT1)
1.8 m, (Z2) 3.9 m, and (Z3) 5.2 m.
CdS/CdTe thin film heterojunction solar cells
Figure 4 for photovoltaic device C1 with higher efficiency
shows a temperature-independent ideality factor value of
~13 in the dark at low voltages and ~5 at voltages 0.2
V<U<UD. These data indicate that the current transport
across
the
junction
is
dominated
by
tunneling/recombination process in this cell in the above
mentioned temperature intervals. This is in agreement with
many other studies [10-12] because CdS/CdTe interface is
heavily dislocated due to the lattice and structural
mismatch. Figure 9 shows the current-voltage
characteristics of the CdS/CdTe thin film heterojunction
2
solar cells with different back contacts under 100 mW/cm

978-1-4244-9965-6/11/$26.00 2011 IEEE

Figure 9. Illuminated current-voltage characteristics

of CdS/CdTe thin film heterojunction solar cells with
different back contacts.
Figure 10 shows the EQE for the CdS/CdTe thin film
heterojunction solar cells with different back contacts. At
short wavelengths, the EQE corresponds to the band gap
of CdS. A stronger decrease in the EQE is observed at all
wavelengths for the cell C2 with Cu at the back. The Te
layers deposited at the back by thermal evaporation show
highest sensitivity at the long wavelength of the EQE of
the CdS/CdTe thin film heterojunction solar cells. The
photovoltaic parameters of the ZnSe/CdTe, CdS/CdTe,
and ZnTe/CdTe thin film heterojunction solar cells are
summarized in Table 1. The data for all thin film
heterojunction solar cells can be fitted to the conventional
expression which includes the effect of series and shunt
resistances:

q (U + JRs
J = J f J 0 [exp
nkT

(U + JRs ) ,

1]
Rsh

(6)

001369

where Jf photocurrent; J0 is the reverse saturation

current density, n is the diode factor, Rs is the series
resistance, and Rsh is the shunt resistance.

Norm.Quantum Efficiency (a.u)

1.0
0.8
0.6
C3
C2
C1

0.4
0.2
0.0
1.0

1.5

2.0

2.5

h, eV

Figure 10. Quantum efficiency of the CdS/CdTe thin

film heterojunction solar cells with different back
contacts

The best CdS/CdTe/Te device exhibited an efficiency of

about 10%. The ZnSe/CdTe thin film heterojunction cells
indicate a relatively high open circuit voltage and,
however, a low efficiency in comparison with CdS/CdTe
solar cells. All types of thin film heterojunction solar cells
seriously suffer fill factor (FF). The series resistance and
dark J-V characteristics determine the fill factor (see Table
2).
CONCLUSIONS
The experimental data are described to show another
configuration, which also exploits the innate material
properties of high-resistivity CdTe. The photovoltaic
parameters, in particular, of ZnSe/CdTe and ZnTe/CdTe
thin film solar cells can be improved by the appropriate
choice of component heterojunction thickness and doping
level. The studies of EQE show that generated carriers are
more efficiently collected at the CdS/CdTe interface of thin
film heterojunction solar cells. The technological regime
for ZnSe/CdTe and ZnTe/CdTe must be optimized,
especially the temperatures of the substrate and source.
An improvement in the performance of all types of thin film
heterojunction solar cells can be ensured via decreasing
the resistance of CdTe and the series resistance of thin
film solar cells.

REFERENCES
[1] F. Meillaud, A. Shah, C. Droz, E. Vallat-Sauvain and C.
Miazza, Efficiency limits for single-junction and tandem
solar cells Solar Energy Materials and Solar Cells. V. 90,
Iss. 18-19, 23, 2006, pp. 2952-2959.
[2] T. Potlog G. Khrypunov. M. Kaelin, H. Zogg, and A. N.
Tiwari. Thin-Film Compound Semiconductor, Book, V.
1012, Photovoltaics-2007, edited by T. Gessert, K.
Durose, C. Heske, S. Marsillac, and . Wada, 2007, pp.
181-188. www.mrs.org/meetings/.
[3] X. Wu,High-efficiency polycrystalline CdTe thin-film
solar cells, Solar Energy 77, 2004, pp. 803-814.
[4] J. Gutowski, N. Presser, G. Kudlek Optical Properties
of ZnSe Epilayers and Films, Physica Status Solidi (a)
V. 120, N.1, , 1990, pp. 1159.
[5] [7] Huanyong Li, Wanqi Jie Growth and
characterizations of bulk ZnSe single crystal by chemical
vapor transport, Journal of Crystal Growth, 257, 2003, pp.
110115.
[6] Swanepoel R. Determination of the thickness and
optical constants of amorphous silicon, J. Phys. E: Sci.
Instrum. 16, 1983, pp. 1214-1219.
[7] M. Orita, T. Narushima, and H. Yanagita Transparent
Conductive Cu-doped ZnSe Film Deposited at Room
Temperature Using Compound Sources Followed by
Laser Annealing, Jpn. J. Appl. Phys. 46, 2007, pp. 976978.
[8] D. T. F. Marple Refractive Index of ZnSe, ZnTe, and
CdTe, J. of Appl. Physics, V.35 , N.3,1964, pp.539-542.
[9] Donnelly J.P., Milnes A.G The photovoltaic response
of nGe-nSi heterodiodes, Sol. St. Electronics, V.9, 1966,
pp.174-178.
[10] H. Bayhan and A. S. KavasoluTunnelling enhanced
recombination
in
polycrystalline
CdS/CdTe
and
CdS/Cu(In,Ga)Se2 heterojunction solar cells, Solid-State
Electronics, V. 49, N. 6, 2005, pp. 991-996.
[11] S. S. Ou and O. M. Stafsudd, B. M. Basol Current
transport mechanisms of electrochemically deposited
CdS/CdTe
heterojunction,
Solid-State
Electronics
V. 27, Nr. 1,1984, pp. 21-25.
[12] C. Ercelebi, A.W. Brinkman, T.S. Furlong and J.
Woods Current transport mechanisms in epitaxial
CdS/CdTe heterojunctions, J. of Crystal Growth
V. 101, Nr. 1-4, 1990, pp. 162-166.

ACKNOWLEDGEMENTS
th

This research was supported by the EU 7 Framework

Program PEOPLE International Research Staff Exchange
Scheme project Development of Flexible Single and
Tandem II-VI Based High Efficiency Thin Film Solar Cells
GA-2008, No. 230861.

978-1-4244-9965-6/11/$26.00 2011 IEEE

001370

DEVELOPMENT OF HETERO-JUNCTION a-Si/c-Si BASED ON

METAL-WRAP THROUGH STRUCTURE SOLAR CELLS
Der-Chin Wu, Jui-Chung Shiao, Chien-Hsun Chen, Chao-Cheng Lin, Chien-Hsi Lin, Wen-Haw Lu,
Chern-Lin Chen
Green Energy and Environment Research Laboratories, Industrial Technology Research Institute, 31040,
HsinChu, Taiwan, R.O.C. E-mail: Derrick_Wu@itri.org.tw

How to further enhance the efficiencies of the silicon solar cells which can be applied
into commercial production for the next generation is a most important issue. On the way to
achieving higher efficiencies, back contact solar cells and heterojunction solar cells seem to
be very promising. Therefore, we proposed a amorphous/crystal hetero-junction with metalwrap structure (HJ-MWT) which combines the advantages of different high-efficiency
concepts such as back contacts, and great heterojunction passivation. The advantages of
MWT are less shadowing loss and lower module electrical power losses. It will increase
1.5-3 % module conversion efficiencies compared to conventional silicon solar cells.
Furthermore, the other advantage of MWT isnt fully limited by the bulk lifetime and carrier
diffusion lengths contrast to IBC solar cells. It doesnt need very-high-quality silicon
wafer(minority lifetime> 1ms).
We proposed amorphous silicon or insulation materials were designed as electrical
insulation layers within the via(wrap-through holes) to avoid shunt current occurring. The
fabricating process flow that realizes HJ-MWT solar cells, with only 2-3 additional
processes steps compared to conventional HIT solar cells. Its suitable to the mass
production associated the cost consideration in the future. Before experiments, we used the
software tool to do the two-dimension simulation with HJ-MWT structures. From theoretical
calculation, we found HJ-MWT solar cell efficiencies increased up to 0.6% compared with
conventional HIT solar cells. We have also done the further experimental investigation on
new proposed HJ-MWT structures and corresponding data will be published in this
conference. It shows the HJ-MWT solar cell has the potential to further increase the
efficiencies of solar cell and solar panels in the future.

ITO
i,p a-Si

TCOa-Si(i,p)

n-Si

a-Si(i n)
Ag

n-Si

n-Si

Ag

n-Si
Ag

a-Si(i,p)

978-1-4244-9965-6/11/$26.00 2011 IEEE

Ag

Ag

Insulation

Ag

001515

DEVELOPMENT OF HETERO-JUNCTION a-Si/c-Si BASED ON METAL-WRAP

THROUGH STRUCTURE
Der-Chin Wu, Jui-Chung Shiao, Chien-Hsun Chen, Chao-Cheng Lin, Chien-Hsi Lin, Wen-Haw Lu, Chern-Lin Chen
Green Energy and Environment Research Laboratories, Industrial Technology Research Institute, 31040, HsinChu,
Taiwan, R.O.C. E-mail: Derrick_Wu@itri.org.tw
How to further enhance the efficiencies of the silicon solar cells which can be applied into commercial production for
the next generation is a most important issue. On the way to achieving higher efficiencies, back contact solar cells and
heterojunction solar cells seem to be very promising. Therefore, we proposed a amorphous/crystal hetero-junction with
metal-wrap structure (HJ-MWT) which combines the advantages of different high-efficiency concepts such as back
contacts, and great heterojunction passivation. The advantages of MWT are less shadowing loss and lower module
electrical power losses. It will increase 1.5-3 % module conversion efficiencies compared to conventional silicon solar
cells. Furthermore, the other advantage of MWT isnt fully limited by the bulk lifetime and carrier diffusion lengths contrast
to IBC solar cells. It doesnt need very-high-quality silicon wafer(minority lifetime> 1ms).
We proposed amorphous silicon or insulation materials were designed as electrical insulation layers within the
via(wrap-through holes) to avoid shunt current occurring. The fabricating process flow that realizes HJ-MWT solar cells,
with only 2-3 additional processes steps compared to conventional HIT solar cells. Its suitable to the mass production
associated the cost consideration in the future. Before experiments, we used the software tool to do the two-dimension
simulation with HJ-MWT structures. From theoretical calculation, we found HJ-MWT solar cell efficiencies increased up to
0.6% compared with conventional HIT solar cells. We have also done the further experimental investigation on new
proposed HJ-MWT structures and corresponding data will be published in this conference. It shows the HJ-MWT solar cell
has the potential to further increase the efficiencies of solar cell and solar panels in the future.
INTRODUCTION
The Global Warming and energy crisis have given fresh
impetus to the photovoltaic industries over past several
years. An effective reducing cost of solar cell production is
to enhance conversion efficiencies and reduce process
costs. There are many companies developing their highefficiency solar cell including p-type wafer based cells of
Suntechs Pluto cells and n-type wafer based cells of
Yinglis Panda cells, SunPowers IBC cells, Solland MWT
cells and Sanyo HIT cells. Among the development of
solar cell technologies, amorphous/crystal silicon (aSi:H/c-Si) heterojunction technology is one of the most
unique structures undoubtedly. Expected Efficiencies up to
25% can be achieved with a-Si:H/c-Si heterojunction
structure concept[1] and Sanyo Electric Company has
demonstrated high-efficiency 23% in laboratory on the socalled HITTM structure on n-type c-Si wafers[2].
HIT cells were developed with a very thin intrinsic aSi:H(i-layer) inserted between p-type a-Si:H and n-type cSi, and such un-doped layer greatly reduced the silicon
surface dangling bonds, and finally enhanced the openvoltage above 700 mV. The conventional HIT structure
solar cells are multi-layer structures which are fabricated
with [Ag grid/transparent conductive oxide(TCO)/aSi:H(p)/a-Si:H(i)/n-c-Si/a-Si:H(i)/a-Si:H(n)/TCO/Ag
grid].
There are various grouping involving in these areas and
do the further investigations on HIT solar cells including
the a-Si/c-Si interface recombination model[3,4],
theoretical current transport [5,6] and experimental
improvements of solar cells[7,8]. Based on the
conventional structure, there are more and more groups
exceeding 20% with this reduplicated structure.

978-1-4244-9965-6/11/$26.00 2011 IEEE

Furthermore, the MWT cell concept offers two main

advantages. First, a gain in active cell and thus an
increase in short current is achieved due to less
shadowing loss which was caused by absence of frontside metal bus bar. Second, the n- and p- back-contacts
simplify the interconnection of the cells and allows the
utilities of new tabbing technologies with the objective of
series resistance reduction[11]. It will increase 1.5-3 %
module conversion efficiencies compared to conventional
silicon solar cells. Furthermore, the other advantage of
MWT isnt fully limited by the bulk lifetime and carrier
diffusion lengths contrast to IBC solar cells. It doesnt need
very-high-quality silicon wafer(minority lifetime> 1ms).
Based on the advantages of HIT and MWT solar cells,
we are the first one to propose the new combined
structures
and
implementation
process
flows.
Amorphous/crystal hetero-junction with metal-wrap
structure (called HJ-MWT) will be discussed with
theoretical simulation and experimental implementation in
the next section.
Device Structure
We proposed two novel structures of HJ-MWT cells
as shown in Fig. 1. The major differences between them
are the insulation materials within the via (wrap-through
holes). Electric isolation between front side emitter and
substrate play an important key determining the
conversion efficiencies. If the via insulation layers are not
deposited well, the shunt current increase and finally lead
to lower efficiencies or module fail. Corresponding process
flow and simulation results are shown in Fig. 2-4
respectively.

001516

TCOa-Si(i,p)

n-Si

a-Si(i n)

n-Si
Ag

Ag

Ag

(a) a-Si(i,p)

ITO
i,p a-Si
n-Si

n-Si
Ag

Ag

Insulation

Ag

Fig. 3: Process flow chart of HJ-MWT with silicon nitride

isolation within via.

(b)

Fig. 1: The proposed novel structures of HJ-MWT cells. The

major differences of these two diagrams are the insulation
materials within the via (wrap-through holes). Electric
isolation between front side emitter and substrate play an
important key determining the conversion efficiencies. (a)
isolation materials using i-, p-amorphous silicon(b)isolation
materials using insulation materials such SiNx.

Table I: Simulation Parameters used in this study.

Fig. 2: Process flow chart of HJ-MWT with amorphous silicon isolation

within via.
978-1-4244-9965-6/11/$26.00 2011 IEEE

001517

Fig. 4: Simulation results chart of HJ-MWT. For simplifying the simulation, no surface texturing structure was used in this
study. We used the software tool to do the two-dimension simulation with HJ-MWT structures. Two proposed devices are
theoretically verified and no shunt current occurs from the simulation results. The summarized efficiency table showed HJMWT solar cell efficiencies increased up to 0.6% compared with conventional HIT solar cells. It shows the HJ-MWT solar
cell has the potential to further increase the efficiencies of solar cell in the future. We have tried to implemented the
experimental results and verified the utilities of HJ-MWT, and the data will be disclosed in PVSC conference.
[1] RM. Swanson, Proceedings of the 31th IEEE PVSC,
Lake Buena Vista, 889 (2005).
th
[2] M. Taguchi et al., Proceedings of the 24 EU-PVSEC,
Hamburg, 2CV.2.78 (2009).
[3]S. D. Wolf et al. Appl. Phys. 93, 032101 (2008).
[4]T. F. Schulze et al. 25th EU-PVSEC, Hamburg, (2010).
[5]L. Zhao et al., Phys. Stat. sol. (a) 205, 1215 (2008).
[6]D. C. Wu et al., 35th IEEE PVSC, Honolulu USA, (2010).
th
[7] D. Munoz et al., 35 IEEE PVSC, Honolulu USA,
(2010).
th
[8] D. Lachenal et al. 25 EU-PVSEC, Hamburg, (2010).
[9]R. Street, Hydrogenated Amorphous Silicon, Cambridge.
University, Cambridge, (1991).
[10]M. Schmidt et al. Thin Solid Film, 515, 7475 (2007).
[11]F. Clement et al. Sol. Energy Mater. Sol. Cells, 93,
1051 (2009).

978-1-4244-9965-6/11/$26.00 2011 IEEE

001518

THERMODYNAMIC LIMITS OF HYBRID PHOTOVOLTAIC SYSTEMS USING

MULTIPLE JUNCTION SOLAR CELLS WITH CARRIER TRANSITIONS AT
INTERMEDIATE BAND
Jongwon Lee and Christiana Honsberg
School of Electrical, Computer and Energy Engineering, Arizona State University, Tempe, Arizona, USA

ABSTRACT
We have proposed the first demonstration of hybrid
thermodynamic approach between multiple junction solar
cells and the intermediate band solar cells. Because of
increasing number of transitions inside each multiple
junction from intermediate band, reduced number of
stacks in this hybrid solar cell can be possible to increase
efficiency compared to more than four stacks of
conventional multijunction solar cells. The maximum
efficiency of two stacks hybrid solar cells has been
enhanced and its value is larger than over 4 stacks of
multijunction solar cells. We have also found optimum
bandgap of each stack and studied optimum transition of
each stack both from valence band to intermediate band
and intermediate band to conduction band.
.
INTRODUCTION
Since, Shockley and Queisser (SQ) have been developed
detailed balance model for single junction solar cell and
calculated its maximum conversion efficiency blackbody
radiation [1]. But, 69% is still the loss of energy conversion
And, its dominant factor is the thermalization loss while
excited electrons have been released on the conduction
band (CB) edge. One way for of reducing loss is splitting
the broad solar spectrum into different energy range and
converting each range with a cell of a well matched
bandgap [2,3]. The other way for enhancing performance
is added another absorption path like isolated energy level
to interact between valence band and conduction band [4].
It is the intermediate band solar cell (IBSC) and its benefit
is that its characteristics shows as tandem solar cells. And,
its maximum efficiency from detailed balance has been
enhanced [4]. If IBSC can be included into each stack of
multijunction solar cell, the performance of this hybrid
multijunction solar cell will be improved because of
increasing multi-transitions from IB effect. For instance, 2
stacks of tandem solar cell would be behaved like more
than 4 stacks of tandem cells. Because of increasing
number of transitions, the reduced number of stacks for
hybrid solar cell can be possible as well as its
performance will be enhanced. For this detailed balance
calculations, we have been devised a new model for this
hybrid system and been investigate the optimum IB level
of each stacks for high conversion efficiency.
THEORY

where Eg1=EC,1-EV (=ECV1), Eg2=EC,2-EV (=ECV2), EIV,1=Ei,,1EV, ECI1=EC,1-Ei,1, EIV,2=Ei,2-EV, ECI2=EC,2-Ei,2,

Figure 1. Basic carrier transitions at intermediateband
and multijunction (or tandem) hybrid solar cells
In this section, we are going to introduce an approach to
set the detailed balance model. It is shown in Figure 1 that
there are 6 carrier transitions of two stacks of hybrid solar
cells. From this picture, we have to set that EIV2 and ECI2
has to be greater than Eg1 to avoid overlapping of photon
energy while considering entire range of photon flux. The
detailed balance model for this hybrid solar cell is based
on combining basic assumptions of both tandem (or
multijunction) and IB solar cells. Compared to detailed
balance of conventional multijunction solar cells, we have
to consider different approach of spectrum splitting. For
instance, two stacks of IBT have two intermediate band
and its whole transitions will be six. In other words, first
stacks of IBT will have ECI1, EIV1, Eg1(=ECV1) and 2nd stack
of IBT is consisted of ECI2, EIV2 and Eg2(=ECV2). And, its
sequences of energies are ECI1<EIV1<Eg1<ECI2<EIV2<Eg2
because of using full spectrum range and avoiding double
counting of photon energy. So, we have derived detailed
balance equations of each stack based on these
conditions and its shows in equation (1) to (5). And, the
next higher energy of bottom stack is E1+. If, ECI2 is lower
than Eg1, the photon energy will start to count the
duplicated range in first stack of IBT. As results of these,
the detailed balance of hybrid solar cells is shown below
(equation 1 to equation 5).

Detailed balance equation of hybrid approach

978-1-4244-9965-6/11/$26.00 2011 IEEE

002082

N(E1,E2 ,T, ) 

2
h 3c 2

E2

e

E2
(E  )/kT

E1

1

dE

(1)

Spectrum splitting of multijunction solar cells and

hybrid solar cells

In addition, there are no carrier extractions at IB if

equation 3 is satisfied
J i  q  {[ fS  C  N( E g( i ) , E g( i )  ,TS ,0 )  N(( E g , E g ( i )  ,Ta ,  IV ( i ) )]
 [ fS  C  N( ECI ( i ) , E IV ( i ) ,TS ,0 )  N(( ECI ( i ) , E IV ( i ) ,Ta , CI ( i ) )]}

(2)

J N  q {[fS  C  N (E g (N ) , , TS ,0)  N ((E g (N ) , , Ta , IV (N ) )]

 [fS  C  N (ECI (N ) , E IV (N ) , TS ,0)  N ((ECI (N ) , E IV (N ) , Ta , CI (N ) )]}

fS  C  N(E CI ( i ) ,EIV ( i ) ,TS ,0 )  N(E CI ( i ) ,EIV ( i ),Ta ,CI ( i ) ) 

fS  C  N(E IV ( i ) ,Eg ( i ),TS ,0 )  N(E IV ( i ),Eg ( i ) ,Ta ,IV ( i ) )

CV ( i )  qV( i )  CI( i )  IV ( i )

(3)
(4)

(5)

Where, Ei is the energy for ith stack, is hither energy

state from Ei, TS is the temperature of the Sun (=6000 K),
Ta is the ambient temperature(=300 K), N is the photon
flux absorbed by or emitted from the semiconductor, i is ith
layer of IBT(=1,2,,N), fS=1/46300 and C is the
concentration of sunlight

where Eg1,low and Eg1,high are start and end photon energy
in the first stack, Eg2,low and Eg2,high are beginning and
nd
ending photon energy point at 2 stack. Eg1,opt and Eg2,opt
are the optimum energy points
(a) Spectrum splitting of multijunction solar cells

There are two approaches for connecting each stack for

IBT. In case of independent connected IBT solar cell, the
maximum power (PM) of each has to calculate and add
each PM to find maximum power of whole hybrid cells. The
advantage of this independent connection is less sensitive
compared to series connected because of current density
matching of each stacks for optimum operations [5,6]. In
the series connection, current density of each should be
the same so it is more sensitive than independent
connection [2,5]. We have summarized these relationships
at table 1.
Connect
Current
Voltage
Maximum
ion
Density
Power
N
Indepen
Jm,i
Vm,i
Pmax 
J m,iVm,i
dent
i 1
Connect
ion
Series
JTotal=J1= VTotal=V1+ Pmax=JTotalVTotal
Connect
=JN
+VN
ion
Table 1. The relationships of finding maximum power
point for both independent and series connected IBT
solar cells.
In this paper, we have simulated the independent
connections of IBT. Based on this, we can setup the
hybrid thermodynamic model and calculate its optimum
bandgaps and efficiencies.

978-1-4244-9965-6/11/$26.00 2011 IEEE

where ECV1,low and ECV1,high are start and end photon

energy in the first stack, ECV2,low and ECV2,high are beginning
and ending photon energy point at 2nd stack. ECV1,opt and
ECV2,opt are the optimum energy points
(b) Spectrum Splitting of hybrid solar cells (IBT)
Figure 2 Spectrum splitting of conventional
multijunction solar cells and hybrid solar cells (IBT)
In Figure 2, when considering spectrum splitting between
conventional multijunction solar cells and hybrid solar
cells, the most important thing is overlapping of photon
energy range. In case of conventional multijunction solar
cell, Eg1,high can be the nearly same with Eg2,low. But, it

002083

cannot be overlapped with Eg2,low. So, the optimum

bandgaps are can be determined below Eg1,high and Eg2,high
for bottom and top cell respectively. In case of IBT solar
cells, it is required margin of photon energy range
between ECV1,high and ECV2,low because of spectrum
overlapping and double counting of photon energy. When
regarding maximum energy points for both ECV1 and ECV2,
the ECI1, EIV1, ECI2 and EIV2 will be also determined.
Furthermore, ECV1 and ECI2 cannot be overlapped due to
spectrum splitting. And, the maximum value of of
equation 2 will be difference between ECV1 and ECI2. We
set as 0.01 eV to make simplification of calculating and
using full spectrum range. This is described in Figure 3.

70
One Sun
Maximum Concentration

60

Efficiency (%)

50

40

30

20

10

0
0.5

1.5

2.5
3
Eg (eV)

3.5

4.5

Figure 4 Efficiency vs bandgaps of conventional

intermediate band solar cells under one sun and
maximum concentration (blackbody radiation), where
Eg is equal to ECV
We have simulated for conventional intermediate band cell
through varying Eg (=ECV). And, the results are
summarized at Figure 4 and Table 4.
Con

Figure 3 Spectrum
positions.

splitting

and

its

bandgaps

RESULTS
In this section, we introduce the results of 2 stacks of
independently connected IBT solar cells. To analyze the
results, we have also compared the efficiencies of
conventional independent connection of multijunction (or
tandem) solar cells and intermediate band solar cells. To
evaluate the performance of IBT, we have considered the
position of IB to maximize the efficiencies. For this
consideration, we have changed IB positions to calculate
local maximum power point of each bandgaps (Eg or ECV).

ECI (eV)

EIV (eV)

ECV (eV)

Efficiency
(%)
One
0.92
1.49
2.41
46.89
Max
0.72
1.25
1.97
63.17
Table 2 Optimum bandgap positions and its maximum
efficiencies under one sun and maximum
concentrated sunlight.
To compare the results for IBT solar cell, we have
summarized the optimum bandgaps and its maximum
efficiencies up to 6 stacks of independent connection of
tandem solar cells in Figure 3 and Table 3 and 4 [2].
N
2

Con
Eg1
Eg2
Eg3
Eg4
Eg5
Eg6
One
0.98
1.87
Max
0.77
1.70
3
One
0.82
1.44
2.26
Max
0.62
1.26
2.10
4
One
0.72
1.21
1.77
2.55
Max
0.52
1.03
1.61
2.41
5
One
0.66
1.07
1.50
2.03
2.79
Max
0.45
0.88
1.34
1.88
2.66
6
One
0.61
0.96
1.33
1.74
2.26
3.00
Max
0.40
0.78
1.17
1.60
2.12
2.87
where N is a number of stack, Con is concentration of
sunlight, One is one sun and Max is maximum
concentration. unit of Egi (i=1,2,3,4,5,6,) is eV [2,3].
Table 3 Optimum bandgaps of each stacks for
independent connection of conventional tandem solar
cells.

978-1-4244-9965-6/11/$26.00 2011 IEEE

002084

N
2

Con
Efficiency (%)
One
42.9
Max
55.9
3
One
49.3
Max
63.8
4
One
53.3
Max
68.8
5
One
56.0
Max
72.0
6
One
58.0
Max
74.4
Table 4 Maximum efficiency for independent
connection of conventional tandem solar cells.
75

One Sun
C=1000
C=10000
Maximum Concentration

70

Efficiency (%)

65

60

55

50

45
0.5

1.5
Eg1 (eV)

2.5

Figure 5 Efficiency vs bandgaps of two stacks

independent connection of IBT
by varying
concentration (blackbody radiation)
Con
ECI1
EIV1
Eg1
ECI2
One
0.54
1.02
1.56
1.57
1000
0.44
0.88
1.32
1.33
10000
0.4
0.82
1.22
1.23
Max
0.4
0.82
1.22
1.23
where con is concentration of sunlight
maximum concentraiton, energy unit is eV.

EIV2
Eg2
2.13
3.7
1.89
3.22
1.79
3.02
1.79
3.02
and Max is

Table 5 Optimum bandgap positions related with IBT

Concentration
Efficiency (%)
One Sun
53.86
1000
64.83
10000
68.52
Maximum concentration
70.83
Table 6 The maximum efficiency (one sun and
maximum concentration).
The maximum efficiencies of both one sun and maximum
concentrations are 53.86 and 70.83% respectively. These
values are similar with between 4 or 5 stacks of
independently connect tandem solar cells. But, optimum
Eg1 and ECI2 are quite closed to each other. We have
varied the concentration which are one sun, 1000, 10000

978-1-4244-9965-6/11/$26.00 2011 IEEE

and maximum concentration. And, these results are

summarized at Figure 5 and Table 5 and 6. The values of
Eg1 and ECI2 are quite closed to from every concentration.
If we start to consider photon flux and its splitting range,
top cells of IBT will have a less limit of photon flux splitting.
In other words, top cell will behave a regular IB cell and
photon flux will count to infinite range. However, bottom
IBT cell will experience suppression of photon energy
splitting because the next higher energy state from bottom
cell is ECI2 and its difference is not large. In other words,
the next higher energy state, ECI2, is a part of 2nd stack of
IBT cell and its energy state can be closed to Eg1. In low
photon energy region, the optimum bandgaps between Eg1
and ECI2 will have a large separation because of enough
margin between Eg1 and ECI2. But, when the Eg1 and ECI2
are going to approach near maximum efficiency range,
there is a small margin to separate between ECI2 and Eg1.
CONCLUSION
We have calculated the detailed balance model for IBT
solar cells of independent connection. The maximum
efficiency for this hybrid solar cell is enhanced and its
maximum efficiencies are closed to four or five stacks of
tandem solar cells. Owing to behaving three transitions of
intermediate band solar cell, we can reduce the number of
stacks for hybrid solar cells and its performance has been
improved. And, we have calculated the maximum
efficiencies by changing concentration of sunlight to
analyze the tendencies of optimum bandgap of for each
transition. The transition from valence to conduction band
in the first stack is quite closed to the transition from
intermediate band to conduction band in top cell. These
tendencies are not changed, even if the concentration is
changed because of spectrum splitting scheme of this
hybrid solar cell.
REFERENCES
[1] W. Shockley and H.J Queisser. "Detailed Balance Limit
of
Efficiency
of
p-n
Junction
Solar
Cells",
J.Appl.Phys.32,1961, pp.510-519.
[2] A. S. Brown and M. A. Green, "Detailed Balance Limit
for The Series Constrained Two Terminal Tandem Solar
Cell", Physica E, 14, 2002, pp. 96 100
[3] A. Marti and G. L. Arafijo, "Limiting Efficiencies for
Photovoltaic Energy Conversion in Multigap Systems",
Solar Energy Materials and Solar Cells, 43, 1996, pp. 203222
[4] A. Luque and A. MartiI., Increasing the Efficiency of
Ideal Solar Cells by Photon Induced Transitions at
Intermediate Levels, Phys. Rev. Lett. 78, 1997, pp 50145017.
[5] S. P. Bremner, M. Y, Levy and C. B. Honsberg.
"Analysis of Tandem Solar Cell Efficiencies Under
AM1.5G spectrum Using a Rapid Flux Calculation Method"

002085

Prog. Photovolt: Res. Appl. 16, 2008, pp 225-233

[6] A De Vos, Detailed Balance Limit of the Efficiency of
Tandem Solar Cells, J. Phys. D: Appl. Phys, 13, 1980, pp
839-846.

978-1-4244-9965-6/11/$26.00 2011 IEEE

002086

STUDIES ON THE INFLUENCE OF EFFECTIVE CARRIER LIFETIME IN THE

PERFORMANCE OF INDUSTRIAL SILICON SOLAR CELLS
Saravanan S, Murty D S, Guru Prasad A, Dinesh Kumar and Prakash Suratkar
Cell Technology and Process Engineering, TATA BP Solar, Bangalore 560100, India
ABSTRACT
Minority carrier lifetime plays a significant role in the
performance of silicon (Si) solar cells, due to its
importance in qualifying the bare silicon wafers. In order to
fabricate the high efficiency industrial Si solar cells, high
life time based silicon wafers are a pre-requisite which
strongly depends on the defects and trap concentrations
of the wafer. In the present work, a systematic study of the
variation in the minority effective carrier lifetime on Si solar
cells has been studied. The solar grade mono silicon
wafers used for this study had bulk carrier life time < 10
secs with bulk resistivity of 0.5 3 cm. The silicon solar
cells were fabricated in the production line by employing
the conventional silicon solar cell process. As expected it
has been observed that a change of effective life time in
bare wafers improved the cell performance. This paper
narrates the electrical performance of the mono crystalline
silicon solar cells with respect to the different effective
carrier life time of the bare wafers. The results are
compared and correlated with the passivation process and
have been explained with respect to the different effective
carrier life time.
Keywords:
Mono crystalline silicon solar cells, Minority carrier life time
INTRODUCTION
Crystalline silicon solar cell technology is continued to be
a dominant and advanced technology in the photovoltaics
revolution compared to other PV technologies because of
the material availability at relatively lower cost (1). Even
though it is an advanced technology the researchers are
continued to work on Silicon based technology in the few
technical and economical patterns. Hence nowadays, PV
industries use the solar grade silicon material for the solar
cell production to reduce the over all cost of the PV
modules. Solar grade silicon is having lower minority
carrier life time compared with electronic grade silicon.
However these materials having higher concentration of
chemical and mechanical defects, including transition
metals, dislocations, and grain boundaries (1) which cause
lower minority carrier lifetime in solar grade Si. Minority
carrier lifetime is an important property of Si, which has
significant influence on the electrical performance of the
solar cells.
Carrier life time can be measured by using photo
conductance decay (PCD) technique and can also be
used as tool to detect effect of various processing steps
on minority carrier lifetime of silicon wafers (2,3). PCD

978-1-4244-9965-6/11/$26.00 2011 IEEE

technique is non-contact, non-destructive, highly sensitive

and faster way of measuring carrier lifetime. Hence this is
the promising metrology for carrier life time
characterization of Si wafers in PV industries.
The measured carrier life time is influenced by the bulk life
time (bulk) and surface recombination life time (surf) as
well and hence the life time is known as the effective
carrier life time (4). The measured carrier life time (meas)
can be expressed as

meas

bulk

diff =

diff

1
+ surf

(1)

d2
2 .D n , p

(2)

d
2.S

(3)

surf =

Where diff is the diffusion time for minority carriers from

the interior of the wafer to the surface determined by the
wafer thickness and the diffusion constant of minority
carriers; d is the wafer thickness, Dn,p is the diffusion
constants of minority carriers and S is the surface
recombination velocity. As expressed in the equation, the
measured carrier life time depends the surface
recombination of the wafers. It is important to note that
while the wafer is not passivated, the surface
recombination velocity is high which implies the measured
carrier life time is mainly indicates the surface
characteristic of the wafer. Where as when the wafer
surface is passivated, the surface recombination velocity
is zero the measured life time will correspond to the bulk
characteristic of the wafer (4).
In this paper, Cz wafers with different effective carrier life
time have been processed by conventional silicon solar
cell process. The electrical results were compared and
correlated to the carrier life time.
EXPERIMENTAL
P - type Boron doped solar grade mono Si wafers of size
125 x 125 mm with bulk resistivity in the range of 0.5 3
ohm/sq and thickness around 190 10 were taken as a
starting material for the solar cell fabrication. These silicon
wafers have been processed in the conventional industrial
silicon solar cell process and is described in brief. As per

002940

the process flow, initially the saw damages and the

residual contamination of the wafers were removed by the
alkali etching followed by the alkali texturization to form
the uniform homogeneous pyramids with crystal
orientation <111>. Subsequently, the p type wafers were
crafted as pn junction by diffusing phosphorus in tube
furnace using POCl3 as precursor. Sheet resistance of n
layer is around ~ 40 / . The phospho silicate glass
(PSG) formed at the edges and also at the surface of the
silicon wafers, were removed by plasma edge isolation
and with HF wash respectively. In order to make use of
maximum incident light and for the hydrogen passivation,
SiNx was coated of thickness ~ 80 nm and refractive index
~ 2.01 by employing PECVD technique. The front and
back contacts were made by using conventional screen
printing technique and fired to realize the ohmic contacts.
Electrical characteristic of the solar cells were measured
under the AM1.5G simulated solar radiation at 25 C.

performance of solar cells depend the carrier life time

along with the process as well.
The carrier life time is plotted against the open circuit
voltage and short circuit current which is shown in Fig 2. It
is evident from Fig 2 that the current and open circuit
voltage is increasing with respect to the carrier life time. In
general, the better carrier life time yields reduced bulk
recombination in the cell which improves the current and
voltage of the cell as well. From Fig 2 it has been noted
that improvement in current is predominant compared to
corresponding improvement in voltage. The better lifetime
indicates the bulk property which directly affects collection
of minority charge carriers. Therefore, improvement in
current is higher compared to the voltage.

The Life time measurements were carried out by using the

Sinton PCD method.
RESULTS AND DISCUSSION
The variation of the solar cell efficiency with effective
minority carrier lifetime is as shown in Fig 1. The value of
the efficiency is the average value of around 5000 wafers
and the carrier life time of these incoming wafers were
carried out as per the sample inspection plan and the
number of wafers checked are around 500 wafers. From
Fig. 1 it has been observed that the effective carrier life
time is directly influences the electrical performance.
Figure 2 Effect of life time on short circuit current and
open circuit current density of cells

17.2

1.2

1.05
16.9
1

629

36
35

627

34
33
32

Voc (mV)

17

1.1

37

Jsc (mA/cm^2)

1.15

631

38

17.1

Efficiency (%)

Carrier life time (micro sec)

1.25

625

31

0.95

16.8
Sample 1

Sample 2

Sample 3

30

623
1.136

Sample

Figure 1 Variation of the cell efficiency with effective

carrier lifetime
Also it has been noted that a change of ~ 10% in life time
enhances the efficiency by 0.6% where as the change of
~20% life time increases the efficiency by ~1.5%
improvement in the cell performance. As a result a small
change in the lifetime affects the cell performance by a
great extent which clearly indicates the electrical

978-1-4244-9965-6/11/$26.00 2011 IEEE

1.06

1.206

Effective carrier life time (micro sec)

Figure 3 Variation of short circuit current density and

Open circuit voltage with effective carrier lifetime
From Figure 1 and 2, it has been found that the efficiency
and open circuit voltage of the solar cells varied
significantly with the minority carrier life time. However the
short circuit current density for all the cells is almost
constant (Figure 3). Although the processing of the all the

002941

wafers are similar, it has been noted that the efficiency

improved with respect to carrier life time.
CONCLUSIONS
Mono crystalline silicon solar cells were fabricated by the
conventional industrial solar cell process. The influence of
minority carrier lifetime of the raw wafers on the electrical
performance was studied. The improvement in the life time
of the bulk material enhances open circuit voltage and
short circuit current of the solar cell which results in the
improvement of solar cell power conversion efficiency. The
fill factor of the cells remains unaffected as it mainly
depends on the sheet resistance of the front emitter. In
conclusion, the effective carrier life time was measured for
the large set of wafers and subsequently the electrical
parameters of the cells fabricated were measured. Even
though there is no much variation in the short circuit
current density, It is confirmed that for the conventional
cell production high open circuit voltage and hence
efficiencies are interpreted with effective carrier life time.
ACKNOWLEDGEMENT
Authors wish to thank Mr. K. Subramanya, CEO, TATA BP
Solar and Mr. Keshav Prasad, VP (Project) for the support
and encouragement in carrying out this work. Authors also
acknowledge Mr. Raghu Tatachar, Head, Cell
manufacturing, for the support. Authors are also indebted
to Mr. C. Nagesh for the sample preparation and
measurements.
REFERENCES
[1] Dieter K. Schroder, Carrier Lifetimes in Silicon IEEE
Transactions on electron devices, 44 (1), 1997, pp 160170
[2] R.K. Ahrenkiela, S.W. Johnston, Lifetime analysis of
silicon solar cells by microwave reflection Solar Energy
Materials & Solar Cells, 92, 2008, pp 830835
[3] LI Feng et al, Influence of surface passivation on the
minority carrier lifetime, Fe-B pair density and
recombination center concentration Chinese Sci Bull, 55
(17), 2010, pp 1828-1833
[4] G. Kumaravelu et al, Minority carrier lifetime in
plasma-textured silicon wafers for solar cells Solar
Energy Materials & Solar Cells, 87, 2005, pp 99106
[5] B. Ohnesorge et al, Minority-carrier lifetime and
efficiency of Cu(In,Ga)Se2 solar cells Applied Physics
Letters, 73(9), 1998, pp 1224-1226

978-1-4244-9965-6/11/$26.00 2011 IEEE

002942

122

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 1, FEBRUARY 2013

Defect Detection in Solar Modules

Using ICA Basis Images
Du-Ming Tsai, Member, IEEE, Shih-Chieh Wu, and Wei-Yao Chiu

AbstractSolar power has become an attractive alternative of

electricity energy. Solar cells that form the basis of a solar power
system are mainly based on multicrystalline silicon. A set of solar
cells are assembled and interconnected into a large solar module to
offer a large amount of electricity power for commercial applications. Many defects in a solar module cannot be visually observed
with the conventional CCD imaging system. This paper aims at
defect inspection of solar modules in electroluminescence (EL)
images. The solar module charged with electrical current will emit
infrared light whose intensity will be darker for intrinsic crystal
grain boundaries and extrinsic defects including micro-cracks,
breaks and finger interruptions. The EL image can distinctly highlight the invisible defects but also create a random inhomogeneous
background, which makes the inspection task extremely difficult.
The proposed method is based on independent component analysis (ICA), and involves a learning and a detection stage. The large
solar module image is first divided into small solar cell subimages.
In the training stage, a set of defect-free solar cell subimages are
used to find a set of independent basis images using ICA. In the
inspection stage, each solar cell subimage under inspection is reconstructed as a linear combination of the learned basis images.
The coefficients of the linear combination are used as the feature
vector for classification. Also, the reconstruction error between the
test image and its reconstructed image from the ICA basis images is
also evaluated for detecting the presence of defects. Experimental
results have shown that the image reconstruction with basis images
distinctly outperforms the ICA feature extraction approach. It can
achieve a mean recognition rate of 93.4% for a set of 80 test samples.
Index TermsDefect detection, independent component analysis, surface inspection, solar module.

I. INTRODUCTION

OLAR power has become an attractive alternative of electricity energy due to growing environmental concerns and
global oil shortage. Solar cells that convert the photons from
the sun to electricity are mostly based on crystalline silicon in
the current market because it can generate good performance in
usable lifespan and conversion efficiency among the currently
available techniques. Multicrystalline solar cells are more popular than monocrystalline solar cells. They dominate the production volume in the photovoltaic industry due to lower manufacturing costs.
The surface of a multicrystalline solar wafer shows multiple
crystal grains of random shapes and sizes in random positions
Manuscript received September 02, 2011; revised October 12, 2011 and February 06, 2012; accepted May 19, 2012. Date of publication July 20, 2012; date
of current version December 19, 2012. Paper no. TII-11-504.
The authors are with Yuan-Ze University, Chungli 320, Tao-Yuan, Taiwan
(e-mail: iedmtsai@saturn.yzu.edu.tw)
Digital Object Identifier 10.1109/TII.2012.2209663

Fig. 1. CCD-captured images of (a) multicrystalline solar wafer; (b) solar cell;
(c) solar module composed of 36 solar cells.

and orientations. It results in a heterogeneous texture in the surface. No two solar wafer surfaces contain completely identical
patterns. A solar wafer is a thin slice of a silicon ingot. It is
further processed and fabricated into a solar cell that forms the
basic unit of a solar power system. A set of solar cells are assembled and interconnected into a solar module to offer a large
amount of electricity power for a variety of commercial applications.
Fig. 1(a) shows the image of a defect-free multicrystalline
solar wafer. It contains random crystal grains on the surface.
Fig. 1(b) is a solar cell image taken by a typical charge-coupleddevice (CCD) camera, where the vertical thin metal strips are
finger electrodes that supply current to the two horizontal thick
bus bars. The bus bars are used for interconnecting solar cells
to make a solar module. Fig. 1(c) demonstrates a CCD-captured
solar module image. It is composed of a matrix of 36 (6 6)
solar cells.
Image processing techniques have become an important tool
for biometric recognition [1], [2], robotic automation [3], [4]
and industrial inspection [5][7]. The quality inspection of
a solar module in manufacturing is very important to ensure
the expected conversion efficiency and durable lifespan. A
few automatic defect detection algorithms in the literature
have been focused on the detection of defects in solar wafers
and solar cells. The inspection in the early manufacturing
stages can prevent delivering defective materials to the subsequent process. However, some of critical defects including
micro-cracks, breaks and finger interruptions may inevitably
occur in solar modules during the transportation, handling and

1551-3203/$31.00 2012 IEEE

TSAI et al.: DEFECT DETECTION IN SOLAR MODULES

assembly. Fuyuki and Kitiyanan [8] have also pointed out that
the point-pressure cracks due to the failure in the metal wire
soldering could occur in the solar module assembly process.
The defects of a solar module may reduce the current or even
consume the power stored in the battery bank. They may
also cause thermal effect that deteriorates the materials and
eventually shorten the lifespan of the solar module. This paper
proposes an efficient machine vision scheme for automatic
defect detection in multicrystalline solar modules. It can detect
subtle defects in solar modules with inhomogeneous surfaces,
and is computationally very fast to implement. It requires no
complicated feature design and feature extraction process for
distinguishing local defects from a wide variety of crystal grain
patterns.
Due to the random crystal grain surface of a solar cell or
interior defects that do not visually appear on the cell surface,
automatic defect detection in solar modules with the typical
CCD camera and lighting system cannot be effectively realized.
In order to highlight the deficiencies that degrade the conversion efficiency of a solar module, the electroluminescence
(EL) imaging technique [8][10] has been proposed in recent
years. In the EL imaging system, current is sent through a
solar module in a darkened room, and then a cooled Si-CCD
or InGaAs camera is used to capture the infrared light emitting
from the excited solar module. Areas of crystal silicon with
higher conversion efficiency exhibit brighter luminescence in
the sensed image. Process deficiencies such as micro-cracks,
breaks and finger interruptions will appear as dark regions
because they are inactive and hardly emit light. The dislocation
and grain boundary of silicon wafers also create dark regions in
the background of the EL image. Because the crystal grain pattern is random and unique for each individual multicrystalline
solar wafer, the crystal grain backgrounds of all solar cells in
a solar module are also randomly present in the EL image.
This causes automatic defect detection of solar modules in EL
images extremely difficult.
Fig. 2(a) and (b) demonstrates the EL images of a defect-free
and a defective solar module, respectively. As seen in Figs. 1(c)
and 2(b), the solar module captured in the CCD image cannot
detect the inner micro-cracks and some subtle local defects,
while the EL image can well present various defects embedded
in the solar module. Fig. 3(a)(c) further shows three enlarged
solar cell subimages in a solar module, which contain respectively micro-crack, break and finger interruption. As seen in
Fig. 2(a), each solar cell in the solar module presents a random
background pattern. The defective solar cells in a solar module
can be visually present in the EL image, as shown in Fig. 2(b).
Because a typical solar module is composed of multiple solar
cells in series, the image size of a solar module is generally
very large. However, the local defects of a small solar cell in
the module image could be very small. The large image size
with a low resolution of defects demands a very efficient and
yet very effective automatic visual inspection scheme for defect
detection of multicrystalline solar modules in EL images.
As aforementioned, most of the machine vision algorithms
for defect detection in the photovoltaic industry are mainly focused on the process levels of solar wafers and solar cells. Fu
et al. [11] presented a machine vision method to detect cracks

123

Fig. 2. Demonstrative EL images of (a) defect-free solar module, and (b) defective solar module.

Fig. 3. Defect types of solar cells displayed in the EL image: (a) micro-crack;
(b) break; (c) finger interruption.

in solar cells by evaluating the geometric features of the defect.

The method can only identify cracks occurred in the cell edges
with distinct gray levels from their surrounding. Ordaz and Lush
[12] analyzed the conversion efficiency of a solar cell based on
the gray-level distribution in the EL image. The gray-level histogram gives only the information about the overall proportion
of dark regions in the EL image. It cannot be used to detect the
presence/absence of local small defects in the EL image. Pilla
et al. [13] used the thermographic technique to highlight cracks
in solar cells.
Recently, Tsai et al. [14] presented an anisotropic diffusion
algorithm for detecting micro-cracks appeared on the solar
wafer surface. The sensed image is obtained from a typical
CCD camera and light source that shows the cracks with low
gray levels and high gradients. The diffusion process smoothes
the suspected defect region and preserves the crystal grain
background. The crack location can then be well detected by
subtracting the diffused image from the original image. The
method is specifically designed to identify micro-cracks by
assuming micro-cracks present darker intensity and higher gradient from their surrounding in the CCD-captured image. The
typical CCD image cannot detect the invisible inner cracks and
the algorithms can only detect the specific outer micro-cracks
in solar wafers. Tsai and Luo [15] further proposed a mean-shift
algorithm to detect three defect types of saw-mark, fingerprint
and contamination in multicrystalline solar wafers. The target
defects generally present high variation of gradient directions,
while crystal grain edges show more consistent gradient directions. The original gray-level wafer image is first converted
into an entropy image, where each pixel defined in a small
neighborhood window is represented by the entropy of gradient
directions. The mean-shit smoothing process is then performed
to remove noise and grain edges in the entropy image. The
final edge points remained in the filtered image are identified as
defective ones. Chiou et al. [16] used a near infrared imaging

124

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 1, FEBRUARY 2013

system to detect micro-cracks in solar wafers. The intensities

of a micro-crack are darker than those of the surrounding grain
background in the sensed image. A local thresholding and
a region-growing segmentation, followed by morphological
post-processing and blob analysis, are used to detect the dark
micro-cracks. The method is performed in the solar wafer
process, but not in the final solar module process. It is assumed
that the sensed micro-crack is significantly darker than the
crystal grains by infrared lighting. A dark, thin elongated
crystal grain in the defect-free solar wafer could be falsely
detected as a defect. It cannot be directly extended for detecting
various defect types found in solar modules.
The currently available defect detection algorithms reviewed
above are generally applied to detect defects with distinct geometric and intensity features, or are computationally intensive
to detect defects in a high-resolution image. Since the EL image
of a solar module under inspection is generally very large in size
and local defects in a solar cell subimage is relatively very small
in a random background, we need a fast detection algorithm to
screen all possible defective solar cells in the large solar module
image.
The proposed method is based on the independent component
analysis (ICA) technique [17]. ICA is a novel statistical signal
processing technique to extract independent sources given only
observed data that are mixtures of unknown sources without any
prior knowledge of the mixing mechanism. It has been applied
in texture analysis [18], [19] that uses either the estimated independent components or the column vectors of the estimated
mixing matrix as features for texture classification/segmentation. The ICA techniques have also been applied for surface
defect detection of textile fabrics [20], [21] and liquid crystal
display (LCD) panels [22] in manufacturing, and for grading
tomato ripening [23] in agriculture. All the ICA-based defect
detection methods are only applicable to non-textured or homogeneously-textured surfaces. They cannot be extended for
defect detection in the EL images with inhomogeneous background patterns.
The proposed ICA-based defect detection scheme involves a
learning process and an inspection process. Each solar module
image is first partitioned into individual solar cell subimages.
In the learning process, a set of defect-free solar cell subimages are randomly chosen from different solar modules. Each
2D solar cell subimage used for training is reshaped as a 1D
signal and forms a row vector in the data matrix. The ICA technique is then used to find a set of statistically independent basis
images from the training data matrix. In the inspection process,
the solar module image is also first divided into individual solar
cell subimages. A test solar cell subimage is reconstructed as a
linear combination of the learned basis images. It is expected
that a defect-free solar cell subimage can be well represented
by the basis images and the reconstructed image of a defective
solar cell subimage will be distinctly deviated from its original
one.
In this study, we evaluate two ICA-based approaches for
identifying the presence or absence of defects in a solar cell
subimage. The first approach is based on feature extraction and
the second approach is based on image reconstruction from
ICA basis images. The first ICA approach uses the coefficients

of the linear combination of basis images as the feature vector.

It is then compared with that of each training sample in the data
set by a distance measure, and the minimum distance among
all training samples is used as the discrimination measure. The
second ICA approach simply calculates the reconstruction error
between the original solar cell subimage under inspection and
its reconstructed image from the basis images. If the distance or
reconstruction error is larger than a predetermined threshold, it
is identified as a defective solar cell subimage. Otherwise, it is
declared as a defect-free subimage. The performance between
these two ICA-based approaches will be evaluated by a set of
EL images of solar modules. The proposed method can fast
detect the presence/absence of defects in solar modules, but it
cannot identify the actual shape and location of a defect in the
image.
The paper is organized as follows: Section II first discusses
the basic ICA model, followed by the process for extracting
independent basis images from a set of training samples. The
feature extraction and image reconstruction from the ICA basis
images for detecting the presence/absence of defects in an inspection image are then presented. Section III evaluates the performance of the proposed ICA-based detection methods and
analyzes the effect of changes in the number of basis images.
The comparative study between the proposed method and the
existing methods of adaptive thresholding and gray-level morphology is also presented in this section. The paper is concluded
in Section IV.

II. ICA BASIS IMAGES FOR DEFECT DETECTION

The EL image of a multicrystalline solar module investigated
in this study present multiple solar cell subimages, each containing randomly-shaped dark regions in the background. The
critical defects found in a solar module are micro-cracks, breaks
and finger interruptions. They are present as dark line- or barshaped regions, or divide a region into a bright and a dark area in
the EL image, as seen in Fig. 3(a)(c). There are no straightforward features based on geometric shapes and intensities to distinguish the difference between defect regions and the dark regions in the background. The proposed method uses ICA to find
a set of representative basis images that can best describe various defect-free solar cell subimages. Each solar cell subimage
partitioned from a large EL image of solar module is then represented as a linear combination of the basis images.
In order to inspect individual solar cells in the large solar
module, the solar cells in the whole EL image must be partitioned into small subimages, each containing exactly one solar
cell. It can be observed from the solar module image in Fig. 2
that the spacing between adjacent solar cells in either horizontal
or vertical direction is straightly and continuously dark in the
whole EL image of a solar module. Horizontal and vertical projections that accumulate the gray levels of each row and column
in the solar module image are first constructed. The local valleys
(i.e., local minima) of the projection then indicate the spacing
between adjacent solar cells. Two adjacent local valleys in the
same projection direction are fine-tuned so that each partitioned
solar cell subimage gives the same size.

TSAI et al.: DEFECT DETECTION IN SOLAR MODULES

125

It is expected that a line- or bar-shaped defect will have a very

small accumulated magnitude. We thus select the SE with the
minimum accumulated gray levels. Let -SE be the corresponding structuring element with a minimum accumulated
magnitude among the three directions, i.e.,
Fig. 4. Three line structuring elements in directions (a) 135 , (b) 45 and (c)
90 for background smoothing.

The proposed defect detection scheme in this paper includes

the off-line learning of ICA basis images and the on-line detection of defective subimages. For either off-line learning or
on-line detection, a solar module must be first partitioned into
solar cell subimages using the gray-level projection described
above. The subsequent learning and detection processes are
all carried out for solar cell subimages. In the off-line learning
stage, the ICA model is used to find a set of independent basis
images from randomly sampled defect-free solar cell subimages. In the on-line detection stage, feature extraction and
image reconstruction from the basis images are used to evaluate the presence/absence of defects in the inspection image.
In the following subsections, a morphological preprocessing
for image smoothing is first present. The basis ICA model
is then briefly discussed, followed by the off-line learning of
independent basis images. The on-line detections with feature
extraction and image reconstruction are finally described.
A. Morphological Smoothing
As seen in Fig. 2, each solar cell subimage presents a random
dark-region pattern in the background. Some of the dark regions are similar to the defect regions. In order to smooth the
dark-region background and preserve the defect shapes in the
EL image, a specific morphological smoothing process is preformed first in the EL image before it is used for off-line learning
and on-line detection. As observed in Fig. 3(a)(c), defects of
micro-cracks, breaks and finger interruptions are mostly lineor bar-shaped. We therefore design three structuring elements
(SEs) in the directions of 45 , 90 and 135 , each of length
pixels and width 1 pixel, as shown in Fig. 4.
The length is generally larger than the largest dimension of
a random dark region but smaller than the defect length so that
most of the dark regions in the background can be smoothed, and
yet the defects can be well preserved in the filtered image. In this
study, the length is given by 13. For a given pixel point
with gray level
, the gray levels in each of the three SEs
are accumulated. Given the origin of the image in the upper-left
corner, we have

Once the structuring element -SE is chosen, the gray-level dilation (i.e., local maximum gray level) is applied to the solar cell
subimage with the structuring element in direction . That is

If -SE contains all defect points, the dilated value will be still
small for a dark region. Otherwise, the dilated value will be large
for a bright region in the background. In this study, we do not
use a horizontal structuring element (i.e., 0 -SE) because we
would like to remove the two horizontal thick bus bars in the
solar cell subimage. Fig. 5(a1)(d1) shows, respectively, a defect-free and three defective solar cell subimages in the EL images. Fig. 5(a2)(d2) presents the corresponding morphological
smoothing results of Fig. 5(a1)(d1). The filtering results show
that the dark regions in the background are better smoothed
while all the defects of micro-crack, break and finger interruption are well preserved. The morphological smoothing process
allows a better extraction of representative basis images in the
ICA learning stage and a better image reconstruction in the detection stage.
B. ICA Model and Basis Images
In the ICA model [17], [24], the observed mixture signals
can be expressed as
(1)
where is an unknown mixing matrix; represents the latent
source signals, meaning that they cannot be directly observed
from the mixture signals . The ICA model describes how the
observed mixture signals are generated by a process that uses
the matrix to mix the latent source signals . The source signals are assumed to be mutually statistically independent. Based
on this assumption, the ICA solution is obtained by finding a
demixing matrix . It is used to linearly transform the observed
mixture signals to yield the estimated independent signals
with
(2)
The separated components in , called independent components (ICs), are required to be as mutually independent as possible. The matrix is an estimate of the latent source signals .
There exist many algorithms performing ICA [25][29]. FastICA [26], [30] is one of the most widely used techniques to
solve for the ICA model due to its computational efficiency
and effectiveness for source separation. The objective of an

126

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 1, FEBRUARY 2013

solar modules are used as the input data set

in the off-line
learning process. In the ICA model, each training sample of a
2D solar cell subimage must be reshaped as a 1D row vector in
the data matrix . Denote by
the gray level at pixel coordinates
of an
subimage. The data matrix consists
of defect-free samples, i.e.,
, where
is the -th training sample, and
with
. The 1D elements of are obtained from the 2D
subimage
, i.e.,
(5)
and
for
the training data matrix is given by

. The ICA model for

The demixing matrix

is obtained from the FastICA algorithm [30]. The row vectors of the estimated sources with
are the basis images. Given a data set
of training subimages, we can obtain up to basis images.
They are used to extract discriminant features or reconstruct the
image by representing the test image as a linear combination of
the learned basis images .
C. On-Line Defect Inspection

Fig. 5. Morphological smoothing on solar cell subimages: (a1) defect-free

sample; (b1)(d1) three defective samples; (a2)(d2) respective smoothing
results of (a1)(d1).

ICA model is to maximize the statistical independency of ICs.

Non-Gaussianity is a commonly used measure for statistical independency. It can be given by the negentropy [17]

For a solar module image under inspection, it is first divided

into individual solar cell subimages by horizontal and vertical
projections. The morphological operation then follows for each
solar cell subimage to smooth the background. Let be a solar
cell subimage in its 1D form for test. The surface characteristics
of this solar cell can be represented by the coefficients of the
linear combination of the basis images
.
That is
(6)

(3)
where
is a Gaussian random vector of the same covariance matrix as .
is the entropy of a random vector .
The negentropy is always non-negative and is zero if and only
if the random variable has a Gaussian distribution. It is well justified as an estimator of the non-Gaussianity of the ICs. Since
the exact probability density of the random variable is unknown,
the exact negentropy computation using (3) above is prohibited.
An approximation of the negentropy is proposed as follows:
(4)
where
ance.

is a Gaussian variable of zero mean and unit variis a non-quadratic function, which can be given by
, as suggested by Hyvarinen et al. [17].
The demixing matrix
with
that maximizes the
negentropy can be efficiently calculated with the fixed-point
search algorithm [26], [27] in FastICA.
To find a set of representative basis images, a collection of
partitioned solar cells randomly chosen from the defect-free

where
is the coefficient vector of the linear
combination. It can be used as the feature vector of the test
sample . The coefficient vector can be obtained by
(7)
is the pseudo-inverse of , and is given by
.
To determine the presence or absence of defects in the test
sample , the cosine distance is used to evaluate the similarity
between the test subimage with feature vector and every
sample
with feature vector
in the training data set .
Hence, the cosine distance
is defined as
where

(8)
where the feature vector

of training sample

is given by
(9)

TSAI et al.: DEFECT DETECTION IN SOLAR MODULES

127

A cosine distance of zero is obtained if the two compared subimages are identical. The final cosine distance of the test sample
is given by the one with the minimum value among all the
training samples in the data set , i.e.,
(10)
is larger than some preIf the minimum cosine distance
determined threshold, the test sample is classified as a defective one. Otherwise, it is claimed to be defect-free.
Since the feature extraction approach is based on the distance
measure of the test sample with respect to every training sample
in the data set, the choice of defect-free solar cell subimages
used in the training data set becomes very critical. We therefore
also propose a self-referential image reconstruction process to
detect the presence/absence of defects in a test image. Any test
sample can be reconstructed from the basis images by
(11)
where is the coefficient vector as defined in (7). The reconstruction error of the test sample is defined as
(12)
where the constant is used as a regularization, and is given by
(13)
If the test sample contains defects, we expect that it cannot be
well reconstructed from the defect-free basis images and the resulting reconstruction error should be distinctly large. If the reconstruction error is larger than some pre-determined threshold,
the test sample is identified as a defective one. Otherwise, it
is classified as a defect-free solar cell subimage.
For self-referential image reconstruction with ICA, there is a
special finding that is worth noting here. Because the demixing
matrix
given by the FastICA algorithm is orthogonalized, it
indicates that
. The reconstruction of a test sample
can be thus given by

computational complexity is
for the coefficient vector
and
for the reconstruction image . Therefore, the
overall computational complexity is
. The
computation can be easily and efficiently executed with matrix
operations.
III. EXPERIMENTAL RESULTS
This section presents the experimental results from a number
of multicrystalline solar modules in EL images to evaluate the
performance of the proposed method. All proposed algorithms
and experiments were implemented on a personal computer
with a Pentium Core 2 Duo 3.00 GHz processor. The whole
image of a solar module is 1250 1250 pixels wide with 8-bit
gray levels. The solar module under test comprises a matrix of
36 solar cells, each of image size 208 208 pixels. The proposed ICA-based defect detection method is highly dependent
on the defect-free training samples for extracting representative
basis images. Especially, the image reconstruction approach
evaluates the reconstruction error directly from the original
training samples. Different defect-free samples used to generate
the representative basis images may produce different detection
results. Therefore, we conducted 10 training sets of randomly
chosen defect-free samples for the experiments. Each training
set contains 30 random defect-free solar cell subimages and,
thus, each data set creates a different combination of 30 basis
images.
The performance of the proposed method under different
scenarios (e.g., input images with and without morphological
smoothing, and feature extraction approach vs. image-reconstruction approach) is evaluated by the recognition rate. It is
defined as
%
where

is the total number of test samples;

is the number of mis-detected defect samples;
is the number of falsely-detected defect-free
samples.

In order to identify defective solar cells with large reconstruc, we use the simple statistical control limit to set
tion errors
the threshold value. It is given by
(14)
In the equation above,
is the coefficient vector for the
linear combination of the original data matrix . Therefore, the
image reconstruction approach for detecting the presence/absence of defects in a test sample does not require the ICA
optimization search, i.e., the learning stage can be omitted. The
original data set itself forms the basis images and the same
reconstruction error can be obtained either from the ICA basis
images or from the original training images .
The proposed method needs to calculate the coefficient vector
using (7) and then reconstruct the inspection image from the
basis images using (11). Given an image of size
and
basis images, the reshaped 1-D image is of size
with
and the matrix
is of size
, the

where
and
are the mean and standard deviation of
from a set of defect-free test samples. The parameter
is a control constant. It is given by 3 in this study to follow
the commonly-used 3-sigma rule in statistical process control.
Likewise, the threshold for cosine distance
is also determined by the statistical control limit. In the experiments, there
are a total of 80 solar cell subimages used as the test samples, of
which 28 are defect-free and 52 are defective. The test samples
used in the experiments were provided by a local solar cell manufacturer. The quality engineers care more about the detection
capability for various defect types that may present randomly in
any locations in solar modules.

128

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 1, FEBRUARY 2013

TABLE I
RECOGNITION RATES OF ICA-BASED FEATURE EXTRACTION AND IMAGE
RECONSTRUCTION APPROACHES

The first experiment was conducted to evaluate the performance between the feature-extraction
and the image-reconstruction
. The morphological preprocessing is applied to both ICA approaches, and the number of basis images
is 30. Table I summarizes the statistics of recognition rates R%
of the two ICA methods for the 10 training data sets. It shows
that the average recognition rate of the image reconstruction approach is 93.4% with a maximum of 98.7%, while the average
recognition rate of the feature extraction approach is only 78.4%
with a maximum of 82.5%. The Std. dev. in the table means
the standard deviation of recognition rate % obtained from
the 10 training sets. Table I shows that most of the 10 training
sets can yield a good recognition rate larger than 92% and the
standard deviation is only 3.2% for the image reconstruction approach. The variation is due to the different basis images derived from individual training samples. If the training samples
contain good representations of various background patterns of
defect-free solar cells, the derived basis images can then sufficiently represent all possible background patterns of the test
samples. Since the training process can be performed off-line,
we can always carry out multiple training sets that contain different defect-free samples in a preliminary experiment and then
select the one that results in the maximum recognition rate for
on-line inspection in manufacturing.
To further compare the performance of these two approaches, the receiver operating characteristic (ROC) curves
are also given in Fig. 6 for the 9th training data set that gives
the best recognition rate. In the ROC plot, the true positive (TP)
rate is shown in the vertical axis and the false positive (FP)
rate is shown in the horizontal axis. The results show that the
ROC curve of the image reconstruction approach rises swiftly
upward and far outperforms the feature extraction approach.
The image reconstruction method uses the whole image (
data points of an
image) to calculate the reconstruction
error. It performs the self-comparison between the input image
and the template image pixel by pixel, where the template
image is generated from the input image itself. However, the
feature-extraction method uses only a small number of data
points, where
is the number of basis images, to calculate
the difference of feature vectors between the inspection image
and each individual defect-free image used in the training set.

Fig. 6. Receiver operating characteristic (ROC) curves of the ICA-based feature extraction and image reconstruction approaches.

Therefore, the feature-extraction method may result in severe

deviation when the inspection image is quite different from any
of the training samples. Since the image reconstruction method
uses more data points and evaluates the errors from the image
itself, it is more stable and reliable to generate a good discrimination measurement to detect the presence/absence of defects
in a solar cell subimage. In the subsequent experiments, only
the image reconstruction approach under different scenarios is
evaluated accordingly.
The second experiment was conducted to evaluate the performance of the proposed image reconstruction approach with and
without morphological smoothing. The number of basis images
used in the experiment is also 30. Table II lists the recognition
rates with the original EL images as the input and those with the
morphologically-smoothed images as the input. It shows that
the morphological smoothing process can slightly improve the
mean recognition rate by 2%. For all 10 training data sets, the
recognition rates of morphologically-smoothed images are consistently larger than those of original EL images. The morphological process tends to smooth (or remove) the dark regions in
the background and generate a more simple and uniform surface
so that the basis images can have a better representation for arbitrary EL images. The dark regions in the background are very
difficult to be smoothed while the defective regions are clearly
retained at the same time. The proposed morphological process
can mildly smooth or blur the dark regions in the background
and, thus, the resulting EL images can be better represented
from a given set of basis images. This gives a slight improvement on recognition rate. The experimental results indicate that
the proposed image reconstruction method can further improve
the recognition rate if a more effective background smoothing
process can be developed in the future.
The last experiment in this paper is to evaluate the effect of
changes in the number of training images (i.e., the number of
basis images). The proposed image reconstruction approach
with morphological smoothing is applied in the experiment.
Table III summarizes the statistics of recognition rates with
varying number of basis images from 20, 25, 30 to 40. It shows
that the recognition rate increases as the number of basis images

TSAI et al.: DEFECT DETECTION IN SOLAR MODULES

TABLE II
RECOGNITION RATES OF IMAGE RECONSTRUCTION APPROACH WITH AND
WITHOUT MORPHOLOGICAL SMOOTHING

TABLE III
EFFECT OF THE NUMBER OF BASIS IMAGES ON THE RECOGNITION RATE

is added up to 30. A small number of basis images of 20 also

generate a good mean recognition of 91.8%. A larger number of
basis images of 40 cannot further improve the recognition rate.
In this study, each individual inspection image is reconstructed
from a linear combination of a given set of basis images. If
the derived basis images cannot sufficiently represent a variety of crystal-grain backgrounds in EL images, the resulting
reconstruction errors could be very large even for defect-free
test images. With less training samples, the limited number
of basis images cannot well represent all possible defect-free
background patterns. Both defect-free and defective EL images
may thus generate large reconstruction errors in a similar range,
and the resulting measurement cannot distinctly discriminate
the defective images from the normal ones. When the number
of training examples is sufficiently large, more training examples give only redundant representations of crystal-grain
background patterns in EL images and cannot further improve
the recognition rate.
Most of the currently available algorithms for solar wafer/cell
inspection such as [11], [16] are either based on adaptive thresholding followed by binary morphology and blob analysis or
based on gray-level morphology followed by region-segmentation and blob analysis. We therefore compare the proposed
method with such thresholding/morphology approaches. The
four solar cell samples demonstrated in Fig. 5 are used for the
performance comparison. Fig. 7(a1)(d1) shows repeatedly the
four test samples. Fig. 7(a2)(d2) presents the detection results

129

from the adaptive thresholding, followed by a 11 11 binary

morphological closing and blob analysis for elongated objects,
where black objects are final detected faulty items and the gray
objects are the intermediate results of binary morphology. The
experiment shows that the finger interruptions can be detected.
It fails to identify the subtle micro-cracks, and presents severe
noise. Threshlding-based methods require a higher defect
contrast than the grain boundary contrast in the EL image.
However, the defective image shown in Fig. 7(d1) contains
a small crack with gray values larger than those of grain
edges. The thresholding method thus mis-detects the crack
and presents some grain edge points as noise in the segmented
image. Fig. 7(a3)(d3) further shows the detection results of
gray-level morphology. The original solar cell subimage is first
smoothed by gray-level morphological closing with 5 13
structuring elements at 0 , 45 , 90 and 135 directions to
remove the possible defects. The original image is then subtracted from the smoothed image. Blob analysis for elongated
objects is finally carried out to identify the defective regions.
The experiment shows that it can detect some of the finger
interruptions and the small micro-crack with severe noise. In
the EL image, the crystal grains could be cloud-shaped and
very large in size in the background. There are no effective
morphological operations that can remove thin elongated defects and preserve cloud-shaped grains at the same time, or
vice versa. Severe false alarms for defect-free solar cells are
thus generated. The proposed method can identify correctly all
these four test samples. The comparative methods of adaptive
thresholding and gray-level morphology are also applied to
the 80 test samples aforementioned in this section. The two
comparative methods perform poorly with recognition rates
lower than 50%, which are worse than random guess.
IV. CONCLUSION
In this paper, we have proposed defect detection methods
based on ICA basis images to detect defective solar cell subimages of a large solar module in the electroluminescence image.
The line- and bar-shaped defects of micro-cracks, breaks and
finger interruptions in the solar module can be well presented
as dark regions in the EL image. However, the EL image also
displays the dislocations and grain boundaries of the multicrystalline solar wafer as dark regions and results in a random inhomogeneous background. The dark regions of defects and those
in the defect-free background can be visually observed in the
EL image, but they are extremely difficult to be distinguished
automatically.
We have evaluated two discrimination approaches based on
ICA basis images. The first feature extraction approach represents the test image under inspection as a linear combination of
the basis images and takes the coefficients as the feature vector.
The minimum cosine distance of feature vectors between the
test image and all training samples is then used as the discrimination measure. The second image reconstruction approach synthesizes the test image by the linear combination of the basis
images. The reconstruction error between the test image and
its reconstructed image is then used as the discrimination measure. Experimental results show that the image reconstruction
approach distinctly outperforms the feature extraction approach

130

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 1, FEBRUARY 2013

Fig. 7. Comparison of the proposed method with adaptive thresholding/morphology approaches: (a1)(d1) EL images of one defect-free and 3 defective solar
cells for testing; (a2)(d2) respective detection results from adaptive thresholding and binary morphology; (a3)(d3) respective detection results from gray-level
morphology.

and it can use the original defect-free training samples as the

basis images without the need of ICA process for learning. The
overall recognition rate of the image reconstruction approach is
93.4% for the inspection of solar cell subimages. The computation time of image reconstruction and measurement of reconstruction error is only 0.03 seconds with 30 basis images for a
208 208 solar cell subimage, or 1.08 seconds for a whole solar
module image containing 36 solar cells. It is computationally
very efficient for a large solar module of image size 1250 1250
pixels.
The additional processing time of the morphological operation is 0.04 seconds for a 208 208 solar cell subimage or
1.44 seconds for a whole solar module image of 1250 1250.
A more efficient smoothing preprocessing that can effectively
remove the random dark regions in the background for better
basis image extraction and representation is worth further investigation.
The proposed method in this study can be treated as a global
approach that detects only the presence/absence of defects in
solar modules. It cannot classify the type of defect. A local
approach that detects the shape and location of a defect in the
EL image using a more complicated spectral technique such as
the Fourier transform or the wavelet transform is currently under
investigation. The local approach generally could be computationally more intensive. Since a solar module involves a very
large image size, the global approach can efficiently screen
the suspected defective solar cells and then the local approach
can effectively verify and detect the actual shape and location
of a defect for the suspected solar cells. The incorporation of
both global and local approaches can then both efficiently
and effectively identify defects of solar modules in EL images.

REFERENCES
[1] Y. Si, J. Mei, and H. Gao, Novel approaches to improve robustness,
accuracy and rapidity of iris recognition systems, IEEE Trans. Ind.
Inf., vol. 8, pp. 110117, Feb. 2012.
[2] S. Jin, D. Kim, T. T. Nguyen, D. Kim, M. Kim, and J. W. Jeon, Design and implementation of a pipelined datapath for high-speed face
detection using FPGA, IEEE Trans. Ind. Inf., vol. 8, pp. 158167,
Feb. 2012.
[3] R. Gerndt, S. Michalik, and S. Krupop, Embedded vision system
for robotics and industrial automation, in IEEE Int. Conf. Ind. Inf.
(INDIN), Caparica, Lisbon, 2011, pp. 895899.
[4] S. Livatino, F. Banno, and G. Muscato, 3-D integration of robot vision and laser data with semiautomatic calibration in augmented reality
stereoscopic visual interface, IEEE Trans. Ind. Inf., vol. 8, pp. 6977,
Feb. 2012.
[5] C.-H. Chan and G. K. H. Pang, Fabric defect detection by Fourier
analysis, IEEE Trans. Ind. Appl., vol. 36, pp. 12671276, 2000.
[6] A. Picon, O. Ghita, P. F. Whelan, and P. M. Iriondo, Fuzzy spectral
and spatial feature integration for classification of nonferrous materials
in hyperspectral data, IEEE Trans. Ind. Inf., vol. 5, pp. 483494, 2009.
[7] D.-M. Tsai, I.-Y. Chiang, and Y.-H. Tsai, A shift-tolerant dissimilarity
measure for surface defect detection, IEEE Trans. Ind. Inf., vol. 8, pp.
128137, Feb. 2012.
[8] T. Fuyuki and A. Kitiyanan, Photographic diagnosis of crystalline silicon solar cells utilizing electroluminescence, Appl. Phys. A, vol. 96,
pp. 189196, July 2009.
[9] T. Fuyuki, H. Kondo, T. Yamazaki, Y. Takahashi, and Y. Uraoka, Photographic surveying of minority carrier diffusion length in polycrystalline silicon solar cells by electroluminescence, Appl. Phys. Lett.,
vol. 86, p. 262108, Jun. 2005.
[10] Y. Takahashi, Y. Kaji, A. Ogane, Y. Uraoka, and T. Fuyuki, LuminoscopyNovel tool for the diagnosis of crystalline silicon solar
cells and modules utilizing electroluminescence, in Proc. 4th World
Conf. Photovoltaic Energy Conv., Hawaii, 2006, pp. 924927.
[11] Z. Fu, Y. Zhao, Y. Liu, Q. Cao, M. Chen, J. Zhang, and J. Lee, Solar
cell crack inspection by image processing, in Proc. Int. Conf. Business Electronic Product Reliability Liability, Shanghai, China, 2004,
pp. 7780.
[12] M. A. Ordaz and G. B. Lush, Machine vision for solar cell characterization, in Proc. SPIE, San Jose, CA, 2000, pp. 238248.

TSAI et al.: DEFECT DETECTION IN SOLAR MODULES

[13] M. Pilla, F. Galmiche, and X. Maldague, Thermographic inspection

of cracked solar cells, in Proc. SPIE, Orlando, FL, USA, 2002, pp.
699703.
[14] D. M. Tsai, C. C. Chang, and S. M. Chao, Micro-crack inspection
in heterogeneously textured solar wafers using anisotropic diffusion,
Image Vis. Comput., vol. 28, pp. 491501, Mar. 2010.
[15] D. M. Tsai and J. Y. Luo, Mean shift-based defect detection in multicrystalline solar wafer surfaces, IEEE Trans. Ind. Inf., vol. 7, pp.
125135, Feb. 2011.
[16] Y. C. Chiou, F. Z. Liu, and Y.-T. Liang, Micro crack detection of
multi-crystalline silicon solar wafer using machine vision techniques,
Sensor Rev., vol. 31, pp. 154165, 2011.
[17] A. Hyvarinen, J. Karhunen, and E. Oja, Independent Component Analysis. New York: Wiley, 2001.
[18] R. Manduchi and J. Portilla, Independent component analysis of textures, in Proc. IEEE Int. Conf. Comput. Vis., Kerkyra, Greece, 1999,
pp. 10541060.
[19] R. Jenssen and T. Eltoft, Independent component analysis for texture
segmentation, Pattern Recognit., vol. 36, pp. 23012315, Oct. 2003.
[20] S. Serdaroglu, A. Ertuzun, and A. Ercil, Defect detection in textile
fabric images using wavelet transforms and independent component
analysis, Pattern Recognit. Image Anal., vol. 16, pp. 6164, Jan. 2006.
[21] O. G. Sezer, A. Ercil, and A. Ertuzun, Using perceptual relation of
regularity and anisotropy in the texture with independent component
model for defect detection, Pattern Recognit. , vol. 40, pp. 121133,
Jan. 2007.
[22] C. J. Lu and D. M. Tsai, Independent component analysis-based defect detection in patterned liquid crystal display surfaces, Image Vis.
Comput., vol. 26, pp. 955970, Jul. 2008.
[23] G. Polder and G. W. A. M. van der Heijden, Tomato sorting using
independent component analysis on spectral images, Real-Time
Imaging, vol. 9, pp. 253259, Aug. 2003.
[24] A. Hyvarinen and E. Oja, Independent component analysis: Algorithms and applications, Neural Networks, vol. 13, pp. 411430, Jun.
2000.
[25] A. J. Bell and T. J. Sejnowski, An information-maximization approach to blind separation and blind deconvolution, Neural Comput.,
vol. 7, pp. 11291159, Nov. 1995.

131

[26] A. Hyvarinen and E. Oja, A fast fixed-point algorithm for independent component analysis, Neural Comput., vol. 9, pp. 14831492, Oct.
1997.
[27] A. Hyvarinen, Fast and robust fixed-point algorithms for independent
component analysis, IEEE Trans. Neural Netw., vol. 10, pp. 626634,
May 1999.
[28] R. Boscolo, H. Pang, and V. P. Roychowdhury, Independent component analysis based on nonparametric density estimation, IEEE Trans.
Neural Netw., vol. 15, pp. 5565, Jan. 2004.
[29] T. Blaschke and L. Wiskott, CuBICA: Independent component
analysis by simultaneous third- and fourth-order cumulant diagonalization, IEEE Trans. Signal Process., vol. 52, pp. 12501256, May
2004.
[30] J. Hurri, H. Gavert, J. Sarela, and A. Hyvarinen, FastICA Package [Online]. Available: http://www.cis.hut.fi/projects/ica/fastica/

Du-Ming Tsai received the B.S. degree in industrial engineering from the
Tunghai University, Taiwan, in 1981, and the M.S. and Ph.D. degrees in
industrial engineering from Iowa State University, Ames, in 1984 and 1987,
respectively.
From 1988 to 1990, he was a Principal Engineer of Digital Equipment Corporation, Taiwan, where his work focused on process and automation research and
development. Currently he is a Professor of Industrial Engineering and Management at the Yuan-Ze University, Taiwan. His research interests include automated visual inspection, object recognition and texture analysis.

Wei-Yao Chiu received the B.S. degree in industrial management from the
Chung-Hua University, Taiwan, in 2005 and the M.S. degrees in industrial engineering and management from the Yuan-Ze University, Taiwan, in 2007, where
he is currently pursuing the Ph.D. degree in industrial engineering and management.
His research interests include automated visual inspection and video surveillance.

S3-1

Invited

High efficiency amorphous and nanocrystalline silicon thin film solar cells on flexible
substrates
Baojie Yan, Guozhen Yue, Jeffrey Yang, Subhendu Guha
United Solar Ovonic LLC
1100 West Maple Road, Troy, Michigan 48084, USA
Phone: +1-248-370-5410

Fax: +1-248-362-4442

E-mail: byan@uni-solar.com

We review the progresses and issues towards manufacturing hydrogenated amorphous silicon (a-Si:H) and nanocrystalline silicon (nc-Si:H)
based thin film multi-junction solar cells using a roll-to-roll process on flexible substrates. United Solar has been heavily involved in the
research and development of high efficiency a-Si:H and nc-Si:H multi-junction solar cells since 2001. We have resolved several critical issues
limiting nc-Si:H solar cell performance, such as nanocrystalline evolution, impurities, and porosity/ambient degradation. We have developed
new cell designs, including the proper optimization of the seeding layer for nc-Si:H growth and proper n/i and i/p buffers for improvement of
cell efficiency. We have optimized Ag/ZnO back reflectors for nc-Si:H cell performance. Combining all of the efforts in the improvement of
material quality and optimization of device structure, we have advanced thin film silicon solar cell efficiency. We reported 16.3% initial
active-area efficiency using an a-Si:H/a-SiGe:H/nc-Si:H triple-junction solar cell. Furthermore, we attained 12.5% stable total area (0.27 cm2)
efficiency using a-Si:H/nc-Si:H/nc-Si:H triple-junction solar cells and 11.3% stable aperture area (800 cm2) efficiency using the same cell
structure, where the efficiencies were measured by NREL and were the records for thin film silicon photovoltaic technology.
Introduction
The photovoltaic (PV) solar energy industry has been growing

used as absorber layer in the bottom cells in multi-junction solar

very fast in the last a few years. Crystal silicon (c-Si) PV panel

light-induced degradation and high absorption coefficients in the

production is the dominant technology because of the high

red and infrared wavelength region. The higher current capability

efficiency and very well developed manufacturing technology.

and better stability of a-Si:H and nc-Si:H based multi-junction

However, with the high demands for solar energy of terawatt

solar cells imply higher stable solar cell and module efficiencies

scales, new technologies capable of providing low production

than a-Si:H and a-SiGe:H based multi-junction solar cells. With the

costs which are environment friendly and have no-limitation in

great efforts of the community, the a-Si:H and nc-Si:H

raw materials are needed. Thin film silicon PV is one of the

multi-junction solar cell efficiency has exceeded the previous

potential technologies to meet the requirements. Hydrogenated

records attained using a-Si:H/a-SiGe:H/a-SiGe:H triple-junction

amorphous silicon (a-Si:H) and silicon germanium alloy

solar cells. The main disadvantage of nc-Si:H as the absorber layer

(a-SiGe:H) based thin film silicon have been used as the absorber

in solar cell is the low absorption in the short and middle

layers in multi-junction solar cells. The high defect density and low

wavelength region such that a thick nc-Si:H intrinsic layer is

carrier mobility in thin film amorphous silicon materials, especially

needed. Therefore, a high deposition rate is essential for cost

a-SiGe:H, cause a-Si:H and a-SiGe:H based solar cells having

effective nc-Si:H solar cell manufacturing. Very high frequency

lower efficiency than c-Si solar cells. Although 14.6% initial and

(VHF) glow discharge has been proven an effective method to

13.0% stable active area efficiencies were attained in 1997 by

increase nc-Si:H deposition rate, while the reduced wavelength of

1.

1)

cells. The advantages of nc-Si:H over a-SiGe:H are low

the a-Si:H/a-SiGe:H/a-SiGe:H solar laminate

VHF signal compared to radio frequency (RF) excitation leads to

product still shows a rated stable aperture-area efficiency around

more technical challenges in large area uniform deposition.

8.2%. The lower efficiency in the product than the lab devices is

Therefore, high efficiency, high rate disposition, and large-area

because of various limitations such as high deposition rate, less

uniformity are still the major tasks for the thin film silicon PV

reflective Al/ZnO back reflector (BR), and non-uniformity over a

community. We have focused on these three aspects. In this paper,

large area deposition.

we mainly review the issues related to efficiency improvements

United Solar,

Since the pioneer work of the Neuchtel group in the early

and the progresses made in United Solar.

2)

2.

90s, hydrogenated nanocrystalline silicon (nc-Si:H), also called

microcrystalline silicon (c-Si:H), has been widely studied and

Experimental Details
We have investigated three methods for nc-Si:H solar cell

AM-FPD '12

67

depositions: RF glow discharge, VHF glow discharge, and

other parameters may also be used to control the crystallinity, such

microwave glow discharge. Although RF glow discharge can

as the VHF excitation power. In depleting deposition regimes,

make nc-Si:H with good quality, the deposition rate is normally

increasing the VHF power can effectively increase atomic

lower than needed in production. Microwave glow discharge can

hydrogen density, hence increases the crystallinity. One can reduce

deposit nc-Si:H at very high rates such as >3 nm/s, but the material

the VHF power during the deposition to control the crystallinity.

quality is not good. A high density of microvoids results in an

Other deposition parameters such as substrate temperature,

ambient degradation without intentional light soaking. Therefore,

pressure, and total gas flow rate may also be used, but no

we have mainly focused on VHF deposition. We use

systematic studies have been made.

multi-chamber systems to make nc-Si:H solar cells. The doped

layers are normally made using RF glow discharge and the nc-Si:H
intrinsic (i) layers are made using VHF glow discharge. n-i-p
structured solar cells are made on Ag/ZnO BR coated flexible
stainless steel substrates. An Indium-Tin-Oxide (ITO) layer is

Table I. J-V characteristics of two pairs of nc-Si:H single-junction solar

cells made using constant (CHD) and profiled (PHD) hydrogen dilution
during the i-layer depositions. The pair of A-C and A-P has poorer
performance than the pair of B-C and B-P.
FF

Eff

Sample

Jsc

Voc

No.

(mA/cm2)

(V)

A-C

21.48

0.482

0.632

6.54

CHD

spectra are measured in the wavelength range from 300 nm to

A-P

25.15

0.502

0.663

8.37

PHD

1200 nm. For large area modules, metal grid wires are bonded on

B-C

26.61

0.504

0.665

8.92

CHD

the ITO layer for effective current collection and encapsulations

B-P

26.73

0.502

0.685

9.12

PHD

deposited on the top of the p layer as a transparent electrode.

Current density versus voltage (J-V) characteristics are measured
under an AM1.5 solar simulator at 25C. Quantum efficiency (QE)

are made for protection.

3.

Results and Discussions

Compared to a-Si:H, nc-Si:H growth has its unique

characteristics, which has significant impacts on solar cell

performance. We address a few key issues below.
Crystalline evolution and dynamic control during deposition
It has been observed that nanocrystalline formation normally
starts with an amorphous incubation layer. The crystallinity

(%)

Comment

Impurities and their correlation to carrier collection

nc-Si:H solar cells are more sensitive to impurities than
a-Si:H solar cells. Normally, a-Si:H solar cells do not show
noticeable effect of O when the O content is below 1-21019
at./cm3. However, when the O level exceeds 1-21019 at./cm3, it
causes a reduction in the QE response in the long wavelength as
shown in Fig. 1(a).4)

(crystallite size and crystalline volume fraction) increases with the

film thicknesses when a constant deposition condition is used. It
causes a harmful effect on solar cell performance. First, the
amorphous incubation layer has a high resistance and reduces the
effective carrier collection. Second, the increased crystallinity
causes a high microvoid density and poor grain boundary
passivation, which normally results in an ambient degradation. To
resolve this issue, we have developed a hydrogen dilution profiling
technique, where the H2/SiH4 ratio decreases with time during the
deposition.3 By changing the dynamics of the hydrogen dilution,
we demonstrated that the crystallinity can be effectively controlled
and can maintain the desired distribution for high solar cell
performance. Table I lists a comparison of nc-Si:H solar cells made
with and without hydrogen dilution profiling. It clearly shows that
the cell performance is improved by the hydrogen dilution
profiling. It also shows that the hydrogen dilution profiling has a
greater effect when the solar cells are poor as compared between
the two pairs.
Because the nc-Si:H deposition is a multi-variable system,

68

AM-FPD '12

Fig. 1. QE spectra of nc-Si:H solar cells with (a) n-type impurity and (b)
p-type impurity contaminations.

Similar to c-Si, O forms weak donors in nc-Si:H, which

(RMS) smaller than 40 nm and a thin ZnO layer of 140 nm gives

moves the Fermi-level towards to conduction band-edge. When

the best nc-Si:H solar cell efficiency.5) Fig. 3 shows the J-V and QE

the weakly n-type doped nc-Si:H is used in an n-i-p solar cell, the

curves of a nc-Si:H solar cell on such a Ag/ZnO BR, in which the

region near the n-i interface becomes n-type and the electric field in

Jsc is over 30 mA/cm2. The QE curves measured under 0V and

this region becomes weak such that the photocarriers generated in

-5.0V show no difference, indicating the texture of this BR does

this region are not easily collected by the external circuit without

not cause any issues in photocarrier collection. Using a highly

an applied reverse bias. Because only long wavelength photons

textured BR, we can obtain high photocurrent density under a

can generate carriers in the n-i interface region, the QE loss is

reverse bias, but a large collection loss causes a poor FF and low

observed in the long wavelength region when the nc-Si:H solar

efficiency. Therefore, it is a great challenge to achieve high light

cells have weak n-type doping.

trapping without causing a negative impact on nc-Si:H material

Due to the same reason, when nc-Si:H solar cells have p-type

quality.
5

doping, a short wavelength loss is observed as shown in Fig. 1(b).

-5
J (mA/cm 2)

previous p-layer deposition. Experimentally, we found that the B

doping is much more sensitive in nc-Si:H solar cells. A level of
16

(a)

The weak p-type doping is normally from residual B from the

10 at./cm can cause a noticeable change in the QE curves and a

Jsc=30.47 mA/cm2
Voc=0.488 V
FF=0.646
Rs=2.33 W.cm2
Eff=9.61%

-10
-15
-20
-25

reduction in fill factor (FF). Both the n-type and p-type impurities

-30

in nc-Si:H solar cells should be carefully reduced to maintain good

-35
-0.1 0.0

performance. In addition, a proper compensation could reduce the

0.1

0.2

0.3

0.4

0.5

V (V)

QE losses, but it is very difficult to control precisely the amount of

1.0

compensation. We have demonstrated that we can reduce the O

(b)

level down to less than 51018 at./cm3 and B level down to less

0.8

than 11016 at./cm3, in this case, the impurities have no effect on

0.6

0V

QE

nc-Si:H solar cell performance.

Proper cell design
In order to achieve a high efficiency, nc-Si:H solar cells

0.4

-5 V

0.2

normally have a complex cell design. For an n-i-p structure, we

0.0

need to add an a-Si:H buffer layer, which prevents any potential P

300

contamination from the n-layer deposition; a high crystalline seed

500
700
900
Wavelength (nm)

1100

layer on top of the a-Si:H buffer layer to ensure a proper growth of

Fig. 2. (a) J-V characteristics and (b) QE curves of a nc-Si:H solar cell. The

the nc-Si:H intrinsic layer; and an a-Si:H buffer layer is inserted

QE curves were measured under (blue) 0V and (red) -5.0 V.

between the nc-Si:H i-layer and the p-layer to reduce the shunt
current and improve Voc.
Light trapping
Because of the indirect bandgap in the crystalline phase,
nc-Si:H solar cells need much more thicker i-layers than a-Si:H

Multi-junction structures
In order to use the solar spectrum effectively and reduce the
light-induced degradation, multi-junction solar cells are commonly
used for high efficiency. We have investigated various

solar cells. In order to get high photocurrent density without a very

multi-junction solar cell structures. We found that the

thick i-layer, effective light trapping is needed. Very textured

a-Si:H/nc-Si:H double-junction structure is relatively simple and

substrates such as Ag/ZnO BRs are used for highly effective light

could be used for production with low cost. However, the

trapping. However, it has been observed that nc-Si:H material

efficiency

quality is reduced with the increase of substrate texture such that

a-Si:H/a-SiGe:H/nc-Si:H triple-junction structures give the highest

one cannot use very textured BRs in nc-Si:H solar cells for high

initial efficiency. Using this cell structure and a dual-function

efficiency. For Ag/ZnO BRs, the textures can be made on the Ag

surface and/or on the ZnO surface. Because the texture on Ag
surface seems more effective than on ZnO surface, we found that a
Ag/ZnO BR with the proper Ag texture of root-mean-square

is

lower

than

triple-junction

structures.

nc-SiOx:H layer between the a-SiGe:H middle cell and the nc-Si:H
bottom cell, we achieved an initial active-area efficiency of
16.3%.6) The J-V characteristics and QE curves of the high

AM-FPD '12

69

to match the current. A thick a-SiGe:H middle cell normally causes

limit the solar cell efficiency, we have improved thin film silicon

a large light-induced degradation. In addition, the a-SiGe:H has to

solar cell efficiency. Fig. 4 shows the progress made by United

be made at a low rate. Therefore, the a-Si:H/a-SiGe:H/nc-Si:H

Solar. We believe that the understanding of the nc-Si:H materials is

triple-junction may not be a viable selection for high efficiency thin

still very limited. The plasma process can be further optimized and

film solar laminates. United Solar decided to use an

new deposition methods may be invented. Therefore, the

a-Si:H/nc-Si:H/nc-Si:H triple-junction cell structure for next

efficiency of thin film silicon solar cells can be further improved to

generation high efficiency product. Although the initial efficiency

make it a viable technology for large volume solar panel

achieved using this structure is not the highest, the light induced

production.

degradation is low and the deposition rates can be high for all three

J (mA/cm2)

efficiency solar cell are shown in Fig. 3. By resolving issues which

component cells. We have recently demonstrated aperture-area

(800 cm2) initial and stable module efficiencies of 12.0% and

-2

11.3%

Voc=2.242 V
FF=0.771
J sc=9.43 mA/cm2
Eff=16.3%

-4

an

a-Si:H/nc-Si:H/nc-Si:H

-6

With further optimization of the nc-Si:H quality, cell structure, and

-8

light trapping, we believe that a 12% thin film solar product is

achievable.
-0.5

0.0

0.5

1.0

Voltage (V)

1.5

2.0

2.5

1.0

28.6 mA/cm2

0.9

4.

Summary
We reviewed the issues and solutions of nc-Si:H solar cells.

0.8

We believe that proper control of the crystallinity along the film

0.7

thickness, reduction of impurity, and improvement in light trapping

0.6

are the major areas for improving nc-Si:H component cells. Using

0.5

a multi-junction solar cell structure is the path to achieve high

9.43

0.4

9.44

0.3

9.73

efficiency. United Solar has made significant progress to address

0.2

the issues and to improve thin film silicon solar cell efficiency. We

0.1

believe that thin film silicon solar cell efficiency will be improved

0.0
300

500

700

900

1100

further and it is a viable technology for large volume PV

Wavelength (nm)

Fig. 3. J-V characteristics and QE spectra of an a-Si:H/a-SiGe:H/nc-Si:H

manufacturing.

triple-junction solar cell.

Acknowledgments
The work is supported by the US DOE SAI under Sub-Contract

17

FC36-07 GO 17053, and by the US DOD through US Army under

16

Initial Cell Eff (%)

using

triple-junction cell structure made under high deposition rates.

-10

QE

respectively,

Sub-Contract No. W9132T-11-C-0007.

15

References

14

1) J. Yang, A. Banerjee, and S. Guha, Appl. Phys. Lett. 70, 2975 (1997).

13

2) J. Meier, R. Flckiger, H. Keppner, and A. Shah, Appl. Phys. Lett. 65,

12

860 (1994).

11
2002

2004

2006

2008

2010

2012

Year

Fig. 4. Initial active-area nc-Si:H based solar cell efficiency reported by

United Solar.

3) B. Yan, G. Yue, J. Yang, S. Guha, D. L. Williamson, D. Han, and C.-S.

Jiang, Appl. Phys. Lett. 85, 1955 (2004).
4) G. Yue, B. Yan, L. Sivec, Y. Zhou, J. Yang, and S. Guha, Sol. Energy
Mater. & Sol. Cells, submitted (2012).
5) B. Yan, G. Yue. L. Sivec, J. M. Ovens-Mawson, J. Yang, and S. Guha,

Large-area multi-junction solar cells

Because the total current in the a-Si:H/a-SiGe:H/nc-Si:H solar

Sol. Energy Mater. & Sol. Cells, submitted (2012).

cell is very high, the a-SiGe:H middle cell has to be thick enough

Lett. 99, 113512 (2011).

70

AM-FPD '12

6) B. Yan, G. Yue, L. Sivec, J. Yang, S. Guha, and C.-S. Jiang, Appl. Phys.

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

17

Increased Upconversion Response in a-Si:H Solar

Cells With Broad-Band Light
J. de Wild, T. F. Duindam, J. K. Rath, A. Meijerink, W. G. J. H. M. van Sark, and R. E. I. Schropp

AbstractHigh-bandgap solar cells such as a-Si:H with a

bandgap around 1.8-eV loose 50% of the incident solar energy
due to transmittance of near infrared photons. Part of these photons can be converted into visible photons with a suitable upconverter and in this way can increase the photo response of solar
cells in the near infrared. In this contribution, we demonstrate increased photo response of a-Si:H solar cells for sub bandgap light
under broad-band light excitation of the upconverter. Broad-band
excitation represents daily life and is more effective because simultaneously different wavelengths can be absorbed and converted.
Bifacial a-Si:H solar cells were made with and without Gd2 O2 S:
Er3 + , Yb3 + upconverter attached at the back of the cells. A set-up
was made to concentrate near infrared light from a solar simulator up to 25 times. IV curves were measured of upconverter and
reference solar cells. The upconverter solar cells showed a nonlinear and stronger increase in current with increasing concentration
than the reference solar cells.
Index TermsAmorphous silicon, infrared response, luminescence, photovoltaic cell.

I. INTRODUCTION
IGH-EFFICIENCY solar cells require absorption of photons of the full solar spectrum followed by effective generation and collection of charge carriers. The high bandgap of
amorphous silicon of 1.8 eV implies that the material is transparent for sub bandgap, near infrared (NIR) light, constituting a
high photon loss. Upconversion (UC) may enhance the response
of solar cells in the infrared [1]. UC is a luminescence process,
whereby two or more low-energy photons are converted to one
higher energy photon. When a layer containing UC species is
attached to the rear of a solar cell, the sub bandgap photons are
absorbed and higher energy photons are emitted; these can subsequently be directed to the solar cell using an optical reflector
behind the cell, where they can be absorbed in the active layer.
An increasing research effort is undertaken on crystalline silicon solar cells as well as thin-film silicon solar cells using

Manuscript received April 25, 2012; revised July 27, 2012; accepted August
9, 2012. Date of publication September 18, 2012; date of current version December 19, 2012.
J. de Wild, T. F. Duindam, J. K. Rath, and R. E. I. Schropp are
with the Faculty of Science, Debye Institute for Nanomaterials Science, NanophotonicsPhysics of Devices, Utrecht University, Utrecht
3508, The Netherlands (e-mail: j.dewild@uu.nl; T.F.Duindam@students.uu.nl;
j.k.rath@uu.nl; r.e.i.schropp@uu.nl).
A. Meijerink is with the Faculty of Science, Debye Institute for Nanomaterials Science, Condensed Matter and Interfaces, Utrecht University, Utrecht
3508, The Netherlands (e-mail: A.meijerink@uu.nl).
W. G. J. H. M. van Sark is with the Copernicus Institute, Utrecht 3584, The
Netherlands (e-mail: W.G.J.H.M.vanSark@uu.nl).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JPHOTOV.2012.2213799

upconverters based on lanthanide ions [2][5]; our work rather

concentrated on UC in a-Si:H solar cells under monochromatic
laser light [4]. In this paper, experiments are performed to investigate the response of a-Si:H solar cells upon broad-band
excitation of the upconverter.
II. THEORY
UC of lanthanide ions is extensively investigated since the
1960 s, and an overview is given by Auzel [6]. Lanthanides are
most commonly found in the trivalent ionized state and have an
4fn 5s2 5p6 electron configuration, where n varies from 0 to 14.
Interactions between the 4f electrons in the partly filled inner 4fn
shell give rise to a rich energy levels structure with absorption
and emission from UV to NIR. Because the 4f electrons are
shielded by the outer 5p6 and 5s2 shells the energy level structure
and optical properties are barely influenced by the surrounding
host lattice.
The upconverter material used is crucial for obtaining any
significant increase in solar energy conversion efficiency. We
applied a commercially available upconverter, Gd2 O2 S: Yb3+ ,
Er3+ , in which Yb3+ absorbs light around 980 nm and Er3+
emits in the visible spectrum (400700 nm). These absorption
and emission wavelengths are very suitable for use with a-Si:H
single-junction solar cells, as the absorption edge of a-Si:H is
between the wavelengths for absorption and emission and the
spectral response is very high in that emission range.
Different mechanisms are responsible for the UC luminescence. The dominant UC mechanism in Gd2 O2 S: Yb3+ , Er3+
is energy transfer upconversion (ETU). The Yb3+ ion has only
one excited state and is an ideal sensitizer for Er3+ because
of the relatively high oscillator strength of the 2 F7/2 2 F5/2
transition and the fact that Er3+ has a state with similar energy
(4 I11/2 ), which is populated by energy transfer from Yb3+ , see
Fig. 1. Population of the first excited state of Er3+ (4 I11/2 ) is
therefore directly proportional to the incoming light intensity.
When UC is the main route, energy transfer from the first excited state (4 I11/2 ) to the second excited (4 F7/2 ) state follows.
After some small energy-relaxation steps, emission is observed
from 4 S3/2 , 2 H11/2 (green) and 4 F9/2 (red) states. The 4 F9/2
can also be reached after energy transfer from the 4 I13/2 state.
These processes are illustrated in Fig. 1. Because two or more
photons are required for upconverted emission, a higher order
dependence of the incoming light intensity is expected
Nn Nn 1 Ns (Ns )n Pinn

(1)

where n is the number of photons needed to excite the upconverted state. Nn is the nth excited state in the Er3+ ion
and Ns the excited state of the sensitizer ion Yb3+ . When a

2156-3381/$31.00 2012 IEEE

18

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

20

1000

cm-1

4I
9/2
4I
11/2

2F
5/2

980
nm

9/2

4F
7/2
2H
11/2
4S
3/2
4F
9/2

15
10

2H

410
nm

2F
7/2

660
nm

1540
nm

Yb3+

525
550 nm
nm

Er3+

4I
13/2

4I
15/2

Fig. 1. UC in the (Yb3 + , Er3 + ) couple. The dashed lines represent energy
transfer, the full lines the radiative decay, and the curly lines indicate multiphonon relaxation processes. The main route is a two-step energy transfer that
leads to excitation to the 4 F7 / 2 state of the Er3 + ion. After relaxation from this,
state emission is observed from the 2 H1 1 / 2 , the 4 S3 / 2 level (green), and the
4F
9 / 2 level (red).

higher energy level saturates, other processes like nonradiative

relaxation to lower energy states occurs and as a consequence,
deviations from the expected power-law dependence (i.e., 2, 3)
are observed [7], [8].
The upconverted emission intensity is thus proportional to
the population of the higher excited state Nn . When an upconverter is applied to the back of a solar cell, the increased
photogenerated current is due to this emission and thus
UC
Pinn
ISC

(2)

UC
the photogenwhere Pin is the incoming light intensity, and ISC
erated short-circuit current in the solar cell. Thus, for current
increase due to UC we expect a quadratic power dependence
on the concentration factor because the upconverted emission
happens after absorption of two photons.

III. EXPERIMENTAL DETAILS

The upconverter was purchased from Phosphor Technology
Ltd. containing Yb3+ and Er3+ in Gd2 O2 S host, which was
confirmed by XRD experiments. From ICP-OES measurements,
the concentration was determined to be 5 1% Er3+ and 10
1% Yb3+ . Upconverted emission spectra were recorded with an
Edinburgh spectrofluorometer and the sample was excited with
a pulsed Opotek Opolette HE 355 II laser at 980 nm. The UC
powder mixture was applied to the rear of the solar cells by first
dissolving it in a solution of PMMA in chloroform after which
it was drop casted. A layer of 200300 m was obtained. As a
back reflector white foil [9] was used.
Standard p-i-n amorphous silicon solar cells were made by
plasma enhanced chemical vapor deposition (PECVD) with an
area of 0.16 cm2 and an intrinsic layer thickness of 350 nm. As
back contact, a 1.5-m ZnO:Al layer from a ZnO:Al2 O3 0.5%
target was applied.

Fig. 2. Transmission of light for flat and textured solar cells and the collection efficiency when light is coming though the n-layer, which is the case for
upconverted light. Most emission of the upconverted light is at 550 nm, where
the quantum efficiency for the textured and flat solar cells are the same.

To show the enhancement of solar cell performance due to

UC, current-voltage measurements were performed. A set-up
was made, which concentrated the light coming from a Wacom
solar simulator. This type of simulator operates using a halogen
and xenon lamp. The typical xenon lamp peaks are suppressed
with a dichroic mirror and most NIR light is coming from the
halogen lamp. With a convex glass lens, the broad-band light
could be concentrated up to 25 times. A long pass 900-nm filter
was added so that only subbandgap light reached to solar cell.
The spectral match of the produced spectrum to the AM1.5
standard spectrum is +14% for the 9001100-nm region [10].
IV. RESULTS
A. Solar Cells
Two types of p-i-n a-Si:H solar cells were made, one on
Asahi glass and one on flat ZnO:Al 0.5% superstrate. The efficiency obtained for the cells is 8% for textured and 5% for flat
solar cells, both without back reflector. Backside illumination,
through the n-layer yields an efficiency of 5% for textured solar
cells and 4% for flat solar cells. These low values for backside
illumination are mainly caused by current losses in the blue part,
due to bad carrier collection. Blue light is absorbed close to the
top of the cell in the n-layer and holes are not collected at the
p-side. The spectral response measured through the n-layer,
shows a quantum efficiency of 0.7 for both textured and flat
solar cells at 550 nm; the spectral response at 660 nm is lower,
i.e., 0.4 for textured cells and 0.15 for flat cells. The transmission for 9001040-nm was 4045% for the textured solar cells
and between 60% and 80% for the flat solar cell, see Fig. 2.
The thickness of the i-layer was chosen so that an interference
maximum occur at 950 nm, increasing the transmission at this
wavelength. Therefore, more light will be absorbed by the upconverter layer in the flat solar cell configuration. Back reflector
adds 11% to the current and, thus, efficiency. The absorption
in the PMMA layer was 1.5%, see Fig. 3.

DE WILD et al.: INCREASED UPCONVERSION RESPONSE IN a-Si:H SOLAR CELLS WITH BROAD-BAND LIGHT

Fig. 3. Upconverted emission and absorption spectra of the upconverter in the

PMMA layer. The emission spectrum is obtained when the upconverter shows
no saturation and only emission peaks from the 4 S3 / 2 , 2 H1 1 / 2 (510560 nm)
and 4 F9 / 2 (650680 nm) states are observed.

19

Fig. 4. Upconverted emission spectra under low- and high-excitation density.

For the low-excitation power, the green state was not yet saturated. The intensities may be compared. New peaks (italic) are assigned: 2 H9 / 2 -> 4 I1 5 / 2
transition at 410 nm, 4 I9 / 2 -> 4 I1 5 / 2 transition at 815 nm, and the intermediate
transition 2 H9 / 2 -> 4 I1 3 / 2 at 560 nm.

B. Upconverter
The absorption and emission spectra are shown in Fig. 3.
The absorption is highest around 950 nm. The upconverter was
excited with a xenon lamp at 950 10 nm and 980 10 nm.
The 4 F7/2 state at 2.52 eV is reached after two times energy
transfer from Yb to Er. The upconverter was already shown to
be very efficient at low light intensities. Saturation was measured
under light intensities of less than 1 W/cm2 [12]. Although the
absorption at 950 (1.31 eV) nm is higher, excitation at 980
(1.26 eV) nm leads to a two times higher upconverted emission
intensity. This may be attributed to the perfectly resonant energy
transfer step of 980 nm (1.26 eV), since the 4 F7/2 state is at
2.52 eV.
The upconverter was for further experiments excited at
980 nm with the Opolette laser. Because UC is a two-photon
process, the efficiency should be quadratically dependent on the
excitation power density. The intensity of the laser light was varied with neutral density filters. UC spectra were recorded in the
range of 400 to 850 nm under identical conditions with varying
excitation powers. Varying the intensity shows that for low light
intensities the red part is less than 6% of the total emission, see
Figs. 3 and 4. Only when the emission from the green emitting
states becomes saturated the red emission becomes more significant and even blue emission from 2 H9/2 state is measured, see
Fig. 4. By comparing the emission intensities, it becomes clear
that the emission intensity is not increasing quadratically with
the excitation power density. Instead emission from higher and
lower energy states are visible. The inset shows the integrated
emission peaks for the green and total emission. With this insert,
it is shown that at very high laser intensities the total emission
is saturated.
C. a-Si:H Solar Cells
The subbandgap response in the NIR due to the band
tails of a-Si:H solar cells is not neglectable [4]. To distin-

Fig. 5. Dark and light IV curves of textured solar cells with upconverter.
The solar cell was illuminated by long wavelength light (>900 nm), which was
concentrated 25 times.

guish between upconverter response and subbandgap response,

intensity-dependent measurements are performed on solar cells
with and without upconverter. IV curves are measured under
concentrated solar light, with wavelengths longer than 900 nm,
and Isc is extracted from these measurements, see Fig. 5. Intrinsic response of the band tails is linearly dependent on the light
intensity, while response due to upconverted light is expected to
be quadratically increasing with the concentration. Fig. 6 shows
the current measured of the different solar cells for the different
concentration factor of the subbandgap light. The slope of the
line fitted to the data yields the value n, as given by formula
1. As expected, the subbandgap response in linearly increasing
with the light intensity and values of n higher than 1 are measured for the UC solar cells. The value is close to 1 because a

20

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

Total current
Flat Upconverter, slope 1.4
Textured Upconverter, slope 1.2
Flat Reference, slope 0.9
Textured Reference, slope 1.0

0.25

-1

10

10

15

20

25

30

Current generation due to upconverter

Flat cell, slope 1.8
Textured cell, slope 1.5

Current (mA/cm )

Upconverter solar cell

Reference solar cell

0.30

Current (mA/cm )

10

0.35

Current (mA/cm )

0.20
0.15
0.10
0.05
0.00
900

-1

10

1100

1300

Wavelength longpass filter

-2

10

10

15

20

25

30

Fig. 7. Current under a concentration of 20 sun with different filters. If other

path would contribute to the current, i.e., direct excitation of Er3 + 2 I1 3 / 2 state
followed by excite state absorption (ESA) from 2 I1 3 / 2 to 2 F9 / 2 the current of
the upconverter should be higher with an 1100-nm filter. Though, there is only
a difference measured when Yb3 + is excited and for longer wavelengths the
current is the same.

Solar Concentration
Fig. 6. Current measured in the solar cells under illumination of subbandgap
light. In the upper figure, the total current of the reference and UC cells are
plotted as a function of the concentration factor; in the lower graph, the current
generated by the upconverter is shown. The slope for subbandgap response is 1
for flat and textured solar cells. The contribution of the upconverter increases
the slope slightly; when corrected for the subbandgap response, the slope is 1.5
for the textured and 1.8 for the flat solar cells.

large part of the total current is due to the subbandgap response

(see Fig. 6, upper graph). When the total current measured for
the upconverter solar cells is corrected for the subbandgap response, the current due to UC only shows a higher value for
n, see Fig. 6 lower graph. Values of n = 1.5 for textured solar
cells and n = 1.8 for flat solar cells are determined. The current
is not increasing quadratically with increasing concentration. It
is unlikely that the upconverter is saturated, because the power
density is far below the saturation level of 0.6 W/cm2 [12]. It is,
therefore, more likely that the deviations are due to decreasing
carrier collection efficiency with increasing concentration. This
effect would play a larger role in textured solar cells, because
they have a higher defect density than flat solar cells. This may
explain why for flat solar cells the value n is closer to 2, than
the value n for the textured solar cells.
D. Comparison of Current Generation Under Broad Light
With Monochromatic Light
The textured solar cells were also measured under laser
light excitation. Monochromatic laser light with wavelength
at 981 nm and power density of 0.2 W/cm2 gave a current
density of 0.14 mA/cm2 for the upconverter solar cells and
0.04 mA/cm2 for the reference solar cells. The contribution of
subbandgap absorption is much smaller, with monochromatic
laser light because only 981-nm light is contributing while for

the broad-band excitation all wavelengths longer than 900 nm

add to the current. More revealing is the current generated in
the upconverter solar cell. The current due to the upconverter is
comparable to the current measured under 20 sun: 0.1 mA/cm2
(see Fig. 6). This is remarkable in two ways. First, the results
are in contrast with previously reported experiments with broadband excitation of c-Si solar cells [14], where the current under
broad-band excitation was much smaller than that under laser
light excitation. However, in [14] another upconverter was applied (NaYF4 ) and different processes are undertaken in the upconverter, namely excited state absorption. In the upconverter
in this study (Gd2 O2 S), ETU is the main upconverter path and
the broad-band absorption of Yb3+ may increase the transfer
between Yb3+ and Er3+ .
Second, the power that is absorbed by Yb3+ is 3.44 mW/cm2
[13], which yields a broad-band power density of 70 mW/cm2
under a concentration of 20 sun. This is three times less than the
power density of the laser. A large difference here is that light
of the solar simulator extends to further than 1600 nm, thus
also the Er3+ : 2 I13/2 state is excited directly. Addition of other
paths that lead to upconverted light may contribute to the current. These paths may be non resonant excited state absorption
between the energy levels of Er3+ or three-photon absorption
around 1540 nm at Er3+ : 2 I13/2 (see Fig. 1). Direct excitation of
the Er3+ 2 I13/2 state followed by excited state absorption (ESA)
from 2 I13/2 to 2 F9/2 results in a visible photon around 650 nm,
while three-photon absorption around 1540 nm results in emission from the 2 F9/2 state too. Wavelengths required for these
transitions are around 1540 and 1200 nm, which are present
within the broad spectrum that the solar cell is illuminated with.
Contribution of these UC routes increases the emission and
thereby the current in the solar cells. To see if there is a contribution of excited state UC, the solar cells are illuminated with

DE WILD et al.: INCREASED UPCONVERSION RESPONSE IN a-Si:H SOLAR CELLS WITH BROAD-BAND LIGHT

900, 1100, and 1300-nm-long pass filters under a concentration

of 20 sun. The results are shown in Fig. 7. There is a distinct
difference in response between the upconverter and reference
solar cell with the 900-nm filter. However, the current measured
under the other filters is the same for the upconverter and reference solar cells. It, therefore, seems that the contribution of the
upconverter is due to absorption of Yb3+ only and its efficient
energy transfer to Er3+ . This transfer may be more efficient
under broad-band light and should be investigated further.
V. CONCLUSION
In this paper, we have shown that UC increases the response of
solar cells for broad-band excitation. Broad-band excitation may
be more effective because simultaneously different wavelengths
can be absorbed and converted. The response of the solar cells
due to upconverted emission of Gd2 O2 S: Er3+ , Yb3+ under
broad-band excitation was compared to that under monochromatic laser light. The response due to the broadband was the
same as that under laser light excitation, though the power density of the laser was three times higher. No other UC paths contribute to the current. It might still be possible that broad-band
excitation increases the transfer between Yb3+ and Er3+ . This
would lead to the higher response measured under broad-band
excitation compared with laser light excitation.
REFERENCES
[1] T. Trupke, M. A. Green, and P. Wurfel, Improving solar cell efficiencies
by up-conversion of sub-band-gap light, J. Appl. Phys., vol. 92, no. 7,
pp. 41174122, 2002.
[2] P. Gibart, F. Auzel, J. Guillaume, and J. K. Zahraman, Below bandgap IR response of substrate-free GaAs solar cells using two-photon upconversion, Jpn. J. Appl. Phys.Part 1: Regular Papers and Short Notes
and Review Papers, vol. 35, no. 8, pp. 44014402, 1996.
[3] A. Shalav, B. S. Richards, T. Trupke, K. W. Kramer, and H. U. Gudel,
Application of NaYF4: Er3 + up-converting phosphors for enhanced
near-infrared silicon solar cell response, Appl. Phys. Lett., vol. 86, no. 1,
pp. 013505-1013505-3, 2005.

21

[4] J. de Wild, J. K. Rath, A. Meijerink, W. G. J. H. M. Van Sark, and

R. E. I. Schropp, Enhanced near-infrared response of a-si:H solar cells
with -NaYF4:Yb3 + (18%), Er3 + (2%) upconversion phosphors, Sol.
Energy Mater. Sol. Cells, vol. 94, no. 12, pp. 23952398, 2010.
[5] G. Shan and G. P. Demopoulos, Near-infrared sunlight harvesting in dyesensitized solar cells via the insertion of an upconverter-TiO2 nanocomposite layer, Adv. Mater., vol. 22, no. 39, pp. 43734377, 2010.
[6] F. Auzel, Upconversion and anti-stokes processes with f and d ions in
solids, Chem. Rev., vol. 104, no. 1, pp. 139173, 2004.
[7] M. Pollnau, D. R. Gamelin, S. R. Luthi, H. U. Gudel, and M. P. Hehlen,
Power dependence of upconversion luminescence in lanthanide and
transition-metal-ion systems, Phys. Rev. BCondens. Matter Mater.
Phys., vol. 61, no. 5, pp. 33373346, 2000.
[8] J. F. Suyver, A. Aebischer, S. Garca-Revilla, P. Gerner, and H. U. Gudel,
Anomalous power dependence of sensitized upconversion luminescence, Phys. Rev. BCondens. Matter Mater. Phys., art. no. 125123,
2005.
[9] H. Knauss, E. L. Salabas, M. Fecioru, H. D. Goldbach, J. Hotzel,
S. Krull, O. Kluth, R. Kravets, P. A. Losio, J. Reinhardt, J. Sutterluti,
and T. Eisenhammer, Reduction of cell thickness for industrial micromorph tandem modules, in Proc. Eur. Photovoltaic Sol. Energy Conf.
Exhib., 2010, pp. 30643067.
[10] Manual solar simulator Model: WXS-140-SUPER, Voss Electronic
GmbH, Munchen, Germany, 2011.
[11] M. J. Keevers and M. A. Green, Extended infrared response of silicon
solar cells and the impurity photovoltaic effect, Sol. Energy Mater. Sol.
Cells, vol. 4142, pp. 195204, 1996.
[12] J. de Wild, A. Meijerink, J. K. Rath, W. G. J. H. M. Van Sark, and
R. E. I. Schropp, Improved upconversion in amorphpus silicon solar
cells, in Proc. 25th Eur. Photovoltaic Sol. Energy Conf., 2010, pp. 255
259.
[13] J. de Wild, A. Meijerink, J. K. Rath, W. G. J. H. M. Van Sark, and
R. E. I. Schropp, Upconverter solar cells: Materials and applications,
Energy Environ. Sci., vol. 4, no. 12, pp. 48354848, 2011.
[14] J. C. Goldschmidt, S. Fischer, P. Loper, K. W. Kramer, D. Biner, and
M. Hermle, Experimental analysis of upconversion with both coherent
monochromatic irradiation and broad spectrum illumination, Sol. Energy
Mater. Sol. Cells, vol. 95, no. 7, pp. 19601963, 2011.

Authors photographs and biographies not available at the time of publication.

(In)GaAsN Materials and Solar Cells for Super-High-Efficiency

Multijunction Solar Cells
Masafumi Yamaguchi, Boussairi Bouzazi, Hidetoshi Suzuki, Kazuma Ikeda, Nobuaki Kojima,
and Yoshio Ohshita
Toyota Technological Institute, Nagoya 468-8511, Japan
ABSTRACT

since FY2008 and EU-Japan Collaborative Research on CPV since

2011. In order to realize 40% and 50% efficiency, new approaches
for novel materials and structures are being studied. InGaAsN is
one

of

EV-Japan Project "NGCPV"

We have been studying concentrator multi-junction

solar cells under Japanese Innovative Photovoltaic R&D program

appropriate

materials

for

4-

or

5-junction

solar

cell

A New Generation of Conccutrntor Photovoltaic Cell .

Modules "ud ySleUIS

Aim of tile Project "NC PV"

... 10 accelerate the move /0 velY

_ [PV

high effiCiency and lower cost

configuration because this material can be lattice-matched to GaAs

and Ge substrates. However, present InGaAsN single-junction solar
cells have been inefficient because of low minority-carrier lifetime
due to N-related recombination centers and low carrier mobility
due

to

alloy

scattering

and

non-homogeneity

of

N.

Present

(In)GaAsN single-junction solar cells have been inefficient because

Co urt esy

of low minority-carrier lifetime due to N-related recombination

: D:iido

centers such as N-H-VGa, (N-N)As and so on, and low carrier

mobility due to alloy scattering and non-homogeneity of N. To
solve above problems we have been developing the CBE technique.
We have focused especially on the N incorporation in GaAsN thin
films grown by CBE as a function of growth temperature and
substrate orientation. We have also proposed a new flow-rate
modulation CBE

(FM-CBE)

method

in

order

to

increase

incorporation and to reduce C and H incorporation in films. By

adapting CBE technique to grow (In)GaAsN thin films, higher
mobility and longer minority-carrier lifetime compared to those
grown by the other growth methods have been achieved. According
to these electrical properties, more than 15% efficiency is expected
in CBE grown homo junction (In)GaAsN solar cell although only
2
7.2% efficiency (Jsc=15.1 mA/cm , Voc=0.662 V, FF=0.65)
GaAsN single-junction cells have been obtained.

To solve above

problems, We have characterized deep levels in grown GaAsN

films by DLTS and defect behaviors have been clarified.

iNDEX TERMS

and

M.Yam
A.
11.
__

"""

(NEOO
;;:::
::

Fra"nhofE[.
-CPV ""

BSQ,

,!I!..J!I.J.:.ft

_ .. , _

Fig. I. Outline of EU-Japan Collaborative Research on CPV.

Although
36.9%
efficiency
at
I-sun
AMI.5G
with
l
InGaP/GaAs/InGaAs 3-junction cell by Sharp . in this program, it is
necessary to develop a new material with bandgap energy of 1.04eV
in order to realize 4- or 5-junction cells with efficiencies of more
2
than 40 % or 50% . In addition, the new material should be lattice
matched to Ge and GaAs. InGaAsN is one of appropriate materials
for 4- or 5-junction solar cell configuration because this material can
be lattice-matched to GaAs and Ge. However, present InGaAsN
single-junction solar cells have been inefficient because of low
minority-carrier lifetime due to N-related recombination centers and
low carrier mobility due to alloy scattering and non-homogeneity of

III-V-N, CBE Growth, Defect Analysis,

N. To solve above problems we have been developing chemical

3
beam epitaxy (CBE) technique We have characterized deep levels

Minority-carrier Lifetime, Concentrator Solar Cells

in grown (In)GaAsN films by deep level transient spectroscopy

(DLTS).

I.INTRODUCTION

This paper presents our major results in the understanding of

III-V compound multi-junction solar cells have great potential

for

space

and

terrestrial

applications

because they

have

high

efficiency potential of more than 50% and superior radiation

resistance. Recently, more than 40% efficiency 3-junction cells were
reported by Fraunhofer ISE, Sharp, Spectrolab, Spire and Solar

majority and minority carrier traps in GaAsN grown by CBE and

their relationships with the poor electrical properties of the material
and high background doping level.
II. CBE GROWTH AND OPTO-ELECTRICAL PROPERTIES OF
GaAsN

Junction. Concentrator 4-junction or 5-junction solar cells have great

potential for realizing super high-efficiency of over 50%.
We have been studying concentrator multi-junction solar cells
under

Japanese

Innovative

Photovoltaic

R&D

program

since

FY2008 and EU-Japan Collaborative Research on CPV since 2011

as shown Fig. I.

In order to attain high minority-carrier lifetime and carrier

mobility in (In)GaAsN, we have been developing the CBE technique.
We have focused especially on the N incorporation in GaAsN thin
films grown by CBE as a function of growth temperature and
substrate orientation. We have also proposed a new flow-rate
modulation

CBE (FM-CBE) method in order to increase

46
incorporation and to reduce C and H incorporation in films

978-1-4673-0066-7/12/$26.00 2011

IEEE

000831

[NJ

'm

.....

"b

We have characterized deep levels in CBE-grown GaAsN films

0.85%, O.4Ilm/h

by deep level transient spectroscopy (DLTS) and FTIS (Fourier

transform infrared spectroscopy). The DLTS measurements of n-type

As-grown

b\ ni

and p-type GaAsN films were carried out from 2SK to 3S0K by

0.92

/3
.....

using GaAsN Schottky junctions, GaAs/GaAsN hetero-junctions.

3.1 Behavior and Origin of Hole Trap H2
A nitrogen-related hole trap acceptor like state H2

...

located

approximately O.IS eV above the valence band maximum (VBM) of

GaAsN has been observed for the first time and its nature has been

Improved!!

clarified in this study. The properties of H2 and its relationship with

ionized acceptors (NA) are investigated using DLTS and on the basis
of the temperature dependence of the junction capacitance. At room

100000
50000
21d(cm')

150000

temperature,

NA is found to show a linear dependence on N

concentration under N- and H-rich growth conditions. Furthermore,

a N-dependent sigmoid increase in junction capacitance is observed

Fig. 2. PL lifetime of CBE-grown GaAsN.

in a specific temperature range fTom 70 to 100 K, which is the same

By adapting CBE technique to grow (In)GaAsN thin films,

as in the case where H2 is recorded. Such behavior is explained by

higher mobility and longer minority-carrier lifetime compared to

the thermal ionization of H2, whose density affects in great part the

those grown by the other growth methods have been achieved as

46
shown in Fig.2 and Fig.3 . According to these electrical properties,

magnitude order of NA, essentially for an N concentration higher

than O.IS %.

30

more than IS% efficiency is expected in CBE grown homo junction

(In)GaAsN solar cell although only 7.2% efficiency (Jsc=IS.1
2
mA/cm , Voc=0.662 V, FF=0.6S) GaAsN single-junction cells with

AR

coating have been obtained. Figure 4 shows dark and light JV

curves of GaAsN solar cell with anti-reflection coating.

T300

&200

high step dens' ysub.

:c
100

-<111-<111.
4.

I\fBE

[1][2]
IVOCVD[1]

-<III

o
0.0

0.5

E20
(J
'"
015
.....

[110
<?

z'r.

-<III

CBE Ga&As-step

25

1.0

composition rIo]

5 '"

",-

()

1 -8
-,

-12

AM1.5G
Cell area: O.2686cnt
Jsc= 15.04 rnlVcnt

hole trap H2 (NHc2) obtained from DLTS measurements

...,.
,
15....

1
"
1:>'

"

1t:J.D. 9
C,LlOI
,

/II
1/1

1/1

,I ..{]
/

-16!:,.--x-..

!--.-'---;h,.--'--,
...
---;h;:
.. 0.2....0.4
...
....
.4 ...-'-- -0.2....
0.0
0.6
0.8
1.0
v [V]
Fig. 4. Dark and light JV curves of GaAsN solar cell with anti

'"

'"

'"

'"

Fig. S. Nitrogen concentration [N] dependence of the density of the

"

Voc= 0.662 V
FF = 0.650
Eff=7.18%

'"

'"

'"

'"

OL-------
0.1
0.2
0.3
0.4
0.5
[N] (%)

Fig. 3. Hole mobility of CBE-grown GaAsN.

o'------+-----------+-----

'"
","'.

'"

'"

'O"[],
"

10

1.40

to

1.45

1000fT(111<1

1.50

0
/

DJd

0/6

r-----

o 400'C
.'oc
420'C
o

5
[N.H] (x10"cm"l

.30C

10

Fig. 6. Relationship between the N and H concentration ([N-H]) and

the hole concentration ([p]) at each growth temperature (Inset:
Arrhenius plot of [p]/[N-H] at each growth temperature.)

reflection coating.
III. DLTS ANALYIS

AND FTIR ANALYSIS OF CBE-GROWN

GaAsN

Concerning its origin, H2 is strongly considered to act as N-H

related acceptor state and to be the main cause of high background
doping level in the film as shown in Fig. S. Due to the fact that the
local vibrational modes (LVMs) are very sensitive to the atomic

978-1-4673-0066-7/12/$26.00 2011

IEEE

000832

structure surrounding N and H, we have investigated the LVMs in p

GaAsN grown by CBE by using FTIR measurements. One peak at
!
around 3098 cmis originating from N-H bond. The hole
concentration has been found to be linearly increased with increasing
the N-H concentration as shown in Fig. 6. The slope increases with
increasing growth temperature (from 400 to 430C). It indicates that

"I""

the hole concentration in GaAsN films is determined by both the

03

number of N-H bond and an unknown defect, such as impurities,

><

vacancies, and interstitials. This defect concentration increased with

increasing growth temperature, suggesting that it is determined by
Arrhenius type reaction. Theoretically, N-H complex related to VGa,
with an energy level within 0.03 to 0.18 eV from the VBM of
8
GaAsN, was predicted to play the role of an acceptor and to be most
9
stable
using first-principles calculations. Therefore, N-H or N-H
V Ga

89
,

defects are thought to be as origin of the H2 center. Figure 7

200

Temperature (K)

shows possible origin of H2 defect center in GaAsN.

Fig. 8. Reduction of DC-DLTS maximum for the N-related electron

traps El under minority carrier injection spectra. The indication is
that EI is a recombination center

Fig. 7. Possible origin of H2 defect center in GaAsN

3.2 Behavior and Origin of Electron Trap EI
The nature of a well-known nitrogen-related electron trap (EI),
located approximately 0.33 eV from the conduction band minimum
(CBM) of GaAsN grown by CBE, has also been clarified in this
study. This level exhibits a high capture cross section compared with
that of native defects in GaAs. Its density increases significantly with
N concentration. In this study, EI has been newly confirmed to act
as a recombination center by comparing its activation energy with
that of the recombination current in the depletion region of the alloy
and by using double carrier (DC)-DLTS as shown in Fig. 8. The
non-radiative e-h recombination process through E1 is verified and
the capture cross sections for electrons and for holes are evaluated.
The physical parameters of E1 are summarized in a configuration
coordinate diagram, in which the energy state of E1 is described as a
function of lattice configuration. Furthermore, the origin of E1 was
investigated based on its sensitivity to arsenic and nitrogen source
flow rates and using the effect of proton implantation. E1 is thought
to be a defect formed from an N and an As in single arsenic site.
lO
Figure 9 shows a possible origin of this defect, (N-As)As Its effect
on the electrical properties of GaAsN can be characterized through
the

estimation

of

minority

carrier

lifetime

using

the

Shockley-Read-Hall (SRH) model for generation-recombination.

Such parameter has been estimated to be about 0.2ns which is in

Fig. 9. Possible origin of EI defect center in GaAsN

3.3. Behavior of Hole Trap HO
DC-DLTS was also used to characterize recombination centers in
p-type GaAsN grown by CBE. Figure 10 shows DLTS spectrum of
p-GaAsN. The DLTS peak height of a shallow hole trap HO, at 0.052
eV above the valence band edge of GaAsN, decreases by the
injection of both majority and minority carriers for various values of
second pulse voltage and duration. Although the trap is shallow, its
capture cross section is relatively large to capture electrons and holes.
The decrease in the number of traps is explained by the electron-hole
recombination process. The HO center has for the first time been
confirmed to be recombination center in p-type GaAsN. Due to its
energy level and on the basis of temperature dependence of capture
cross section, the recombination process is found to be radiative. As
shown in Fig. 11, three different peaks PI, P2, and P3 were observed
by photoluminescence (PL) measurements at low temperature. By
comparing the thermal activation energy of HO to that of the three

be due to free to acceptor (HO) recombination, more deep analYSIS IS

required to discuss this point.

accordance with theoretical and experimental results. For realizing

higher efficiency (In)GaAsN solar cells, it is very important to
reduce EI defect in (In)GaAsN.

978-1-4673-0066-7/12/$26.00 2011

IEEE

peaks, only P2 can be assimilated to HO. Although P2 is thoug t o

000833

2.0r------I1'---,

3)

A hole trap HO center (Ev+0.052eV) newly observed in p-GaAsN

has been found to also act as acceptor state and to act as a
radiative recombination center. The HO center has been found

1.5

to correlate with photoluminescence peak.

Figure 12 shows summary of our results for several traps in

S'

GaAsN observed by DLTS measurements.

C> 1.0

GaAS

'(ij
(/)

>

00.5

'"

W
>.

30

35

150

Temperature(K)

Q)
I:
W
C
o

300

Fig. 10. DLTS spectrum of p-type GaAsN sample, recorded in the

temperature range 20 -330 K, under a reverse bias VR
pulse voltage Vp
-0.5 V, an emission rate window erw
1 s.
and a filling pulse width tp
=

-3 V, a
24.4

S-I,

r-------------------...,
--20k
--30k
--40k
--50k

CBM

1.29

I.

GaAsN: [NJ 0.5"/(

Eg 1.29 eV
=

P rin ci pal cause of poor minority carriers'

life time Non-radiative recombination center

0.99'

IE1 ({NAs).,) Non-radiative recombination center

0.890.84-

:.;:::;
I1l
>
:;::::
CJ

GaAs

E3

0.59'

(Ni

0.48 .

EL5 type defect

E2 EL2 type

H5

defect (electron trap)

Double

state of EL2 (+1++)

(hole trap)

<t

III

II.

0.15 -

...

Principal cause of High background doping

H2

0.05
- HO
O

I-

Fig.

12.

(N.H.V..) Hole trap acceptorlike state

III. Radiative recombination center

VBM

--.......-

Summary of our results for several traps in GaAsN

observed by DLTS measurements.

However, further study is necessary to understand details of
origins and behavior of those defects and to control them.
ACKNOWLEDGMENTS
This work is partially supported by the New Energy and
Industrial

Development

Organization

under

the

Ministry

of

Economy, Trade and Industry, Japan. The authors wish to express

sincere thanks to Dr. T. Takamoto, Dr. K. Araki, Profs. Y. Nakano,

Fig.

II.

1.33
Photon Energy (eV)

Y. Okada, H. Amano, K. Kakimoto, Y. Kangawa, K. Nishioka, A.

Luque, Drs. A. Bett, G. Siefer for their fruitful collaboration.

PL spectra of p-type GaAsN at 20, 30, 40, and 50 K

REFERENCES
IV. SUMMARRY
1. A. Yoshida, T. Agui, K. Nakaido, K. Murasawa, H. Juso, K.

In order to realize 40% and 50% efficiency, new approaches for

Sasaki and T. Takamoto, Technical Digest of 2]" International

novel materials and structures are being studied. We have obtained

Photovoltaic Science and Engineering Conference, Fukuoka, Japan,

promising results:

I)

36.9% efficiency InGaP/GaAs/InGaAs 3-

junction cells, 2) high quality (In)GaAsN material with higher

Nov. 28 - Dec. 2 (2011).

2. S.R. Kurtz, D. Myers and J.M. Olson, Proceedings of the 26th

mobility by chemical beam epitaxy compared to those grown by the

iEEE Photovoltaic Specialists Conference,

other growth methods. However, current GaAsN single-junction

1997) p.875.

solar cells have been inefficient because of low minority-carrier

lifetime and low carrier mobility.
GaAsN films by DLTS. In conclusion, the following results have
been obtained in this study:
A hole trap H2 center (Ev+0.15eV) observed in p-GaAsN has
concentration. N-H or N-H-VGa is thought to be a possible
origin of the H2 center.

Yamaguchi, Thin Solid Films, 515, 5008 (2007).

Yamaguchi, Jpn. J Appl. Phys. 47, 2072 (2008).
6. K. Saito, K. Nishimura, H. Suzuki, Y. Ohshita and M. Yamaguchi,
Thin Solid Films 516, 3517 (2008).
7. B. Bouzazi, H. Suzuki, N. Kojima, Y. Ohshita and M. Yamaguchi,

A well-known electron trap E2 (Ec-0.33eV) center in n-GaAsN

and p-GaAsN has been for the first time confirmed to be non
radiative recombination center by using double-carrier pulse
DLTS. (N-As)As is thought to be a possible origin of the EI

center. For realizing higher efficiency (In)GaAsN solar cells, it

is very important to reduce EI defect in (In)GaAsN.

978-1-4673-0066-7/12/$26.00 2011

4. H. Suzuki, K. Nishimura, H.S. Lee, Y. Ohshita, N. Kojima and M.

5. K. Nishimura, H. Suzuki, K. Saito, Y. Ohshita, N. Kojima and M.

been found to also act as acceptor state and correlate with N-H

2)

3. M. Yamaguchi, T. Warabisako and H. Sugiura, J Crystal Growth

136, 29 (1994).

To solve above problems, We characterized deep levels in grown

I)

(IEEE, New York,

IEEE

Appl. Phys. Express, 3, 51002 (2010).

8. A. Fleck, B.l Robinson and D.A. Thompson, Appl. Phys. Lett., 78,
1694 (2001).
9. A. Janotti, S-H. Wei, S.B. Zhang, S. Kurtz and CG. Van de Walle,
Phys. Rev. B67, 161201 (2003).
10. S.B. Zhang and S.-H. Wei, Phys. Rev. Lett. 86, 1789 (2001).

000834

Industrial Development of Silicon Hetero-junction Back Contact

Solar Cells
'
'
'
'
'
2
Jin-Ho Choi , Jong-Chul Lee , Sang-Kyun Kim , Hun Park , Ki-Hyun Kim , and Won-Jae Lee
'

Green Energy Research Institute, Hyundai Heavy Industries, Co., Ltd., Yongin, 446-716, KOREA
2

Green Energy Business Division, Hyundai Heavy Industries, Co., Ltd., Eumseong, KOREA

Abstract
Si:H/c-Si

Silicon hetero-junction (SHJ) with the a

structure

is

one

of

the

candidates

for

high

efficiency solar cells. Despite SHJ solar cells show more than

22% efficiency, SHJ solar cells are limited by optical losses

on the front surface. To overcome these limitations, silicon

Figure 1 described the process flow of HBC solar cells.

Si wafers

(6

inch,

mono-crystalline,

n-type,

0.53.0

n'cm) were sequentially cleaned and textured with acidic

solutions. After wet processes, antireflection coating layer

hetero-junction back contact (HBC) solar cells are suggested.

was deposited on front side. The a-Si thin film layers

HBC solar cells combined the advantages of silicon hetero

were

junction (SHJ) and Interdigitated Back Contact (IBC) solar

deposition

cells. In this study, HBC solar cell structures were designed

by simulation methods and optimized fabrication processes
for

industrial

application.

Dimensions

of

back

contact

deposited

by

plasma-enhanced

(PECVD)

on

back

chemical

side.

vapor

Deposition

temperature, RF power density and H2/SiH4 gas flow rate

were varied. After p/i-a-Si deposition, p/i-a-Si layers in n

patterns were design by TCAD ATLAS software. Based on

type electrode region were removed and then nli-a-Si

simulation results, p-type and n-type back contact patterns

layers

were fabricated.

formed

Index Terms
hetero-junction

ATLAS simulation, back contact, silicon

were
on

deposited.

Finally,

both

and

p-

metal

n-type

contacts

were

electrodes.

J-V

characteristics of HBC samples were measured under AM

1.5G condition.

I. INTRODUCTION

Texturing process

SHJ solar cells have received great interest due to the

advantages of high open circuit voltages
processing temperature

(Voe),

low

200 C) and high efficiency up

Front AR coating

Back a-Si:H(i)

back a-Si:H(i)

Back a-Si:H{n)

to 23.7% by Sanyo [1]-[3]. However, the front layer

absorption losses stem from a-Si:H emitter, transparent
conducting oxide (TCO) layer and metal grids give rise to
reduced short circuit current density (Jse) [4].
By introducing IBC concepts to SHJ solar cells, high

Back electrode

Voe and Jsc can be obtained coincidentally. Because all

metal contacts are formed on back side, shading loss by
metal grids can be avoided and also trade-off between
antireflectioin (AR) coating and series resistance in the
front TCO layer can be circumvented. It is expected that
high efficiency over 24% can be obtained by combining

Fig. I.
Back contact pattern formation processes for heterojunction back contact (HBC) solar cells
In HBC solar cell structure, the ratio of p-type emitter
to n-type BSF region on the back side was important to

interdigitated

achieve high efficiency. The commercial TCAD ATLAS

In this study, HBC solar cell structures were designed

simulation was based on parameters of AFORS-HET

silicon

heterojunction

(a-Si:H/c-Si)

and

back contact concepts [4]-[7].

and

fabricated.

Especially

deposition

conditions

of

(Silvaco) was used to design back contact patterns. The

software for modeling HBC solar cells.

amorphous Si and low cost patterning processes for back

contacts were investigated for industrial application.

II. EXPERIMENTAL METHODS

III. RESULTS
To obtain high efficiency HBC solar cells, the intrinsic a
Si layer with good interface passivation properties on back
side should be deposited after texturing process. Low

978-1-4673-0066-7/12/$26.00 2011 IEEE

001023

recombination velocity at interface leads to high Voc. For

improving interface passivation properties, the deposition
parameters (deposition temperature, RF power density
and H2/SiH4 flow rate) of intrinsic a-Si layer were
investigated.
Figure 2 showed variations of measured implied Voc as
a function of deposition temperature, RF power density
and H2/SiH4 flow rate. As a result of this study, we
selected the optimized condition for textured n-type c-Si
wafers. Measured implied Voc over 710mV was achieved.
power

parameters adopted for HBC solar cell simulation were

shown in table I. Most of parameters were confIrmed by
silicon hetero-junction (SHJ) solar cell simulation as
shown in fIgure 3.
40

25
i
'120
lS

10

]2D'lm",.tI

Measured

... power B, 220'C

Voc(mv)

power B. 240'C

Fig. 3.
Comparison with measured
simulation result of SHJ solar cell.

power C. 240'C

I-V

data

and

2D

I k M ..,...----.....,

N type c-Si

20
Hz/SiH.

I
I

A. 220C

power B. 200'C

f-

+-

.-Si:H
(P)

flow rate (R)

Fig. 2.
Variations of implied Voc as a function of deposition
condition for the intrinsic a-Si layers.

To fabricate HBC solar cells, intrinsic and p-type

amorphous silicon layers should be deposited on back
side sequentially and then regions to deposit intrinsic and
n-type amorphous silicon should be perfectly removed.
Area ratio and pitch between p-type emitter and n-type
BSF are very important parameters to increase Jsc and FF
in HBC structure. ATLAS software was used to optimize
back contact pattern design for HBC solar cell. Some

Parameters

Fig. 4.
Schematic structure of HBC solar cell used for
ATLAS 2D simulation.

Figure 4 showed HBC structure using in ATLAS 2D

simulation with parameter values shown in table I. Area
ratio of p-type emitter to n-type BSF was changed from
0.6 to 0.8. The effIciency of HBC structure was calculated
as a function of emitter region. As shown in fIgure 5,
calculated effIciency was increased, when the pitch (unit

TABLE I INITIAL SIMULA nON PARAMETERS FOR HBC SOLAR CELL

c-Si(n)
a-Si(n)
a-Si(p)

a-Si(i)

Thickness (urn)

180

10

Dielectric constant

11.8

11.8

11.8

11.8

Band gap (eV)

1.12

1.74

1.74

1.72

Effective conduction band density (cm-3)

2.8xl0'9
1.04x102O

0
lx102
0
lxl02

Effective valence band density (cm-3)

Mobility (cm-2V-'S-')

0
lxl02
20
Ix10

0
lx102
0
lxl02

Hole mobility (cm-2V-'S-')

1040

412

Thermal velocity of electron (cmS-')

lxlO'7

lxlO'7

lxl0'7

lxlO'7

Thermal velocity of hole (cmS-')

lxl0'7

lxlO'7

lxl0'7

Doping concentration of acceptors (cm-3)

lxl0'7
0
lx102

Doping concentration of donors (cm-3)

lxl0'6

978-1-4673-0066-7/12/$26.00 2011 IEEE

20
Ix10

001024

cell of p-n pattern) was decreased up to 800f.!m. Moreover,

in the case of the same pitch condition, smaller gap
between emitter and BSF region was effective to increase
efficiency. In this simulation results, HBC solar cell with
23% efficiency was achieved when the structures were
optimized.

By combining these two methods, back contact area of

HBC solar cells was made successfully as shown in figure
7. I-V characteristic data of these samples will be shown
in the poster presentation.

24 r----,

23.5

..
'u

23

::
w

9ap lOGum. emftter ratio 0.6

22.5

-+-gap lOGum. emttter ratio 0.7

'-'-gap lOOum. emJtter ratio 0.'

22

==
9.
p
10
U

..
to
l
0
. a__________

aoo

900

1000

1100

1200

Fig. 7.
cell.

Back side electrode patterns of 5x5cm2 size HBC solar

Pitch (um)

Fig. 5.
Variations of efficiency for HBC solar cells as a
function of the pitch distance (The ranges of the area ratio of p
type emitter to n-type BSF are 0.60.8, gap 10Ilm & 100Ilm).
=

A few methods were used to make pin patterns on the

back side of HBC solar cells. In this study we used two
representative methods in patterning processes. Firstly,
etching pastes can be used to remove TCO or a-Si layers
as shown in figure 6 (left).
Etching pastes were screen-printing followed by heat
treatment. Patterning by etching pastes was very simple
and cost effective process. However, it is not easy to
remove residues of etching pastes after heat-treatment.
Secondly, we demonstrated patterning processes by
polymer resists as shown in figure 6 (right). As a
chemical resist, polymer resists were screen-printed and
dried. Because polymer resists were usually resistant to
acidic solution, pin patterns were easily made by
immersing samples in solution. After removing polymer
resists, etched area and non-etched area were simply
obtained.

IV. CONCLUSION
The deposition conditions of amorphous silicon were
optimized to fabricate HBC solar cells. Also the ATLAS
2D simulation was used to design back contact patterns.
Patterning processes were performed by using etching
paste and polymer resist, respectively. 5 x 5 cm2 size
HBC solar cells were manufactured successfully. I-V
characteristic data of HBC samples will be shown in the
poster presentation. Optimization of pattern design and
development of processes are ongoing project in our
group.
REFERENCES
[I] T. Mishima et aI., Solar Energy Mater. Sol. Cells 95 (2010)
18
[2] H. Fujiwara and M. Kondo, Appl. Phys. Lett. 90, 013503
(2007).
[3] T. H. Wang,E. Iwaniczko,M. R. Page,D. H. Levi,Y. Yan,
H. M. Branz, and Q. Wang, Thin Solid Films 501, 284
(2006).
[4] M. Lu, U. Das, S. Bowden,S. Hegedus, R. Birkmire, Prog.
Photovolt: Res. Appl. 19 326 (2011)
[5] N. Mingirulli, J. Haschke, R. Gogolin, R. Ferre, T.F.
Schulze, J. Dusterhoft, N. Harder, L. Korte, R. Brendel, B.
Rech,Phys. Status Solidi RRL 5,No. 4,159 (2011)
[6] M. Tucci,L. Serenelli,S.D. Iulliis,M. Izzi, G.D. Cesare,D.
Caputo,Phys. Status Solidi C 8,No. 3,932 (2011)
[7] M. Lu, S. Bowden, U. Das, R. Birkmire, Appl. Phys. Lett.
91,063507 (2007)

Fig. 6.
Patterning processes of amorphous silicon by using
etching paste (left) and polymer resist (right)

978-1-4673-0066-7/12/$26.00 2011 IEEE

001025

DEVELOPMENT OF HIGH EFFICIENCY MONO-CRYSTALLINE SILICON SOLAR CELLS: OPTIMIZATION OF

REAR LOCAL CONTACTS FORMATION ON DIELECTRICALLY PASSIVATED SURFACES
Kapila Wijekoon*, Hemant Mungekar, Michael Stewart, Prabhat Kumar, Jeff Franklin, Mukul Agrawal, Kalyan Rapolu,
Fei Yan, Yi Zheng, Andie Chan, Manoj Vellaikal, Xuesong Lu, Damanjot Kochhar, Lin Zhang,
David Tanner, Vikas Dabeer and Hari Ponnekanti
Applied Materials, 3535 Garrett Drive, Santa Clara, CA95054
kapila_wijekoon@amat.com
ABSTRACT
An integration process was developed for the fabrication
of rear passivated point contact solar cells achieving
19.36% conversion efficiency by using 156x156mm,
pseudo square, p-type single crystalline silicon wafers.
This is a significant improvement when compared to
unpassivated, full area aluminum back surface field solar
cells, which exhibit only 18.64% conversion efficiency on
the same wafer type. The rear surface was passivated
with a Al2O3 layer and a SiNX capping layer. The
thicknesses of individual films were optimized to obtain
maximum minority carrier lifetimes. The rear surface
contact pattern was created by laser ablation and the
contact geometry was optimized to obtain voids free
contact filling resulting in a uniform back surface field.
Internal quantum efficiency and reflectance measurement
show significant improvement in rear passivated cells in
the infrared wavelength region in comparison to reference
cells.
The rear surface internal reflectivity for the
passivated cell was 93% while that for the reference cell
was only about 73%. The rear surface recombination
velocity for the rear passivated cell was about 52 cm/s
while that for the reference cell was about 300 cm/s. The
efficiency gain in rear passivated cells over the reference
cells is mainly due to improved short circuit current and
open circuit voltage. However, rear passivated solar cells
show lower fill factors due to increased series resistance.

surface, resulting in significant improvements in solar cell

efficiency. In the PERC approach the rear surface is
passivated with an appropriate dielectric layer. The
presence of the dielectric layer not only reduces the
surface recombination but also improves the internal
reflectivity, hence reducing absorption losses at the rear
aluminum contact.
Various firing stable surface
passivation dielectric layers such as SiO2, SiCX, SiNX and
Al2O3 have been shown to be suitable for fabrication of
PERC type solar cells [1]. For the fabrication of metal
contacts, local areas of passivation layers are removed
before screen printing of aluminum paste. The localized
removal of the dielectric layer is accomplished by using
either laser ablation or chemical etching. In the case of
chemical removal, etching paste is screen printed on
passivation films. Several different parameters, such as
geometry of the contact opening, reactivity of the
aluminum paste and thermal budget, seem to substantially
affect the formation of a robust local metal contact.
In this work, a process flow for fabrication of PERC solar
cells yielding 19.36% cell efficiency was developed. A thin
Al2O3 surface passivation layer with a thicker SiNX capping
layer was used as the passivation film stack. Local
contact areas in the passivation layer were opened by a
laser ablation process. Following that both front and rear
conductors were screen printed and co-fired to form metal
contacts. A brief discussion of the development of the
PERC solar cells will be presented in this paper.

INTRODUCTION

EXPERIMENTAL

Most of the industrial solar cells manufactured today apply

a full area aluminum back surface field (Al-BSF) at the
rear surface of the cell to repel minority carriers. In this
process front side silver grid and rear side aluminum
conductor is screen printed and co-fired at elevated
temperatures to form ohmic contacts. Although the full AlBSF forms an ohmic contact, the rear surface passivation
is moderate with effective surface recombination velocities
as high as 600 cm/s on p-type silicon substrates [1, 2].
Also, with full Al-BSF only about 70% of infrared light
reaching the aluminum rear contact is reflected back into
the bulk silicon.
In order to further improve the cell
efficiency, passivation and optical quality of the rear
surface needs to be improved. This would reduce the
surface recombination velocity, preferably below 100 cm/s,
and increase the back surface reflection in the range of
90%. Solar cells based on the passivated emitter and rear
cell (PERC) concept can achieve remarkably improved
surface passivation and optical reflectivity at the rear

P-type Czochralski wafers (156 x 156mm pseudo square)

having a bulk resistivity of 1-3 cm were used for the
experiments described in this study. All the wafers were
subjected to single side texturization process by using an
IPA free texture etching process [3-5]. After POCl3
diffusion the rear emitter was removed and a front ARC
layer (SiNX) was deposited. Front ARC layers were
deposited in PECVD thin film deposition tool. The wafers
were then split in to two groups. The group with no rear
passivation layers was used for the fabrication of
reference cells (RC). The second group was used for
fabrication of PERC cells. On wafers in this group rear
passivation films were deposited and local contacts were
opened with laser ablation. The rear passivation film stack
consists of a thin Al2O3 layer and a thicker SiNX capping
layer and was deposited using a single pass PECVD
deposition system. Contact opening was performed using
a high throughput laser ablation tool. Both the wafer
groups were then combined and randomized. The same

978-1-4673-0066-7/12/$26.00 2011 IEEE

001158

front conductor silver paste was screen printed on all

wafers. Then wafers were again separated into two
groups and different aluminum pastes were used for each
wafer group. A specially formulated aluminum paste with
a modified firing process was used on the PERC cells.
For the RC wafer group the standard baseline aluminum
paste and firing profile was used. However in both cases
the same peak firing temperature was maintained in order
to retain the integrity of the front metal contact.
RESULTS AND DISCUSSION
Figure 1 shows a SEM picture of a local via formed after
ablation of the dielectric film. It is evident that the via
formation process results in clean opened features with no
debris in the opened area. EDX spectra showed only Si
was in the opened contact area. In addition, the opened
contact area was inspected with surface Auger microcopy
(Figure 2). The center, edge and outside of the opened
contact were inspected and only silicon was found to be
present in the opened area. There were no indications for
the presence of either SiNx or Al2O3 residue inside the
opened via, indicating a very clean via formation. Outside
of the contact area only SiNX (peak at 86eV) was found as
expected. The inside and edge of the contact area mainly
elemental Si (Peak at 95eV) was found. Traces of native
silicon oxide also detected in these areas (peak at 83eV).

During these studies it was found that the quality of the

rear dielectric layer, physical geometry of the contact
pattern, aluminum paste composition and firing conditions
all played a significant role in the fabrication of a robust
local contact. For example with the same paste if the
firing conditions and geometry of the contacts are not
optimized severe void formation can occur. In some
cases voids as deep as 30m were observed (Figure 3)
after etching of Aluminum in HCl. This behavior was
almost identical for various aluminum pastes evaluated for
fabricating rear local contacts. Nevertheless, with the
correct contact geometry and optimized firing conditions a
complete surface alloying was achieved in contact areas.
However, it should be noted that paste formulation plays a
significant role in the surface alloying process and BSF
formation. During contact alloying silicon dissolution will
take place within the opened contact area [6-10]. Some of
the silicon will be diffused away from the contact site. The
alloying process will complete only if enough silicon is
available for the alloying process. If there is not sufficient
amount of silicon available for alloying process an
unpassivated contacts will be formed (Figure 4).

Figure 3. Void formed due to non-optimized firing

condition reveals incomplete BSF alloying. Depth of the
void is 30m. The aluminum in the via-contact was etched
away with HCl
Figure 1. Top-down SEM picture and EDX spectrum of
local via formed after rear AlOX and SiNX passivation
layers are removed. Only elemental silicon was found in
the opened contact area.

Figure 4. SEM image of a local contact with incomplete

alloying process.
Figure 2. Elemental silicon map and Auger spectra of
contact area.

978-1-4673-0066-7/12/$26.00 2011 IEEE

When an aluminum paste printed wafer is subjected to a

thermal firing process, silicon atoms diffuse into the
aluminum matrix during the heating cycle. The aluminum

001159

then melt at the elevated temperatures (the melting point

o
of aluminum is ~660 C) and dissolution of silicon into
melted aluminum begins to takes place, forming an Al-Si
liquid phase. Some aluminum atoms are transported out of
the particles to the wafer surface. At the peak firing
temperature the maximum amount of silicon is dissolved
and the entire wafer surface is covered with liquid Al-Si [610]. A higher peak firing temperature increases the
amount of silicon dissolved in aluminum, causing silicon
atoms to diffuse deeper into the aluminum layer. During
the cooling cycle, silicon is ejected from the liquid Al-Si
phase and re-crystallizes epitaxially at the wafer interface.
The silicon is doped with aluminum at its solubility limit
+
creating a p (BSF) layer. During Further cooling eutectic
o
temperature (~577 C) is reached and the remaining Al=Si
liquid phase solidifies, completing the alloying process.
In the case of local contacts only the opened contact
areas can provide silicon for the Al-Si liquid phase
formation. Silicon concentration in liquid aluminum is high
above the contact points while other areas contain melted
aluminum. This concentration gradient will initiate lateral
diffusion of silicon away from contact site. The lateral
diffusion of silicon away from the contact will deplete the
silicon concentration that can contribute to BSF formation
resulting in very thin Al-BSF layers. However, for a given
contact geometry, these difficulties can be overcome with
modifications to paste composition and modifications to
firing conditions. Optical micrograph and cross section
SEM shown in Figure 5 and Figure 6 show the voids free
contact alloying with the formation of a uniform Al-BSF

Alloyed contacts were further characterized with Acoustic

Micro Imaging (AMI) technology (Figure 7). This method
detects the presence of buried voids in the alloyed
contacts. A detailed analysis of AMI shown in Figure 7
indicates more than 80% of the alloyed contacts are
completely filled without any voids.

Figure 5. Optical microscope image of completely filled

contacts fabricated with specially formulated aluminum
paste and modified firing conditions. The sample was
cleaved and polished before taking the image.

978-1-4673-0066-7/12/$26.00 2011 IEEE

Figure 6. An image of cross sectional SEM of a fully

alloyed contact. The BSF is clearly visible. The sample
was polished and then etched in a mixture of acids
(HF/HNO3/CH3COOH) before taking the SEM image.

Figure 7. Acoustic Micro Image of rear local contacts.

More than 80% of contacts are completely alloyed without
voids.
The champion PERC cell fabricated in this study resulted
in conversion efficiency of 19.36% compared to the
champion RC of 18.64%. Clearly the gain in the PERC
cell is due to improved JSC and VOC values. Figure 8 and
Figure 9 show the analysis of IV characteristics and
conversion efficiencies of the PERC and RC. Each wafer
group contains 25 cells which were fabricated as
described previously. It should be noted that during rear
emitter removal step back side of the wafers get polished
and texturization from the back side is removed.
Therefore, both types of cells (PERC and RC) have the
identical rear side reflections after parasitic emitter
removal. The average gain in conversion efficiency in
PERC cells are about 0.7% as shown in the Table 1. The
2
average JSC gain in PERC cells is about 1.2 mA/cm
compared to the reference cells. Also, the PERC cells
show significantly improved VOC over the RC. However,
the PERC cells show relatively lower fill factors due to
increased series resistance [1].

001160

refection measurements is shown in Table 2. The

improved light confinement in the PERC cell is also
contributing for improvement in JSC in rear passivated cells.
The RC with full Al-BSF absorbs more light compared to
the PERC cells which have dielectrically passivated rear
surface.

Figure 8. Comparison of the PERC and the reference cell

IV curves.

Cell
Reference
PERC

JSC
(mA/cm2)

VOC
(mV)

FF
(%)

Eta
(%)

36.7
37.9

639
646

79.4
79.2

18.64
19.36

Table 1. Comparison of average values of parametric for

two cell lots. Each lot contains 25 cells.

Figure 10. Internal quantum efficiency and reflectance of

the PERC cell (blue) and the reference cell (red).

Figure 11. The results of Laser Beam Induced Current

(LBIC) mapping of the IQE of the PERC cell at 973nm.

Figure 9. Cell conversion efficiency for the five PERC and

Reference 9RC) cell lots demonstrating 0.6 to 0.7% gain
in cell efficiency with PERC structure over conventional
We further analyzed and compared the PERC cells and
RC by measuring their reflectance and internal quantum
efficiencies (IQE) shown in Figure 10. In the short
wavelength region both IQE and reflectance were almost
identical for the two types of cells. In the infrared region
the PERC cell shows significant improvement in IQE and
rear surface internal reflectance due to dielectric
passivation. The IQE of the PERC cell at infrared region
was measured with the LBIC (Laser Beam Induced
Current) measurement (Figure 11). The measured IQE
results show efficiency higher than 95% with very good
uniformity. The gain in JSC in passivated cells can be
attributed to reduced rear surface recombination. Internal
reflectivity of the front and rear surfaces as well as rear
surface recombination velocity extracted from IQE and

978-1-4673-0066-7/12/$26.00 2011 IEEE

Reflectivity
Front (%)

Reflectivity
Rear (%)

BSRV
(cm/s)

Reference

93

73

300

PERC

93

93

53

Cell

Table 2. Comparison of the rear surface recombination

velocity (BSRV) and internal reflection of the PERC cell
with the RC.

We further analyzed the defects in the PERC RC with

Photoluminescence (PL), Electroluminescence (EL) and
Dark Lock-in-Thermography (DLIT) measurements. The
results of those measurements are shown in Figures 1214.
As seen in PL measurements (Figure 12) no
significant defects are observed in either cell types. Also,
the EL measurements (Figure 13) show that there are no
micro cracks in these cells and the PERC cells are forming
good local contacts.
Furthermore, the DLIT

001161

measurements (Figure 14) show that there are no major

shunt regions in both types of cells. The minor shunting
paths seen at the edge of both cells may be due to
remaining parasitic emitter in these cells.

voltage. The PERC cells have lower fill factor due to

relatively higher series resistance of the cell.
ACKNOWLEDGEMENT
Authors thank Steve Johnston at NREL for the help in PL,
EL and DLIT imaging measurements.
REFERENCES
[1]. S. Gatz, T. Dullweber and R. Brendel: Evaluation of
Serious Resistance Losses in Screen-Printed Solar Cells
with Local Rear Contacts IEEE Journal of Photovoltaics,
1, 2011, pp. 37-42.

Figure 12. Photo-luminescence measurements of PERC

cell (right) and reference cell (left). Mark at the edge of
the PERC cell image is from the handling tweezers.

[2]. T. Lauermann, T. Luder, S. Scholz, B. Raabe, G. Hahn

and B. Terheiden: Enabling Dielectric Rear Side
Passivation for Industrial Mass Production by Developing
Lean Printing-Based Solar Cell Processes, IEEE PVSC,
2010, pp. 28-33.
[3]. K. Wijekoon, T. Weidman, S. Paak and K. Mac
Williams: Production Ready Novel Texture Etching
Process for Fabrication of Single Crystalline Silicon Solar
Cells IEEE PVSC, 2010, pp. 3635-3641.
[4]. K. Wijekoon, P. Kumar, D. Tanner, H. Ponnekanti and
C. Gay: Direct Texturization of as Sawed monoCrystalline
Silicon Solar Wafers: Solar Cell Efficiency as a Function of
Total Silicon Removal IEEE PVSC, 2011, pp.2216-2220.

Figure 13. Electro-luminescence measurements of PERC

cell (right) and reference cell (left). Mark at the edge of
the PERC cell image is from the handling tweezers.

[5]. K. Wijekoon, P. Kumar, D. Tanner, H. Ponnekanti and

C. Gay: Effect of Additive and Etchant Concentration in
Surface Morphology of Mono-Crystalline Silicon Solar cells
Textured with non-Alcoholic Chemical Formulations
EU PVSEC, 2011 pp. 2015-2019.
[6]. E. Urrejola, K. Peter, H. Plagwitz and G. Schubert:
Silicon Diffusion in Aluminum for Rear Passivated Solar
Cells APPL. Phys. Lett. 98, 2011, pp.153508
[7]. E. Urrejola, K. Peter, H. Plagwitz, G. Schubert: Al-Si
Alloy Formation in Narrow p-type Si Contact Areas for
Rear Passivated Solar Cells J. Appl. Phys 107 (2010)
124516

Figure 14. DLIT measurements of PERC cell (right) and

reference cell (left).
SUMMARY
A process flow for manufacturing PERC solar cells
yielding 19.36% conversion efficiencies have been
developed for industrial applications.
The average
efficiency gain in PERC cells over the conventional solar
cell is 0.7%. The efficiency gain in PERC cells are mainly
due to improved short circuit current and open circuit

978-1-4673-0066-7/12/$26.00 2011 IEEE

[8]. T. Lauermann, A. Zuschlag, S. Scholz, G. Hahn and B.

Terheiden: The Influence of Contact Geometry and Sub
Contact Passivation on the Performance of Screen Printed
th
Al2O3 Passivated Solar Cells 26 EU PVSEC, 2011 pp.
1137-1143.
[9]. M. Rauer, C. Schmiga, K. Ruhle, R. Woehl, M. Hermle,
and S. Glunz: Investigation of Aluminum-Alloyed Local
Contacts for Rear Surface-Passivated Silicon Solar Cells,
IEEE Journal of Photovoltaics, 2011. pp. 22-28.
+

[10]. J. Krause, R. Woehl, D. Biro, Analysis of local Al-P

Layers for Solar Cells Processed by Small Screen-Printed
th
Structures 25 EU PVSEC, 2010 pp.1899-1904.

001162

Towards Light-Trapping Free Amorphous Si Only Multi-Junction

Solar Cells
i
2
3
Seung Jae Saik , Koeng Su Lim , and Jeong Chul Lee
I
Department of Electrical, Electronic, and Control Engineering, Hankyong National University,
Anseong-si, Gyeonggi-do, 456-749 Korea (e-mail: sjbaik@hknu.ac.kr)

2
0epartment of Electrical Engineering, Korea Advanced Institute of Science and Technology,
Yuseong-gu, Oaejeon, 305-701 Korea

KIER-UNIST Advanced Center for Energy, Korea Institute of Energy Research, Yuseong-gu,
Daejeon, 305-343 Korea

Abstract
enables

We propose a thin film solar cell structure that

ultrathin

film

absorber

application

and

single

material multi-junction structure. This proposal is targeting

amorphous Si only multi-junction solar cells on a patterned
transparent

electrode,

which

might

perform

high

light

harvesting efficiency and high output voltage at the same

time. In addition, photon management of this proposed
structure does not rely on traditional light trapping scheme
but rather on lateral collection scheme with 'sidewall solar
cells'. To implement the proposed solar cells, a suitable
process sequence is suggested, where a challenging node
separation process between sidewall solar cells is included.
In this work, we present a-Si solar cell on patterned TeO
without

node

separation.

We

found

the

signature

of

absorption path elongation by sidewall solar cells, which is

one of the essential principles of our proposed solar cell. This
proposal

would

provide

promising

route

to

high

performance and low cost a-Si only photovoltaics.

Index Terms
cells,

photovoltaic

amorphous silicon, mUlti-junction solar

cells,

photon

management,

Staebler

Wronski effect, lateral collection.

I. INTRODUCTION
In thin film Si solar cells, it is widely accepted that an
innovative light trapping technology is crucial for further
efficiency improvement [1]. The importance of light
trapping in thin film Si solar cells originates from
inherent light induced degradation property of amorphous
Si (a-Si) [2], because we cannot increase stabilized
efficiency simply by increasing the absorber layer
thickness. In addition, micromorph double junction
scheme [3] is regarded as a mandatory technology for a
large scale commercialization because of low power
conversion efficiency. However, low throughput of
microcrystalline Si deposition process [4] still limits cost
reduction opportunities.

978-1-4673-0066-7/12/$26.00 2011 IEEE

A possible strategic implementation to overcome above

issues is presented in Fig. I(e). This structure possesses
following features: (1) Thinner layer of a-Si can be used

without the loss of light absorption capability, because

perpendicular collection scheme [5] is used. (2) Multi
junction can be realized using single material with equal
thickness. Therefore, in this scheme, a-Si/a-Si double
junction produces doubled open circuit voltage of single
junction a-Si with automatically current-matched.
Moreover, photocurrent can be increased by simply
increasing the height of the front electrode patterns (H)
without loss of open circuit voltage (Voc). To fabricate
this structure, following fabrication sequence is possible:
(1) patterned TeO preparation (2) coating of bottom
isolation dielectric (3) deposition of solar cell layers with
isotropic etch for node separation (4) coating of top
isolation dielectric (5) back Teo and electrode deposition.
These are graphically illustrated in Fig. I (a) - (e).
Processing steps (2) and (4) are not commonly used in
photovoltaic industry, and they seem challenging for mass
production. However, solution processing seems to be a
promising method [6], as it was successful In
semiconductor manufacturing.
Prior to a full demonstration our original proposal, in
this work, we have performed thin film solar cell
fabrication on patterned Teo without node separation
process. We have focused on the conformality of PEeVD
a-Si deposition, characteristics of sidewall-formed solar
cells, and performance limitation of the structure without
node separation [7].

II. EXPERIMENTAL

001182

Patterned Teo fabrication

Bottom isolation dielectric

FrontTCO

FrontTCO

(a)

(b)

Deposition of solar cell

Illvers & Node separation

TABLE I
VARIATION OF STEP COVERAGE WITH
DIFFERENT PROCESS CONDITIONS IN PECVD
DEPOSITED A-Si.
PECVD Condition

Sidewall Coverage

Normal Condition

29%

Higher Pressure

38%

Higher Flow

34%

Top isolation dielectric

fabricated using photo-lithographic technique, and solar

cell was deposited by plasma-enhanced chemical vapor
deposition (PECVD). gTCO is fabricated as follows:
insulating SiN is deposited on Indium doped tin oxide
(ITO) thin film, and SiN is patterned by photo
lithography followed by dry etching, and then, AI doped
ZnO is deposited on the surface of patterned SiN. The
image of the patterned substrate, and transmittance
characteristics are shown in Fig. 2. For the solar cell layer,
p type amorphous SiC (p-a-SiC), intrinsic amorphous Si
(i-a-Si), and n type amorphous Si (n-a-Si) thin films are

Front Teo

Front TeO

(c)

(d)

subsequently deposited on gTCO.

Back Teo and Metal

Thin film deposition by PECVD is not conformal on a

three dimensionally patterned substrate, and therefore, the
thickness of p-a-SiC was optimized to maximize open
circuit voltage (Voc) and fill factors (FF).

Back TCO

4-....

4- ....
0

4-....
o

nip

nip

pin

III. RESULTS AND DISCUSSION

To fabricate high performance solar cells on patterned
substrates, film coverage on sidewall surfaces should be
as thick as that on top surfaces. In PECVD, deposition
characteristics are not conformal, that is, step coverage is

4- ....
0

less than 100%. Table 1 summarizes the variation of step

coverage with different process conditions of PECVD. In
our experiment, 40% step coverage was the best value we

pin

(b)
Front TCO

(e)
Fig. 1.

(a)

(e)

Fabrication

sequence

of

proposed

200

structure. (e) shows the final structure proposed in this

400

600

800

1000

Wavelength (nm)

1200

work for the high efficiency and low cost implementation

Fig. 2 (a) Scanning electron microscopy image of patterned

of thin film Si solar cells. Yellow areas represent insulating

substrate. (b) Total and diffuse transmittance of as-prepared

layer, and the directions of electron-hole separation (lateral

gTCO.

collection) are indicated by arrows.

Patterned TCO (gTCO: grating patterned TCO) was

978-1-4673-0066-7/12/$26.00 2011 IEEE

001183

have obtained. This may not be of critical importance in

the structure with node separation illustrated in Fig. 1, but

12 ---,-----,--.

in our experiment without node separation process, we

have chosen the condition showing the best step coverage.
The image of a fabricated solar cell is shown in Fig. 3,
where sidewall and bottom TCO surface of the vertical
image shows thinner a-Si due to non-ideal step coverage
characteristics.
To demonstrate the effect of sidewall solar cells, the

N
E

10

(a)

u
(/)
--,

short circuit current density (Jsc) is plotted with

increasing i layer thickness as shown in Fig. 4(a). The Jsc
of the solar cell on gTCO exceeds the Jsc of the solar cell
on a planar substrate. This Jsc gain is due to the solar
cells on sidewalls. The areal ratio of sidewall vertical
solar cell is around 13%, which is similar to the
photocurrent improvement. Sidewall solar cell also
contributes to the long wavelength light absorption as
indicated in Fig. 4(b).
The impact of light scattering may also contribute to the
photocurrent improvement as expected from the finite
diffuse transmittance shown in Fig. 2 (b). Further analysis
to separate the effect of sidewall solar cells and light
scattering is under consideration.
The output characteristics of solar cells on gTCO show
smaller Voc than those on planar ZnO (Fig. 5). This is
due to the non-uniform coverage of p-a-SiC due to
deposition characteristics of PECVD. By increasing the
thickness of p-a-SiC, Voc can be increased but

100

planar TCO
gTCO

200

300

Average i layer thickness (nm)

>.
(,)

iE

0.8

(b )..::.. ..
,

E 04

\WI

400

Sidewall
solar cell

\:;:.-+

___ gTCO

(average ilh=270n ....

--- Planar TCO (ith = 360nm)

0.0
400

600

....

..::.

Wavelength (nm)

800

Fig. 4 (a) Short circuit current variation versus average i layer

thickness (b) External quantum efficiency of solar cell on
gTCO and planar ZnO.

accompanying increase in series resistance (see the diode

on-current shown in Fig.5(b)) degrades solar cell
performance. Therefore, conformal deposition of thin
films is critical in performance enhancement of solar cells
on patterned substrates. In a separate experiment, we have
found that device performance on patterned substrate is
improved with photochemical vapor deposition method,
which exhibited improved step coverage characteristics.
In addition, the Voc values with thicker p-a-SiC are still
smaller than that of solar cells on planar ZnO. The
relatively higher diode leakage shown in Fig. 5(b) is the
signature of this Voc degradation. This is attributed to the
effect of 'corner cell' indicated in Fig. 6.
In this experiment, the a-Si solar cell on gTCO can be
modeled with parallel connection of three solar cells that

Fig. 3 Vertical image of a-Si solar cell on gTCO. Target

thickness of a-Si is around 360nm (top thickness). Sidewall and
bottom thickness is around 40% of the top thickness. The
refractive index of SiN is around 2, which is similar to that of
ZnO or ITO.

978-1-4673-0066-7/12/$26.00 2011 IEEE

have different operation mechanisms. Top cell indicated

in Fig. 6 has the same operation mechanism with planar
cell, and sidewall cell operates like radial junction solar
cell [5]. In addition, in corner cell, electrons or holes
crowding occurs as indicated in Fig. 6. This current
crowding accompanies field concentration at each corner,
and thus, resulting in the enhancement of carrier
recombination. This is an intrinsic drawback of this

001184

structure, which may not be avoided even with conformal

n

thin film deposition.

Our experiment demonstrates the effect of sidewall
solar cells, which motivates further development of our

Top cell

Sidewall cell

proposed solar cell structure with node separation process.

In node-separated structure, conformal deposition is no

Corner cell

Teo

more necessary because solar cell layers on top and

bottom areas are removed by anisotropic etch after
deposition processes. In addition, the effect of corner
solar cells may not be significant in this node-separated
structure.
Further works on the electrical and optical modeling of
the proposed device as well as the process development
for node-separated structure formation are under
consideration.
IV. CONCLUSION
We have proposed thin film solar cell structure that
enables ultrathin film deposition for absorber layer and
single material multi-junction structure. This structure has
particular importance in a-Si solar cells, because light

12 .---------,--,_--__.

NE

'iii
c
(])
"0

::J

1a

r---=-

(a)

As a result of this prototype fabrication, we have

found that we have found the signature of absorption path
elongation by 'sidewall solar cells', which is an essential
part of our proposed solar cell principle. In addition, even
though this prototype device has processing difficulties
and intrinsic drawbacks, these would not be expected in
the node-separated device. Further research and
development of the proposed structure would provide a

ACKNOWLEDGEMENT

4
2

- gTC O (average ith=270nm)

- - - planar TCO (ith=360nm)

OL-________L-________
0.0
0.4
0.8

O -------.--.--.
1nm(gTCO)
- 2nm(gTCO)
- 3nm(gTCO)
- 1nm(planar TCO)

This work was partly supported by the next generation

phase change memory project funded by the Ministry of
Knowledge Economy of Korean government.
REFERENCES
[1] H. A. Atwater and A. Polman, "Plasmonics for improved

photovoltaic devices" , Nat. Mat., vol. 9,pp. 205-213,2010

induced degradation can be effectively suppressed by

using ultrathin a-Si layer.
To gain a further insight and motivation, we have
performed a-Si solar cell fabrication on patterned TCO
without node separation.

Voltage (V)

-5

Fig. 6 Modeled representation of solar cell on gTCO in this

experiment.

novel route to a high performance and low cost a-Si only

photovoltaics.

hv

[2] M. Stutzmann, W. B. Jackson, and C. C. Tsai, "Light

induced metastable defects in hydrogenated amorphous silicon:
A systematic study",Phys. Rev. B., vol. 32,pp. 23-47, 1985

_
-

[3] A. V. Shah,M. Vanecek, J. Meier,F. Meillaud, J. Guillet,D.

Fischer, C. Droz, X. Niquille, S. Fay, E. Yallat-Sauvain, Y.

C)
.2
-10

Terrazzoni-Daudrix, and J. Bailat,

"Basic efficiency limits,

recent experimental results and novel light-trapping schemes in

-1

(b)

5 u------
0.0
0.7
-0.7

Voltage (V)
Fig. 5 (a) Photo J-Y characteristics of solar cell on gTCO and
planar ZnO. (b) Dark I-V characteristics of solar cells on gTCO
with various p-a-SiC thicknesses and solar cell on planar ZnO.

978-1-4673-0066-7/12/$26.00 2011 IEEE

a-Si:H, Ilc-Si:H

and

'micromorph

tandem

solar cells", J

Noncryt. Solids., vol. 338-340, pp.639-645, 2004

[4] O. Vetter!, F. Finger, R. Carius, P. Hapke, L. Houben, O.

Kluth, A. Lambertz, A. Muck, B. Rech, and H. Wagner,
"Intrinsic

microcrystalline

silicon:

new

material

for

photovoltaics", Solar Energy Materials and Solar Cells, vol. 62,

pp. 97-108,2000
[5] E. Garnett and P. Yang, "Light trapping in silicon nanowire

solar cells" , Nano Lett., vol. 10,pp. 1082-1087,2010

001185

[6] D. Kim, S. Lee, T. K. Oh, S. Y. Cha, S. J. Hong, and B.

Kang,"Influence of dummy active patterns on mechanical stress
induced by spin-on-glass-filled shallow trench isolation in n

[7] M. Vanecek, O. Babchenko, A. Purkrt, J. Holovsky, N.

Neykova, A. Poruba, Z. Remes, J. Meier, and U.

Kroll,

"Nanostructured three-dimensional thin film silicon solar cells

MOSFETs",Microelectronic Engineering, vol. 88,pp. 882-887,

with very high efficiency potential" , Appl. Phys. Lett., vol. 98,

2011

163503,2011

978-1-4673-0066-7/12/$26.00 2011 IEEE

001186

Silicon Nanowire Development for Solar Cell Devices

l2
l
l
I
2
Gopal G. Pethuraja , , Harry Efstathiadis , Caitlin Rouse , Manisha V. Rane-Fondacaro , Ashok K. Sood
l
and Pradeep Haldar
I

Energy and Environmental Applications Center (E2TAC), College ofNanoscale Science and Engineering,
University at Albany, State University ofNew York, Albany, NY 12203, USA
2

Magnolia Solar Inc., 251 Fuller Road, CESTM-B250, Albany, NY 12203, USA

Abstract

Silicon thin films are promlsmg optical absorber

and stainless steel [1]. While the reported efficiencies of

layers for inexpensive thin film solar cells that are fabricated on

nanocrystalline Si based thin film solar cells have been higher

foreign substrates, such as glass and stainless steel. While the

reported efficiencies of nanocrystalline Si based thin film solar
cells have been higher than the known stabilized efficiency of
amorphous silicon based thin film solar cells, they don't compare

than the known stabilized efficiency of amorphous silicon

based

thin

film

solar

cells,

they

don't

compare

to

the

efficiencies reported for wafer based silicon solar cells, mainly

to the efficiencies reported for wafer based silicon solar cells.

due

This is mainly due to the recombination of minority carriers at

boundaries and intra-grain defects that are present in the

to

the

recombination

of

minority

carriers

at

grain

grain boundaries and intra-grain defects that are present in the

nanocrystalline-Si absorber layer [2].

nanocrystalline-Si absorber layer. Therefore, a device geometry

geometry that allows the benefits of nanostructures and lowers

that allows the benefits of nanostructure and also lowers grain

grain boundary defects should greatly enhance efficiency,

boundary defects should greatly enhance efficiency, reduce the

cost of the final solar cell and help increase market penetration.
Vertically

aligned

silicon

nanowire

(SiNW)

solar

cells

have

Therefore, a device

reduce the cost of the final PV cell and help increase market
penetration. Vertically aligned silicon nanowire (SiNW) solar

already been shown theoretically to be much less sensitive to

cells have already been shown theoretically to be much less

impurities, have low reflective losses and significantly higher

sensitive to impurities [3] and have low reflective losses [4]

optical absorption compared to planar Si Solar cells.

A number

of devices based on SiNW building blocks have been suggested;

however,

high

performance

and

scale-up

utilizing

low

cost

processing approaches to produce effective devices remain a

challenge.

compared to planar Si solar cells.

SiNWs have received a lot of attention in the last few years
because of their attractive electrical and optical properties, as
well as their compatibility with current processing approaches
used in the semiconductor nanoelectronics industry. A number

A high volume and large sample area catalytic chemical vapor

of devices based on SiNW building blocks have already been

deposition system has been designed and built to synthesize

suggested [5]-[8], however, high performance and scale-up

silicon nanowires using the vapor-liquid-solid growth process.

utilizing low cost processing approaches to produce effective

The

prototype devices remain challenges that need to be overcome.

nanowires

substrates.

were

fabricated

on

gold

coated

p-type

Si

The formation of nanowires was confirmed by

secondary and transmission electron microscope analysis. Design

of experiments was carried out to identify the SiNW process
window for solar cell applications. The catalytic chemical vapor
deposition at 550C and 150 torr process pressure yield nanowire

SiNWs are grown from the well-known vapor-liquid-solid

(VLS) reaction [9], where Au particles on a silicon substrate
are generally used as the mediating solvent to direct the
growth of SiNWs.

Other methods,

such as pulsed laser

deposition (PLD)[10],

oxygen-assisted-growth (OAG) [11,

were fabricated by growing i-type SiNWs on a p-type wafer

12],

(VSS)[13]13,

followed by forming a top emitter layer using n-type spin-on

etching [15-17] have also been developed to synthesize

dopant. Further process optimization is on going for realizing

SiNWs.

growth with higher populations.

P-i-n type solar cell devices

high efficiency silicon nanowire solar cells.

Index Terms

vapor-solid-solid

Usually

Silicon nanowires, Photovoltaic cells, Catalytic

chemical vapor deposition.

SiNWs

grown

[18]

by

14]

using

and

VLS

electroless

or

VSS

techniques have shown to be advantageous compared to OAG

and electroless etching techniques in terms of well-defmed
surface and well-controlled diameters and features. The basic

I. INTRODUCTION

issue is to realize controllable synthesis of large-scale high

Thin-film structures can reduce the cost of solar power by

using inexpensive substrates and a lower quantity and quality
of

semiconductor

material.

Recently

there

has

been

an

increased interest in silicon thin films as absorber layers in thin

film solar cells fabricated on foreign substrates, such as glass

978-1-4673-0066-7/12/$26.00 2011 IEEE

quality SiNWs. Moreover, the VLS and VSS techniques can

prepare ordered vertically standing epitaxial nanowires on
single crystal substrates with better control of growth direction
as a device platform [19-22]. Unfortunately, metals, especially
gold and copper, used as catalyst to direct the growth of

001911

SiNWs by VLS or VSS techniques may induce deep levels in

system and the exhaust line. The reaction zone is consisted of

SiNWs, which would trap electrons and holes in Si, and then

a horizontal quartz tube with a 2" diameter placed inside a

depress the desired optical and electronic properties of the

high temperature furnace with a 24" heated zone capable of

nanowires. Achieving uniform doping in SiNWs remains a key

reaching up to l100e. A thermocouple, mounted at the center

issue to fabricate high performance SiNW-based devices. At

the present time, the doping methods of SiNWs include

Throttle Valve

thermal diffusion, ion implantation and in-situ doping during

Heating

the growth process by co-flowing doping gas with silane or

tetrachlorosilane. These methods still have disadvantages, such

Catalyst coated substrates

--

as inducing defects and doping fluctuation, especially at high

concentrations.

Precise

nanomaterials

remains

doping
a

of

challenge.

semiconductor
Recently,

researchers successfully synthesized SiNWs via the solid

liquid-solid

(SLS)

method,

relatively

'\...: Quartz chamber

CNSE

'--

Vacuum Gauge

straightfordward

technique without a gas phase precursor such as SiH4 or SiCI4

[23].
The physical properties including electrical, photoelectrical,
and mechanical properties of SiNWs employed as building
blocks

for

devices

make

their

incorporation

extremely

Fig. 1.
Schematic of catalytic CVD system that employs
vapor-liquid-solid mechanism to grow SiNWs.

attractive. The energy gap of SiNWs can be tuned to increase

with decreasing SiNW diameter. Other structural parameters

of the reaction zone on the outer wall of the quartz tube, was

(i.e. growth directions, cross section) also affect band structure

used to continuously monitor the temperature during the

and thus physical properties of SiNWs, especially for ultra tiny

nanowire growth process. On this end of the quartz tube is

SiNWs [24]. The high carrier mobility of small diameter

conceded to a vacuum pump system.

SiNWs is essential for fabricating high performance solar

The substrates were loaded into the reaction zone through a

cells. Moreover, SiNWs have high surface-to-volume ratio

quick access door that was installed on the one end of the

which ensures that mass carriers can be controlled easily by

quartz tube. On the same side of the tube, two gas delivery

weak electrical fields. SiNW-arrays were shown to be more

lines were constructed and interfaced with the associated gas

efficient light absorbers than bulk silicon [25]. Recent reports

delivery control systems. The flow of the precursor and inert

indicated that photo excited minority carriers can diffuse a

gases on the reaction zone was controlled by MKS mass flow

relatively long distance without recombination and contribute

controllers. Mixture of silane/argon (SiH4/Ar) was used as the

to more carrier collection if the p-n junction is aligned along

the radius, as the diameter of a SiNW is comparable with
minority carrier diffusion length, [3] making them suitable for
high efficiency solar cells.
A number of devices based on SiNW building blocks have
already

been

suggested,

however,

its

feasibility

of

high

efficiency photovoltaic devices remains a challenge that needs

to be overcome. In this paper, we studied the influence of
process parameter on the growth and properties of SiNW. The
nanowires grown at 550C, 150 torr process pressure yields
higher population of nanowires.

II. EXPERIMENTAL
A high volume and large sample area catalytic chemical

WO=

5mm

TiH Angle:. 0.0

vapor deposition system was designed and built to synthesize

silicon nanowires, using the vapor-liquid-solid growth process.
Silicon nanowire layers were fabricated on catalyst coated Si
and

stainless

chemical

steel

vapor

substrates

deposition

using

system.

the

in-built

Figure

Fig. 2.
Typical SEM image of catalytic nanoparticels formed
after pre-treatment.

catalytic
shows

Si source and high purity argon (Ar) gas was used for

schematic of the high volume and large sample area catalytic

annealing the catalyst material prior to the growth of SiNWs.

chemical vapor deposition system. The catalytic CVD system

In the reactor geometry employed, reaction by-products and

consists of reaction zone of the reactor, the gas delivery

excess gas were removed from the other end of the reactor.

978-1-4673-0066-7/12/$26.00 2011 IEEE

001912

Before loading the substrate into the reactor, the substrates

were cleaned using acetone, methanol, isopropanol and de-

aluminum were used as catalysts in this study.

Catalytic

chemical vapor deposition employing the vapor-liquid-solid

(VLS) growth mechanism is used to grow intrinsic type Si

n+

Al

emitter

i-SiNW

pSi

Diameter: -100nm

nanoparticles on the substrate.

pre-treatment at 550C.

Fig. 3. Device design for solar cell testing

Native oxide layer on silicon substrate were

stripped using diluted HF solution. After the cleaning process,

nm

Fig. 2 shows the typical SEM

image of catalytic nanoparticles that are formed during the

AI

approximately 5

dat 450 or 550 C in Ar ambient for 5 to 10 minutes for

agglomeration of the catalytic film and formation of catalytic

substrate

on the substrate.

silane gas in to the reactor. The pre-treatment was carried out

Height: 5-lOum

0.1-2 Oem, 200-400J.1m

ionized water.

nanowires. The substrates were pre-treated before introducing

tJ:Hek of the catalytic film was deposited

Gold, nickel, gold-palladium alloy and

The catalytic nano-particles act as a

nucleation site for the VLS growth of silicon nanowires.

Then, the silane source gas was introduced in the reactor to
grow the nanowires. The growth temperature varied from 450
to 650 C and growth time varied from 30 to 60 min.

The

SiNW samples were characterized by Secondary Electron

Microscope

(SEM),

Transmission

Electron

Microscope

(TEM) and Energy Dispersive Spectroscopy (EDS) analysis.

Figure 3 shows a schematic of the Si nanowire test solar
cell. Device fabrication begins with growing nanowires on a p-

Fig. 4.
SEM micrographs of Si nanowires grown by catalytic CVD at various process temperatures using Au as a catalyst. (a) 450 C;
(b) 550 C; (c) 650 C; (d) 550 C and textured Si substrate.

978-1-4673-0066-7/12/$26.00 2011 IEEE

001913

Fig. 5.

SEM micrograph of silicon nanowires grown by catalytic CVD at various process pressures. (a) 150 torr; (b) 170 torr; (c) 250
torr; (d) high resolution image of a nanowire in figure (c),shows the presence of gold catalyst on top.

type Si substrate.

Then the gold catalysts were etched by

dipping the nanowire sample in to mixtures of nitric acid and

hydrochloric acid (in a mixing ratio of 1 :
temperature. N-type spin-on dopant,

i) light I-V curves under standard test conditions, ii) dark I-V
curves.

3) at room

Phosphorofilm from

III. RESULTS AND DISCUSSION

Emulsitone company, Whippany, New Jersey, USA was used

to form the nanowire emitter. Recently, we have demonstrated

Silicon nanowire layers were fabricated on gold, nickel and

the spin-on-dopant approach for fabricating pn type silicon

gold palladium coated Si substrates using catalytic chemical

solar cells with 13% efficiency [26]. Based on these results,

vapor deposition method.

the incorporation of SiNWs could further increase the cell

electron microscopy micrograph of the nanowire grown at

efficiencies.

various process temperatures ranging from 450 to 650 C at

After spin coating the dopant, the samples were

Figure 4 shows typical secondary

The annealed

150 torr pressure using Au catalytic films. The CVD process at

samples were subjected to oxide layer etching, the RCA1 and

550C yielded highly populated nanowires, while the process

annealed at 850 C for 1 hour in vacuum.

RCA2

cleaning

process.

The

cell

was

completed

by

temperature 450 C and 650 C didn't yield Si nanowires.

depositing Al metal contact or an Al doped ZnO transparent

EDS

conductive oxide layer contact.

nanowires are Si (not shown here). The nanowires are smooth

The fabricated devices were

compositional

analysis

indicated

that

the

observed

tested in a standard AM 1.5 solar simulator. Prior to device

and the diameter varies from 5 to 100

measurements a Si reference cell calibrated by the National

the spacing between nanowires, the Si substrate was textured.

nm .

In order to control

Renewable Energy Laboratory (NREL) was measured. The

The texture in the Si substrate didn't influence the structure

solar cells were characterized by the following measurements:

and growth of the nanowires (see Fig. 4.(d)).

978-1-4673-0066-7/12/$26.00 2011 IEEE

001914

An investigation has been performed to study the effect of

process pressure on growth properties of nanowires. Figure 5
shows the SEM micrographs of silicon nanowires grown at

pressure, the thermal CVD is more favorable, resulting in less

population of nanowires.
Compositional analysis of the nanotubes shown in Figure

150, 175 and 250 torr pressure, 550C for 60 minutes using

5(d) were carried out by EDS.

gold catalyst film.

EDS mapping analysis.

At higher flow rates, nanowire population

Figure 6 shows the results of

The Si and Au elemental mapping of

is significantly reduced. Figure 5(d) shows the presence of the

the silicon nanowire show the droplet observed on top of the

gold catalyst at the top of the nanowire.

nanowire is gold.

resolution

image

of

the

nanowire

morphology on the nanowire surface.

CVD of silicon on the nanowires.

In addition, the high

shows

layer by layer

The Si K-alpha spectra image shows

presence of Si all over the sample except the top droplet

This indicates thermal

region observed in the electron micrograph. This is due to the

During the deposition

presence of the Si layer, deposited by thermal CVD, in

process, two different processes take place.

The first process

addition to the observed nanowire.

But the Au M-alpha

Fig. 6.
SEM micrograph and EDS elemental mapping analysis of a silicon nanowire. (a). Micrograph showing the nanowires. (b) Si
Ka spectral image; (c) Au Ma spectral image.

(a)

0.2 m
Fig.7.
TEM micrographs of silicon nanowires grown at 550 C using gold catalyst. (a). Micrograph showing presence of defects in the
nanowires. (b) High resolution micrograph shows nanotubes and gold catalyst cap are in single crystalline structure.
At

spectra image shows the presence of gold only at the droplet

150 torr pressure, VLS mechanism induced catalytic CVD is

region. The gold signal from the substrate layer (catalytic film)

more pronounced than the thermal CVD. However, at higher

is reduced by the growth of the Si layer.

is thermal CVD and the second one is the catalytic CVD.

978-1-4673-0066-7/12/$26.00 2011 IEEE

001915

Further analysis on the structure of nanotubes was carried

out by HRTEM.
nanowire.

Figure 7 shows the HRTEM micrograph of a

The micrograph shows that the nanotubes and the

gold catalyst are in single crystalline.

It also shows the

presence of defects in the nanotubes.

In addition, Si nanowire solar cells were designed and
fabricated.

These test devices were used to optimize the

Catalytic

CVD

process

parameter.

Further

process

optimization is required to realize higher efficiency solar cell.

IV. SUMMARY
A high volume and large sample area catalytic CVD system
was built and silicon nanowires were successfully grown on Si
and stainless steel substrates.

Catalytic CVD process at 550

C and 150 torr pressure yield highly populated silicon

nanowires.

SiNW solar cells were successfully fabricated

using spin-on-dopants and used to optimize the catalytic CVD

process parameters. Further optimization is required to realize
high efficiency silicon nanowire solar cells.

ACKNOWLEDGEMENT
This work is supported by National Science Foundation.

[14] H.F. Yan, Y.1. Xing, Q.L. Hang, D.P. Yu, Y.P. Wang, J. Xu,
Z.H. Xi, and S.Q. Feng, Chern. Phys. Lett. vo1.323, p.224,
2000.
[15] K.Q. Peng, Y.1. Yan, S.P. Gao, J. Zhyu, Adv. Mater. vol. 14,
pp.II64-1167,2002.
[16] K.Q. Peng,Y.1. Yan, S.P. Gao, J. Zhyu,Adv. Funct. Mater. 13,
2003, pp. 127-132.
[17] K.Q. Peng, 1.1. Hun, Y.1. Yan, et. al. Adv. Funct. Mater. 16,
2006, pp.387-394.
[18] Y. Wang, V. Schmidt, S. Senz, and U. Gosele, Nature
Nanotechnology, vol. I,p. 186,2006.
[19] N. Elfstrom, 1. Linnros, Appl. Phys. Lett., vo1.91, p. 103502,
2007.
[20] K.Q. Peng, X. Wang, S.T. Lee, Appl. Phys. Lett., vol.92,
p. 163103,2008.
[21] A.1. Hochbaum, R. K. Chen, R.D. Delgado, et.al. Nature
vo1. 451,pp. 163-168,2008.
[22] A. Asenov, A.R. Brown, 1.H. Davies, et al. IEEE Trans Elec
Dev., vo1.50,pp. 1837-1852,2003.
[23] 1.H. Lee, P.H. Rogers, M.A. Carpenter, E.T. Eisenbraum, Y.
Xue, and R.E. Geer, Proc. 81h IEEE conference on
Nanotechnology, (IEEE NANO'08) (Aug.,2008).
[24] T. Vo, A.J. Williamson, G. Galli, Phy. Rev. B, vo1.74, p.
045116,2006.
[25] L. Tsakalakos, J. Balch, J. Fronheiser,B.A. Korevaar, Appl.
Phys. Lett., vo1. 91,p.233117,2007.
[26] A.V. Nguyen, H. Efstathiadis, Z. Zhouying, A. Fioramonti, D.
Morrissey,P. Haldar, Proc. 341h IEEE PVSC, 2009.

REFERENCES

[I]

[2]
[3]
[4]
[5]
[6]
[7]
[8]

[9]
[10]
[11]
[12]
[13]

K. Yamamoto, A. Nakajima, M. Yoshimi, T. Sawada, S.

Fukuda, T. Suezaki,M. Ichikawa,Y. Koi,M. Goto, T. Meguro,
T. Matsuda, M. Kondo, T. Sasaki, and Y. Tawada, Solar
Energy, vol.77, p. 939,2004.
R.B. Bergmann, Applied Physics A: Materials Science and
Processing, vo1.69,p. 187, 1999.
Kayes, B. M.; Atwater, H. A.; Lewis, N. S. 1. Appl. Phys.
vo1.97, p. 114302,2005.
Erik Garnett, Peidong Yang, Nano Letters, vol.lO, pp. 10821087. 20 10.
1.H. Lee,M.A. Carpenter,and R.E. Geer,1. Mater. Res., vo1.26,
pp. 2232-2239,20 11.
1.H. Lee, LN. Lund, E. T. Eisenbraun and R.E. Geer,
Nanotechnology, vo1.22, p. 085603, 20 10
Y. Wu and P. Yang,1. Am. Chern. Soc. vol.l23, p.3165,200 1.
S. Hofinann, R. Sharma, C.T. Wirth, F. Cervantes-Sodi, C.
Ducati, T. Kasama, R. E. Dunin-Borkowski, J. Drucker, P.
Bennett,and 1. Robertson,Nature Materials, vol.7, p.372, 2008.
R.S. Wagner, W.C. Ellis, Appl. Phys. Lett. vol.4, pp.89-90,
1964.
Y.F. Zhang, Y.H. Tang, N. Wang, D.P. Yu, C.S. Lee, I. Bello,
and S.T. Lee,Appl. Phys. Lett., vol. 721, p.835, 1998.
R. Q. Zhang,Y. Lifshitz, and S.T. Lee: Adv. Mater., vol.l5, 635
(2003).
R. Q. Zhang, T.S. Chu, H.F. Cheung, N. Wang, and S.T. Lee,
Phys. Rev. B, vo1.64,p.I13304,200 1.
M. Paulose, O.K. Varghese, and C.A. Grimes, 1. Nanosci.
Nanotechnol., vol.3, p.341,2003.

978-1-4673-0066-7/12/$26.00 2011 IEEE

001916

Development and Comparison of Small and Large Area Boron Doped

Solar Cells in n-type and p-type Cz-Si
Izete Zanesco, Adriano Moehlecke, Jaqueline Ludvig Pinto, and Moussa Ly
Solar Energy Technology Nucleus NT-Solar, Faculty of Physics Pontifical Catholic University of Rio
Grande do Sul PUCRS, Av. Ipiranga, 6681, P. 96A Porto Alegre RS Brazil CEP 90619-900
Abstract The first step to introduce an evolutionary
approach is the development in the laboratory scale and, then, in
pilot plant and mass production. Boron diffusion forms the
emitter in n-type substrates and the back surface field (BSF) in
p-type silicon wafers. N-type Cz-Si solar cells have been
investigated due to the potential to produce high efficiency solar
cells and boron BSF may provide a better passivation than
aluminum BSF in p-type wafers. The goal of this paper is to
present the development and comparison of small and large area
solar cells, processed in Czochralski silicon wafers by using spinon dopant to obtain the boron emitter in n-type Si substrates and
the boron BSF in p-type ones. The average efficiency of boron
emitter and BSF cells was of around 15.0 % and 13.4 % for small
and large area solar cells, independently of the substrate type.
The reduction of the efficiency is due to the short-circuit current
density that falls of around 5 mA/cm2 when the area is enlarged.
The low fill factor obtained in p+nn+ cells is due to the high
resistivity of Ag/Al paste deposited in the boron emitter cells and
the soldering of the Ag/Sn/Cu ribbon increased this parameter
from 0.70 to 0.75 in large area cells.
Index Terms silicon solar cells, boron-emitter, boron-BSF.

I. INTRODUCTION
An important goal of the PV industry is to reduce the PV
electricity cost in order to achieve the grid parity. This
objective is influenced by many factors as the introduction of
evolutionary technologies in order to increase the efficiency
and/or reduce the production cost. The first step to introduce
an evolutionary approach is the development in the laboratory,
followed by the pilot scale and mass production [1]. In lab.
scale, 4 cm2 silicon solar cells are usually developed and when
the area is enlarged technical problems may be overcome.
The industry of silicon solar cells is based on the aluminum
back surface field (Al-BSF) formed by the conventional
screen printing metallization, using an aluminum paste
deposited on Czochralski-grown (Cz-Si) p-type substrate. The
average efficiency of industrial solar cells is about 16.5 % [1].
The Al-BSF rear passivation technology limits the efficiency
mainly in thin substrates [2]. Moreover, the screen printed Al
paste produces bowing of the wafer, despite of the evolution
of low-bow pastes. To overcome these challenges, the p+ layer
can be produced with boron doping.
In p-type Si substrates, the boron diffusion is used to
produce the back surface field (B-BSF) and the challenge is to
obtain the surface concentration that fits in well the standard
screen printing metallization and without contamination of the
silicon substrate during the high temperature diffusion process

978-1-4673-0066-7/12/$26.00 2011 IEEE

[2]. Boron BSF silicon solar cells have been developed with
dielectric passivation and the efficiency of 20 % was achieved
in 4 cm2 cells and with screen printing metallization [3].
The boron diffusion in n-type silicon material forms the
emitter (B-emitter). Beyond of the above mentioned
challenges, the Al/Ag pastes currently used on p+ regions
present higher resistivity than the Ag paste to form the front
metal grid. Nevertheless, since the last decade n-type Cz-Si
solar cells are being investigated due to the potential to
produce high efficiency solar cells. Substrate doped with
phosphorus is more stable and have higher minority carrier
lifetimes than p-type silicon [4].
Sun Power developed the interdigitated contact solar cell
and the average efficiency in industrial production in float
zone substrate (FZ-Si) was above 22 % [1]. With other
structure and front boron emitter, Mihailetchi et al [5]
developed large area solar cells with efficiency of 17.4%. In
multi-crystalline silicon wafers (MC-Si), solar cells with an
area of 156.25 cm2 were processed with boron and
phosphorus simultaneous diffusion and the efficiency of 16.4
% was obtained [6]. The influence of the substrate in boron
doped emitter cells was experimentally studied and the
efficiency of 14.7 %, 15.9 % e 17.1 % was achieved with MCSi, Cz-Si and FZ-Si, respectively [7]. Solar cells were also
processed with boron doped p+ layer using spin-on dopant
instead the conventional BBr3 dopant. In FZ-Si n-type
material, the efficiency of 15.9 % was obtained with
SiO2+SiNx:H passivation.
The aim of this paper is to present the development and
comparison of small and large area screen printed silicon solar
cells, processed by using spin-on dopant to form the boron
emitter in n-type Cz-Si substrates and B-BSF in p-type
wafers.
II. SOLAR CELL PROCESS
The baseline process used to develop the solar cells was the
following: texture etching, RCA cleaning, boron spin-on
deposition, boron diffusion in a quartz tube furnace, resist
deposition and oxide etching, RCA cleaning, phosphorus
diffusion, borosilicate glass etching, TiO2 antireflection
coating deposition, screen-printing metallization in both sides
and edge isolation. Solar cells were developed in 1 .cm 20
.cm n-type and p-type solar grade Cz-Si wafers with
thickness of 200 m. The boron dopant was spun onto one

002284

side of the wafer and the diffusion was carried out in a quartz
tube furnace at 1000 C. The n+ layer was produced by
phosphorus diffusion using POCl3. The oxide to protect the p+
layer of the phosphorus diffusion was grown in the same
thermal step of boron diffusion. A specific passivation was
not implemented.
The boron and phosphorus diffusions were independently
and experimentally optimized for n-type and p-type material
as well as the metal grid firing process. The metal grid was
formed with two busbars in large area cell and with only one
busbar in small area devices, in order to compare the typical
lab. cells with an area of 4 cm2 with the large area (61.58 cm2)
silicon solar cells.
All solar cells were characterized under standard conditions
(100 mW/cm2, AM1.5G and 25C) in a solar simulator
calibrated with a small (4 cm2) and a large (61.58 cm2) area
silicon solar cell previously measured at CalLab - FhG-ISE
(Fraunhofer-Institut fr Solare Energiesysteme), Germany.

Ag/Al paste used to form the metal grid on the frontal face.
With the same metal grid, but using an Ag paste on the frontal
surface, fill factors of 0.78 were obtained in n+pp+ solar cells
doped with phosphorus and aluminum.

III. BORON-EMITTER SOLAR CELLS

The boron and phosphorus profiles measured by ECV

(electrochemical capacitance-voltage) technique are presented
in the Fig. 1 and Fig. 2, respectively. The junction depth for p+
emitter was about 1 m, similar to the thickness of BSF
region doped with phosphorus (~ 1.1 m). The surface
concentration (CS) of boron was higher than 1x1019
atoms/cm3, as Fig. 1 shows. On the other hand, the CS found
with phosphorus profile was of around 5x1020 atoms/cm3.

TABLE I
VOC, JSC, FF AND EFFICIENCY OF THE 4 CM2 B-EMITTER (P+NN+) SILICON
SOLAR CELLS PROCESSED IN THE WAFER 3.
JSC
FF
(%)
Cell
VOC (V)
(mA/cm2)
3-A
598.4
35.1
0.72
15.1
3-B
586.3
34.7
0.74
15.0
3-C
591.5
35.4
0.73
15.2
3-E
587.5
35.1
0.74
15.2
3-F
597.5
35.9
0.72
15.4
3-G
595.3
35.5
0.70
14.8
3-H
596.9
36.0
0.70
15.1
Aver.
5935
35.40.5
0.720.02
15.10.2

The average FF was of 0.72 and 0.74. These values are

lower than the expected probably due to the resistivity of

978-1-4673-0066-7/12/$26.00 2011 IEEE

Dopant Concentration (atoms/cm)

In the p+nn+ solar cells, the diffusion processes formed a 43

/ p+ emitter and a 25 / n+ layer. The relative standard
deviation was about 6 % in both sheet resistances. Table I and
Table II present the open circuit voltage (VOC), short-circuit
current density (JSC), fill factor (FF) and efficiency () of the 4
cm2 solar cells, processed in the wafer 3 and 7, respectively.
The cells were processed in the same batch in order to
compare the influence of the different wafers of the same
supplier. The average efficiency of 15 % is similar to the cells
processed on both wafers, but the VOC is higher for cells of the
wafer 3. This result can be attributed to a lower base
resistivity or a higher minority carrier lifetime of the wafer 3.
The average efficiency is similar, because the reduced value
of the FF found in cells of wafer 3 was compensated by the
high value of VOC. Open circuit voltage and short-circuit
current are limited due to the lack of a high quality
passivation.

TABLE II
ELECTRICAL CHARACTERISTICS OF THE 4 CM2 B-EMITTER (P+NN+)
SILICON SOLAR CELLS PROCESSED IN THE WAFER 7.
JSC
VOC
FF
(%)
Cell
(mV)
(mA/cm2)
7-A
584.2
35.0
0.74
15.2
7-B
576.5
34.6
0.73
14.7
7-C
585.5
35.2
0.74
15.3
7-D
577.7
34.6
0.74
14.8
7-F
577.9
34.7
0.73
14.7
7-G
588.7
35.5
0.74
15.5
7-H
580.2
34.9
0.72
14.5
7-I
589.7
35.2
0.73
15.3
Aver.
5825
35.00.3
0.740.08
15.00.4

1E+21
1E+20
1E+19
1E+18
1E+17
1E+16
1E+15
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Depth (um)
+

Fig. 1. Boron profile of p nn

measured by ECV technique.

(B-emitter) silicon solar cells

The electrical characteristics of large area solar cells

processed in two batches are presented in Table III. The small
area devices were fabricated in the same batch of B2-A and
B2-B cells. The average efficiency of large cells was (13.4
0.2) %. Comparing the values presented in the Table III and
the Table I, we observed that VOC and FF are similar and the
main difference occurs in the JSC. The value of JSC is reduced
of around 11% when the area of the cells is increased, causing
an efficiency reduction of approximately 1.5 % (absolute).
This result cannot be explained only by the area of metal grid

002285

Dopant Concentration (atoms/cm)

in the frontal face. In large and small area cells, 9.4 % and 6.0
% of the area is covered by the metal grid, respectively. The
p+nn+ cells were simulated by using PC-1D device modeling
program and the 3.4% higher metal grid coverage in large area
cells implied in a JSC reduction of only 3.9%. For instance, it
is worth to comment that the difference between JSC of small
area and large area cells was also observed in n+pp+ cells
doped with phosphorus/aluminum, produced in the PUCRS
and measured at CalLab - FhG-ISE. In this case, the
difference was of around 7%.
1E+21

TABLE IV
ELECTRICAL CHARACTERISTICS OF THE 4 CM2 B-BSF (N+PP+) SILICON
SOLAR CELLS PROCESSED IN THE WAFER 8.
JSC
FF
(%)
Cell
VOC (V)
(mA/cm2)
8-F
580.6
33.3
0.74
14.3
8-G
580.5
32.7
0.73
13.8
8-I
580.9
33.6
0.73
14.3
Aver.
580.80.3
33.20.5
0.730.02
14.10.8

1E+20
1E+19
1E+18
1E+17
1E+16
1E+15
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Depth (um)

Fig. 2.

The electrical characteristics of 4 cm2 cells are shown in

Table IV and Table V. The small cells were processed in two
wafers. The devices produced in wafer 8 presented lower VOC
and JSC than that obtained with wafer 14. With wafer 14, the
average efficiency of cells was 15.0 %, similar to that
achieved in small p+nn+ solar cells.
The results obtained with large area cells are presented in
Table VI. The average efficiency was 13.3 %, 1.7 %
(absolute) lower than the efficiency achieved with 4 cm2 cells.
Again, a worse efficiency of large area cells is due to lower
short-circuit current density.

Phosphorus profile of p+nn+ (B-emitter) silicon solar cells.

TABLE III
VOC, JSC, FF AND EFFICIENCY OF THE LARGE AREA B-EMITTER (P+NN+)
SILICON SOLAR CELLS AND AVERAGE VALUES.
JSC
Cell
VOC (V)
FF
(%)
(mA/cm2)
B1-A
596.2
31.0
0.73
13.5
B1-B
590.4
31.4
0.73
13.5
B2-A
596.1
31.6
0.70
13.2
B2-B
599.4
31.6
0.71
13.4
Aver.
595.51.7
31.40.5
0.720.4
13.40.2

In large area cells, fill factor was also lower than expected.
High metal grid resistivity could explain the low FF obtained.
To investigate the influence of the Ag/Al paste resistivity,
Ag/Sn/Cu ribbons were soldered on the front busbars and
cells were measured at standard conditions. Fill factor rose
from 0.70 to 0.75 causing an increase of 13.2 % to 14.9 % in
the efficiency. PV modules with 36 cells connected in series
were fabricated and FF reached values ranging from 0.74 to
0.76.
IV. BORON-BSF SOLAR CELLS
The process was also implemented to obtain n+pp+ (B-BSF)
cells. The sheet resistivity of the phosphorus emitter and the
B-BSF were 30 / , with the relative standard deviation of
20 % and 9%, respectively. Cells with different areas were
processed in the same batch.

978-1-4673-0066-7/12/$26.00 2011 IEEE

TABLE V
MEASURED ONE-SUN PARAMETERS OF 4 CM2 B-BSF (N+PP+) SILICON
SOLAR CELLS PROCESSED IN THE WAFER 14.
JSC
Cell
VOC (mV)
FF
(%)
(mA/cm2)
14-A
596.1
36.6
0.71
15.4
14-B
594.8
37.3
0.60
13.4
14-C
591.6
36.4
0.71
15.3
14-D
591.8
35.4
0.73
15.4
14-E
594.5
36.7
0.71
15.4
Aver.
593.72.0
36.50.7
0,690.05
15.00.9
TABLE VI
VOC, JSC, FF AND EFFICIENCY OF THE LARGE AREA B-BSF (N+PP+)
SILICON SOLAR CELLS AND AVERAGE VALUES.
JSC
Cell
VOC (V)
FF
(%)
(mA/cm2)
A
601.7
30.8
0.72
13.4
B
599.8
30.9
0.69
12.9
C
602.3
31.4
0.72
13.7
Aver.
601.31.3
31.00.3
0.710.02
13.30.4

Fig. 3 and Fig. 4 present the boron and phosphorus profiles

measured by ECV. Comparing the doping profiles presented
in Fig. 1 and Fig. 2 to those presented in Fig. 3 and Fig. 4, we
observed that the depth of pn junction or BSF layer thickness
are similar for boron doped regions. However, the boron
surface concentration is higher in n+pp+ solar cells than in
p+nn+ ones and the CS for phosphorus doped region is lower.
In other words, the surface concentration of the emitter is
lower than the CS of the BSF layer, taking into account the
same dopant. The pn junction depth of n+pp+ cells was of
around 0.6 m, lower than n+ layer thickness of the p+nn+
devices. This difference in phosphorus doping profiles are due

002286

1E+21
1E+20
1E+19

TABLE VII
AVERAGE VALUES OF VOC, JSC, FF AND OF B-EMITTER AND B-BSF
SILICON SOLAR CELLS WITH DIFFERENT AREAS.

1E+18
1E+17

B-emitter B-emitter
B-BSF
B-BSF
small cells large cells small cells large cells

1E+16
1E+15
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Depth (um)
+

Boron doping profile of n pp (B-BSF) silicon solar cells.

5935

JSC
(mA/cm2)

35.40.5

31.40.5

36.50.7

FF

0.720.02

0.720.4

0.690.05 0.710.02

(%)

15.10.2

13.40.2

15.00.9

1E+21

595.51.7 593.72.0 601.31.3

31.00.3

13.30.4

1E+20

40

1E+19
1E+18
1E+17
1E+16
1E+15
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Depth (um)

Fig. 4.

VOC (mV)

Current Density (mA/cm)

Dopant Concentrtion (atoms/cm)

Fig. 3.

with the CP4 etching, varied from 40 s to 160 s. The bulk

lifetime found in the p-type substrate was in the range of 25
s to 50 s. The cells were simulated by using the PC-1D.
Considering a poor surface passivation for p+ and n+ regions
(a surface recombination velocity of 1x106 cm/s) and fitting
the simulated VOC to the experimental ones, base lifetime for
p+nn+ and n+pp+ cells was of 80 s and of 60 s, respectively.

Phosphorus doping profile of n+pp+ (B-BSF) solar cells.

V. COMPARISON OF THE RESULTS

Table VII shows the average one-sun parameters
concerning to the results of Table I, III, V and VI. The
average efficiency of B-emitter and B-BSF cells was of
around 15 % and 13.3 % for small and large area solar cells,
respectively. This difference is caused by the JSC, which
decreases from 35 mA/cm2 - 36 mA/cm2 to 31 mA/cm2 when
the area is increased. The average values of VOC and FF are
similar to solar cells with both areas and types of substrate.
In Fig. 5 and Fig. 6, the current density as a function of
applied voltage (J-V) of the best solar cells is presented. The
efficiency achieved with 4 cm2 cells was of 15.5 % for n-type
and p-type Si-Cz. Cells with this efficiency presented the best
JSC values. The best efficiency obtained with large area cells
was of 13.5 %, similar to the average value.
Comparing the results, we can conclude that the type of
substrate did not affect the efficiency of cells, despite of a
higher initial minority carrier lifetimes measured in n-type
substrates. In this material, the minority carrier lifetime
measured in as-cut wafers, after the etching of 10 m - 20 m

978-1-4673-0066-7/12/$26.00 2011 IEEE

30

20

10

& & & & & & & & & & & & & & & & & &
&
&
,, , , , , , , , , , , , , , , , , ,
&
,,
, &
Voc = 596.2 mV
,,
,, &
,,
Jsc = 31.0 mA/cm2
,,&
Voc = 588.7 mV
,,&
FF = 0.729
,,
2
Jsc = 35.5 mA/cm
,,&
= 13.4 %
,,
FF = 0.740
,,&
2
,,
61.58 cm
= 15.5 %
,&,
,,
,,
2
4 cm
&,
,,
,
&,,
,,
& ,,

0.1

0.2

0.3

0.4

0.5

p+nn+

0.6

0.7

Voltage (V)

Fig. 5. Current density as a function of applied voltage, measured

under standard conditions of the best small and large B-emitter
silicon solar cells.
40
Current Density (mA/cm)

Dopant Concentration (atoms/cm)

to the nitrogen and oxygen flows used in each process (p+nn+

and n+pp+) and because in the process to produce p+nn+ solar
cells there is a drive-in step.

30

20

10

n+pp+

&& & & & & & & & & & & & & & &
&&&&&
&&&
&&
&&
,, , , , , , , , , , , , , , , , ,
, , &&&
,,, &
,, &&
Voc = 602.3 mV
,, &
,, &
,,&
Jsc = 31.4 mA/cm2 Voc = 596.1 mV
,,&
,,&
FF = 0.724
2
Jsc = 36.6 mA/cm
,,&
= 13.7 %
,&,
FF = 0.707
,&,
2
&,
61.58 cm
= 15.4 %
&,,
&,
&,,
2
4 cm
&,,
&,,
&,
&,,
,
&&,

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Voltage (V)

Fig. 6. Current density as a function of applied voltage, measured

under standard conditions of the best small and large B-BSF silicon
solar cells.

002287

VI. CONCLUSIONS
+

[2]
+

Similar efficiency was found in p nn and n pp cells,

independently of the area. The low fill factor in boron emitter
cells was related to the high resistivity of Ag/Al paste and the
soldering of the Ag/Sn/Cu ribbon improved this parameter
from 0.70 to 0.75 in large area cells. Nevertheless, in large
area cells the efficiency falls more than 1.5 % (absolute) due
to the JSC reduction.
In summary, the type of substrates does not affect the solar
cell efficiency for the processes based on boron deposited by
spin-on, followed by the quartz tube diffusion, but the
increasing of the area reduces the efficiency. Consequently,
from lab. to pilot scale fabrication, efforts have to be made to
overcome this undesirable feature.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the financial support of
the Brazilian financing agency FINEP (Financiadora de
Estudos e Projetos) under contract n 01.08.0635.00 and of
the National Council for Scientific and Technological
Development (CNPq).
REFERENCES
[1]

P. Fath, S. Keller, P. Winter, W. Joo, and W. Herbst, Status

and perspective of crystalline silicon solar cell production,
34th IEEE PVSC, 2009, pp. 2471-2476.

978-1-4673-0066-7/12/$26.00 2011 IEEE

[3]

[4]

[5]

[6]

[7]

[8]

J. Jourdan, Y. Veschetti, S. Dubois, T. Desrues, and R. Monna,

Formation of boron-doped region using spin-on dopant:
investigation on the impact of metallic impurities, Progress in
Photovoltaics: Research and Applications, vol. 16, 5, 2008, pp.
379-387.
A. das, V. Meemongkolkiat, D. S. Kim, S. Ramanathan, and A.
Rohatgi, 20 % efficient screen printed boron BSF cells using
spin-on dielectric passivation, 34th IEEE PVSC, 2009, pp.
477-481.
J. Schmidt et al., n-type silicon - the better material choice for
industrial high-efficiency solar cells?, 22th European
Photovoltaic Solar Energy Conference, 2007, pp. 998-1001.
V.D. Mihailetchi et al., 17.4 % efficiency solar cells on large
area and thin n-type silicon with screen-printed aluminiumalloyed rear emitter, 22th European Photovoltaic Solar
Energy Conference, 2007, pp. 837-840.
V.D. Mihailetchi et al., High efficiency industrial screen
printed n-type mc-si solar cells with front boron emitter, 22th
European Photovoltaic Solar Energy Conference, 2007, pp.
1581-1585.
T. Buck et al., Industrial screen printed n-type silicon solar
cells with front boron emitter and efficiencies exceeding 17
%, 21th European Photovoltaic Solar Energy Conference,
2006, pp. 1264-1267.
T. Desrues, J. Jourdan, Y. Veschetti, and R. Monna, N-type Si
solar cells with B-doped emitter using spin-on dopants, 22th
European Photovoltaic Solar Energy Conference, 2007, pp.
1209-1212.

002288

Recent Progress of Spectrolab High-Efficiency Space Solar Cells

Daniel C. Law, X.Q. Liu, J.C. Boisvert, E.M. Redher, C.M. Fetzer, S. Mesropian, R.R. King,
K.M. Edmondson, B. Jun, R.L. Woo, D.D. Krut, P.T. Chiu, D.M. Bhusari, S.K. Sharma, and N.H. Karam
Spectrolab Inc., a wholly owned subsidiary of Boeing, Sylmar, CA 91342, U.S.A.
Abstract High-efficiency Inverted Metamorphic (IMM)
multi-junction solar cells are being developed at Spectrolab for
use in space and near-space applications. Recently, large-area
(26-cm2) IMM3J cells achieved a 1-sun, AM0 conversion
efficiency of 32% with an open-circuit-voltage of 3.04 V, a shortcircuit current-density of 16.7 mA/cm2, and a fill factor of 0.84.
In addition, IMM4J cells (1-cm2) reached a 1-sun, AM0
conversion efficiency of 33%. The 4-junction cell achieved an
open-circuit-voltage of 3.42 V, a short-circuit current-density of
15.8 mA/cm2, and a fill factor of 0.82. Both the inverted
metamorphic 3J and 4J cells reveal excellent component subcell
quantum efficiency. Improvements in crystal growth, as well as
more optimal subcell current density balance, will further raise
the 1-sun, AM0 efficiency of the 3J and the 4J IMM cells to 33%
and 35% respectively. The latest status of the 72-cm2 XTJ (based
on 150-mm Ge substrate) Supercells will also be discussed.
Index Terms III-V multijunction, Ge-based lattice-matched
solar cells, inverted metamorphic solar cells, space solar cells.

SuperCell, 72.44 cm2, that provide equal electrical and

mechanical performance as the XTJ LEONE cells.
SuperCell is currently undergoing engineering confidence
tests per Spectrolab heritage requirements. Over 300 XTJ
SuperCells were built thus far (Fig. 2). Electrical performance
of SuperCells from multiple MOVPE reactor systems over a 6month period showed an excellent result and met the light I-V
parameter specifications from both BOL and after radiation.
Note that the illuminated current-voltage performance
specifications are same for both LEONE (100-mm Ge-based)
and SuperCell (150-mm Ge-based) as shown in Table 1.

I. INTRODUCTION
Recent progress in III-V multijunction space solar cell
technology has led to production GaInP/GaAs/Ge triplejunction space cells with an average 1-sun efficiency close to
30% (AM0, 28C, 1-sun).
The latest lattice-matched
multijunction solar cell, XTJ, typical-size, 26.62-cm2, and
LEONE, 59.65 cm2, shown in Fig. 1 passed qualification per
AIAA S-111-2005 and AIAA S-112-2005 for CIC assembly
level qualification and manufacturing change readiness review.

Fig. 2. SuperCells max. power efficiency distribution (color

represents different cell builds).
TABLE I
XTJ CELLS LIGHT I-V PARAMETER SPECIFICATION
Parameter

Fig. 1. Evolution of XTJ space solar cells.

After XTJ LEONE cells were successfully introduced,
Spectrolab continue on the strategic cost reduction plan by
establishing a 150-mm based manufacturing line for XTJ

978-1-4673-0066-7/12/$26.00 2011 IEEE

Jsc[mA/cm2]
Voc [V]
Vload
Jload@2.31V [mA/cm2]
Jload @ 2.27V [mA/cm2]
Jmp [mA/cm2]
Vmp [V]
Eff @Pmp [%]
Eff@2.31V [%]
Eff@2.27V [%]
Fill Factor
Absorptance [Bare]
EOL]
Absorptance [CIC, 5 mil
AR]
Pm @1 MeV electron, 5E14
Pm @1 MeV electron, 1E15

2 Per XTJ (100mm)

1 Per XTJ (100mm)

SuperCell (150 mm)

(26.62cm2~32.3 cm2)

(53.3cm2~59.65cm2)

(70 cm2 ~72.44cm2)

17.76
2.633
2.310
17.14
N/A
17.02
2.348
29.54%
29.26%
N/A
0.85
0.88
0.87

17.76
2.633
2.270
N/A
17.14
17.02
2.300
28.91%
N/A
28.76%
0.84
0.88
0.87

17.76
2.633
2.270
N/A
17.14
17.02
2.300
28.91%
N/A
28.76%
0.84
0.88
0.87

0.90
0.89
0.85

0.90
0.89
0.85

0.90
0.89
0.85

003146

Future high-performance space solar cells will likely utilize

new technology pathways such as 3- to 6-junction cell
structure, highly metamorphic materials, and inverted crystal
growth to achieve the desired bandgaps for 1-sun, AM0
efficiency of 33% and beyond. Inverted metamorphic 3junction and 4-junction solar cells are envisioned to be the
next space solar cell architecture beyond the current Ge-based
cells. Fig. 3 shows a cross-section schematic of the IMM3J
and the IMM4J cells.

gr
ow Ga
th As
su
bs
tra
te

gro
wt GaA
hs s
ub
str
ate

to current Ge-based space solar cells two large-area IMM cells

can be yielded from each 100-mm substrate. Fig. 5 reveals the
illuminated current-voltage characteristic of a two-per largearea IMM3J cells tested under 1-sun, AM0 simulator
illumination. The solar simulator used was calibrated with
IMM3J LEAR Jet-flown calibration standards. One of the
IMM3J cells measured a short-circuit current density (JSC) of
16.7 mA/cm2, an open-circuit voltage (VOC) of 3.04 V, and a
fill factor of 0.84 for a 1-sun, AM0 efficiency of 31.5%.
Spectral response measurements of the IMM3J cells are shown
in Fig. 6. Note that all three subcell components of the
IMM3J cell achieved or exceeded 90% external quantum
efficiency. The current density values of the limiting subcell is
also consistent with the measured JSC data shown in Fig. 5.

metal
gridline

metal
gridline

1.95-eV AlGaInP cell 1

1.95-eV AlGaInP cell 1

1.43-eV GaAs cell 2

1.47-eV AlGaAs cell 2

transparent buffer

1.00-eV GaInAs cell 3

transparent buffer

1.05-eV GaInAs cell 3

transparent buffer

0.70-eV GaInAs cell 4

Fig. 3. Cross-sections schematic of the IMM3J and the

IMM4J space solar cells.

Spectrolabs IMM space solar cells are grown in an inverted

configuration using production MOVPE reactors on 100-mm
GaAs (or Ge) substrates. The highest bandgap alloy (Subcell
1) is grown first followed by the second highest bandgap
materials (Subcell 2), and depending on whether the space
solar cell is a 3J or a 4J structure, one or two pairs of
metamorphic buffers and the corresponding low bandgap
subcells at the buffer lattice constants are deposited. The
IMM cells discussed in this work have bandgap combinations
of 2.0, 1.4, and 1.0-eV for IMM3J cells and 2.0, 1.5, 1.1, and
0.7-eV for IMM4J cells. Subsequent wafer processing
converts the inverted cell structure to an upright configuration
and transfers the device semiconductor layers to a final cell
handle. Device fabrication processes typical of standard, highvolume space solar cell are then used to complete the device
fabrication. Cell-Interconnect-Coverglass (CICs) are then
assembled based on conditions nearly-identical to typical
production space solar cell assembly processes.
III. FIGURES AND TABLES
Fig. 4 shows a picture of two large-area (>24-cm2) IMM3J
space solar cells taken right before final cell isolation. Similar

978-1-4673-0066-7/12/$26.00 2011 IEEE

Fig. 4. Picture of a large-area (>24-cm2) IMM3J solar cells

taken right before final cell isolation.
0.50
0.45
0.40

AM0 (AX25) Calibrated with Lear

Jet Standards

0.35
0.30
Current (A)

II. EXPERIMENTAL DETAILS

Area = 26 cm2

0.25
0.20

Jsc = 16.7 mA/cm2

Voc = 3.04 V
FF = 84%

0.15
0.10
0.05

AM0 Efficiency = 31.5 %

0.00
0.0

0.5

1.0

1.5
2.0
Voltage (V)

2.5

3.0

3.5

Fig. 5. Illuminated current-voltage characteristic of a largearea IMM3J cell tested under 1-sun, AM0 simulator
illumination.
Fig. 7 plots the illuminated current-voltage characteristic of
recent Spectrolab IMM 4-junction cells. One of the IMM4J
cells measured an VOC of 3.42 V, a JSC of 15.8 mA/cm2, a fill
factor of 0.82, or a 1-sun, AM0 efficiency of 32.7%. Again,
the solar simulator was calibrated using LEAR Jet-flown

003147

or exceed quantum efficiencies of 90%. In comparison, the

highly lattice-mismatched 0.7-eV GaInAs subcells achieved an
internal quantum efficiency of over 85%.
100
90
80
70
60
QE (%)

IMM3J calibration standards. As a measure of the inverted

solar cell quality, it is desirable to achieve a small bandgapvoltage offset, [(Eg/q) VOC], such that the open-circuit
voltage of the IMM4J is as close to the device material
bandgap as possible. The overall (Eg/q) - VOC offset of the
IMM 4-junction cell was calculated to be 1.75 V or an average
offset of 437 mV per subcell. It is estimated that high
performance IMM 4-junction cells with target efficiency of
35% 1-sun, AM0 efficiency will likely have an average offset
of 413 mV per junction. The fill factor of the IMM4J cells
can also be increased with further improvement in the current
collection and the current balance among the four subcells.

50
40
30

100

20

90

10

80

0
350

60
50

550

750

950
1150
1350
Wavelength (nm)

1550

1750

Fig. 8. Internal quantum efficiency data of an IMM4J space

solar cells.

40
30
20

IV. CONCLUSION

10
0
350

550

750

950

1150

1350

Wavelength (nm)

Fig. 6 Spectral response measurements of an IMM3J space

solar cell.
0.020
0.018

Current Density (A/cm2)

EQE (%)

70

0.016

Spectrolab continues to develop high-efficiency latticematched and inverted metamorphic solar cell technology for
space and near-space applications. Large-area IMM3J space
solar cells with 1-sun, AM0 efficiency of 32% have been
demonstrated. In addition, IMM4J cells (1-cm2) have reached
a 1-sun, AM0 conversion efficiency of 33%. Improvements in
crystal growth, as well as more optimal subcell current density
balance among subcells, will further raise the 1-sun, AM0
efficiency of the IMM3J as well as the IMM4J cells to
efficiencies of 33% and 35%, respectively.

0.014
0.012

ACKNOWLEDGEMENT

0.010

The authors would like to thank the entire space

photovoltaic R&D team at Spectrolab. The financial support
from the U.S. Government, the U.S. Air Force Research
Laboratory Space Vehicles Directorate under Contracts
FA9453-09-C-0373, FA9453-04-2-0042, and Boeing Internal
Research and Development funds is gratefully appreciated.

0.008

13e-2097-4_04 3.401 V, 0.0158 A, 32.7 %, 0.823

13e-2097-4_09 3.424 V, 0.0158 A, 32.5 %, 0.811

0.006

13e-2097-4_15 3.422 V, 0.0159 A, 32.2 %, 0.803

0.004

13e-2097-4_16 3.415 V, 0.0157 A, 32.0 %, 0.808

13e-2097-4_19 3.368 V, 0.0157 A, 32.1 %, 0.823

0.002

13e-2097-4_24 3.408 V, 0.0157 A, 32.0 %, 0.812

13e-2097-4_33 3.385 V, 0.0155 A, 32.2 %, 0.828

0.000
0

Voltage(V)
Fig. 7. Illuminated current-voltage characteristic of IMM4J
cells (1-cm2) tested under 1-sun, AM0 simulator illumination
calibrated to LEAR-jet flown standards.

The internal quantum efficiency (IQE) data of the IMM4J

component subcells are presented in Fig. 8. The IQE data
showed that the two lattice-matched subcells, AlGaInP and
AlGaAs, as well as the ~1-eV MM GaInAs subcells achieved

978-1-4673-0066-7/12/$26.00 2011 IEEE

REFERENCES
[1] M. W. Wanlass, S. P. Ahrenkiel, R. K. Ahrenkiel, D. S. Albin,
J. J. Carapella, A. Duda, J. F. Geisz, S. Kurtz, T. Moriarty, R. J.
Wehrer, and B.Wernsman, Proceedings of the 31st IEEE
Photovoltaic Specialists Conference, p. 530 (2005).
[2] R. R. King, D. C. Law, C. M. Fetzer, R. A. Sherif, K. M.
Edmondson, S. Kurtz, G. S. Kinsey, H. L. Cotal, D. D. Krut, J.
H. Ermer, and N. H. Karam, Proc. 20th European Photovoltaic
Solar Energy Conference, p. 118 (2005).
[3] Yoon, Hojun; Haddad, Moran; Mesropian, Shoghig; Yen,
Jason; Edmondson, Kenneth; Law, Daniel; King, Richard R.;
Bhusari, Dhananjay; Boca, Andreea; Karam, Nasser H.,

003148

Proceedings of the 33st IEEE Photovoltaic Specialists

Conference, p. 1 (2006).
[4] J. Boisvert, D. Law, R. King, D. Bhusari, X. Liu, A. Zakaria,
W. Hong, S. Mesropian, D. Larrabee, R. Woo, A. Boca, K.
Edmondson, D. Krut, D. Peterson, K. Rouhani, B. Benedikt,
and N.H. Karam, Proceedings of the 35th IEEE Photovoltaic
Specialists Conference, p. 123 (2010).

978-1-4673-0066-7/12/$26.00 2011 IEEE

003149

340

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 2, APRIL 2013

Optimization of Subcell Interconnection for

Multijunction Solar Cells Using Switching
Power Converters
Mohammed Khorshed Alam, Student Member, IEEE, Faisal Khan, Member, IEEE, and
Abusaleh M. Imtiaz, Student Member, IEEE

AbstractA multijunction solar cell can extract higher solar energy compared to a single junction cell by splitting the solar spectrum. Although extensive research on solar cell efficiency enhancement is in place, limited research materials are available to identify the optimum interconnection of multijunction solar subcells
using power electronic circuits. Multijunction solar cells could be
grouped into two main categories: vertical multijunction (VMJ)
solar cells and lateral multijunction (LMJ) solar cells. In this paper,
a detailed study to identify the optimum interconnection method
for various multijunction solar cells has been conducted. The authors believe that the conducted research in this area is very limited, and an effective power electronic circuit could substantially
improve the efficiency and utilization of a photovoltaic (PV) power
system constructed from multijunction solar cells. A multiple input
dc-to-dc boost converter has been used to demonstrate the advantage of the proposed interconnection technique. In order to ensure
maximum power point (MPP) operation, a particle swarm optimization (PSO) algorithm has been applied needing only one MPP
control for multiple solar modules resulting in cost and complexity
reduction. The PSO algorithm has the potential to track the global
maxima of the system even under complex illumination situations.
A complete functional system with the implementation of the proposed algorithm has been presented in this paper with relevant experimental results.
Index TermsInterconnection, maximum power point tracking
(MPPT), multijunction solar cell.

I. INTRODUCTION

INGLE junction solar cells utilize a fraction of the solar

spectrum depending on the band gap of the material used.
On the other hand, multijunction (MJ) solar cells have evolved
to extract energy from a larger energy band of the solar spectrum
[1][23]. Multijunction solar cells could be grouped into two
main categories depending on the spectrum splitting technique
and orientation of the intermediate junctions: vertical multijunction (VMJ) solar cells and lateral multijunction (LMJ) solar cells
[7], as illustrated in Fig. 1. In VMJ solar cells, beam splitting is
performed by the subcells itself and the efficiency can be as high
as 40.7% under concentrated illumination [3]. VMJ solar cells

Manuscript received October 26, 2011; revised August 05, 2012; accepted
September 29, 2012. Date of publication November 16, 2012; date of current
version March 18, 2013.
The authors are with the University of Utah, Salt Lake City, UT
84112 USA (e-mail: khorshed.alam@utah.edu; faisal.khan@utah.edu;
as.imtiaz@utah.edu).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSTE.2012.2223493

Fig. 1. (a) VMJ solar cell. (b) LMJ solar cell structure.

can be either monolithically grown or mechanically stacked. In

general, VMJ solar cells are more expensive compared to single
junction cells due to the expensive fabrication steps involved in
growing layers of materials with substantial lattice mismatch.
Moreover, strain and interface defects significantly affect the
fabrication yield and performance of the VMJ solar cells [16].
Monolithically grown VMJ solar cells suffer from current
mismatch, and mechanically stacked VMJ solar cells may overcome this limitation by inserting nonconductive tunnel junctions between subcells. On the other hand, optical splitter such
as dichroic beam splitting technique is used in LMJ solar cells
in order to split the beam. The subcells in an LMJ solar cell
may be connected externally; therefore, LMJ solar cells and mechanically stacked VMJ solar cells provide greater flexibility in
terms of interconnection of subcells and material choice issues
and yield higher efficiency.
The power output from the solar cells varies with temperature, illumination, and the electric load connected to the cell.
Therefore, maximum power point tracking (MPPT) is almost
an indispensible part of an efficient solar cell system [24].
Multijunction solar cells are already very expensive, and a
photovoltaic (PV) system designed around multijunction cells
without having an MPPT system is far from an economically
viable solution. It should be noted that the interconnection
technique presented here is applicable for crystalline MJ solar
cells. Therefore, throughout this paper, the term MJ solar cell
will refer to the crystalline MJ solar cell only. A particle swarm
optimization (PSO)-algorithm-based MPPT technique has been
used in this paper where the duty cycles of the dc-to-dc converters are periodically updated with a single MPPT controller
to track the global maximum power point (MPP) of the system.
This paper presents the advancement of the concept described in [1] and [2]. In addition, a modified MPPT technique
that significantly improves the overall system performance
has been implemented. Therefore, this paper is organized in

1949-3029/$31.00 2012 IEEE

ALAM et al.: OPTIMIZATION OF SUBCELL INTERCONNECTION FOR MULTIJUNCTION SOLAR CELLS

341

the following manner: a) present state-of-the-art LMJ solar

cells and VMJ solar cells is presented in Sections II and III,
respectively; b) the proposed interconnection technique for
MJ solar cells is presented in Section IV; c) the detail of the
PSO technique is discussed in Section V; d) the simulation
and experimental results are presented in Sections VI and VII,
respectively; e) a brief comparison with conventional solutions
is shown in Section VIII; f) the effect of partial shading is
discussed in Section IX; and g) the summary of the paper is
presented in Section X.

II. LMJ SOLAR CELL INTERCONNECTION

The design of the subcells in the LMJ solar cell can be independent [7], [12][15] or interconnected [6]. Two independent solar cells with their characteristics under full spectral illumination and with split beam illumination are shown in
Fig. 2. The GaAs solar cell has higher energy band gap than
the silicon solar cell. Therefore, the GaAs solar cell is intended
to absorb the energy from the higher energy photons and Si
solar cell absorbs the lower energy photons. The amount of
power gain achieved using the beam splitting technique is also
shown in Fig. 2. An LMJ solar cell fabrication process using
Zn Cd
S Se
has been demonstrated in [6], and a spatial
composition grading is achieved across a single wafer (
,
). This resulted in a band gap from 1.7 eV (CdSe)
to 3.6 eV (ZnS).
As the beam splitting is performed by the optical instruments,
the LMJ solar cell is free from lattice and current mismatches.
The theoretical absorption efficiency could be 100% with infinite band splitting. The efficiency could be as high as 69% for
three subcells, and this efficiency figure is astonishingly 86%
for ten subcells. Because the efficiency does not increase in proportion with the number of subcells, a compromise is made between the efficiency and cost ($/W). However, the lateral splitting is more feasible in space applications if combined with solar
flux concentration techniques [14]. Due to the logarithmic relationship between the increase in efficiency and solar concentration, it is possible to achieve reasonable efficiency with the
same number of subcells even at low concentration (between
10 and 100 ) [12], and the optical splitter may be integrated
with the static solar concentrator [7]. In addition, the need for
tunnel junction is eliminated as the subcells in the LMJ solar
cell can be connected independently.
Because unique contact metallurgy is required for each element, the design-complexity could be greatly reduced by eliminating tunnel junctions. Despite these advantages, the interconnection technique for LMJ solar cell has not been optimized
yet, and this is possibly because of the lack of information exchange between the solar cell designers and researchers in the
power electronics community. In [15] and [25], an interconnection technique has been proposed where subcells constructed
from the same materials are connected in series strings, and
strings of subcells constructed from different materials are connected in parallel to achieve a nearly voltage-matched interconnection. In spite of achieving several advancements, this solution does not optimize the MPPT operation.

Fig. 2. (a) characteristics and (b) power curve of a GaAs subcell. (c)
characteristics and (d) power curve of a Si subcell. (e) Photograph of the actual
multijunction cell assembly (courtesy of Prof. R. Menon and his research group
at University of Utah).

Through the length of this paper, the interconnection issues

will be explained with greater detail. Although researchers have

342

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 2, APRIL 2013

Fig. 3. VMJ structures (a) monolithically stacked and (b) mechanically

stacked.

already fabricated LMJ cells in the laboratory environment, a

complete optimal solution with an efficient power electronic solution has not been reported yet.
III. VMJ SOLAR CELL INTERCONNECTION
VMJ solar cells can be fabricated in two different ways: either
mechanically stacked or monolithically, as shown in Fig. 3 [19],
[20]. In mechanically stacked VMJ solar cells, multiple subcells
of different materials are manufactured on separate substrates,
and they are stacked vertically resulting in a multiterminal device. Monolithically grown VMJ solar cells have a series of subcells directly grown on one substrate, and interconnected in series by electrically conductive tunnel junctions [3][5], [8], [9],
[11].
A monolithically grown VMJ solar cell can be either lattice
matched or metamorphic in design [3], [8], [9]. The metamorphic structure provides higher current matching among the different subcells in the VMJ cell structure. However, a better current matched structure has been reported in the form of an inverted metamorphic structure in [11]. Despite of the above facts,
a complete current matching is unattainable because the subcells
of different materials are connected in series. Moreover, there
are a few combinations of group IIIIV materials that are suitable for monolithic VMJ solar cells. It is apparent that the mechanically stacked structure provides more flexibility in terms
of material selection and interconnection of subcells where MPP
can be achieved for individual subcells. Therefore, this arrangement would be more efficient than the monolithic structure at the
expense of greater system complexity [19], [20]. The complex
fabrication steps involved in mechanically stacked VMJ solar
cells and recent advancement of the key fabrication processes
for monolithically stacked cells have led researchers towards
the monolithic structures. However, the mechanically stacked
VMJ solar cell is still a promising technology and the subcells
can be connected in a similar fashion as in LMJ subcells where
the greatest advantage would be the exclusion of optical splitter
which incorporates a significant cost for the LMJ structure [7].

Fig. 4. Hybrid MJ structures.

Combining the advantages of both LMJ and VMJ solar cells,

a hybrid structure has been proposed in [12], [17], [22], [23], and
[37]. In these structures, the beam splitting may be performed in
different stages, as shown in Fig. 4. The hybrid structure individually facilitates the optimization needed for the monolithically
grown VMJ solar cells and the optical elements. The efficiency
can be as high as 50% because optical losses and the fabrication
constrains are optimized in this hybrid layout [12]. Moreover,
the interconnection can also be optimized in system level where
the subcells are mechanically stacked as presented in [22].
Although, a substantial effort is in place to improve the efficiency of multijunction solar cells, only a few research initiatives were taken to identify the most efficient interconnection of the subcells using suitable power electronic converters.
Multijunction solar cells having independent subcells can be interconnected to accomplish a completely matched interconnection. In addition, subcells made from similar materials can be
connected in series to form a string to build up the voltage for
suitable power electronic conversion, and identical strings can
be connected in parallel to increase the power output.
In this paper, an MJ solar cell with four different subcells
consisting of GaInP, GaAs, GaInAsP, GaInAs have been considered, and this paper presents a completely matched interconnection in terms of terminal voltage and string current using a
multiple input dc-to-dc boost converter. Although the design has
been considered for four specific subcells, it is compatible with
any number of subcells, and can be extended for higher numbers of subcells without any apparent limitations.
IV. PROPOSED INTERCONNECTION TECHNIQUE
Solar cells in a power generation system are usually interconnected to achieve the desired open circuit voltage
and
short circuit current
compatible with the power converters
used [31]. Typically, solar cells are connected in series (typically
3650 cells) to form a string, and multiple strings are connected
in parallel to form a higher current rating module, as shown in
Fig. 5. The output voltage and current of the module are very
sensitive to solar cell mismatches and partial shading. The main

ALAM et al.: OPTIMIZATION OF SUBCELL INTERCONNECTION FOR MULTIJUNCTION SOLAR CELLS

343

TABLE I
SUBCELL MATERIAL, BAND GAP ENERGY, AND SUMMERY OF THE PROPOSED INTERCONNECTION

Fig. 5. (a) String and module formation; (b) interconnection of modules for
higher power output.

disadvantage of series connection of solar cells is the inability to

operate at optimum point during partial shading, and the performance of the module significantly degrades. As an example, if
3650 solar cells are connected in series as shown in Fig. 5, and
only one cell is partially shaded by 50%, the overall power may
decrease to 50% [38]. However, the insertion of bypass diodes
gives some advantages of protecting the shaded cell from excessive heating that may cause a permanent damage to the cell.
Usually one bypass diode is connected across every 510 solar
cells in the string [32].
Power generated from the nonshaded cells can pass across
the shaded cells through the bypass diode arrangement [41].
Although this method provides protection to a certain level
and helps to prevent drastic rise in temperature in that string,
it is not completely immune to partial shading, and therefore
suffers from degraded module performance. Several methods
to address this problem have been demonstrated in [32][34].
Although highly paralleled configurations exhibit superior performance in partial shading or fluctuating irradiance conditions,
the comparatively low open circuit voltage
or MPP
voltage
is not suitable for efficient power conversion
for power generation, and their use is typically limited to
portable or low power systems. Therefore, the interconnection
shown in Fig. 5(a) is capable of supplying small amount of
power suitable for standalone operation or charging a battery.
In order to produce higher power from a PV system, the PV

modules are interconnected in a series-parallel combination to

boost the output power, and this concept is similar to interconnecting solar cells [31], [32]. There may be number of PV
modules in series in each of number of parallel connections
in a solar panel as shown in Fig. 5. In case of a completely
matched situation, the terminal voltage would be times the
voltage across each module, the current would be
times of
each module, and the power would be
times each
modules power. However, this theoretical limit is not typically
achieved in practice. There is a diode in series of each string
of the modules to confirm unidirectional power flow, and it
prevents power flow from one string of modules to another.
These diodes introduce conduction loss and degrade the overall
efficiency.
In [22], the individual subcell characteristics of an MJ solar
cell system are given. Based on this reference and considering
a 20-cm surface area of each subcell, Table I has been formed.
If all subcells with different band-gap energies are connected in
series, current mismatch will reduce the efficiency and will result in a low short-circuit current of the system ( 0.1616 A). In
addition, connecting all subcells in parallel will cause voltage
mismatch and very low output voltage for power conversion
( 0.336 V). In order to avoid current mismatch and to achieve
higher terminal voltage, the subcells constructed from the same
material can be connected in series to form a string and the
strings with similar characteristics can be connected in
parallel without any mismatch, and therefore a module can be
formed. This interconnection will end up having completely
matched modules. Moreover, a highly parallel configuration is
desired [25] to compensate for the effect of partial shading.
Therefore, similar modules should be connected in series-parallel combination as shown in Fig. 5(b) for higher power output.
In order to ensure the MPPT operation of individual modules,
multiple dc-to-dc converters are needed to dynamically maintain the operating point of the converters when external conditions (temperature, illumination, and loading) are altered. Fig. 6
shows the proposed configuration with four modules connected
to a four input dc-to-dc boost converter ensuring global maxima
of the system, and the multiple input dc-to-dc converter shown
in this figure can be designed for any number of modules. This
feature is important for an LMJ, mechanically stacked or hybrid
MJ solar cell systems. Moreover, this configuration allows simpler gate drive circuit for the switches.
Similar interconnection systems for connecting conventional
solar panels have been discussed in [26][30], and they are
termed as multistring systems. A high efficiency (95%) multistring converter using similar multiple input dc-to-dc boost converter topology is commercially available [29], [42]. However,

344

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 2, APRIL 2013

Fig. 6. Proposed interconnection: each module is connected to a separate input of the multiple input dc-to-dc converter.

the interconnection of the subcells of MJ solar cell with suitable

power electronic solution has not been reported yet.
V. THE PSO ALGORITHM
The output characteristics of a PV array depend on several
factors such as temperature, insolation, and the electrical load
connected across the array. Therefore, an MPPT circuit is essential to maximize the power extraction from a PV array, especially for already expensive MJ solar cells. In conventional
MPP tracking circuits, separate MPP trackers are usually implemented for each dc-to-dc converters. In [24], various MPPT
techniques have been discussed with a comparative study. Most
of the MPPT circuits work with a single converter, and this proposed interconnection will require four MPPT circuits to harness power from four modules of subcells. Eventually, these
multiple MPPT circuits would require four current sensors to
localize the individual MPPs. In order to reduce the cost and
complexity of the system, the global MPP can be achieved by
using only one current sensor at the final output rather than individual sensors located at the output of each module. This can
be achieved by using advanced control algorithms. PSO is a
member of swarm intelligence (SI) and appropriate for global
optimization over an -dimensional surface of possible solutions. A PSO approach using only one MPPT instead of two
has been discussed in [36], where the voltage information of
the dc-to-dc converters is continuously updated. Moreover, PSO
has the capability of bypassing the local MPPs and tracking the
global MPP under complex illumination [36]. The PSO algorithm used here can be applied for interconnections among subcells of LMJ solar cells, mechanically stacked VMJ solar cells
and hybrid MJ solar cells.
The PSO algorithm applied in the proposed circuit updates
the duty cycle of four single input dc-to-dc converters (individual inputs of a multiple input dc-to-dc converter) connected
to the corresponding modules, and each module consists of subcells with matching characteristics. Although the subcells
in a module are electrically matched, the characteristics of

individual modules could widely vary, and the multiple input

dc-to-dc converter is used to address this variation. The duty ratios ( ) of the single input dc-to-dc converters are varied so that
the global maxima of the system can be tracked. PSO is a computational technique, where global maximum is searched by a
number of agents with a continually updated velocity controlled
by previous search of that agent and social interaction among
the agents. Therefore, every PSO implementation requires extensive and iterative computations to decide the number of required agents and the weights to be used to update the velocity
of the agents.
The flowchart of the PSO algorithm used in the proposed
system is shown in Fig. 7. Five agents are used to search the
global maxima. The initialization of the agents is shown in
Table II.
Duty cycles of an agent are grouped in a position vector
as follows:
(1)
The subscript
indicates the duty cycles of the th agent at
the th iteration and is the duty cycle of the th converter for
agent , and
is the total number of agents. Equation (2) has
been used to update the position of the th agents:

(2)
where

is the position vector corresponding to the

maximum power achieved by the agent , and

is the
position vector corresponding to the maximum power ever
achieved by any agent.
are weight constants and
are random values (0 to 1).
is the inertia coefficient
of the agent, and
and
are factors that decide how fast
the agents will move towards the agents maximum point and

ALAM et al.: OPTIMIZATION OF SUBCELL INTERCONNECTION FOR MULTIJUNCTION SOLAR CELLS

345

Fig. 8. Simulation results: (a) variation of output power versus time; (b) variation of duty ratio of the dcdc converters versus time.
TABLE III
CONSTANTS FOR PSO ALGORITHM

Fig. 7. PSO algorithm.

TABLE II
INITIAL DUTY CYCLES FOR DIFFERENT AGENTS

the global maximum point. Therefore, assigning values for

constant
is crucial for searching global maxima and
convergence of the algorithm. Variation in duty cycle is limited
to 0.2 to 0.8, and the global as well as agents maximum values
are updated once the change in the output power is larger than
the allowed maximum power variation.
VI. SIMULATION RESULTS
Simulation is performed for four solar modules in PSIM with
a functional model of solar cells as demonstrated in Table I. The
MPP voltages and currents are estimated by matching the fill
factor of the corresponding subcells. The MPPT algorithm is
written in inside PSIM using the simplified block, and the
obtained simulation results are shown in Fig. 8. In this simulation, the values of the constants are set as shown in Table III. The
parameters of different components of the power converter were
taken from the datasheet of the components used for building the
prototype. From Fig. 8(a), it is apparent that all the agents are
moving towards the peak power and they settle down near the
global maxima. The duty cycle variation of the dc-to-dc converters is also shown in Fig. 8(b).

VII. EXPERIMENTAL SETUP

As a proof of concept, four solar modules were built using
commercially available crystalline silicon solar cells to mimic a
four-junction MJ cell system. The experimental setup is shown
in Fig. 9. A 60-V 3-A Schottky diode (MBR360) was placed in
series of each of these four modules to block the reverse current
flow, and synchronous rectification topology was used to minimize the power loss across the diode of the boost converters.
IRFI4410ZPbF MOSFETs and MBR360 diodes were used as
switching devices in the proposed circuit, and an IRS2108 bootstrap gate driver provided the necessary gate signals for the
high and low side MOSFETs in each switching pair. Inductors
and
in Fig. 6 were implemented using 390 H inductors, and both
and
were 780 H in order to ensure continuous conduction. All the inductors were implemented using
2100LL-391-H-RC.
The algorithm was implemented in dSPACE 1104 system,
and the switching frequency of the converters was set at 5 kHz.
The duty ratio was updated every 0.2 s. The output voltage was
kept constant at 6.5 V using a switch mode power supply, and a
fixed resistor of 2.5 was used as the load. In this way, a dc bus
was created with two sources and a load. The output current of
the converter was measured using an LTS 6 current transducer,
and the experimental results are shown in Fig. 10. The measured voltages and currents of the solar modules are listed in
Table IV, and the maximum power delivered by all four modules was 14.28 W. It should be noted that the straight line shown
in Fig. 10(a) indicates the summation of the maximum powers

346

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 2, APRIL 2013

Fig. 10. Experimental results: (a) variation of output power versus time;
(b) variation of duty ratio of the dcdc converters versus time.

Fig. 9. (a) DCDC converters with control circuits; (b) experimental setup.

delivered by four solar modules. The efficiency of the MPPT algorithm has not been considered here, and the reason has been
explained in the later part of this section. Moreover, the simulation results shown in Fig. 8 incorporate the MJ solar cell system
summarized in Table I, and the experimental results shown in
Fig. 10 incorporate the proof of concept modules built using
the commercially available crystalline solar cells. These results
have been summarized in Table IV. The measured efficiency
of the converter with this loading was 84.6% at
steady-state.
The multiple input dc-to-dc converter shown in Fig. 6 consists
of four single input dc-to-dc boost converters, and the output
terminals of these single input boost converters are connected in
parallel. The efficiency of the multiple input dc-to-dc converter
depends on the operating points of the individual single input
dc-to-dc converters. The overall efficiency can be written as
(3)
is the efficiency of the multiple input dc-to-dc
where
converter, and is the efficiency of the th single input dc-to-dc
boost converter. and
are the input power of the th single
input dc-to-dc converter and summation of the input power of
the individual single input dc-to-dc converters, respectively.
is the output power of the multi-input dc-to-dc converter.
is the number of single input dc-to-dc converters with paralleled output.
It is possible to design a dc-to-dc boost converter so that it
operates near its peak efficiency at the MPP of the connected

TABLE IV
MEASURED VOLTAGE AND CURRENT OF THE SOLAR MODULES
DURING THE EXPERIMENT

solar module. The efficiency

of the individual single input
dc-to-dc boost converters were measured at different load currents, although the output voltage was maintained at a steady
6.5-V level. The efficiency versus load current of individual
single input dc-to-dc converters are shown in Fig. 11. The conduction loss in the series resistance of the inductor and the conduction losses of diodes and MOSFETs increase as the load
current increases. Therefore, the efficiency of individual single
input dc-to-dc converter decreases with increased load current.
Converters 3 and 4 have two inductors in series resulting in
higher overall conduction loss. Therefore, the efficiency of Converters 3 and 4 decreases more rapidly compared to Converters
1 and 2 for the same variation in the load current. From the observed results it is apparent that the overall efficiency can be
improved by a better hardware implementation.
The experiment was performed outdoor with continuously
varying illumination. Therefore, it was not possible to measure
the steady-state efficiency of the PSO algorithm when it was implemented with real solar cells. However, emulators were used
in a controlled environment to determine the efficiency of the
algorithm. Each of the solar modules listed in Table IV was emulated using a bench top power supply in series of a resistor,

ALAM et al.: OPTIMIZATION OF SUBCELL INTERCONNECTION FOR MULTIJUNCTION SOLAR CELLS

347

TABLE V
SUMMARY OF SERIES INTERCONNECTION OF ALL SUBCELLS

SUMMARY
Fig. 11. Efficiency versus load current curve of individual single input dc-to-dc
boost converters.

and the voltage and resistance of each emulator was selected

such that the MPP of this orientation matches the values shown
in Table IV. The efficiency of the PSO algorithm thus measured
using emulators was found to be 96.7%. This efficiency is highly
dependent on the value of the constants , , and , and this
MPP tracking efficiency may vary by a fair amount for the original MJ solar cell implementation. Therefore, the MPP tracking
efficiency can be maximized by selecting proper values of these
constants.
Due to the unavailability of the commercially available MJ
solar cells, experiments were performed using commercially
available crystalline silicon (single junction) solar cells. Therefore, the results obtained here may not match the dynamic behavior of the true MJ solar cell system. However, the method
presented in this paper explores the feasibility of the architecture proposed in Fig. 6which represents a true MJ PV system.
An effective implementation of this system can be realized
using many commercially available microcontrollers [44]. The
cost of the microcontroller is considered as a fixed cost and has
small contribution towards the total system cost. Moreover, a
high efficiency (95% peak) multistring power converter with
built-in MPPT is already commercially available by SMA, and
it is named the Sunny Boy Multistring [29], [42] system. The
true implementation of subcell interconnection of different MJ
solar subcells and interfacing with the power converters can be
left as a future research. It is possible to observe the dynamic
performance of the MJ solar cells with the future implementation as well.
VIII. COMPARISON WITH CONVENTIONAL INTERCONNECTION
TECHNIQUES
In order to make a fare comparison with other possible interconnections, the components of the power converter in PSIM
have been simulated as ideal ones. The resulted output voltage
and current ratings by series connecting all the subcells of each
module listed in Table I are shown in Table V. In PSIM, these
four modules are connected in series in order to build a system
where all the subcells are connected in series, and the maximum
power output from the series interconnection of all the subcells
was found to be 11.88 W. In contrast, the power output from the
proposed interconnection using PSIM simulation was found to
be 17.47 W with PSO applied. Therefore, in the proposed interconnection, it yielded 47.05% additional power compared to
the conventional nonoptimized solution. Considering 6 hours of
average insolation, it will result in approximately 681 kWh of

OF

TABLE VI
OPEN CIRCUIT VOLTAGE MATCHED INTERCONNECTION
OF ALL SUBCELLS

energy gain per year for a 1-kW unit. In the second phase, the
subcells constructed from the same material were connected in
such a way that the open circuit voltages of four modules match
the best [15], [25]. The resulted voltage and current ratings of
these modules are shown in Table VI, and these modules were
connected in parallel. The maximum power output of the system
from PSIM simulation was found to be 15.78 W, and therefore,
the proposed interconnection still provides 10.71% power gain.
This amount of efficiency enhancement is really significant in
solar power systems in critical applications such as in a battlefield of scientific expedition. The above comparisons are performed using PSIM simulations considering ideal dc-to-dc converters. In practice, the efficiency of the converter will affect the
power gain magnitudes.
IX. EFFECT OF PARTIAL SHADING
The effect of partial shading in an MJ solar cell system
might be different from multistring solar systems constructed
from single junction solar cells. The subcells of a mechanically stacked MJ solar cell or LMJ solar cell share the same
incident solar beam. Therefore, any partial shading will affect
all the subcells of the MJ cell under partial shading. As the
subcells are part of different modules, all the modules will be
affected. This phenomenon is illustrated in Fig. 12. Therefore,
partial shading will have more power degrading impact on the
MJ solar cell system discussed here compared to multistring
system of single-junction solar cells. However, most of the
concentrated PV (CPV) MJ solar cells system comes with a
concentrator system with homogenizer. The homogenizer is
used to improve flux uniformity, and it compensates for the
misalignment, reduces chromatic aberration, and compensates
for focusing errors [43]. The layout of a typical homogenizer
system has been shown in Fig. 13 which can potentially solve
the partial shading problem.
X. CONCLUSION
The efficiency of crystalline multijunction solar cells is much
higher than the single junction solar cells, and the use of MJ

348

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 2, APRIL 2013

ACKNOWLEDGMENT
The authors would like to acknowledge Prof. R. Menon of
University of Utah for his valuable suggestions regarding MJ
solar cells.
REFERENCES

Fig. 12. Effect of partial shading on proposed MJ solar cell system. Partial
shading will affect all the modules.

Fig. 13. MJ solar cell system with homogenizer in order to mitigate the effect
of partial shading.

solar cells is gradually increasing due to new material combinations, advancement in concentrator materials and optical splitting techniques, and progress in fabrication processes at the cost
of reasonably increased price. In this paper, a subcell interconnection technique for different crystalline MJ solar cells has
been presented. Due to higher $/watt, the application of these
MJ solar cells is limited to aerospace application and concentrated PV (CPV) systems [39], [40]. The ongoing research on
MJ solar cells is in place to enhance the efficiency at the cell
level. To the knowledge of the authors, there is no significant
work done on the optimal subcell interconnection of MJ solar
cells with efficient power electronic solution. The identification
of the optimum subcell interconnection for multijunction solar
cells using suitable power electronic circuit is the main consideration of this work. It has been shown that the proposed interconnection will result in achieving a completely matched subcell
interconnection system for different multijunction solar cells.
The possible partial shading scenarios have also been discussed
with tentative solutions.

[1] M. K. Alam, F. H. Khan, and A. M. Imtiaz, Interconnection and

optimization issues of multijunction solar cellsA new mitigation
approach using switching power converters, in Proc. IEEE Applied
Power Electronics Conf. (APEC), Feb. 2012, pp. 583589.
[2] M. K. Alam, F. H. Khan, and A. S. Imtiaz, An efficient power electronics solution for lateral multi-junction solar cell systems, in Proc.
IEEE IECON 2011, pp. 43734378.
[3] R. R. King, A. Boca, W. Hong, D. Law, G. Kinsey, C. Fetzer, M.
Haddad, K. Edmondson, H. Yoon, P. Pien, and N. Karem, High-efficiency multijunction photovoltaics for low-cost solar electricity, in
Proc. Ann. Meeting IEEE Lasers and Electro-Optics Society, Nov.
2008, pp. 23.
[4] I. Bhattacharya and S. Y. Foo, Effects of gallium-phosphide and indium-gallium-antimonide semiconductor materials on photon absorption of multijunction solar cells, in Proc. 1st Asia Symp. Quality Electronic Design, 2009 (ASQED 2009), pp. 237241.
[5] J. F. Wheeldon, C. E. Valdivia, A. Walker, G. Kolhatkar, T. J. Hall, and
K. Hinzer et al., AlGaAs tunnel junction for high efficiency multijunction solar cells: Simulation and measurement of temperature-dependent operation, in Proc. 34th IEEE Photovoltaic Specialists Conf.
(PVSC), 2009, pp. 106111.
[6] C. Z. Ning, A. L. Pan, and R. B. Liu, Spatially composition-graded
alloy semiconductor nanowires and wavelength specific lateral-multijunction full-spectrum solar cells, in Proc. 34th IEEE Photovoltaic
Specialists Conf. (PVSC), 2009, pp. 14921495.
[7] A. G. Imenes and D. R. Mills, Spectral beam splitting technology for
increased conversion efficiency in solar concentrating systems: A review, Solar Energy Mater. Solar Cells, vol. 84, no. 14, pp. 1969,
2004.
[8] T. Takamoto, T. Agui, and A. Yoshida et al., Worlds highest efficiency triple-junction solar cells fabricated by inverted layers transfer
process, in Proc. Photovoltaic Specialists Conf. (PVSC), 2010, pp.
412417.
[9] R. K. Jones, P. Hebert, P. Pien, R. R. King, D. Bhusari, R. Brandt, O.
A. Taher, C. Fetzer, J. Ermer, A. Boca, D. Larrabee, X. Q. Liu, and
N. Karam, Status of 40% production efficiency concentrator cells at
Spectrolab, in Proc. IEEE Photovoltaic Specialists Conf. (PVSC), Jun.
2010, pp. 189195.
[10] M. Yamaguchi, H. Suzuki, Y. Ohshita, N. Kojima, and T. Takamoto,
Recent R&D topics on concentrator multi-junction solar cells and materials under innovative solar cellss project, in Proc. 35th IEEE Photovoltaic Specialists Conf. (PVSC), 2010, pp. 237242.
[11] J. Boisvert, D. Law, R. King, D. Bhusari, X. Liu, A. Zakaria, W. Hong,
S. Mesropian, D. Larrabee, R. Woo, A. Boca, K. Edmondson, D. Krut,
D. Peterson, K. Rouhani, B. Benedikt, and N. H. Karam, Development of advanced space solar cells at Spectrolab, in Proc. 35th IEEE
Photovoltaic Specialists Conf. (PVSC), 2010, pp. 123127.
[12] A. Barnett, C. Honsberg, D. Kirkpatrick, S. Kurtz, D. Moore, and D.
Salzman et al., 50% efficient solar cell architectures and designs, in
Proc. IEEE 4th World Conf. Photovoltaic Energy Conversion, 2006,
pp. 25602564.
[13] X. Wang, N. Waite, P. Murcia, K. Emery, M. Steiner, and F. Kiamilev
et al., Improved outdoor measurements for very high efficiency solar
cell sub-modules, in Proc. 34th IEEE Photovoltaic Specialists Conf.
(PVSC), 2009, pp. 409414.
[14] W. Blocker, High-efficiency solar energy conversion through flux
concentration and spectrum splitting, Proc. IEEE, vol. 66, no. 1, pp.
104105, Jan. 1978.
[15] C. R. Lewis, W. M. Phillips, V. B. Shields, and P. M. Stella,
Multi-bandgap high efficiency converter (RAINBOW), in Proc.
IECEC 1997, pp. 401406.
[16] J. H. Karp and J. E. Ford, Multiband solar concentrator using transmissive dichroic beamsplitting, in Proc. SPIE, 2008, vol. 7043, pp.
70430F70430F-8.
[17] A. Barnett, D. Kirkpatrick, C. Honsberg, D. Moore, M. Wanlass, and
K. Emery et al., Very high efficiency solar cell modules, Prog. Photovoltaics: Res. Applicat., vol. 17, pp. 7583, 2009.

ALAM et al.: OPTIMIZATION OF SUBCELL INTERCONNECTION FOR MULTIJUNCTION SOLAR CELLS

[18] M. A. Green and A. Ho-Baillie, Forty three per cent composite splitspectrum concentrator solar cell efficiency, Prog. Photovoltaics: Res.
Applicat., vol. 18, pp. 4247, 2010.
[19] L. Zhao, High Efficiency Mechanically Stacked Multi-Junction
Solar Cells for Concentrator Photovoltaics, Ph. D. dissertation,
Katholieke Universiteit LeuvenFaculteit Ingenieurswetenschappen
Arenbergkasteel, Belgium, Mar. 2011.
[20] L. L. Kazmerski, Solar photovoltaics R&D at the tipping point: A
2005 technology overview, J. Electron Spectroscopy and Related Phenomena, vol. 150, pp. 105135, 2006.
[21] National Renewable Energy Laboratory News Release NR-1907 [Online]. Available: http://www.nrel.gov/news/press/2007/524.html
[22] J. D. M. Cambridge, M. A. Steiner, B. L. Unger, K. A. Emery, E. L.
Christensen, and M. W. Wanlass et al., Compact spectrum splitting
photovoltaic module with high efficiency, Prog. Photovoltaics: Res.
Applicat., vol. 19, no. 3, pp. 352360, May 2011.
[23] X. Wang and A. Barnett, One lateral spectrum splitting concentrator
photovoltaic architecture: Measurements of current assemblies and
analysis of pathways to 40% efficient modules, in Proc. 35th IEEE
Photovoltaic Specialists Conf. (PVSC), 2010, pp. 27452750.
[24] T. Esram and P. L. Chapman, Comparison of photovoltaic array maximum power point tracking techniques, IEEE Trans. Energy Convers.,
vol. 22, no. 2, pp. 439449, Jun. 2007.
[25] A. L. Lentine, G. N. Nielson, M. Okandan, W. C. Sweatt, J. L. CruzCampa, and V. Gupta, Optimal cell connections for improved shading,
reliability, and spectral performance of microsystem enabled photovoltaic (MEPV) modules, in Proc. 35th IEEE Photovoltaic Specialists
Conf. (PVSC), 2010, pp. 30483054.
[26] L. Gao, R. A. Dougal, S. Liu, and A. P. Iotova, Parallel-connected
solar PV system to address partial and rapidly fluctuating shadow conditions, IEEE Trans. Ind. Electron., vol. 56, no. 5, pp. 15481556,
May 2009.
[27] M. Calais, J. Myrzik, T. Spooner, and V. G. Agelidis, Inverters for
single-phase grid connected photovoltaic systemsAn overview, in
Proc. Power Electronics Specialists Conf., 2002, pp. 19952000.
[28] J. M. A. Myrzik and M. Calais, String and module integrated inverters
for single-phase grid connected photovoltaic systemsA review, in
Proc. Power Tech Conf., 2003, vol. 2, p. 8.
[29] M. Meinhardt, G. Cramer, B. Burger, and P. Zacharias, Multi-stringconverter with reduced specific costs and enhanced functionality,
Solar Energy, vol. 69 (Suppl.), no. 16, pp. 217227, 2000.
[30] S. A. Khajehoddin, A. Bakhshai, and P. Jain, A novel topology and
control strategy for maximum power point trackers and multi-string
grid-connected PV inverters, in Proc. APEC 2008, pp. 173178.
[31] M. Meinhardt and G. Cramer, Past, present and future of grid connected photovoltaic- and hybrid-power-systems, in Proc. IEEE-PES
Summer Meeting, 2000, vol. 2, pp. 12831288.
[32] P. A. Lynn, Electricity From Sunlight an Introduction to Photovoltaics,
1st ed. Hoboken, NJ: Wiley, 2010.
[33] L. Gao, R. A. Dougal, S. Liu, and A. P. Iotova, Parallel-connected
solar PV system to address partial and rapidly fluctuating shadow conditions, IEEE Trans. Ind. Electron., vol. 56, no. 5, pp. 1548156, May
2009.
[34] L. Gao, R. A. Dougal, S. Liu, and A. Jotova, Portable solar systems
using a step-up power converter with a fast-speed MPPT and a parallelconfigured solar panel to address rapidly changing illumination, in
Proc. APEC 2007, pp. 520523.
[35] N. E. Ball, Parallel and Virtual Parallel Interconnection of Solar Cells
in Solar Panel, U.S. Patent 7521630, Apr. 21, 2009.
[36] M. Miyatake, M. Veerachary, F. Toriumi, N. Fujii, and H. Ko, Maximum power point tracking of multiple photovoltaic arrays: A PSO
approach, IEEE Trans. Aerosp. Electron. Syst., vol. 47, no. 1, pp.
367380, Jan. 2011.
[37] B. Mitchell, G. Peharz, G. Siefer, and M. Peters et al., Four-junction spectral beam-splitting photovoltaic receiver with high optical efficiency, Prog. Photovoltaics: Res. Applicat., vol. 19, no. 1, pp. 6172,
2011.
[38] N. E. Ball, Parallel and Virtual Parallel Interconnection of Solar Cells
in Solar Panels, U.S. Patent 7,521,630 B2, Apr. 2009.
[39] D. Crisp, A. Pathareb, and R. C. Ewell, The performance of gallium
arsenide/germanium solar cells at the Martian surface, Acta Astronautica, vol. 54, no. 2, pp. 83101, Jan. 2004.

349

[40] NREL, Concentrating Photovoltaic Technology [Online]. Available:

http://www.nrel.gov/csp/concentrating_pv.html
[41] C. Deline, Partially shaded operation of a grid-tied PV system, in
Proc. IEEE PVSC, Jun. 2009, pp. 001268001273.
[42] Sunny Boy 5000TL Multi-String [Online]. Available: http://www.esedracoop.it/documenti/rete/sunny_5000.pdf
[43] D. C. Miller, M. D. Kempe, C. E. Kennedy, and S. R. Kurtz, Analysis
of transmitted optical spectrum enabling accelerated testing of CPV designs, Proc. SPIE, pp. 74070G74070G-12, 2009, DOI: 10.1117/12.
826251.
[44] Texas Instruments [Online]. Available: http://www.ti.com/lit/sg/
sprb176l/sprb176l.pdf

Mohammed Khorshed Alam (S11) received the

B.S. degree in electrical and electronic engineering
from Bangladesh University of Engineering and
Technology, Dhaka, Bangladesh, in 2009. Since
2011, he is working toward the Ph.D. degree in
electrical engineering at the University of Utah, Salt
Lake City.
He was Lecturer at Ashanullah University of Science and Technology, Dhaka, Bangladesh from 2009
to 2010. His research interest includes renewable energy harvesting, high power switched-capacitor converters, fault detection, and reliability analysis of circuits.

Faisal Khan (S01M07) received the B.Sc., M.S.,

and Ph.D. degrees from Bangladesh University of
Engineering and Technology, Arizona State University, and University of Tennessee, Knoxville, in
1999, 2003, and 2007, respectively, all in electrical
engineering.
From 2007 to 2009, he has been with Electric
Power Research Institute (EPRI) as a senior power
electronics engineer. Since 2009, he is with the
Electrical and Computer Engineering Department,
University of Utah as an Assistant Professor. His
major area of interest is high-power capacitor-clamped converters. However,
since his appointment at the university, he has extended his research into the
field of power converter reliability prediction and cell level power converter
design for photovoltaics, especially multijunction solar cells. In addition, he is
also involved with renewable energy research including wind energy harvesting
using split-phase induction generators and grid-tied energy storage.
Prof. Khan is a member of the IEEE Power Electronics Society, Industry Applications Society, and Industrial Electronics Society. He is the recipient of the
2007 IEEE IAS first prize paper award for his contribution to high power modular multilevel dcdc converters. He is the award chair of IEEE ECCE 2012 and
the general chair of IEEE COMPEL 2013 in Salt Lake City.

Abusaleh M. Imtiaz (S11) received the B.S.

degree in electrical engineering from Bangladesh
University of Engineering and Technology (BUET),
Dhaka, Bangladesh, in 2009. He is currently working
toward the Ph.D. degree in electrical engineering at
University of Utah.
He is the author or a coauthor of more than
10 peer reviewed publications including journals
and conference proceedings. His research interests
include fabrication and characterization of the
state-of-the-art power semiconductor devices and
solar cells, energy management of plug in hybrid vehicles and design of power
converters for renewable energy sources.

548

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

Application of Si Nanowires Fabricated by

Metal-Assisted Etching to Crystalline Si Solar Cells
Mustafa Kulakci, Frat Es, Baris Ozdemir, Husnu Emrah Unalan, and Rasit Turan

AbstractReflection and transmission through a solar cell can

be significantly reduced using light-trapping structures. This approach can be applied to both crystalline and thin-film solar cells
to improve the light absorption and conversion efficiency of the
cell. In this study, vertically aligned Si nanowires were fabricated
over a large area via a metal-assisted etching technique. Following
a detailed parametric study, nanowires were applied to industrialsize (156 mm 156 mm) Si solar cells. The reflectivity from the
device surface was reduced to less than 5% for the entire visible
spectrum (350750 nm), including the blueviolet region. Standard
solar cell fabrication procedures were employed to fabricate cells
with and without Si nanowires, and the results showed that the
efficiencies of solar cells with nanowires were similar to those of
standard pyramid-textured cells, revealing the potential of the proposed concept. A systematic study of the dependence of the solar
cell parameters on the length of the nanowires was performed. The
quantum efficiency of the cells exhibited relatively poor performance in the blue-ultraviolet range of the spectrum, and enhancement in carrier generation was observed in the red-infrared region
especially for shorter nanowires.
Index TermsLight management, monocrystalline, nanowires,
silicon solar cells, texturing.

I. INTRODUCTION
URRENTLY, crystalline Si solar cells dominate the photovoltaic market. However, high material and processing
costs are still the most significant drawbacks of these systems.
To reduce the cost and improve the overall efficiency of the
cell, new-design-based thinner silicon wafers with improved
light-trapping structures should be used. Because Si is an indirect bandgap material, when the wafer thickness is less than
200 m, Si wafers are not capable of absorbing the entire solar
spectrum. To overcome this problem and improve the efficiency
of the cell, absorption throughout the cell must be improved by

Manuscript received August 5, 2012; revised October 11, 2012 and

October 28, 2012; accepted October 31, 2012. Date of current version December 19, 2012. This work was supported by the State Planning Organization of
Turkey under Contract BAP-08-11-DPT-2009-K121030 and the Distinguished
Young Scientist Award of the Turkish Academy of Sciences.
M. Kulakci and F. Es are with the Center for Solar Energy Research and
Applications, Middle East Technical University, 06800 Ankara, Turkey (e-mail:
kulakcimustafa@gmail.com; esfirat@metu.edu.tr).
B. Ozdemir and H. E. Unalan are with the Center for Solar Energy
Research and Applications and the Department of Metallurgical and Materials
Engineering, Middle East Technical University, 06800 Ankara, Turkey (e-mail:
ozbaris@metu.edu.tr; unalan@metu.edu.tr).
R. Turan is with the Center for Solar Energy Research and Applications and
the Department of Physics, Middle East Technical University, 06800 Ankara,
Turkey (e-mail: turanr@metu.edu.tr).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JPHOTOV.2012.2228300

increasing the path length of the light and the number of scattering events of the incident photons by trapping. Surface texturing
and antireflection (AR) coatings have been traditionally used for
such purposes [1], [2].
In recent years, nanostructures such as Si nanowires have attracted considerable attention due to their unique properties and
functionalities. Si nanowires can be vertically aligned and arranged in integrated arrays on the solar cell surface, which leads
to remarkable light-trapping strength with reflectivity values
of less than 1% around 500 nm [3]. The light-trapping ability of Si nanowires is particularly superior in the blueviolet
part of the spectrum compared with conventional pyramidal
textures [3]. Several different methods for the synthesis of Si
nanowire arrays have been employed, such as vaporliquid
solid [4], laser ablation [5], direct reactive ion etching [6], and
metal-assisted etching (MAE) [7][9]. Among these, MAE is
the most appealing approach, because it can be conducted at
room temperature and does not require expensive production
equipment. Thus, the nanowire formation via MAE contributes
little to the overall processing costs. Moreover, contrary to the
conventional nanowire fabrication methods, doping is not an issue in the MAE method. It is because the nanowires are etched
down from bulk Si, where the doping level and growth direction
of the fabricated Si nanowires would be the same as the starting wafer. Thus, cost-effective Si nanowire fabrication methods
that enable large-scale and large-area production in a controlled
manner could be adapted to the industrial-scale fabrication of solar cells. Some reports focusing on the photovoltaic properties
of Si nanowire solar cells have already been published [10]
[13]. However, realization of a full-scale solar cell has not been
demonstrated yet. In addition to the light-trapping applications,
Si nanowires can be applied to several different types of innovative devices, including radial p-n junctions [14]. In a recent
work, we have already demonstrated a remarkable improvement
in the photovoltaic conversion efficiency in the heterojunction
solar cells employing nanowires on the surface [15].
In this study, we fabricated industrial-size (156 mm
156 mm, with an area of 239 cm2 ) monocrystalline Si solar
cells with vertically aligned nanowire arrays formed on their
surface by MAE. To the best of our knowledge, this is the first
demonstration of the utilization of nanowires in a solar cell of
commercial size in this scale. The effect of the Si nanowire
length on the solar cell characteristics was worked out in a systematic way. The light-trapping and antireflective properties of
the Si solar cells were investigated and compared with that of the
conventional texturing technique. External quantum efficiency
(EQE) measurements were also conducted for comparison. The
technical procedures presented herein are fully compatible with
the present monocrystalline Si solar cell production lines.

2156-3381/$31.00 2012 IEEE

KULAKCI et al.: APPLICATION OF SI NANOWIRES FABRICATED BY METAL-ASSISTED ETCHING TO CRYSTALLINE SI SOLAR CELLS

549

II. EXPERIMENTAL DETAILS

For solar cell fabrication, Czochralski-grown, 200-m-thick,
p-type Si wafers in (1 0 0) crystallographic direction, with a resistivity of 4 cm and with dimensions of 156 mm 156 mm,
were used. To eliminate residual surface damage induced by
mechanical sawing, fabrication began with alkaline-based saw
damage etching. Subsequently, hydrochloric acid (HCl) and hydrofluoric acid (HF) were used to remove metallic contaminants and silicon oxide formed on the surface, respectively. Two
separate groups of substrates were prepared to compare the
effect of conventional surface texturing and nanowire formation on the solar cell performance. Pyramidal surface texturing
was conducted on wafers in the first group. This procedure
was performed in hot potassium hydroxide (KOH)/isopropanol
solution, which facilitated the formation of random pyramids.
Subsequently, the aforementioned cleaning and KOH treatment
were performed. The second group of wafers was not subjected
to an alkaline texturing process. Instead, vertically aligned Si
nanowire arrays were formed on the surface by MAE, as described elsewhere [3]. In brief, an etching solution of 0.02 M
silver nitrate (AgNO3 ) and 4.6 M HF was prepared at 20 C,
and the wafers were kept in this solution for 1050 min, depending on the desired length of the nanowires. At the end of
the process, silver dendrite layer was removed using nitric acid
(HNO3 ). After pyramidal texturing and nanowire formation, the
emitter layer was formed by phosphorus diffusion, resulting in
a sheet resistance of 50 / on a standard wafer. Following the
removal of phosphorus glass through dipping the substrates into
a 5% HF solution, an 80-nm-thick silicon nitride (Si3 N4 ) layer
was deposited via plasma-enhanced chemical vapor deposition
for surface passivation. The silicon nitride layer also acts as an
AR coating. The front silver (Ag) and back aluminum (Al) metal
contacts were screen printed and cofired. Finally, edge isolation
was performed by laser scribing.
The resulting surface-texturing features were observed by
scanning electron microscopy (SEM). Cross-sectional SEM images were obtained from the cleaved edges of the Si substrates.
Reflection measurements were conducted through a calibrated
integrated sphere using a Si detector. Device characterization
was performed by measuring the currentvoltage characteristics
in the dark and under solar simulated light (AM1.5G). Minority
carrier lifetime measurements were made using a Sinton lifetime
tester. EQE measurements were performed using an apparatus
equipped with a monochromator, chopper, lock-in amplifier system, and calibrated light detector.
III. RESULTS AND DISCUSSION
A. Nanowire Synthesis on Crystalline Si Surfaces
MAE process was applied to monocrystalline wafers for Si
nanowire formation. In this process, homogeneously dispersed
Ag ions extracted from the solution come into contact with
the substrate surface, where galvanic reactions occur simultaneously. As a result, Si diffuses into the solution as silicon
hexafluoride (SiF6 ), leaving pits on the wafer surface. These
reactions only occur on areas where etching was initiated. Due

Fig. 1. Cross-sectional SEM images of Si nanowires on mono-c cells fabricated at an etching time of (a) 10 min, (b) 20 min, (c) 30 min, (d) 40 min, and
(e) 50 min. (f) Nearly linear relationship was observed between the etching time
and nanowire length.

to the isotropic and selective growth of holes along the surface,

the remaining structure is transformed into vertically aligned
arrays of Si nanowires. We have extensively studied the relationship between the length and distribution of Si nanowires
and determined their relationship with the etching time, solution temperature, concentration, wafer orientation, resistivity,
and doping type [3].
In this study, we incorporated nanowire formation as a batch
process into a standard crystalline Si solar cell fabrication line,
and samples with five different nanowire lengths were studied.
As shown in the SEM images (see Fig. 1), vertically aligned and
well-controlled Si nanowire arrays were formed on the device
surface. Detailed SEM analysis revealed that the length, shape,
and distribution of the nanowires were uniform for all of the
wafers. SEM images of nanowires fabricated with etching times
of 10, 20, 30, 40, and 50 min are shown in Fig. 1(a)(e), respectively. The nanowire length as a function of the etching time is
plotted in Fig. 1(f). In most cases, the diameter of the nanowires
was within the range of 50200 nm.

B. Reflection From the Si Surface

The optical reflection spectra and AM1.5G weighted effective reflectivities of pyramid/nanowire-textured wafers before

550

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

Fig. 3. Images of fully processed solar cells. A significant color change was
observed between the conventionally textured cells (a) and the nanowiretextured cells etched for 50 min (b) due to enhanced absorption at wavelengths
less than 450 nm.

Fig. 2. Reflection spectra and AM1.5G weighted effective reflectivities of

samples (a) before and (b) after Si3 N4 deposition.

and after Si3 N4 deposition are shown in Fig. 2(a) and (b),
respectively.
As the nanowire length increased, the reflectivity of the wafers
decreased accordingly. When an AR coating was not present on
the surface, all of the nanowire-coated wafers exhibited lower
optical reflection than the standard pyramid-textured wafer, except for the wafer fabricated at an etching time of 10 min.
However, following AR deposition, the pyramid-textured wafer
revealed lowest reflectance for the wavelengths greater than
450 nm. In contrast, nanowire-textured wafers did not show
significant differences following AR deposition, except for the
nanowires fabricated with an etching time of 10 min. Therefore, the use of Si3 N4 layer on Si nanowires can have a passivation effect without any significant AR property. For the
wavelengths lower than 450 nm, the samples with nanowires
exhibited low reflectivity values and excellent light-trapping
properties. This feature was evidenced in the color of the fully
processed monocrystalline Si solar cells, as shown in Fig. 3. The
surface of cells was black due to the presence of nanowires.
C. Solar Cell Device Performance
The sample with the shortest nanowire arrays (formed with
an etching time of 10 min) revealed similar performance with
the pyramid-textured device. Although further optimization is

Fig. 4. IV characteristics of different monocrystalline samples under standard testing conditions (AM1.5G, 100 W/cm2 , 25 C).

needed to improve the performance, our results prove the successful implementation of Si nanowires to standard solar cells
at an industrial scale.
Currentvoltage (IV) characteristics of the solar cells are
provided in Fig. 4. The solar cell device parameters, such as the
photovoltaic conversion efficiency, fill factor (FF), short-circuit
current density Jsc , series resistance, and open-circuit voltage
Vo c extracted from the IV curves, are shown in Fig. 5.
Jsc and Vo c decreased with an increase in the nanowire length,
leading to a reduction in the overall power conversion efficiency.
These results can be attributed to an increase in surface recombination along the nanowire surfaces. Due to MAE etching, broken/dangling Si bonds were generated on the nanowire surfaces,
which act as trap energy levels for photogenerated carriers. The
presence of broken Si bonds becomes more of an issue for longer
nanowires. It is because the number of recombination sites increases with the nanowire length with an overall increase in the
total surface area.
In order to get more insight into the current mechanisms, we
have measured dark IV characteristics of the solar cells, not
shown here. We have determined the ideality factors and the
saturation currents and seen that the ideality factor of the cells
is relatively good for the samples with pyramid texture (ideality
factor n of 2.03 and saturation current Jo of 7 107 A/cm2 )

KULAKCI et al.: APPLICATION OF SI NANOWIRES FABRICATED BY METAL-ASSISTED ETCHING TO CRYSTALLINE SI SOLAR CELLS

551

Fig. 7. EQE of samples with different nanowire lengths compared to those of

a pyramid-textured cell.

Fig. 5. Effect of nanowire length on various solar cell parameters. The pyramid textured solar cell parameters are inserted as a reference (red horizontal
lines) on the efficiency and Jsc graphs.

Fig. 6. Effective minority carrier lifetimes of undoped samples coated with

SiN. Lines are for visual aid.

and short nanowires (n = 1.92 and Jo = 6 107 A/cm2 ). We

believe that the recombination-generation current results in the
high ideality factors in the samples with longer nanowires. This
is indeed supported by the systematic increase in the current
with the nanowire length for low forward bias voltage where the
recombination generation current is important.
The effect of the Si nanowire length on the minority carrier
lifetime was also correlated with the solar cell parameters. Fig. 6

shows the effective minority carrier lifetimes of the samples.

Due to increased surface recombination, lower minority carrier
lifetimes were obtained for solar cells with longer Si nanowire
arrays. The observed decrease in the minority carrier lifetime
led to a reduction in the VOC values as the nanowire length
increased. Better cell performance is expected with improved
passivation of the nanowire surfaces.
Although the cells fabricated for an etching time of 20 min or
longer exhibited low reflectance values, the Isc values obtained
from these cells were low compared with those fabricated for
an etching time of 10 min and compared with alkaline-textured
cells. This result was attributed to one or more of the following
possible effects: 1) Contact formation over long nanowires was
poor; 2) bulk absorption for wafers with longer wires was lower
due to decreased bulk thickness; or 3) the blue response of the
cells was poor due to high recombination at the surface of long
nanowires.
To investigate the effect of nanowire length on the Isc values and the dependence of the solar cell performance on the
wavelength, EQE measurements were conducted, and results
are provided in Fig. 7. The EQE results were in good agreement
with the solar cell parameters presented previously. However,
response of the nanowire-textured cells decreased rapidly within
the visible range, indicating that photogenerated carriers were
readily lost in this region of the spectrum. This is because the
visible part of the solar radiation was absorbed at the uppermost
region of the nanowires, and carriers generated in this area must
be transferred through the total length of the nanowires in order
to reach the bulk region where the junction was formed.
Since the nanowires possess a high surface-to-volume ratio,
the diffusion length of the carriers was low due to high surface recombination. Although the AR properties were superior
for blue light (see Fig. 3), the absorbed photons were not effectively used for the generation of photocurrent in solar cells
with nanowires. It is clear that the improved surface passivation
and better control over nanowire doping will lead to better cell
performance in the visible region of the spectrum.

552

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 1, JANUARY 2013

As revealed from the SEM images, although Ag metal paste

formed the proper metallization on the standard solar cell, it
did not adhere properly to the solar cells with nanowires due
to the low effective contact area and hydrophobic nature of the
nanowires. The poor adhesion of the paste could be the main
reason for the observed reduction in the FF with an increase in
the nanowire length (see Fig. 5). The series resistance, which
was calculated from the IV curves, is plotted in Fig. 5. As
seen in the figure, the series resistance increased with nanowire
length, which is in agreement with the SEM results. Thus, the
observed increase in the series resistance can be attributed to
poor emitter contact formation in standard screen print metallization. Hence, the generated carriers must travel along the
nanowires to reach the front contact, which increases the series resistance. To improve the device characteristics, effective
charge collection through optimized electrical contacts with new
metallization routes and materials should be investigated. The
buried contact approach could solve the metallization problem.
IV. CONCLUSION

Fig. 8. SEM images revealing front-side Ag contact formation on (a) a surface

etched with saw damage and (b) and (c) a nanowire-textured surface. The Ag
finger sticks to the surface properly in the cell shown in (a), while the finger
does not diffuse in between the nanowires, as shown in (b) and (c). Rather, as
shown in (b) and (c), the finger floats on the surface.

In this study, nanowires produced by the MAE method have

been applied to industrial-size (156 mm 156 mm) monocrystalline Si wafers, and fully operational solar cells have been
fabricated. It has been demonstrated that the nanowire surface
texturing with this process can be successfully integrated into
the conventional solar cell production line. With an optimized
production process, Si nanowires could be effectively used as
a light-trapping layer. However, this excellent light-trapping
behavior did not improve the solar cell efficiency. The photon
conversion efficiency and EQE values decreased with an increase in the nanowire length due to the carrier recombination
on the surface of the nanowires. The charge collection efficiency
can be improved through the optimization of the surface passivation, dopant density, and nanowire distribution. Screen print
metallization using standard Ag paste led to formation of a poor
contact on top of the nanowires due to the small contact area.
Thus, a new metallization scheme, possibly with a more suitable
paste material, must be developed to improve metallization on
the nanowire-structured surfaces.
Besides the conventional Si homojunction solar cell applications, nanowires fabricated through the MAE method would offer advantages in c-Si/a-Si:H heterojunction solar cells through
the introduction of a 3-D p-n junction in the form of core/shell
structure. This also orthogonalizes the photon absorption and
carrier transport. Therefore, the heterojunction structure could
significantly improve the device efficiency.
REFERENCES

The quality of the metal contact is important for efficient

charge collection in solar cells. Due to the nanostructured surface of the cells, a new approach for the metallization scheme
was required. For this reason, metal contact formation was studied by cross-sectional SEM images of saw damage-etched samples and samples with nanowires, as shown in Fig. 8 (a)(c),
respectively.

[1] M. A. Green, Z. Jianhua, A. Wang, and S. R. Wenham, Very high efficiency silicon solar cells-science and technology, IEEE Trans. Electron
Devices, vol. 46, no. 10, pp. 19401947, Oct. 1999.
[2] J. Zhao, A. Wang, and M. A Green, 245% efficiency silicon PERT cells
on MCZ substrates and 247% efficiency PERL cells on FZ substrates,
Progr. Photovoltaics: Res. Appl., vol. 7, pp. 471474, 1999.
[3] B. Ozdemir, M. Kulakci, R. Turan, and H. E. Unalan, Effect of electroless etching parameters on the growth and reflection properties of silicon
nanowires, Nanotechnol., vol. 22, pp. 155606-1155606-7, 2011.

KULAKCI et al.: APPLICATION OF SI NANOWIRES FABRICATED BY METAL-ASSISTED ETCHING TO CRYSTALLINE SI SOLAR CELLS

[4] T. Stelzner, M. Pietsch, G. Andra, F. Falk, E. Ose, and S. Christiansen,

Silicon nanowire-based solar cells, Nanotechnol., vol. 19, pp. 2952031295203-4, 2008.
[5] A. M. Morales and C. M. Lieber, A laser ablation method for the synthesis of crystalline semiconductor nanowires, Science, vol. 279, pp. 208
211, 1998.
[6] Y. Q. Fu, A. Colli, A. Fasoli, J. K. Luo, A. J. Flewitt, A. C. Ferrari, and
W. I. Milne, Deep reactive ion etching as a tool for nanostructure fabrication, J. Vac. Sci. Technol. B: Microelectron. Nanometer Struct., vol. 27,
no. 3, pp. 15201526, 2009.
[7] K.-Q. Peng, Y.-J. Yan, S.-P. Gao, and J. Zhu, Synthesis of large-area
silicon nanowire arrays via self-assembling nanoelectrochemistry, Adv.
Mater., vol. 14, no. 16, pp. 11641167, 2002.
[8] K. Peng, Y. Xu, Y. Wu, Y. Yan, S. T. Lee, and J. Zu, Aligned singlecrystalline Si nanowire arrays for photovoltaic applications, Small, vol. 1,
no. 11, pp. 10621067, 2005.
[9] V. Sivakov, G. Andra, A. Gawlik, A. Berger, J. Plentz, F. Falk, and
S. H. Christiansen, Silicon nanowire-based solar cells on glass: Synthesis,
optical properties, and cell parameters, Nano Lett., vol. 9, pp. 15491554,
2009.
[10] C. Chen, R. Jia, H. Yue, H. Li, X. Liu, D. Wu, W. Ding, T. Ye, S. Kasai,
H. Tamotsu, J. Chu, and S. Wang, Silicon nanowire-array-textured solar
cells for photovoltaic application, J. Appl. Phys., vol. 108, pp. 0943181094318-5, 2010.

553

[11] H. Yue, R. Jia, C. Chen, W. Ding, Y. Meng, D. Wu, D. Wu, W. Chen,

X. Liu, Z. Jin, W. Wang, and T. Ye, Antireflection properties and solar
cell application of silicon nanostructures, J. Vac. Sci. Technol. B: Microelectron. Nanometer Struct., vol. 29, no. 3, pp. 031208-1031208-5,
2011.
[12] Y. Dan, K. Seo, K. Takei, J. H. Meza, A. Javey, and K. B. Crozier, Dramatic reduction of surface recombination by in situ surface passivation of
silicon nanowires, Nano Lett., vol. 11, pp. 25272532, 2011.
[13] C. Chen, R. Jia, H. Li, Y. Meng, X. Liu, T. Ye, S. Kasai, H. Tamotsu, N. Wu,
S. Wang, and J. Chu, Electrode-contact enhancement in silicon nanowirearray-textured solar cells, Appl. Phys. Lett., vol. 98, pp. 143108-1
143108-3, 2011.
[14] E. C. Garnett and P. Yang, Silicon nanowire radial p-n junction solar
cells, J. Amer. Chem. Soc, vol. 130, pp. 92249225, 2008.
[15] B. Ozdemir, M. Kulakci, R. Turan, and E. H. Unalan, Silicon nanowire
Poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) heterojunction
solar cells, Appl. Phys. Lett., vol. 99, pp. 113510-1113510-3, 2011.

Authors photographs and biographies not available at the time of publication.

Simulation of aSi:H Dual Junction Solar Cells

A. GarciaRivera, E. Comesana, A. J. GarciaLoureiro, R. Valin

J. A. Rodrguez, M. Vetter

Centro Singular de Investigacion en Tecnoloxas

da Informacion (CITIUS), Univ. Santiago de Compostela (USC),
Rua Jenaro de la Fuente Domnguez s/n,
Campus Vida, 15782 Santiago de Compostela, Spain
Email: angela.garcia.rivera@usc.es
Telephone: +34 881816475

Dept. de Tecnologa, Innovacion & Desarrollo

de TSolar Global S. A., Rua de Vigo 5,
San Cibrao das Vinas (Ourense), Spain

AbstractIn this paper the behaviour of the current density

voltage (JV) characteristic curves of hydrogenated amorphous
silicon (aSi:H) tandem solar cells (aSi:H/aSi:H) with heavy
doped tunnelling junction of hydrogenated microcrystalline silicon (cSi:H) n+ /p+ is analysed. The multijunction solar cells
are being used as an alternative to single junction (SJ) solar
cells trying to reduce the StaeblerWronski Effect (SWE) and to
increase the solar cell stability and efficiency. The dualjunction
(DJ) solar cell is formed by a stack of two pin aSi:H solar
cells (top and bottom solar cells) whose JV characteristic curves
are also studied. Important in this kind of device is an exact
adjustment of the current generation in top and bottom cells since
both cells are connected in series and therefore must generate the
same short circuit current density (Jsc ). The Jsc could present a
missmatching due to the different absorption of light in deeper
parts of the device if the thickness of top and bottom cell are not
well adjusted. Changing the relation between the thickness of the
active layers of the tandem solar cell, the missmatching in the
Jsc value between top and bottom solar cells can be minimized.
The here developed computer simulations of the electrical and
optical properties of the tandem device will help us to get a better
understanding of tandem solar cells in order to design in a future
work multijunction solar cells made of different materials.

I. I NTRODUCTION
In recent years, the interest of hydrogenated amorphous
silicon (aSi:H) tandem solar cells (aSi:H/aSi:H) has increased since these devices present reduced the Staebler
Wronski Effect (SWE) regarding to the aSi:H singlejunction
(SJ) device [1]. Tandem solar cells provide better stability and
increased efficiency than aSi:H SJ solar cell [2] [3]. The
tunneljunction, which acts as joint in the tandem structure
between the top and the bottom aSi:H cells in tandem
structure, is an important design factor of tandem solar cells
because of its impact on the solar cell efficiency. Different
materials have been investigated in literature with the objective
to improve recombination rates in the the tunnel junction and
to reduce the light absorption in the doped layers [4] [5] [6].
The structure of the tandem solar cell analysed is basically
composed by a stack of two pin junction solar cells based
on aSi:H. The structure of this typical aSi:H solar cell with
a pin superstate configuration is depicted in Fig. 1. It is a
consecutive deposition of a p type layer (player), an intrinsic
type layer (ilayer) and n type layer (nlayer), on a float glass
with a transparent conductive oxide (TCO). The TCO layer

Fig. 1. Typically structure of a pin aSi:H SJ solar cell with flat interfaces.

acts as a front contact and the Aldoped zinc oxide (AZO) and
aluminium (Al) metal layers, which are deposited on the pin
structure, form the back contact of the solar cell. In this work,
we will study the current densityvoltage (JV) characteristic
curve of aSi:H/aSi:H tandem solar cell with a heavy doped
tunnelling junction of hydrogenated microcrystalline silicon
(c-Si:H) n+ /p+ layers.

373

II. P HYSICAL

MODEL AND SIMULATION METHOD

The Synopsys TCAD Sentaurus is a commercial software

package able to simulate the electrical, thermal and optical
behaviour of semiconductor devices [7]. It is considered an
advanced tool to achieve a faster development of semiconductor devices, enabling the simulation of new geometries and
material devices. One challenge related to the simulation of
the aSi:H SJ solar cell and as consequence related to the
simulation of the aSi:H/aSi:H dualjunction (DJ) solar cells
is to introduce properly in the simulation the physical models
which describe the behaviour of these devices. Some models
are already defined in this package but the continuous distribution of density of states (DOS) in aSi:H and the physical
model to take into account the contribution from all bandgap
states to the recombinationgeneration (RG) processes are
not included [8]. They must be considered to carry out an
accurate simulation of the aSi:H/aSi:H tandem solar cell.

Fig. 2. Representation in linear scale of the continuous distribution of the

density of states in aSi:H.

A. Continuous Distribution of the DOS in aSi:H and RG

processes
Disorders and defects of the atomic structure of aSi:H
make the difference in electrical and optical properties with
respect to the crystalline silicon (cSi). The aSi:H requires
to introduce a new distribution of states which is called continuous distribution of the density of states. This distribution
takes into account that the bandgap which exists between the
valence band (VB) and the conduction band (CB) is defined
in a different way than in cSi approaches. In this model,
such as it is shown in Fig. 2, the VB and the CB have a
parabolic form and the disorder in the atomic structure of
aSi:H causes a spread of the energy states of VB and CB
into the bandgap, forming the regions called band tails states
which follow exponential decay distributions. The amphoteric
states caused by dangling bonds (DB+/0 and DB0/-) have
been also considered. These are permitted energy states located
between the VB and the CB, specifically near the middle of
the bandgap. These states have been included in our simulation
as two equal Gaussian distributions separated between them
by the correlation energy (U), such as it is presented in Fig. 3
in logarithmic scale. The states which follow the parabolic
VB and CB distributions are known as extended states and
they are nonlocalized states. In the bandgap of aSi:H there
is a considerable density of states between the energy levels
that separate the extended states from the localized states.
These energy levels are known as the VB (Evmob ) and the CB
(Ecmob ) mobility edges and this region is called the mobility
gap [9]. The localized states within the mobility gap are
formed by tail states and dangling bond states which have
different nature. The tail states act either as acceptor states
(CB tail states) or as donor states (VB tail states). To take
into account the RG processes for band tails and DB, the
ShockleyReadHall (SRH) and SahShockley mechanisms
have been considered in our simulation model [10].

Fig. 3. Representation in logarithmic scale of the continuous distribution

of the density of states in aSi:H and amphoteric states caused by dangling
bonds.

B. Simulation
The simulated structure of the DJ solar cell analysed in
this paper is a stack of two pin junction aSi:H solar cells
with flat interfaces such as is shown in Fig. 4. Between the
top and the bottom solar cells a tunneljunction is included
with heavy doped cSi:H p+ /n+ layers. Table I shows the
baseline input parameters of these two pin top and bottom
cells which form the structure of aSi:H/aSi:H tandem solar
cell [11]. The behaviour of the JV characteristic curves
for the tandem structure and also for the top and bottom
cells was obtained from a coupled solution of Poisson and
continuity equations. The performed simulation considers that
the photocurrent of the cells is only generated by the active
layers and it assumes that photons which are absorbed only
generate only one electronhole pair. The optical model to
simulate the the illumination of the solar cell is based on the
transfer matrix method (TMM) using a wavelength range from
300nm to 800nm under the standard AM 1.5 spectrum [12]. A
nonlocal bandtoband tunneling model is used to simulate
the tunneljunction in DJ solar cell.

374

TABLE I
I NPUT PARAMETERS USED IN THE SIMULATION FOR P I N TOP AND
BOTTOM CELLS WHICH FORM THE DJ SOLAR CELL .

player

ilayer

nlayer

3 1018

1015

8 1018

Relative permittivity

7.2

11.9

11.9

Electron affinity (eV)

3.90

4.00

3.99

1.95

1.78

1.80

Electron mobility (cm /Vs)

20

20

20

Hole mobility (cm2 /Vs)

Bulk material
Doping (cm3 )

Mobility gap (eV)

2

Effective DOS in CB (cm

Effective DOS in CV (cm3 )

20

10

20

10

1020

1020

1020

1020

0.144

0.144

0.144

Dangling bond states

Standard deviation
3

Max. density NDB (cm

+/0

18

8 10

16

5 10

2 1019

EDB (eV) (from CB)

-0.70

-0.89

-1.40

Correlation energy, U (eV)

0.20

0.20

0.20

CB mobility edge

2 1021

8 1021

1021

VB mobility edge

1021

4 1021

2 1021

CB tail energy (eV)

0.180

0.032

0.070

0.090

0.047

0.160

Tail states
DOS (cm3 /eV) at:

VB tail energy (eV)

2

Capture cross section (cm ):

Fig. 4. Structure of the simulated aSi:H/aSi:H DJ solar cell with flat

interfaces.

electron/hole in CB tail

1017 /1015

electron/hole in VB tail

1015 /1017

III. R ESULTS
The aim of our computer simulation is to obtain the behaviour of the JV characteristic curve, under the standard
AM1.5 sunlight sprectrum, for the aSi:H/aSi:H DJ solar cell
presented in Fig. 4.
The schematic of the band diagram of this tandem solar cell
under the AM 1.5 spectrum generated by the simulation tool
when the applied voltage is 0V, including the cSi:H tunnel
junction simulated with a heavy doped p+ /n+ layers is shown
in Fig. 5. In Fig. 6 the JV characteristic curve for the top
and bottom cells is presented. The Jsc and Voc values for the
top cell are higher than the bottom cell since in this example
the layer properties are not the optimized ones. The electronic
properties of the bottom solar cell play in this example the
critical role in the global structure of the DJ solar cell because
e.g. the Jsc of the tandem solar cell is currentlimited by this
bottom cell. In Fig. 7 the JV characteristic curve of DJ solar
cell is depicted. This figure shows that the Jsc value of the
tandem solar cell is limited by the bottom cell. Nevertheless,

Fig. 5. Band diagram of tandem solar cell under the AM 1.5 spectrum when
the applied voltage is 0V.

375

which form the tandem cell. The simulation of different layer

stacks with different top and bottom cell properties will help
us to get a better understanding of tandem solar cells in order
to design in a future work multijunction solar cells made of
different materials.
ACKNOWLEDGMENT
This work has been supported by FEDER funds and Xunta
de Galicia under projects IN841C 2011/231 and IN841D
2010/14 (HELATHIS Cofunding of the FP7 EUProject
HELATHIS, Grant Agreement No.241378) and contracts
2010/28, by Spanish Government (MCYT) under project
TEC2010-17320 and under contract Torres Quevedo PTQ10
03524.
R EFERENCES
Fig. 6. JV characteristic curve for the top and the bottom cells which form
the DJ solar cell.

Fig. 7.

Characteristic JV curve for the DJ solar cell.

the Voc of the DJ solar cell is nearly the sum of the Voc value
of the top and bottom solar cells.
IV. C ONCLUSION
In this paper the behaviour of the JV characteristic curves
of the aSi:H/aSi:H DJ solar cell and also the top and bottom
cells of the stack individually are analysed. The top cell of the
here analysed layer stack with 5080nm thickness of the top
cell and 200300nm thickness of the bottom cell presents a
higher value of Jsc than the bottom cell value. For this reason,
the Jsc value of the bottom cell works as a currentlimited
value of the tandem solar cell. The missmatching in the Jsc
value between top and bottom solar cells may be improved
changing the relation between the thickness of the active layers
of the tandem solar cell. The Voc value of the DJ solar cell is
nearly the sum of the Voc value of the top and bottom cells

[1] D. L. Staebler, and C. R. Wronski, Reversible conductivity changes

in dischargeproduced amorphous Si, Appl. Phys. Lett., vol. 31, pp.
292294, 1977.
[2] A. Kolodziej, C. R. Wronski, P. Krewniak, and S. Nowak, Silicon thin
film multijunction solar cells, OptoElectron. Rev., vol. 8, no. 4, pp.
339345, 2000.
[3] H. Jingya, X. Jianping, and K. Frank, Nonlocal recombination in tunnel
junction of multijunction amorphous Si alloy solar cells, Mat. Res. Soc.
Symp. P., vol. 336, pp. 717722, 1994.
[4] J. Y. Hou, J. K. Arch, S. J. Fonash, S. Wiedeman, and M. Bennett,
An examination of the tunnel junctions in triple junction aSi:H based
solar cells: modelling and effects on performance, Proccedings of the
Photovoltaic Specialist Conference, pp. 12601264, Las Vegas, October
1991.
[5] D. S. Shen, R. E. I. Schropp, H. Chatham, R. E. Hollingsworth, J.
Xi, and P. K. Bhat, High efficiency a-Si:H/a-Si:H tandem solar cells,
Proccedings of the Photovoltaic Specialist Conference, pp. 14711474,
Bhat, May 1990.
[6] J. K. Rath, F. A. Rubinelli, and R. E. I. Schropp, Effect of oxide
treatment at the microcrystalline tunnel junction of a-Si:H/a-Si:H tandem
cells, Journal of NonCrystalline Solids, vol. 266269, pp. 11291133,
2000.
[7] Sentaurus device user guide.
Version D2012.06, Mountain View,
Synopsys, Inc., 2012.
[8] C. Lee, H. Efstathiadis, J. E. Raynolds, and P. Haldar, Twodimensional
computer modeling of single junction aSi:H solar cells, Proc. 34th
IEEE Photovoltaic Specialists Conference (PVSC), pp. 001 118001 122,
June 2009.
[9] R.E.I. Schropp, and M. Zeman, Amorphous and microcrystalline silicon
solar cells: modeling, materials and device technology. Kluwer Acad.
Publ., 1998.
[10] C. T. Sah and W. Shockley, ElectronHole Recombination Statistics in
Semiconductors through Flaws with Many Charge Conditions, Physical
Review, vol. 109, no. 4, pp. 11031115, 1958.
[11] J. A. Rodriguez, P. Otero, M. Vetter, J. Andreu, E. Comesana, and A.J.
GarciaLoureiro, Simulation of the effect of player properties on the
electrical behaviour of aSi:H thin film solar cells, Spanish Conference
on Electron Devices, 2011.
[12] P. Yeh, Optical Waves in Layered Media (Pure & Applied Optics). John
Wiley & Sons Inc., 1988.

376

Solar Power Enhancement Using Concentration

Method, Tracking System and Fuzzy Based Control
System
Muhammad Salahuddin Kabir,
Rahnuma Rifat Chowdhury
and M. Nabil Amin
Department of Electrical, Electronic
and Communication Engineering,
Military Institute of Science and
Technology, Dhaka, Bangladesh.
Email:
muhammad salahuddin kabir@yahoo.com,
rahnumarifatchowdhury@yahoo.com,
m.nabilamin@yahoo.com

Deen Md Abdullah
and Wali Md Abdullah

AbstractThis research paper carries out a realistic experimental approach to enhance the solar output power to a
significant level with the application of solar concentrator and
tracking system and also proposes a fuzzy based control system
for the utilization of the duration for which solar power is
available. By using only a solar panel, the theoretical value is
supposed to be around 40% more. The method is arranged
using Double Sun Technologyand Solar Tracker. In addition
to Double Sun Technology, four flat mirrors have been used
to increase the output power. A remarkable enhancement in the
output power is achieved with this method. Rather than using
only a solar panel, the output power increases by 58.32% by
using a photovoltaic (PV) panel with four flat mirrors with solar
tracker system. The concentrators are designed to collect the
incident light from the sun and concentrate them on the PV panel
so that the solar irradiance is more. The experiment is conducted
in four different setups. In the first setup, the output power is
calculated without the use of Double Sun and solar tracker. In the
second setup, power is measured with Double Sun but without
solar tracker. Both Double Sun and solar tracker are used in
the third setup. Then four flat mirrors and solar tracker are
considered. Output power has been calculated using open circuit
voltage and short circuit current. Finally, the proposed system
improves the total output by maximum utilization of available
photovoltaic energy and maximum utilization of the duration for
which solar power is available based on fuzzy logic.
Index TermsPhotovoltaic (PV) panel, passive device, Double
Sun Technology, solar tracker, concentrator, fuzzy logic.

I. I NTRODUCTION
Power generation is going to take a complete new shape in
twenty first century. Fossil fuels like coal, gas, petroleum are
not going to be plenty for electricity generation. Renewable
energy presents a better alternative to fossil fuels with greener
impact on environment. Though hydroelectricity offers a potential amount of renewable energy to the world, yet it is not
available to all the places of the globe. Solar energy on the

978-1-4673-4836-2/12/$31.00 2012 IEEE

Muhammad Abdullah Arafat

Department of Electrical, Electronic

and Communication Engineering,
Department of Computer Science
Bangladesh University of Engineering and
and Engineering,
Military Institute of Science and Technology (BUET), Dhaka, Bangladesh.
Email: abdullah arafat@eee.buet.ac.bd
Technology, Dhaka, Bangladesh.
Email: rabi.mist@yahoo.com,
shashi.mist@yahoo.com

other hand has all the opportunity to take over the existing
power generation system [1]. Solar energy can be used for
small industries and in a very small scale for large industries;
like garment industries where the solar energy can be used as
secondary source for lighting and ventilation, reducing power
consumption from the main power line. But the output power
of photovoltaic (PV) panel is considerably low [2]. Extensive
research shows that output power can be increased in either of
two ways - by fabrication or by using passive devices [3], [4],
[5], [6], [7]. As fabrication is costly and hard, passive devices
are widely used to increase the efficiency as well as fill factor.
Concentrators with two flat mirrors and solar tracker are used
to enhance solar power. Concentration with two flat mirrors is
known as Double Sun Technology.
Fuzzy control is a methodology for constructing control
system in which the mapping between real-valued input and
output parameters is represented by fuzzy rules. Fuzzy control
has been very successful in commercial products such as
automatic transmission, video cameras etc.
In this paper, experimental results have been given which
shows the enhancement of output power by using concentrator
and solar tracker with a solar panel. The experiment is
conducted in four different setups. In the first setup, the output
power is calculated without the use of Double Sun and solar
tracker. In the second setup, power is measured with Double
Sun but without solar tracker. Both Double Sun and solar
tracker are used in the third setup. And finally four flat mirrors
and solar tracker are considered. An increase of 33.50% solar
power is found with the Double Sun and solar tracker, the
theoretical value of which is 40% [8]. 58.32% increase in
solar energy is observed by applying four fat mirrors with
solar tracker. Then we have proposed a scheme as [9], [10]
based on the use of conventional DC-DC converter, where

464

Fig. 1.

Solar PV panel with two flat mirrors (Double Sun Technology).

combination of voltage and power feedback control system is

implemented with the use of artificial intelligence algorithm
that makes tracking response faster and maximizes the power
extracted from the solar module.
II. D OUBLE S UN AND F OUR S UN T ECHNOLOGY
The amount of radiation that falls upon the modules in this
system is usually characterized by the concentration factor.
The most common definition is the geometric concentration
factor (C), which, for this system, is calculated by the ratio
between the effective area (the area seen by the sun) and the
active cell area as defined by the following equation [8], [11]:

C=

Aef f
Amirrors + Amodule
Amirrors
=
=1+
Amodule
Amodule
Amodule

Fig. 2.

Solar PV panel with four flat mirrors (Four Sun Technology).

Amirrors = 2 (10 14 + 10 20.5)sq.inch

= 690sq.inch
Amodule = (12.5 18.5)sq.inch = 231.25sq.inch
C = 3.98x
According to the definition, the geometrical concentration
factor using four flat mirrors is 3.98x ( 4); whereas it is 1.8x
( 2) for Double Sun Technology [8]. Hence four flat mirrors
is termed, hereafter, as Four Sun Technology in this paper.
Fig. 1 and Fig. 2 show the experimental setups of Double Sun
and Four Sun Technology respectively. However, the value
of C is higher than the real concentration factor, since the
efficiency of the mirrors, reflective losses and the percentage
of direct radiation are disregarded. Such factors were taken
into account by the theoretical model developed by Reis [12].
III. F UZZY L OGIC I MPLEMENTATION

Here, Aef f is the effective area (the area seen by the sun),
Amirrors is the area of the mirrors that is seen by the sun and
Amodule is the total area of the modules. In this experiment,
for Double Sun Technology

Amirrors = 2 (10 20.5)sq.inch = 410sq.inch

Amodule = (12.5 18.5)sq.inch = 231.25sq.inch
C = 1.8x
And for four flat mirrors,

978-1-4673-4836-2/12/$31.00 2012 IEEE

A. Fuzzy Logic Controller Structure

Our designed controller takes the measured value of the light
strength received by the sensor as the feedback and implements
control using many rounds of modifications.
The basic fuzzy control system structure is shown in Fig. 3.
B. Implementation
In our proposed control system the whole fuzzy controller
can be designed in five steps.
Definition: We define the input variables, output variables
and linguistic variables.

465

Fig. 3.

Solar Energy Fuzzy Control System Structure.

The input variable is Error e

The output variable is the number of seconds between the rotation and reversion adjustment of the
stepper motor.
For the domain of the output and input variables we
shall use normalized discrete domain to set domain
scope.
For linguistic items, we can define five linguistic items for each linguistic variable, i.e. e =
N B, N S, ZE, P S, P B
NB: Negative Big
NS: Negative Small
ZE: Zero
PS: Positive Small
PB: Positive Big
Build Membership Input Functions: Based on the
defined linguistic variables, we can build the input membership function.
Set up Fuzzy Rules Base: The setup of the fuzzy
rules base is crucial because all states must be operated
based on the rules defined in the rules base. To set up
a fuzzy rules base smoothly, we shall adopt five fuzzy
control rules expressed with IF THEN statements for five
linguistic items.
Define Fuzzy Inference Engine: There are many fuzzy
inference methods and different results are inferred with
different methods. In our proposed control system we use
the center of gravity method proposed by Mamdani as
the defuzzification tool because the method is easy and
reliable.
Defuzzify Result: There are many methods for defuzzification. In this proposed control system we use center of
gravity for defuzzification to obtain actual operation. The
defuzzification method can also be easily implemented
with a digital circuit.

Fig. 4.

Solar output power without concentrator and without solar tracker.

Sun and with solar tracker. Open circuit voltages and short
circuit currents are measured for seven hour period starting
from 9 : 00 am to 4 : 00 pm with half hour interval and
power is calculated from the data. Maximum power is achieved
when the sun ray is at right angle to the solar panel surface.
By implementing fuzzy control system we can ensure that the
sun ray will be at right angle to the solar panel surface during
our defined period.
A. Without Concentrator and without Tracker
The flat solar panel has been exposed to the sun. No solar
tracker and no mirror is used in this stage. The solar panel is
inclined at a fixed position for the entire time period. Fig. 4
shows the variation of the experimentally measured power with
time. Data are collected for three different days of similar
weather condition.
B. With Double-Sun but without Solar Tracker
In this setup, data are taken with the Double sun Technique,
i.e. using two side mirrors, but no tracking has been done.
The power versus time graphs for the same days are shown in
Fig. 5.
C. With Double-Sun and with Solar Tracker
In third setup, open circuit voltages and currents are measured, with the Double Sun Technique and using solar tracker.
The power versus time graphs are shown in Fig. 6.
D. With Four Sun and with Solar Tracker

IV. E XPERIMENTAL S ETUP

Finally, with the Four Sun Technology and solar tracker,

open circuit voltages and currents are measured. The power
versus time graphs for the same three days are shown in Fig. 7.

To verify the energy enhancement by the concentration

methods and tracking system, data have been taken for three
days for four different conditions: without concentrator and
without solar tracker, with Double Sun and without solar
tracker, with Double Sun and with solar tracker and with Four

The average output power for the four experimental setup is

given in Table I and Fig. 8 shows the variation of the average
output power with time for the four conditions. In the first case
i.e. only with flat PV panel at fixed position, the average output

978-1-4673-4836-2/12/$31.00 2012 IEEE

V. DATA A NALYSIS

466

Fig. 5. Solar output power with Double Sun concentrator but without solar
tracker.

Fig. 6.
tracker.

Solar output power with Double Sun concentrator and with solar

Fig. 7. Solar output power with Four Sun technology and with solar tracker.

Fig. 8.

Average output power in four experimental setups.

power of approximately 16.99 watt (W) has been achieved.

With Double Sun but without tracker, the average output power
has been increased by approximately 12.14% than the first
case. Thirdly, with Double Sun and solar tracker, the average
output power has been increased by approximately 33.50%
than the initial condition. Finally, with Four Sun Technology
and solar tracker, the average output power has been increased
by approximately 58.32% than the initial condition and the
average power is found to be 26.90 W.

solar tracker can lead to an improvement of 58.32% in the

total average output power. This paper has also proposed
a fuzzy based control system to improve the total output
by maximum utilization of available photovoltaic output and
maximum utilization of the duration for which solar power is
available. Therefore, the proposed Four Sun method with solar
tracker and fuzzy control system is an easy and cost effective
way of increasing energy received from the solar panel.

VI. C ONCLUSION

The authors would gratefully like to acknowledge the endless support by Military Institute of Science and Technology
(MIST), Dhaka, Bangladesh for supporting this research-work.

This paper has shown the experimental results of the performance of the Double Sun Concentration and Four Sun Concentration PV systems with geometric concentration factors
of 1.8x and 3.98x, respectively, using conventional silicon
modules and no active cooling. The result for the Double
Sun Technology with solar tracker is increased by 33.50%.
It also demonstrates that the use of four flat mirrors and

978-1-4673-4836-2/12/$31.00 2012 IEEE

ACKNOWLEDGMENT

R EFERENCES
[1] F. Kreith and D. Y. Goswami, Handbook of Energy Efficiency and
Renewable Energy. CRC Press, Taylor & Francis Group, 2007.
[2] A. McEvoy, T. Markvart, and L. Castaner, Practical Handbook of
Photovoltaic: Fundamentals and Applications. Academic Press, 2011.

467

TABLE I
COMPARISON OF THE FOUR EXPERIMENTAL SETUPS

Time
(Hour)

9:00 AM
9:30 AM
10:00 AM
10:30 AM
11:00 AM
11:30 AM
12:00 PM
12:30 PM
1:00 PM
1:30 PM
2:00 PM
2:30 PM
3:00 PM
3:30 PM
4:00 PM
Pavg
Percent
Increase

Average Output Power (W)

With
With
Without
Double
Double
With Four
concentrators Sun but
Sun and
Sun and
and without
without
with solar
with solar
solar tracker
solar
tracker
tracker
tracker
12.16
15.61
20.44
24.80
13.00
16.29
21.69
27.38
14.59
17.53
21.87
27.60
16.20
18.41
23.18
27.91
19.05
21.80
23.70
28.45
20.11
22.45
24.56
28.14
20.18
22.99
24.55
29.25
20.90
22.98
23.18
28.46
19.79
22.27
23.03
27.58
19.20
21.97
23.60
27.83
18.43
20.30
22.83
25.79
17.26
17.83
22.67
25.49
16.22
15.64
22.38
25.10
14.81
15.17
22.36
25.43
13.02
14.63
20.29
24.38
16.99
19.05
22.68
26.90
Base

2.14%

33.50%

978-1-4673-4836-2/12/$31.00 2012 IEEE

[3] M. Brogren, Optical efficiency of low-concentrating solar energy systems with parabolic reflectors, Doctoral thesis, 2004.
[4] R. M. Swanson, The promise of concentrators, Progress in Photovoltaics: Research and Applications, vol. 8, no. 1, pp. 93111, 2000.
[5] A. Luque and S. Hegedus, Handbook of Photovoltaic Science and
Engineering. John Wiley & Sons, 2011.
[6] R. B. Pettit and E. P. Roth, Solar Mirror Materials: Their Properties
and Uses in Solar Concentrating Collectors. Solar Materials Science,
1980.
[7] A. Luque, G. Sala, and I. L. Heredia, Photovoltaic concentration at
the onset of its commercial deployment, Progress in Photovoltaics:
Research and Applications, vol. 14, no. 5, pp. 413428, 2006.
[8] F. Reis, V. Corregidor, M. C. Brito, R. Rodrigues, J. Wemans, and
G. Sorasio, Power generation and energy yield using doublesunr
photovoltaic solar concentration, EUPVSEC, pp. 803806, 2009.
[9] S. A. Khan and M. I. Hossain, Design and implementation of microcontroller based fuzzy logic control for maximum power point tracking
of a photovoltaic system, 6th ICECE, Dhaka, Bangladesh, 2010.
[10] M. S. Hussain, D. M. Abdullah, W. M. Abdullah, and L. Ali, Development of a fuzzy logic based smart solar system for irrigation, 9th
ICME, Dhaka, Bangladesh, pp. RT050, 2011.
[11] S. P. Sukhatme and J. K. Nayak, Solar Energy: Principles of Thermal
Collection and Storage. Tata McGraw-Hill Education, 2008.
[12] F. Reis, Development of solar concentrators for large photovoltaic
plants, M. Sc. thesis, IST, Personal communication, 2008.

58.32%

468

Design & Simulation of Solar DC Pump in Simulink

P.Badari Narayana

B.R. Sanjeeva Reddy

Director, Green Life Energy Solutions LLP,

Secunderabad, A.P, India
E-mail: badari.p@gmail.com

Research Scholar, NIT Warangal, A.P, India

Email: sanjeev.reddy@nitw.ac.in
Sanjay Dubey

Prasad Motepalli

Professor, ECE-Department,
BVRIT,
Narsapur, Medak (Dist), A.P, India
E-mail: sanjay.dubey@bvrit.ac.in

Final B.Tech-ECE, BVRIT,

Narsapur, Medak (Dist), A.P, India
Email: prasu.459@gmail.com
Abstract- Solar DC pump system is easy to install and operates
full automatically without watching. Solar DC pump needs less
solar panels to run than the AC pump. A simulation model is
necessary for the manufacturer to assess the performance of
solar pump controller. Solar DC Pump consists of solar PV
Modules, solar DC controller, and submersible DC pump. Solar
cells have nonlinear I-V characteristics, efficiency of PV module
is low and output power depends on solar isolation Level and
ambient temperature. Main point to consider, there will be great
loss of power due to mismatch of source and load. So to extract
the

maximum

power

to

load

from

PV

panel

MPPT

is

implemented in the boost converter circuit using PWM and

Microcontroller. The proposed design optimizes the solar power
System to deliver the maximum power to load from source and is
presented in the developed Simulink model. This model acts as a
template

for

designing

Solar

DC

pump

of

required

rated

capacity. In this work, Parameter extraction, Model evaluation,

design & analysis of PV panel characteristics, modeling boost
converter

and

DC

pump

motor

are

demonstrated

using

MATLAB/Simulink model.

Keywords: Solar Photovoltaic module, Boost converter, MPPT

algorithm, PWM, Microcontroller, Buck converter, DC pump

I.

INTRODUCTION

Renewable energy especially solar photovoltaic is

important alternative source of energy for the future, and
research and development done on this area attained
importance in using solar cells to power the electrical systems.
The two inherent problems in solar technology is low
conversion
efficiency
(lO%to
16%
efficiency
for
commercially available amorphous silicon solar cells) and
next one is the presence of highly non-linear I-V
characteristics. The problem gets worse due to dependence of
characteristics of solar cell on temperature and insolation
level. Further, due to mismatch between the operating point
and maximum power point (MPP) of the solar cells, the power
available. From the solar cell is not fully extracted. In order to
extract the maximum power from the source PV panel must be
capable of tracking the solar panel unique maximum power
point that varies with irradiance and temperature. MPPT is

978-1-4673-6150-7/13/$31.00 2013 IEEE

implemented by using any one of the algorithms perturb &

observe (P&O), incremental conductance. This is represented
in the current simulink environment through a PWM generator
that has the same value of duty cycle [1] [2].
I-V characteristics of solar cell play an important role
in designing a solar powered device. The two major
parameters of solar cell are Voe (Open circuit voltage) and Ise
(short circuit current). Current from the solar cell is calculated
by the characteristic equation of solar cell which involves Ise,
Voe, Rs (series resistance) and Temperature. For the present
design exercise, modules such as boost converter, buck
converter, DC motor are modeled in Simulink.
II.

DESIGN CALCULAnONS OF SOLAR PUMP

A simple model is chosen for the MATLAB modeling and

simulation. A simple PV cell is represented by a light
dependent current source (lsc) in antiparallel with a diode
driven by current Id[2] and a series resistance in the current
path through semiconductor material, the metal grid, contacts
and current collecting bus [4]&[5].
A simple power calculation is given below for designing a
solar DC Pump for a required capacity.
Step 1: Amount of water required per day is 25,000 liters/day.
Step 2: Total Dynamic Head (TDH). It depends on two
factors.
Total Vertical Lift.
Total frictional losses.
TDH= Vertical lift + Frictional losses.
Vertical lift = Elevation+ standing water
level+ Draw down.
Frictional losses = 0.5% of Total vertical lift
for the given specifications,
Vertical lift= 16 m.
Frictional losses= 16*0.051100= 0.008m
Then TDH=16 m+0.008 m =16.008m.

Step 3: Hydraulic energy required to raise the water level per

day is ...
= Mass* g*TDH.
= (Density*volume)*g*TDH.

429

(1000kg/m3*25m3/day)*9.8m/s2*16.008m.
1089.43 watt-hour/day.
Step 4: Solar radiation data
6hours/day. (Peak
=1000wattim2).
Step 5: Size and number of solar PV modules required are ...
Total wattage of PV panel = Total hydraulic energy/no.
of
hours.
= 1089.43/6 =18l.57 watt.
System losses = Total PV panel wattage/ (pump efficiency *
mismatch factor).
=181.57 watt/ (0.3*0.85) = 712.03
Considering operating factor = Total PV panel wattage after
losses/operating factor.
PV panel capacity = 712.03/0.75 =949.38watt.
Number of PV panels required = 949.38/240 4.
1 HP DC motor can be run with a solar capacity of 960Wp
with the available 240Wp model solar panel.

( )

Rs = _

dV
d1

nKT I q
KT
lo eQ(v+1Rs)ln

-1]

(6)

Solar cell now is a direct element of simelectronics library part

of MATLAB 2012's simulink environment.

\ I

Figure I: Equivalent circuit for solar cell in MATLAB simulation

III.

MODELLING AND VALlDFA nON IN SIMULINK

Solar Panels, Boost converter, DC pump are separately

modeled and unified in Multisim and simulink environments
as shown below.
A.

The two important curves which provide the complete

information about solar cell are I-V curve and P-V curve
which are drawn for the rated solar panel available in the
industry (240Wp).

Simulation of Solar Panel

A simple modeling is chosen for MATLAB modeling and

simulation. A simple PV cell is represented by a light
dependent current source (lsc) in antiparallel with a diode
driven by current Id[2] and a series resistance in the current
path through semiconductor material, the metal grid, contacts
and current collecting bus [4]&[5]. The presence of diode
determines the output V-I characteristics of solar cell.
The parallel resistance associated with a small
leakage of current through a resistive path in parallel with an
intrinsic device [5] & [6] is very large, so its effect is very less
and is neglected. The equations which govern the
characteristics of solar cell are:

Q(v+fRs)lnKT
1= Isc- lo e

Isc (;:0 )ISC

IG

o IT = 1SC ITref

Isc IT = IsQTref

[1

-1]
IGo

( J
Tref

The values of the power curves generated in simulink are in

agreement with the selected solar panel of 240Wp [9].
...,.::.
3OO ,---.-:- .::.--,

:
150

(1)

100
50

31

n
e

+ a T - Tref

+IRs )lnKT
1= Isc_1o eQ(v

Fig 2. Simulink model for Solar Panel

(qEg)(

- --;;-K y:1 - Tr1ef

Fig 3.T-V curve

(2)
(3)

-1]- V

(4)

Rp

Quality factor, N:

(5)

60

80

8.42

1000

W/m'2

Open-circuit voltage, Voc:

Irradiance used for
measurements,IrO:

+ IRs

40

100

Fig 4.P-V curve.

PV module designed using solar cells connected in series and

the parameters choosen for solar cell are given below.
Parameterize by:
IBy sic current and ole voltage, 5 parameter
1
Short-circuit current, Isc:

)]

20

1.5

Fig 5.parameters

Amount of irriadiation = 1000 watt/m 2.

Short circuit current (lsc)=8.42 amp
Open circuit voltage(Voc)=38V.

430

Quality factor(N)= 1.5.

Series resistance=O.OOl n.
B.

Simulation

of

Boost

converter

in

Multisim

environment

Boost converter is an Electronic DC to Dc converter which

allows us to implement the MPPT algorithm by varying the
duty cycle of PWM which is given to IGBT, PWM is
generated by Microcontroller, through MPPT algorithm
imple?ted in microcontroller.

Fig 6. Boost converter.

C.

Simulink model for DC pump.

Here, DC motor of 1 HP is run by PV array through boost

converter circuit to extract the maximum power from the
source. The simulated model for DC motor is presented below.

Solar module designed by solar cells is connected to DC

motor through Boost converter circuit that has MPPT
algorithm [7]. The modeled Solar DC pump displays the
designed speed and voltage specifications.

IV.

CONCLUSIONS

Solar DC pumps are gaining popularity as they are easy to

install for rural agricultural applications. Solar DC Pump is an
integral product of Solar PV Panels, Controller and
Submersible DC pump. Solar panels with effects of variations
in temperature, insolation power output are observed and
analyzed for MPPT design. The modeling of each component
of PV system simulated in MATLAB gives the feasibility of
the system under real weather conditions. Solar pump model
using the equivalent circuit of moderate complexity was
chosen for simulation and found to be in good matching with
the available solar DC pump in the market.
The designed simulink module can be used as a
template to design solar DC powered water pumping system
of required capacity. This current exercise of simulation of
solar DC pump is very helpful to the solar pump manufacturer
for optimizing the design of solar DC pump controller.
REFERENCES
[I]

Geoff Walker

"Evaluating

MPPT

Converter

topologies

using

MATLAB PV module," Journal of Electrical & Electronics Engg.

Australia; Volume 21, Tssue1:2001; 49-55
[2]

Roberto faranda, Sonia Leva "Energy comparison of MPPT techniques

for PV system," WEASTransction on Power Systems, ISSN: 1790-5060,
Issue6, Volume3, June 2008

[3]
1.\t 1k"". ,,,
.J.
:- EA2ICII I R.IA'ItL-.rJCf'..! ",n.n,,"
.

Edition, CRC press, 2005

[4]

http://www.solarsemiconductor.com/asialssi-m6.pdf

[5]

N.Mohan,

TM.Undeland,

and

WP.Robbins,

"Power

Electronics

Converters, Applications and Design," John Wiley & Sons, Inc., NY,

Fig 7. DC motor model

D.

nd

Messenger, Roger& Jerry Ventre, "Photovoltaic systems Engg", 2

2003

Simulink model for Solar DC Pump

[6]

Luis Castafler, Santiago Silverstre, "Modeling photovoltaic systems

[7]

B.R.Sanjeeva Reddy,P.Badari Narayana,Praveen Jambholkar,Prof. K.

using PSpice," John Wiley and Sons, Jan 2003.

Srinivasa

Reddy,"MPPT

Photovoltaic

module

Algorithm

using

Implementation

Microcontroller",

India

for

Solar

Conference

(INDICON), 20II Annual IEEE.

[8]

Mukesh Kumar

Gupta, Rohit

Jain, "MPPT

Simulation with DC

Submersible Solar Pump using Output Sensing Direct Control Method

and Cuk Converter",INTERNATIONAL JOURNAL of RENEWABLE
ENERGY RESEARCH, Vol.3, No.1, 2013
[9]

http://www.mygreenenergv.in/60 Cell Module Data sheet MGE1.pdf

Fig 8. Solar DC Pump model in simulink

431

1298

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 4, OCTOBER 2013

Electrochemical Potentials (Quasi-Fermi Levels) and

the Operation of Hot-Carrier, Impact-Ionization,
and Intermediate-Band Solar Cells
A. Mart, Member, IEEE, and A. Luque, Senior Member, IEEE

AbstractIn the framework of the so-called third generation

solar cells, three main concepts have been proposed in order to exceed the limiting efficiency of single-gap solar cells: the hot-carrier
solar cell, the impact-ionization or multiple-exciton-generation solar cell, and the intermediate-band solar cell. At first sight, the
three concepts are different, but in this paper, we illustrate how all
these concepts, including the single-gap solar cell, share a common
trunk that we call core photovoltaic material. We demonstrate
that each one of these next-generation concepts differentiates in
fact from this trunk depending on the hypotheses that are made
about the physical principles governing the electron electrochemical potentials. In the process, we also clarify the differences between
electron, phonon, and photon chemical potentials (the three fundamental particles involved in the operation of the solar cell). The
in-depth discussion of the physics involved about the operation of
these cells also provides new insights about the operation of these
cells.
Index TermsHot-carrier, impact-ionization, intermediate
band, solar cells, thermodynamics.

I. INTRODUCTION
OT-CARRIER [1], intermediate-band [2], impactionization [3], and multiple-exciton-generation solar cells
[4] are novel solar cells proposed to exceed the efficiency limit
of single-gap solar cells [5]. Their fundamental models have
been reviewed several times and, each time, our understanding
of their operation improves or their practical implementation
becomes more feasible [6][12]. However, a unified treatment
of all these cells at a fundamental level is lacking, and this paper
is intended to fill this gap. Electrochemical potentials (or quasiFermi levels) will serve as the basis for this unified treatment.
A single-gap solar cell [see Fig. 1(a)] absorbs photons with
energy higher than the semiconductor bandgap EG . Its efficiency is limited, on one hand, by the fact that photons with
energy lower than the bandgap are not absorbed. On the other
hand, photogenerated electrons quickly relax to the bottom of
the bands by giving away their energy to phonons. The output voltage of the single-gap solar cell V, equals the electron
and hole quasi-Fermi level split (e,C e,V ) divided by the

Manuscript received May 2, 2013; revised June 14, 2013; accepted July 18,
2013. Date of publication August 6, 2013; date of current version September
18, 2013. This work was supported by the European Commission through the
NGCPV Project under Grant 283798.
The authors are with the Instituto de Energa Solar, ETSI Telecomunicacion, Universidad Politecnica de Madrid, Madrid 28040, Spain (e-mail:
amarti@etsit.upm.es; a.luque@upm.es).
Digital Object Identifier 10.1109/JPHOTOV.2013.2274381

Fig. 1. Mechanism involved in the operation of the (a) single-gap

solar cell, (b) HCSC, (c) IISC/MEGSC, and (d) IBSC.

electron charge. The cell emits photons whose chemical potential also equals this split. Throughout this paper, eV will be the
electron charge multiplied by the output voltage of the cell.
To fight the energy losses of the single-gap solar cell, the hotcarrier solar cell (HCSC) proposes that photogenerated carriers
are extracted from the cell before they recombine [see Fig. 1(b)].
This leads to a model in which, contrary to the intuition gained
from the operation of single-gap solar cells, there is no electron
and hole quasi-Fermi level split in the material that absorbs the
light at the time the cell is still capable of producing output
voltage.
Aiming also to optimize the use of the energy provided by the
absorption of high-energy photons, the impact-ionization solar
cell (IISC), and also the so-called multiple-excitation-generation
solar cell (MEGSC), pursues that the absorption of a high-energy
photon creates more than one electronhole pair [see Fig. 1(c)].

2156-3381 2013 IEEE

MARTI AND LUQUE: ELECTROCHEMICAL POTENTIALS (QUASI-FERMI LEVELS) AND THE OPERATION

As in the single-gap solar cell, the output voltage of this cell,

multiplied by the electron charge, equals the electron and hole
quasi-Fermi level split. However, the chemical potential of the
emitted photons is not equal to this split but M times, M being
the number of electronhole pairs generated.
The intermediate-band solar cell (IBSC), on the other hand,
aims to exploit the energy of the below-bandgap energy photons
that are not absorbed in a single-gap solar cell. For that, it
postulates a material defined by the existence of an electronic
band (the intermediate band, IB) additional to the conduction
band (CB) and valence band (VB). As illustrated in Fig. 1(c),
the IB allows the absorption of below-bandgap energy photons
through electronic transitions from the VB to the IB and from
the IB to the CB. In this cell, associated with each of the bands,
three quasi-Fermi levels exist instead of two.
In spite of their apparently different operation, we shall show
that these cells share a common trunk that we call core photovoltaic material (described in Section II). In subsequent sections, we will discuss how the novel solar cells sprout from this
trunk as branches depending on the hypotheses made. For that,
we will assume local equilibrium thermodynamics valid so that
the description of the systems by thermodynamic variables such
as temperature and electrochemical potentials still makes sense
even in nonequilibrium [13][15]. The process will also lead
us to better understand why the relationships between electron
and hole quasi-Fermi levels, cell output voltage, and emitted
photons electrochemical potentials are apparently so different
from one cell to another.

1299

(a)

(b)

(c)

(d)

II. CORE PHOTOVOLTAIC MATERIAL

The core photovoltaic material is described by the set of
energy states available to the electrons. It will be useful to think
of this material as being semiconductor-like so that one set of
these energy states is grouped into a VB and another set, with
higher energy, is grouped into a CB. The two bands are separated
by the bandgap EG . In this core material, we will assume that
electrons can only interact among themselves. This interaction
is symbolically represented by the following reaction:
e1 + e2 e3 + e4

(e)
Fig. 2. Illustration of different electronelectron interactions. Two electrons
interacting within (a) the CB and (b) the VB. (c) Auger recombination and
(d) generation (also called impact ionization). (e) Auger thermalization.

(1)

where ei means an electron at an electronic state with energy

i . We first address the most general case and assume that the
energy levels in (1) can be located anywhere; for example, e1
can be in the CB and e2 in the VB, or both can be in the CB or
both in the VB. In this way, (1) is suitable for describing all the
processes illustrated in Fig. 2. We refer to the processes illustrated in Fig. 2(a) and (b) as intraband scattering processes. In
these intraband processes, all the interacting electrons remain in
the same band after the interaction. Processes such as the ones
illustrated in Fig. 2(c)(e) are referred to as interband scattering processes. They are characterized by their involvement with
electrons in both the CB and the VB. In particular, the process
illustrated in Fig. 2(c) would correspond to Auger recombination, and the process illustrated in Fig. 2(d) would correspond to
an Auger generation or impact-ionization process. The process
illustrated in Fig. 2(d) is referred to as Auger thermalization.

Assuming the electron interactions from (1), we can write:

1 + 2 = 3 + 4

(2)

where i represents the electrochemical potential or quasiFermi level of an electron at energy state i that can be degenerated without loss of generality in our argumentation. Energy
conservation implies that
1 + 2 = 3 + 4 .

(3)

Hence, because (2) must be fulfilled for any set of energies

satisfying (3) [6], [15], i must be a linear function of i
i = ai + b

(4)

where a and b are constants. This implies that if electrons in

the semiconductor are at temperature Tc (cold electron temperature), the probability fFD of whether they occupy the state i

1300

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 4, OCTOBER 2013

follows the Fermi-Dirac distribution given by



fFD =
exp

1
i a i b
kTc

=
+1

exp

1
i e
kTh

(5)
+1

where
Th = Tc /(1 a)

(6)

e = b/(1 a).

(7)

These equations where first deduced by Wurfel [6] and imply that our electron distribution can be visualized, instead of
as a distribution at temperature Tc and an energy-dependent
electrochemical potential, as a distribution at an equivalent
(hot) temperature Th and an energy-independent electrochemical potential e . Notice that, under this constant-hot-electrontemperature description, the difference in the electrochemical
potential between electrons in the CB and VB would be zero
because e is the same for electrons in both bands. The basis
of this conclusion is the fact that electrons in the CB have been
allowed to interact with electrons in the VB. This interaction
can consist of not only Auger generation recombination [see
Fig. 2(c) and (d)] but also Auger thermalization [see Fig. 2(e)].
To make a solar cell, our core material must interact with photons. This interaction takes place through radiative generationrecombination processes and implies that a photon can be absorbed by pumping an electron from the VB to the CB (generation) and also that an electron can recombine with a hole by
emitting a photon (luminescent photon). Because this photon
can be reabsorbed again, luminescent photons and electron
hole pairs stay in equilibrium through the reaction
ei,VB + hv ei,CB

(8)

where hv represents a photon and ei,VB and ei,CB an electron in

the VB and CB, respectively. This equation leads us to visualize
the photons in the core material as a gas at the hot-electron
temperature Th and chemical potential hv
hv = e e = 0

(9)

because according to (7), e is the same for electrons in the VB

and the CB.
To this point, we have not discussed electron interaction with
phonons. This interaction is introduced in the later stages of
this study as our description of the different solar cell models progresses. However, in advance of that discussion, notice
at this stage that if phonons are assumed to have zero chemical potential, phonon = 0, we can write the following reaction
representing electronphonon interaction:
ei + phonon ej

(10)

where, given the low value of the phonon energy, ei and ej both
represent, for example, electrons in the CB. We arrive at
aC i + bC + phonon = aC j + bC aC = 0

(11)

which implies [based on (6)] that electrons are not hot anymore
but at the temperature Tc , which can now be interpreted as the
lattice temperature or room temperature. Similar equations
would hold for electrons in the VB and leading to aV = 0. For

Fig. 3.

Illustration of energy-selective contacts in the HCSC.

the moment, electronphonon interactions are assumed to not

occur.
III. HOT-CARRIER SOLAR CELL
Having a material capable only of interacting with photons
is not sufficient to make an actual solar cell. Selective contacts
allowing only particles with some prescribed characteristic to
pass through must be placed on this material. One type of conceivable, but perhaps unusual, contact is the energy-selective
contact. Its use was proposed by Wurfel [6] to preserve the physical consistency of the HCSC model. As illustrated in Fig. 3,
the energy-selective contacts allow only electrons with energy
e in the CB and holes with energy h to exit the core material toward the metallic contacts. The purpose of these contacts
is to reversibly cool the electrons from their hot temperature
Th to the room temperature of the contacts Ta by increasing
their electrochemical potential. In this way, electrons and holes

in the metal acquire an electrochemical potential, +

F and F ,
respectively, given by


e +
e e
Ta
Ta
+
F
=
F = 1
(12)
e e
kTh
kTa
Th
Th


h
h e
Ta
Ta

F
=
F = 1
h e . (13)
kTh
kTa
Th
Th
The output voltage of the cell is then given by


 1
1 +
Ta
F
=
V =
(
e h ) .
1

F
e
e
Th

(14)

It is important to stress why monoenergetic contacts allow

for reversibly converting electrons from the hot temperature Th
to the contact temperature Ta . In this respect, notice that (13)
and (14) can be satisfied only for a single energy value. This
mechanism is called reversible cooling at the contacts.
As stated in Section I, we emphasize that, in this cell, eV =

+
F F is not equal to the difference of the electrons and hole
electrochemical potentials in the core material, which is zero.
Elaborating on this model, it is possible to demonstrate that
the efficiency limit for solar energy conversion for this device
is 86.5% (assuming that the sun is at 6000 K and the cell at
300 K) and is obtained for a bandgap EG = 0 [6], [15]. This

MARTI AND LUQUE: ELECTROCHEMICAL POTENTIALS (QUASI-FERMI LEVELS) AND THE OPERATION

model for the HCSC is known as the impact-ionization model,

and it can be shown that the currentvoltage characteristics of
the cell from which this efficiency is obtained are formulated
from energy conservation principles with the only additional
condition being the reversible cooling at the energy contacts
previously mentioned [6], [15].
However, in the original paper in which Ross and Nozik
[1] introduced HCSCs, the authors assumed that nonradiative
generation-recombination mechanisms linking the CB and the
VB did not exist. These mechanisms include, in particular,
Auger recombination and impact ionization [the mechanisms
illustrated in Fig. 2(c) and (d)]. We propose herefor the first
time, to our knowledgethat in order for the following arguments to hold, it is also necessary to assume that Auger thermalization [the mechanism illustrated in Fig. 2(e)], which is
also a mechanism linking electrons in the CB to electrons in
the VB, does not exist either. Hence, for generality, instead of
nonradiative recombination processes, we should assume that
interband processes do not exist or occur at a much slower
rate than intraband processes. This is perhaps what one would
naturally expect to be the case in conventional semiconductors,
but, as explained in the following paragraphs, this leads to some
difficulties within the HCSC model.
Following the same arguments that led to (4), the absence of
interband processes implies that the electrochemical potential of
the electrons in the CB, i,C , and the electrochemical potential
of the electrons in the VB, i,V , are related to the energy of the
states by
i,C = aC i + bC

(15)

i,V = aV i + bV

(16)

which implies that electrons in the CB can be regarded as having

an energy-independent electrochemical potential and hot-carrier
temperature given, respectively, by
e,C = bC /(1 aC )

(17)

Th,C = Tc /(1 aC ).

(18)

Similarly, electrons in the VB can be described by an energyindependent electrochemical potential and hot-carrier temperature given by
e,V = bV /(1 aV )

(19)

Th,V = Tc /(1 aV ).

(20)

This means that electrons and holes have, in principle, not

only different electrochemical potentials (e,C and e,V ) but
also different hot-carrier temperatures (Th,C and Th,V ). We
emphasize that this reasoning is based on the fact that electrons
in the CB are isolated from the electrons in the VB due to
the absence of interband scattering processes; therefore, each
set gets its own constants a (namely, aC and aV ) and b
(namely, bC and bV ) that are not necessarily equal. However, as
explained in the discussion that motivated (8), to have a solar
cell, electrons must interact with photons and this interaction
connects the VB with the CB; therefore, electrons in the VB
and the CB have a temperature equal to that of the photons. We

1301

can simply designate this common temperature as Th , which

implies that aC = aV . However, because electrons in the CB
and VB now have different electrochemical potentials (because,
in general, bC = bV ), from (8), we now obtain
hv = e,C e,V

(21)

which is not equal to zero. Therefore, under this model, called

the particle-conservation model, photons have a chemical potential that equals the electrochemical potential difference between electrons and holes. This difference is not equal to eV
because the energy-selective contacts are still necessary to reversibly cool the electrons from their hot temperature to room
temperature. The currentvoltage characteristics of the cell cannot be obtained solely from energy-conservation rules; therefore, another equation is needed to calculate hv . This new
equation (not included here), the one that gives the model its
name, states that the difference between absorbed and emitted
photons must equal the number of electrons flowing through
the external circuit. This model is explained in detail in [7] by
Wurfel, who found peculiar results, including that the hot-carrier
temperature of the emitted photons exceeds the suns temperature at some points along the currentvoltage characteristic of
the cell as well as situations in which a complete currentvoltage
curve cannot be obtained. Given these unrealistic results, it is
preferable to drive the HCSC toward operating conditions in
which the impact-ionization mechanism applies. These conditions would be those in which intraband scattering processes
are slowed to the point that their rate is comparable with that of
the interband processes. By contrast, having e,C e,V = 0
in the particle conservation model introduces an extra degree of
freedom that results in an efficiency limit for the HCSC that is
slightly higher using this model than using the impact-ionization
model under maximally concentrated light.
Until now, electron interactions with phonons have been neglected. As discussed at the end of Section II, allowing the
electrons to interact with phonons leads to aC = 0 and aV = 0.
If this interaction is allowed, the HCSC concept is not possible
because, according to (11), the electrons are not hot anymore
and Th = TC . There are two methods to avoid this. One, obviously, is to assume the existence of a material in which electron
phonon interaction has been suppressed or made much slower
than electron-electron scattering. However, it must be remembered that we also wanted electronelectron interaction to be
slowed in order to create the HCSC impact-ionization model.
Thus, if we decrease the electronelectron interaction rate, it
is likely that electronelectron interaction will conflict at some
point with electronphonon interaction because not all interactions can remain slow simultaneously.
The second method is to assume that electrons interact, not
with all types of phonons but only with optical phonons [16].
Optical phonons are approximately stationary modes because
their energy does not depend on the wave vector; thus, their
group velocity in an ideal situation is zero. Therefore, they do
not transport heat to the cell walls (this condition should ideally
hold for all optical phonon branches). By contrast, the dispersion
curve of acoustic phonons is not flat; thus, they can transport heat
toward the cell walls. Therefore, an acoustic phonon would be

1302

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 4, OCTOBER 2013

in thermal contact with these walls, in contact with the ambient

or with heat sinks, and would be at room temperature. Because
optical phonons would be isolated from acoustic phonons, optical phonons could be set at a higher temperature. In fact, from
(10), they can be at the electron temperature Th while remaining consistent with phonons having a chemical potential of zero.
Interaction of optical phonons with acoustic phonons would be
suppressed by avoiding mechanisms such as Klemens decay by
the creation of phononic bandgaps. Allowing electrons to interact directly with acoustic phonons at room temperature would
collapse the HCSC model.

(a)

(b)

Fig. 4. (a) Illustration of the mechanism leading to the generation of multiple

electronhole pairs after the absorption of a single photon. (b) Direct radiative
recombination of a high-energy electronhole pair.

IV. IMPACT-IONIZATION SOLAR CELL

In the HCSC particle conservation model, we assumed that the
impact-ionization rate (in fact, all interband processes) would
be much slower than electronelectron interaction within bands
(intraband processes). Instead of energy-selective contacts,
we now assume that the cell is contacted with contacts selective to electrons and holes instead, i.e., with contacts that allow
only electrons or holes to pass through. The resulting solar cell
model will be the IISC or MEGSC. The practical difference
between the two is that the IISC model is applied to bulk semiconductors [17] and MEGSC is applied to quantum dots [4], but
fundamentally they can be considered the same model.
Electron- and hole-selective contacts have a wide range of
energies in which electrons can pass freely from the semiconductor to the contact and vice versa. Consequently, electrons at
the semiconductor CB must have the same Fermi-Dirac distribution function as the electron-selective contact and electrons
at the VB must have the same Fermi-Dirac distribution function
as the hole-selective contact. In particular, they are all at the
contact temperature Ta (aC = aV = 0). However, the electrochemical potential of the VB and its contact may be different
from that of the CB and its contact. By setting aC = aV = 0 in
(18) and (20) and using (17) and (19), we obtain in fact
e,C = bC

(22)

e,V = bV .

(23)

In this cell, however, the electronphonon interaction is not

suppressed and this gives an additional reason to consider aC =
aV = 0.
The voltage between terminals of this cell is eV = e,C
e,V , but hv = e,C e,V because (8) does not govern the
photon interaction. Instead, it is governed by
e1 + e2 + + eM + hv() eM +1 + eM +2 + + e2M
(24)
where M is the number of electronhole pairs that a photon
with energy is capable of generating. This interaction between
photon and electrons is shown in Fig. 4(a). Thus, we obtain
hv = M (e,C e,V ) .

(25)

This result was obtained in [3] and [18] using other thermodynamic arguments. Obtaining the result in (25) from (24)
provides new insight because (24) is conceptually different from

(8). The difference is that in (8), one high-energy photon can

generate a high-energy electronhole pair and this electronhole
pair can then recombine to generate a high-energy photon [see
Fig. 4(b)], whereas in (24), one high-energy photon can generate one high-energy electronhole pair, but this electronhole
pair cannot recombine to generate a high-energy photon unless
the electronhole pair has been created by a multiple-impactionization process. Equation (24) and Fig. 4(a) would then represent a multiple-particle collision process and not a succession
of events in time, and would become more and more unlikely
as the number of particles involved increases. Thus, in the same
way that nonradiative recombination is a loss mechanism for
cells working at the radiative limit, radiative recombination of
high-energy electronhole pairs becomes a loss mechanism for
cells based on impact-ionization or multiple-exciton generation.
Equation (25) also leads to a possible means of experimental
verification of the operation of an IISC/MEGSC because the
luminescent emission of these high-energy photons, well above
the bandgap and following the relationship in (25), should be
detectable experimentally. An additional argument for considering the phenomena represented in Fig. 4(a) as a multiple-particle
collision process is that, otherwise, we should accept interband
processes as the ones in Fig. 2(d) and (c) as occurring in the same
time scale what would take us, following the arguments that lead
us to (7), to the impossibility for having different electron and
hole electrochemical potentials and, therefore, to a solar cell
with an output voltage different from zero.
V. CONVENTIONAL SINGLE-GAP SOLAR CELL
The single-gap solar cell model, whose efficiency at the radiative limit was first studied by Shockley and Queisser (S&Q) [5],
can now be obtained from the IISC/MEGSC model by setting
M = 1. In a conventional single-gap solar cell, electrons and
holes have cooled to room temperature after conventional contacts selective to electrons and holes have been placed. Electrons
also cool down if the electronphonon interaction is allowed [see
(10)]. As in an IISC/MEGSC, electrons in the CB and VB obtain
their own electrochemical potential. The S&Q model is perhaps
the one most widely used to study the efficiency limit of solar
cells, and we do not elaborate on it here. To calculate this limit,
nonradiative recombination is assumed to not exist. One of the
authors of this paper reviewed it extensively in [19]. The S&Q
model is also the building block used to study the efficiency

MARTI AND LUQUE: ELECTROCHEMICAL POTENTIALS (QUASI-FERMI LEVELS) AND THE OPERATION

(a)

1303

(b)

Fig. 5. (a) Energy bandgap diagram of an IBSC. (b) Simplified energy-band

structure arising from the implementation of the IBSC concept with a quantum
dot where several energy levels emerge.

limit of multijunction solar cells [20], [21]. Traditionally, the

IISC/MEGSC and HCSC models were obtained from the singlegap solar cell model by following a bottom-up approach. Here,
on the other hand, we have followed a top-down approach in
which the single-gap solar cell model has been obtained starting
from a general core photovoltaic material.
VI. INTERMEDIATE-BAND SOLAR CELL
The IBSC model [2] can now be considered a generalization
of the single-gap solar cell model by assuming the existence
of a material with two bandgaps, EG 1 and EG 2 , instead of one
[see Fig. 5(a)]. Energy levels are now grouped into the CB,
VB, and IB. Electronphonon interaction within each band is
allowed so that electrons are at room temperature. In an ideal
cell, interband processes are assumed to be much slower than
intraband processes so that each band acquires its own electrochemical potential. As with a single-gap solar cell, this hypothesis is equivalent to neglecting nonradiative recombination. In
the particular case where the IBSC model is implemented with
quantum dots [22] [see Fig. 5(b)], our arguments in previous
sections also indicate the plausibility that each energy level in
the QD, 1 , 2 , . . . , n , acquires its own electrochemical potential [23] if these levels are spaced far enough apart in terms
of energy to prevent electronphonon interactions from causing electrons to transit from one state to another. To make an
actual cell, selective contacts are made to electrons in the CB
and VB, while the IB is left noncontacted. This can be achieved
by simply sandwiching the IB material between conventional
p- and n-type semiconductors [9]. Noncontacting the IB is an
additional argument for allowing energy levels in the QD having
their own electrochemical potential.
VII. SUMMARY AND CONCLUSION
We have revisited how electrochemical potentials for electrons and photons are distributed in the HCSC, IISC/MEGSC,
single-gap solar cell, and IBSC models. We started our discussion with a core material in which only electronelectron
interactions existed and arrived at the other cell models by

Fig. 6. Summary of the novel cells discussed in this study and how electron,
hole, and photon electrochemical potentials are distributed: IBSC, S&Q model,
IISC/MEGSC, and HCSC.

imposing different restrictions on this electronelectron interaction, assuming energy-selective contacts or electron-selective
contacts and turning ON and OFF different electronphonon
and electronphoton interactions. Throughout this paper, we
have explained the mechanisms by which electrons in each band
can be described by a single electrochemical potential, whether
electrons can be considered hot and whether the photon electrochemical potential equals eV. The discussion presented in this
paper is summarized in the form of a tree in Fig. 6. The trunk
is the core photovoltaic material and the branches are various
novel solar cells that stem from the hypotheses made.
When selective energy contacts are attached to the core
material and electronacoustic phonon interaction is neglected,
we obtain the HCSC. Still, there are two models for this cell:
the impact-ionization model [6] and the particle conservation
model [7]. In the first, electrons and holes share the same electrochemical potential; in the second, they do not. The chemical
potential of the photons emitted by the cell is equal to the electron and hole electrochemical potential difference; therefore,
this chemical potential is equal to zero in the impact-ionization
model and, in general, different from zero according to the

1304

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 3, NO. 4, OCTOBER 2013

particle conservation model. In both cases, and unlike the usual

single-gap solar cell, the electron and hole electrochemical potential difference in the core material is not equal to eV.
When we make selective contacts for the flow of electrons
and holes (instead of selective energy contacts), we arrive at
IISC/MEGSC. In this case, electrons and holes have different
quasi-Fermi levels, and their split in the core material is now
equal to eV. The chemical potential of the emitted photons is
M times eV, where M is the carrier multiplication factor (the
number of electronhole pairs generated per absorbed photon).
When selective contacts to electrons and holes are made and
the electronphonon interaction is not neglected, we arrive at the
conventional single-gap solar cell model [5]. Electrons and holes
have different electrochemical potentials, and their difference
in the core material equals eV. The chemical potential of the
emitted photons also equals eV. When several single-gap solar
cells are grouped, they form a multijunction solar cell. When
we add a third band (the IB) instead of just the CB and VB, we
obtain the IBSC in which electrons in each band have their own
electrochemical potentials.
In reviewing these topics, we have found some subtleties that,
to our knowledge, were previously disregarded. Thus, we have
found that Auger thermalization processes must be minimized to
implement an HCSC based on the particle-conservation model.
We have also found that in an IISC/MEGSC cell, radiative recombination of the high-energy electronhole pairs (or excitons)
in one step [see Fig. 4(b)] is a loss mechanism for this cell and
suggest that the absorption of a high-energy photon and creation
of multiple electronhole pairs in this cell should be considered
a multiple-particle process rather than a succession of steps in
order to explain why the photon chemical potential in this cell
equals M times eV.
REFERENCES
[1] R. T. Ross and A. J. Nozik, Efficiency of hot-carrier solar energy converters, J. Appl. Phys., vol. 53, pp. 38133818, 1982.
[2] A. Luque and A. Mart, Increasing the efficiency of ideal solar cells by
photon induced transitions at intermediate levels, Phys. Rev. Lett., vol. 78,
pp. 50145017, 1997.
[3] J. H. Werner, R. Brendel, and H. J. Queisser, Radiative efficiency limit
of terrestrial solar-cells with internal carrier multiplication, Appl. Phys.
Lett., vol. 67, pp. 10281030, Aug. 1995.
[4] A. J. Nozik, Quantum dot solar cells, Physica E, vol. 14, pp. 115120,
2002.
[5] W. Shockley and H. J. Queisser, Detailed balance limit of efficiency of
p-n junction solar cells, J. Appl. Phys., vol. 32, pp. 510519, 1961.

[6] P. Wurfel, Solar energy conversion with hot electrons from impact ionisation, Solar Energy Mater. Solar Cells, vol. 46, pp. 4352, Apr. 1997.
[7] P. Wurfel, A. S. Brown, T. E. Humphrey, and M. A. Green, Particle conservation in the hot-carrier solar cell, Progr. Photovoltaics: Res. Appl.,
vol. 13, pp. 277285, 2005.
[8] G. J. Conibeer, D. Konig, M. A. Green, and J. F. Guillemoles, Slowing
of carrier cooling in hot carrier solar cells, Thin Solid Films, vol. 516,
pp. 69486953, 2008.
[9] A. Luque and A. Mart, A metallic intermediate band high efficiency
solar cell, Progr. Photovoltaics: Res. Appl., vol. 9, pp. 7386, 2001.
[10] R. Strandberg and T. W. Reenaas, Limiting efficiency of intermediate
band solar cells with spectrally selective reflectors, Appl. Phys. Lett.,
vol. 97, pp. 031910-1031910-3, 2010.
[11] R. J. Ellingson, M. C. Beard, J. C. Johnson, P. R. Yu, O. I. Micic,
A. J. Nozik, A. Shabaev, and A. L. Efros, Highly efficient multiple exciton generation in colloidal PbSe and PbS quantum dots, Nano Lett.,
vol. 5, pp. 865871, May 2005.
[12] O. E. Semonin, J. M. Luther, S. Choi, H.-Y. Chen, J. Gao, A. J. Nozik, and
M. C. Beard, Peak external photocurrent quantum efficiency exceeding
100% via MEG in a quantum dot solar cell, Science, vol. 334, pp. 1530
1533, Dec. 16, 2011.
[13] D. Kondepudi and I. Prigogine, Modern Thermodynamics. Chichester,
U.K.: Wiley, 1998, p. 87.
[14] M. l. Bellac, F. Mortessagne, and G. G. Batrouni, Equilibrium and
Non-Equilibrium Thermodynamics. Cambridge, U.K.: Cambridge Univ.
Press, 2004, p. 336.
[15] A. Luque and A. Mart, Theoretical limits of photovoltaic energy conversion and new-generation solar cells, in Handbook of Photovoltaic
Science and Engineering, A. Luque and S. Hegedus, Eds. Chichester,
U.K.: Wiley, 2011.
[16] G. J. Conibeer, J. F. Guillemoles, and M. A. Green, Phononic band gap
engineering for hot carrier solar cell absorbers, in Proc. 20th Eur. Photovoltaic Solar Energy Conf., 2005, pp. 3538.
[17] R. Brendel, J. H. Werner, and H. J. Queisser, Thermodynamic efficiency
limits for semiconductor solar cells with carrier multiplication, Solar
Energy Mater. Solar Cells, vol. 4142, pp. 419425, Jun. 1996.
[18] A. Luque and A. Mart, Entropy production in photovoltaic conversion,
Phys. Rev. B, vol. 55, pp. 69946999, 1997.
[19] G. L. Araujo and A. Marti, Absolute limiting efficiencies for photovoltaic
energy conversion, Solar Energy Mater. Solar Cells, vol. 33, pp. 213240,
1994.
[20] A. Devos and H. Pauwels, On the thermodynamic limit of photo-voltaic
energy-conversion, Appl. Phys., vol. 25, pp. 119125, 1981.
[21] A. Mart and G. L. Araujo, Limiting efficiencies for photovoltaic energy
conversion in multigap systems, Solar Energy Mater. Solar cells, vol. 43,
pp. 203222, 1996.
[22] A. Mart, L. Cuadra, and A. Luque, Quantum dot intermediate band solar
cell, in Proc. 28th IEEE Photovoltaic Spec. Conf., 2000, pp. 940943.
[23] A. Luque, P. G. Linares, E. Antoln, E. Canovas, C. D. Farmer,
C. R. Stanley, and A. Mart, Multiple levels in intermediate band solar cells, Appl. Phys. Lett., vol. 96, pp. 013501-1013501-3, 2010.

Authors photographs and biographies not available at the time of publication.

2-2

Invited

Current Status of Thin-Film Solar Cells and Future Prospects

Makoto Konagai
Tokyo Institute of Technology
2-12-1-NE15, O-okayama, Meguro-ku, Tokyo 152-8552, JAPAN
Phone: +81-3-5734-2554

Fax: +81-3-5734-2897

E-mail: konagai.m.aa@ m.titech.ac.jp

An overview of the recent status of photovoltaic (PV) power generation is first presented. Next, the Japanese roadmap for the research and
development (R&D) of PV power generation and the progress in the development of various solar cells are explained. In addition, the
present status and future prospects of Si thin-film solar cells, which are expected to enter the stage of full-scale practical application in the near
future, are described. Thus far, a-Si/c-Si tandem solar cell modules with conversion efficiency exceeding 13% have been reported. In
addition, triple-junction solar cells are introduced, as well as innovative thin-film full-spectrum solar cells and Si nano-wire solar cells, whose
target year of realization is 2030-2050.
1.

Introduction
In 2011, the cumulative capacity of installed PV systems in

Japan was more than 4 GW.

for business use (14 yen/kWh) by 2017 and to that feasible for
industrial use (7 yen/kWh) by 2025. To realize these electricity

For simplicity, assuming the

costs, numerical targets were set for the manufacturing cost,

average annual hours of solar radiation to be 1000 h/year, the

lifetime, and conversion efficiency of PV modules; for example, a

above value is converted into an annual power generation of 2109

kWh (4 GW 1000 h). Assuming the basic unit of lifecycle CO2
emission due to PV power generation to be 0.05 kg-CO2/kWh, the
reduction of CO2 emission per kilowatt hour achieved by installing

conversion efficiency of 25% is targeted for practical PV modules

by 2025.
Several breakthroughs must be achieved to attain the targets in

PV systems is calculated to be approximately 0.3 kg-CO2/kWh.

PV2030+. Today, possible materials for solar cells are being

When this is multiplied by the above value of 0.3 kg-CO2/kWh,

investigated in many research fields. Such materials, regardless

we obtain 1.210 kg-CO2. This is still only about 0.1% of the

of their type, are being developed as a result of research

current CO2 emission (1.3 billion tons) in Japan.

Japan aims to

breakthroughs to achieve the goals in PV2030+, and the

have a cumulative capacity of PV systems of 54 GW by 2030.

discontinuation of the development of other materials is currently

By a simple calculation assuming proportionality, this would lead

not being considered. The R&D of a wide range of materials in

to a 1.2% reduction of CO2 emission.

The technological development of PV power generation
started in 1974 when the first oil crisis occurred, and the main aim
in those days was to develop an energy alternative to oil.
Currently, however, the main purpose of developing PV systems is
to reduce CO2 emission. Moreover, by 2100, energy resources
on Earth will be almost exhausted and solar energy will have to
serve as the main energy source for humans. Thus, the research
and development (R&D) of PV power generation is an
ultralong-term grand project for developing a new stable energy
source that is and will be ongoing in the near and more distant
future. A long-term perspective and continuous and ambitious
R&D strategies will be indispensable.
2.

PV Roadmap
Figure 1 shows PV2030+ presented by NEDO. The milestones

in this roadmap are the reduction of electricity cost to that feasible

both organic and inorganic materials is ongoing because current

material technology is expected to make significant breakthroughs.
3. Type, features, present technological status, and production
capacity of solar cells
Table 1 summarizes confirmed terrestrial cell efficiencies
measured under the global AM1.5 spectrum1). Bulk Si solar cells
account for approximately 85% of solar cells manufactured for
power generation. The conversion efficiency of their commercial
large-area modules is 15-18%. On the other hand, thin-film solar
cells are expected to enter full-scale practical application from now
on.

The conversion efficiency of commercial amorphous Si

(a-Si) solar modules is approximately 6-10%.

Among the

thin-film solar cells, the development of a-Si/microcrystalline Si

(c-Si) double-junction tandem solar cells and Cu(InGa)Se2 solar
cells is being prioritized. In addition, dye-sensitized solar cells

AM-FPD '13

21

Fig.1 Roadmap for development of PV system

ms, PV2030+. This is a revised version of PV2030 preparedd in 2004 (from NEDO website).
and organic semiconductor solar cells are beinng developed with

sunlight and is reversible uponn thermal treatment at a high

their practical application targeted for 2020 or aft

fter.

temperature.

The light inducedd degradation of a-Si solar cells

From the viewpoint of ultrahigh convversion efficiency,

has been reduced through the clarification of the mechanism

triple-junction solar cells with a semiconductorr comprising III-V

behind the SW effect and thhe development of preventive

compound semicondutors are outstanding; their conversion

technologies; however, the compplete prevention of light induced

efficiency exceeds 40%. They are used in conncentrator systems,

degradation has not yet been achieeved.

in which they exhibit their superior conversion efficiency.

e

Amorphous Si solar cellss have achieved a stabilized

Statistical data on the production of each typpe of solar cell and

conversion efficiency of 6-8% in

i practical large-area modules,

future plans for the installation of manufactturing facilities by

although the problem of light induced degradation remains.

worldwide companies are described in detail in the PV Status

However, their module conversionn efficiency is still unsatisfactory,

Report. The data are provided by the Joint Reesearch Center, EU

and therefore, their application is mainly

m
limited to megawatt solar

Commission, and can be freely downloaaded.

systems. To apply a-Si solar ceells to residential PV systems, a

The total

worldwide production capacity was 30 GW in 2011.

2

higher module conversion efficieency (at least 12%) is required.

Increasing the quality of a-SiGe ussed in a-Si solar cells is a difficult

4. Present Status of Development of Si Thin-F

Film Solar Cells
In general, a-Si solar cell films arre fabricated by

challenge. With the aim of devveloping high-efficiency devices,

high-frequency plasma chemical vapor depositioon (CVD) at 13.56

double-junction or a-Si/c-Si-baseed triple-junction solar cells.

the focus of development haas been shifting to a-Si/c-Si

MHz; however, very high frequency (VHF) plaasma CVD at about

The development of c-Si is now

n rapidly progressing from this

60 MHz is considered effective for high rate deposition

d
and has

viewpoint. Microcrystalline Si iss a material containing an a-Si

already reached the phase of practical applicaation. To increase

phase and a polycrystalline Si phhase. The band gap of c-Si is

the quality of films, it is important to control the proportion of SiH3

considered to be equivalent to thaat of single-crystal Si, i.e., 1.1 eV.

and atomic hydrogen (H) on their growing surfaace.

The absorption coefficient of c-Si is greater than that of

Amorphous Si single-junction solar cells baasically comprise a

single-crystal Si because of thhe contribution of amorphous

glass substrate, a transparent conducting oxide (T

TCO) layer, a p-i-n

components. The absolute valuue of the absorption coefficient

structure, a backside reflection layer, and an elecctrode. SnO2 and

depends on the percentage of amoorphous components.

ZnO films with a textured surface for confininng light are used as
the TCO layer.

When using c-Si as a materiial in solar cells, a light-absorbing

layer with a thickness of 1-3 m
m is required to allow a sufficient

Amorphous Si exhibits a light induuced degradation

amount of photons to be absorbedd because its absorption capability

phenomenon called the Staebler-Wronski (SW

W) effect. The light

is low in the visible-light regionn. This value is approximately

induced degradation of a-Si solar cells is generally

g
stabilized

10-fold greater than the thicknesss of the layer in practically used

within approximately three months from the sttart of exposure to

a-Si solar cells; thus, the develoopment of high rate deposition

22

AM-FPD '13

Table 1 Confirmed terrestrial celll efficiencies measured under the global AM1.5 spectrum (10000W/m2) at 25.

technology with a deposition rate of 1-5 nm

m/s is required to
increase the throughput of the fabrication processs.

completely suppressed.
t
solar cells include
The development targets of triple-junction

Currently, an efficiency of 14.7% has beeen achieved by an

those for a-Si/c-Si/c-Si and a--Si/a-SiGe/c-SiGe structures as

a-Si/c-Si tandem solar cell, in which a TCO innterlayer is inserted

well as for the a-Si/a-SiGe/c-S

Si structure. The stabilized cell

in the tunnel junction between the top and botttom cells, although

conversion efficiency of 13.4% (V

( oc=1.963 V, Isc=9.52 mA/cm2,

this figure is the initial conversion efficiency. The role of the

FF=0.719, area=1 cm2) hasn bee achieved by LGE for

TCO interlayer is to reflect part of the incident ligght towards the top

a-Si/c-Si/c-Si. The total photoocurrent in the triple-junction cell

cell by utilizing the difference in refractive inddex at the junction

is 28.5 mA/cm2, revealing that thhe sensitivity of the bottom cell

and to increase the photocurrent components from the top cell.

(c-Si) satisfactorily covers the lonng-wavelength region.

On the basis of these achievements, the produuction of a-Si/c-Si

The improvement of light absorption performance is the

top-priority task in the realization of triple-junction solar cells with

tandem solar cells has been launched.

In PV2030+, the target module conversionn efficiencies of Si

the maximum theoretically possible performance.

thin-film solar cells are 14% for 2017 and 188% for 2025. To
attain these goals, conventional a-Si/c-Si doubble-junction tandem
solar cells must be superseded by triple-junnction solar cells.
Theoretically, it is very difficult to attain the above goals using
conventional double-junction solar cells.

However, it will

become possible if the light induced degradatiion of a-Si can be

5. Establishing Global Researrch Centers to Develop Solar

Cells
Currently, I am acting as a grroup leader of two collaborative
universityindustry research consortiums that focus on innovative

AM-FPD '13

23

solar cells. One is the thin-film full spectrum solar cells project

possible material.

of METI/NEDO and the other is the Si nannowire solar cells

project of MEXT/JST. In each project, whichh involves a study
period of 57 years, we are cooperating with domestic research

Silicon Nano-Wire Solar Cells

Si is a semiconductor with 1.1 eV band gap, way different

organizations as well as foreign countries with

w a budget that

from 1.5 eV band gap which is logically the most suitable band

exceeds four billion yen over the entire periodd. The themes of

gap for converting solar energy. The

T energy conversion efficiency

both projects are challenging and we require a significant change

of silicon solar cells is about 25%

% and getting closer to 28% to

in mindset to achieve our goals, but the fruit of our

o success will be

29%, which is considered the theeoretical limitation. This research

sweet. I discuss the organizations and funcctions of the two

is aiming to realize a unique technnology of Bandgap control of Si

collaborative universityindustry research consoortiums.

using nano-wire to break the phyysical limitation of band gap and

drastically improve the performannce of silicon solar cells. At the

Thin Film Full Spectrum Solar Cells

same time, this research tackles ann extremely difficult challenge of

Where is the upper limit for the converrsion efficiency of

realizing ultra-high efficiency with

th a tandem system of nano-wire

thin-film solar cells? Theoretically, it is possible to achieve a

silicon solar cell and heterojuncttion silicon solar cell. Based on

conversion efficiency of 35% or higher for quinttuple-junction solar

these technologies, we aim to devvelop silicon solar cells with 30%

cells if high-quality light-absorbing layers with different

d
band gaps

energy conversion efficiency that breaks the conventional

from 0.6 to 2.2 eV can be fabricated. As an

a example, Fig. 2

wisdom(Fig.3).

shows a schematic of a thin-film quintuple-junction structure

This project is implementeed with the goal to develop the

currently being developed by myself, the grroup leader of the

open and the most advanced reseaarch and development base in the

NEDO Innovative PV Technology -Thin Fiilm Full Spectrum

world on renewable energies undeer the Basic Policy for Recovery

Solar Cells with Low Concentration Ratios Note

N that although

and Reconstruction of Fukushim

ma. Many excellent researchers

the connection method for each cell and the opptical coupling are

from Japan and overseas will gathher at the research facility that will

not explained in the figure, not only material devvelopment but also

be constructed in Koriyama, Fukuushima, by the end of fiscal year

solutions to these and other issues are neeeded to improve

2013 and engage in unique andd challenging research to realize

conversion efficiency.

nano-wire solar cells. The researrch outcomes of this facility in

When a quintuple-junction structure is coomposed of only

Fukushima will be widely transmiitted throughout the world.

Si-based thin films, the development targets of the first cell are Si
thin films with a band gap of 2 eV or higher,, such as a-Si1-xOx.
Amorphous Si has strong potential as the mateerial for the second

7. Conclusions
Today, the strongest rival to thhin-film solar cells is bulk Si solar

cell. For the third cell, with a band gap of appproximately 1.4 eV,

cells produced by Chinese compaany. Bulk Si solar cells can be

a-SiGe and Si-based quantum dots are powerfful candidates. For

produced at a very low cost off 0.5$/W in China. A marked

the fourth cell, c-Si is suitable. For the fifth cell as the bottom

improvement of energy converrsion efficiency is essential for

layer, c-SiGe is highly suitable, and Ge thiin film is another

thin-film solar cells to compete wiith bulk-Si solar cells in the future.
Si thin-film solar cells are exppected to achieve a conversion
efficiency of 20% by exploiting new materials developed in the
future and by applying a multijuncction structure. Now is the time
to bring together the experiencee and know-how of researchers
working in the area of photovoltaaics to speed up the development
Research
plan for develooping high efficiency Si NW/c-Si
ofFig.3
materials
and devices.
tandem solar cells for FY2012-20016.
References

1) Martin A. Green1, Keith Emeryy, Yoshihiro Hishikawa, Wilhelm

Warta and Ewan D. Dunlop, Progrress in Photovoltaics, , 2012;
Fig.2 Example of quintuple-junction thin-fiilm structure with
theoretical conversion efficiency of 35% or higgher.

24

AM-FPD '13

20:606614

The Influence of Bi and Al Content on the Deep Levels in

InGaP and AlInGaP Solar Cell Structures
A. Khan1, A. Naik,1 J. Gou2, W. Al- Assadi1, A. Gapud3
1
Department of Electrical and Computer Engineering, University of South Alabama, USA.
2
Department of Mechanical & Aerospace Engineering, University of Central Florida, USA.
3
Department of Physics, University of South Alabama, USA.

Abstract InGaP/InGaAs/Ge 3-juction solar cells have proved

to the most effective solar cells for space applications. Recently,
AlInGaP was proposed as a top layer, which has the future prospects of super high efficient solar cells. However, the efficiency
of the AlInGaP based multi-junction solar cells that are under
investigation is not up to the expected levels. In order to understand the reason behind this low efficiency, we have explored
role of Al content on native defects in the performance of
AlGaInP solar cells. In addition, we have also explored the interaction of Bi on p-type InGaP and p-type AlInGaP solar cell
structure. The Deep level transient spectroscopy (DLTS) was
used to investigate the different characteristics of these observed
defects and their possible role in low lifetime/efficiency of the
solar cells. The four types of solar cell structures under investigation showed four electron traps and two hole traps. Detailed
depth profile analysis showed that the hole traps generated in the
Bi doped InGaP and Bi doped AlInGaP act as strong recombination centers. In view of this study, it has been concluded that the
deep levels induced by Al Impurity and Bi impurity plays an
important role in the low efficiency of InGaP / AlInGaP solar
cells.
Index Terms DLTS, InGaP, AlInGaP, Solar cells, Defects,
Traps.

I. NTRODUCTION
InGaP/InGaAs/Ge 3-juction solar cells have proved
to the most effective solar cells for space applications [16]. However, these cells are not ideally suited to Air mass
Zero solar spectrum. The top layer in these cells is current
rich whereas the middle layer is current poor. So in order
to compensate this current, the top layer is made thin, allowing unabsorbed sunlight to penetrate to the lower junction. This increases the current in the middle layer and
hence contributing to the overall current in the multijunction cell. A preferable, but more challenging approach
is to find a higher band-gap material to replace the InGaP
material, thereby increasing the current in the lower junction and increasing the top-cell open-circuit voltage [2].
Recently, AlInGaP was proposed as a top layer, which has
the future prospects of super high efficient solar cells.

978-1-4799-3299-3/13/$31.00 2013 IEEE

Table I SAMPLES USED IN THE IS STUDY

Sample Studied

Al
Composition

Bi Doped

InGaP

0%

No

AlInGaP

4%

No

Bi doped InGaP

0%

Yes

Bi doped AlInGaP

4%

Yes

However, the efficiency of the AlInGaP based multijunction solar cells that are under investigation is not up to
the expected levels. In order to understand the reason behind this low efficiency, we have explored role of Al content on native defects in the performance of AlGaInP solar
cells. In addition, we have also explored the interaction of
Bi on p-type InGaP and AlInGaP solar cell structures. The
Deep Level Transient Spectroscopy (DLTS) was used to
investigate the different characteristics of these observed
defects and their possible role in low lifetime/efficiency of
the solar cells. In view of depth analysis it has been concluded that the deep levels induced by Bi and Al play an
important role in governing the efficiency of Bi doped
InGaP and Bi doped AlInGaP solar cells.
II. EXPERIMENTAL
All samples investigated during the course of this
study were obtained from the EMCORE Corporation as a
part of collaboration with the University of South Alabama. Four samples of solar cells, as shown in table-1
with two different compositions of aluminum were studied
using the DLS-83D spectrometer.

2770

The device structures of solar cell samples are shown in

the Fig. 1. The samples studied in the research have an
n++ Si doped AlxInGaP emitter. The base layer is Zn

tion to 1.15 x 1015 cm-3. Hence, it can be concluded that

the addition of Al is introducing the new defects

500A n++ Si: GaAs contact

500A nid GaAs
AlInGaP

275 A n+ Si: AlInP window

800A n++ Si: AlxGaInP emitter
1m 1 x 1017 Zn: AlxGaInP base
1m 3 x 1017 Zn: GaInAs buffer
p-type GaAs substrate
Fig. 1. Structure of P-AlInGaP solar cell used in this study

doped AlxInGaP with a doping concentration of 1 x 1017

cm-3. The buffer layer is Zn doped GaInAs with a doping
concentration of 3 x 1017 cm-3. Since the base is a p-type
layer, these samples are referred as p-type InGaP solar
cells. Four samples were studied in this research. The first
sample is a p-type InGaP. The second sample is a p-type
AlInGaP solar cell with 4% Al content. The third sample
is p-type Bi-doped InGaP. The fourth sample is Bi doped
AlInGaP with 4% Al.
Conventional current-voltage (I-V) and deep level
transient spectroscopy (DLTS) measurements were performed, coupled with capacitance-voltage (C-V) measurements in order to determine the doping concentration
in the p-type base layer. The characteristics of the defects
were then extracted from the DLTS spectra.

Fig. 2. DLTS spectra of p-type InGaP, p-type AlInGaP

(4%).

closely related to native defects, which could be the cause

of the lowering the efficiency of the AlInGaP solar cell
compared to the InGaP.
B. Influence of Bi on p-type InGaP solar cell structures
The addition of Bi in the p-type InGaP solar cell
structures has produced interesting new results. Under the similar bias conditions, the DLTS spectra appear to be as shown in the Fig. 3. As shown in the
Fig.3, doping of Bi completely compensated the minority electron trap (E1). In addition, Bi introduced a
new electron peak E3, with activation energy of (Ec0.17 eV) below the conduction band and trap concentration of 4.6 x 1015 cm-3.

III. RESULTS AND DISCUSSION

A. Influence of Al on p-type InGaP solar cell
structures
Doping of Al in InGaP solar cell structure affected
the deep levels, as shown in the Fig. 2. The addition of Al
noticeably enhanced the concentration of the peak (E1),
with a minor shift in the general position of the peak. An
electron defect (E1), with approximately similar energy
(Ec-0.22 eV) below the conduction band, has been reported previously [4]. The measured activation energy of the
defect in p-type AlInGaP (E2) was found to be (Ec-0.27
eV) below the conduction band. The trap (E1) concentration in p-type InGaP was found to be 2.9 x 1014 cm -3.
Addition of Al increased the intrinsic trap (E2) concentra-

978-1-4799-3299-3/13/$31.00 2013 IEEE

Fig. 3. DLTS spectra of p-type InGaP, p-type Bi doped InGaP

(4%)

2771

C.

Hole Levels

A significant difference was observed in the DLTS

spectrum of as grown InGaP solar cell structure and Bi
doped InGaP as shown in the Fig. 3. The addition of Bi in
the InGaP solar cell induced a hole trap (H1) with activation energy (Ev+0.77 eV) above the valence band and
concentration of 1.7 x 1016 cm-3. The broad DLTS spectrum of H1 suggests that several overlapping deep levels
are involved in the origin of the H1 defect.
The depth profile of the trap H1 revealed that the defect is concentrated near the emitter-base junction, so it
captures electrons from the emitter side. In addition, concentration of H1 is considerable in the base region as well,
this indicates that the defect also traps holes injected during the majority carrier pulse. The above mentioned characteristics indicate that trap H1, is capable of interacting
with both electrons and holes. Thus, trap H1 in Bi doped
InGaP is acting as a recombination center and limiting the
solar cell performance reaching its optimum efficiency.
D.

Influence of Al on p-type Bi doped InGaP solar cell

structures

The doping of Al in Bi doped InGaP solar cell structure affected the deep levels as shown in the Fig. 4. The
addition of Al noticeably enhanced the concentration of
the peak (E3), with a minor shift in the general position of
the peak. The effect of Al in Bi doped InGaP is found to
be similar to the one discussed in section A.
The trap concentration of electron peak E4 increased
from 4.6 x 1015 cm-3 to 1.7 x 1016 cm-3, with the addition
of Al. Hence, it can be concluded that addition of Al is
introducing the new defects closely related to native defects, which could be the cause of lowering the efficiency
of Bi doped AlInGaP solar cell compared to the Bi doped
InGaP.
Addition of Al, did not affect the majority carrier
traps observed in Bi doped InGaP solar cell structure. As
shown in Fig 4, trap H1 reappeared as H2, with an activation energy of (Ev+0.79 eV) above the valence band with
concentration of 2.6 x 1016 cm-3.
E.

Bi-InGaP

Fig. 4. DLTS spectra of p-type Bi doped InGaP,

p-type Bi doped AlInGaP.

the InGaP, AlInGaP, Bi doped InGaP and Bi doped

AlInGaP were found to be 1.7 x 1017 cm3, 5.4 x 1016 cm3,
4.8 x 1015 cm3, and 2.4 x 1014 cm3, respectively. The total
concentration of the H1 and H2 traps for the Bi doped
solar cell samples was found to be too low compared to
the reduction in the free carrier concentration, which indicates that other traps, which have not been detected by
DLTS measurements, are also acting as compensating
centers. As stated in [5], when both majority and minoritycarrier captures are present during the capture process,
some important defects may remain undetected by the
DLTS.

Carrier removal effect in Al and Bi doped InGaP

From the capacitance-voltage measurements shown in

Fig. 5. The shallow level concentrations of the solar cell
samples were found to decrease one order of magnitude
with the addition of Al and two orders of magnitudes with
the addition of Bi to the p-type InGaP solar cell structures.
This indicates that the hole trap H1 may also act as a
compensating center and plays a major role in the carrier
removal effect, which can significantly affect the solar cell
efficiency. The observed free carrier concentrations for

978-1-4799-3299-3/13/$31.00 2013 IEEE

Fig. 5. Shallow level concentration of the p-type

solar cell samples

2772

VI. SUMMARY
In this study the influence of Al and Bi content on deep
levels in p-type InGaP has been reported. DLTS temperature scans were performed with the aim to investigate these deep levels. Four electron traps E1, E2, E3 and E4 were
observed in four different structures of solar cells.
Two hole traps H1and H2 were observed in Bi doped
InGaP and Bi doped AlInGaP solar cell structures. The
activation energies of the traps E1, E2, E3 and E4 were
calculated to be 0.22 eV, 0.27 eV, 0.17 eV, and 0.19 eV,
respectively. The activation energies of majority carrier
hole traps H1 and h2 were calculated to be 0.77 eV and
0.79 eV, respectively.
The DLTS temperature spectra of traps E1 and E2 were
compared to analyze the impact of Al on the solar cells.
Based on this comparison, it was observed that addition of
Al is noticeably enhancing the intrinsic traps in p-type
InGaP solar cell structure and thus limiting the AlInGaP
solar cell from attaining its maximum efficiency.
In addition to the impact of Al on the solar cell, traps E1
and E3 were compared to evaluate the impact of Bi on the
solar cell. Based on this comparison, it was observed that
Bi introduced additional deep levels which were found
closer to the emitter-base junction. These results suggested that the trap level H1 in Bi doped InGaP solar cell is
acting as a recombination center and is thus limiting the Bi
doped InGaP device from attaining its maximum efficiency.
The DLTS temperature spectra of traps E3 and E4
were compared to observe the impact of aluminum on the
on Bi doped InGaP solar cell solar cell. Based on this
comparison, it was observed that Al enhanced the native
defect concentration in Bi doped InGaP and thus the dominant peak E4 is seems responsible for lower efficiency in
Bi doped AlInGaP solar cell as well.
In addition to the above results, the shallow level
concentrations of the solar cell samples were found to
decrease one order of magnitude with the addition of Al
and two orders of magnitudes with the addition of Bi to
the p-type InGaP solar cell structures. This indicates that
the hole trap H1 may also act as a compensating center
and plays a major role in the carrier removal effect, which
can significantly affect the solar cell efficiency.

REFERENCES
[1] F. Dimroth, C. Baur, M Meusel, S. van Riesen, and A.W.
Bett. 5-junction III-V solar cells for space applications, 3rd
World Conference and Exhibition on Photovoltaic Energy Conversion, 2003.
[2] T. Takamoto et al., Multi-junction solar cell technologies high efficiency, radiation resistance and concentrator applications, 3rd World Conference and Exhibition on Photo voltaic
Energy Conversion, 2003.
[3] Chiang P. K., Kurt D. D., Cavicchi B. T., Bertness K. A.,
Kurtz S. R., and Olson J. M., Large-area InGaP/GaAs/Ge
multijunction solar cells for space applications, proc. 1st World
Conf. Photovoltaic Energy Conversion, pp. 2120-2123 (1994).
[4] A. Khan, S. Marupaduga, S. Anandkrishnan S. Alam M. et
al, Radiation response analysis of wide-gap P-AlInGaP for
super high-efficiency space photovoltaics, Applied physics
letters, Vol. 85, pp. 5218-5220 (2004).
[5] A. Khan., Jacques C. Bourgoin., and Tatsuya Takamoto.,
Thermal annealing study of 1 MeV electron-irradiation induced
defects in n+/p InGaP diodes and solar cells, Journal of Applied
Physics, Vol. 91, pp.2391-2397(2002).

[6] A. R. Mohammad., F. Bastiman., C. J. Hunter., S. J.

Sweeney., The effect of Bi composition to the optical
quality of GaAs1-xBix, Applied Physics Letters, Vol. 99,
0402107 (2011).

ACKNOWLEDGEMENT
Special note of recognition goes to Mr. Aurthur B.
Cornfeld (EMCORE) for providing the samples and solar
cell data.

978-1-4799-3299-3/13/$31.00 2013 IEEE

2773

Qualification and Production of Emcore ZTJ Solar Panels for Space

Missions
Navid Fatemi1, John Lyons2 and Mike Eskenazi3
1

Emcore Photovoltaics, Albuquerque, New Mexico 87122, USA

NASA Goddard Space Flight Center (GSFC), Greenbelt, Maryland 20771, USA
3
ATK Space Systems, Goleta, California 93117, USA

Abstract Emcores latest generation InGaP/InGaAs/Ge ZTJ

triple-junction space-grade high-efficiency solar cells have been in
volume production since 2009, with over 300,000 flight cells
produced to power more than 35 separate satellites. The ZTJ
cells, CICs (Coverglass-Interconnected-Cell) and solar panels
have also been characterized and qualified to both the AIAA-S111 and AIAA-S-112 standards. More than 10 life-cycle coupon
panels have been thermal cycled to temperature extremes
representing varied orbital conditions such as GEO
(geosynchronous), LEO (low-earth) and highly elliptical orbits.
In its larger 1-cell-per-wafer form factor, the ZTJ solar cell has
been used to manufacture solar panels for a dozen NASA and
other commercial spacecraft. Currently Emcore Photovoltaics is
under contract to NASAs Goddard Space Flight Center (GSFC)
to build and deliver the solar panels for the Magnetospheric
Multiscale (MMS) Mission and to ATK Space Systems for the
Commercial Resupply (CRS) mission to the International Space
Station (ISS) and the AMOS-6 commercial telecommunication
satellite. In this paper, the results of the qualification, production
and testing of the life-cycle coupons and the solar panels for these
missions is presented. In addition, the initial on-orbit telemetry
performance results of NASAs NuStar (Nuclear Spectroscopic
Telescope Array) spacecraft solar array will be presented. The
NuStar spacecraft was launched in June of 2012 and built by
Orbital Sciences Corporation It is powered by an ATK solar
array populated with ZTJ solar cells.
Index Terms Multi-junction, space solar cells, MMS, CRS,
AMOS-6, NuStar

I. INTRODUCTION
Solar arrays populated with high-efficiency III-V based
multi-junction solar cells have been powering the payloads of
earth orbiting satellites and interplanetary spacecraft for well
over a decade. Due to their higher performance compared to
silicon cells, multi-junction cells have enabled the realization
of higher power and larger capacity satellites.
The conversion efficiency of the InGaP/InGaAs/Ge triplejunction (3J) solar cells, under Air-Mass Zero (AM0)
illumination at 28C, has been steadily improved from about
23% to approximately 30% at beginning-of-life (BOL) since
the year 2000 [1-2]. The vast majority of the scientific,
military, exploration and telecommunication spacecraft now
use the highest efficiency 3J solar cells available. The power
requirements for these spacecraft range from ~1 kW to as
much as 25 kW at end-of-life (EOL).

978-1-4799-3299-3/13/$31.00 2013 IEEE

One of the highest performance solar cells in volume

production is Emcores ZTJ cell, with an average AM0
conversion efficiency of ~29.5%. More than 300,000 ZTJ
flight cells (or ~350 kW at BOL) have been manufactured
since 2009 to power more than 35 spacecraft. An AM0
efficiency histogram of the ZTJ cell is shown in Fig. 1.

28.0

28.5

29.0

29.5

30.0

30.5

AM0 Efficiency

Fig. 1 AM0 (one-sun at 28C) Efficiency histogram of ZTJ

solar cells in production
Historically, the solar cells used to populate flight solar
panels have been of 2-per-wafer form factor, meaning that two
large-area cells are produced from a 100-mm dia. germanium
wafer. The area for these cells typically ranges from ~20 cm2
to ~30 cm2. In the past few years, however, larger 1-per-wafer
solar cells have been gaining acceptance with users due to
their lower cost of assembly and integration on a $/W basis.
These cells are typically square or nearly square and have
areas as large as ~61 cm2 [3].
To date, nearly 60 kW of 1-per-wafer ZTJ cells have been
manufactured at Emcore to power a dozen spacecraft. The ZTJ
1-per-wafer cell is expected to outpace the ZTJ 2-per-wafer
cell production in the coming years, as more missions begin to
take advantage of the cost saving benefits of larger area cells.
An example of an Emcore 1-per-wafer ZTJ solar cell is shown
in the photograph of Fig. 2.

2793

III. SOLAR PANEL AND ARRAY CONFIGURATIONS

MMS
32 flight solar panels and 2 spare panels are built for the
four MMS satellites, capable of a total output power exceeding
5 kW at EOL at normal sun incidence. The solar array for
each spacecraft consists of 8 body-mounted panels. Each
panel is populated with 9 strings of 18, 1-per-wafer ZTJ solar
cells, for a total of 162 cells. Greater than 90% panel packing
factor is achieved using the ZTJ cells shown in Fig. 3.

Fig. 2 Emcore 1-per-wafer ~8x8 cm ZTJ solar cell

II. FLIGHT MISSIONS BACKGROUND
Emcore is currently under contract to supply solar panels for
12 separate satellites populated with the 1-per-wafer ZTJ solar
cells. These include panels for NASAs Magnetospheric
Multiscale (MMS) mission, managed by the Goddard Space
Flight Center (GSFC). The MMS mission is a Solar Terrestrial
Probes Program within NASAs Heliophysics Division. It
consists of four identically instrumented spacecraft in a highly
elliptical orbit. It will use Earth's magnetosphere as a
laboratory to study magnetic reconnection, which is a
fundamental plasma-physical process that taps the energy
stored in a magnetic field and converts into heat and kinetic
energy in the form of charged particle acceleration and largescale flows of matter.
Emcore is also under contract with ATK Space Systems to
deliver solar panels for two other missions. These are
NASAs 4th through 8th Commercial Resupply Service (CRS)
spacecraft to the International Space Station (ISS) and the
Affordable Modular Optimized Satellite (AMOS-6)
commercial GEO telecommunication satellite. ATK Space
Systems UltraFlex solar arrays are used for the CRS 4-8
spacecraft.
Under the CRS program, Orbital Sciences Corporation is to
provide a fleet of pressurized cargo missions to the ISS. CRS
will provide a U.S.-produced and operated automated cargo
delivery service for ISS logistics support, to complement the
existing Russian, European and Japanese cargo vehicles.
Israels AMOS-6 is a 5-ton class GEO satellite. The
communication payload includes 45 transponders in 3
frequency bands (Ku, Ka, S), which enable the satellite to
provide communication services that can include DTH (directto-home) satellite internet.
NuStar (Nuclear Spectroscopic Telescope Array) is a NASA
Explorer mission and is the first orbiting space-based X-ray
telescope to focus high energy X-rays from astrophysical
sources, especially for nuclear spectroscopy, and operates in
the range of 6 to 79 keV. It was successfully launched by
Orbitals Pegasus XL rocket to orbit on June 13, 2012.

978-1-4799-3299-3/13/$31.00 2013 IEEE

Fig. 3 MMS solar panel

Due to MMS highly elliptical orbit, the panels were
qualified to the hot and cold extreme temperatures of +70C
and -150C, respectively.
The solar panels were also
manufactured to be electrostatically and magnetically clean
using a proprietary process.
CRS
The CRS solar arrays consist of 2 circular wings, with 10
triangle-shaped UltraFlex flexible blanket/gores per wing. A
total of 100 gores are built to power 5 Cygnus satellites. The
solar array for each spacecraft produces about 3.5 kW in orbit.
Due to the UltraFlex flexible gore, the CRS arrays are one of
the lightest space-qualified structures available in the market.
A photo of a CRS gore is shown in Fig. 4.

Fig. 4 ZTJ 1-per-wafer solar cells on ATKs UltraFlex CRS

panel

2794

AMOS-6
The AMOS-6 solar array consists of 2 wings of 6 panels per
wing. The ZTJ solar cell assemblies are laid down on
conventional kapton-covered aluminum honeycomb core
substrates with CFRP Facesheet. Space qualified, proven
processes are applied to the panels to mitigate and protect the
array against ESD (electrostatic discharge) events in the GEO
environment. More than 6,000 ZTJ cells (cell area ~60 cm2)
were laid down on 12 panels to meet the >10 kW EOL power
requirement for AMOS-6s 15-year mission.
NuStar
The NuStar solar panels that were delivered to ATK were
populated with 2-per-wafer ZTJ solar cells. The solar array
consisted of 1 wing of 5 panels. A photograph of the stowed
wing inside the Pegasus rocket is shown in Fig. 5.

followed by thermal cycling in vacuum (TVAC) or in ambient

pressure nitrogen atmosphere (APTC). A partial list of ZTJ
flight panel acceptance testing is shown in Table III.
TABLE I
ZTJ 2-PER-WAFER LIFE-CYCLE COUPONS
Orbit

Number of Cycles
Temperature Extremes
(APTC)

Estimated
Completion

GEO

8 TVAC
2,000

-175C to +80C
-175C to +80C

GEO

22,000

-165C to +125C

Completed

GEO

4,000

-165C to +130C

Completed

Completed

GEO

2,500

-170C to +145C

Completed

LEO

25,500

-100C to +125C

Completed

LEO

1,500

-106C to +88C

Completed

LEO

5,500

-75C to +135C

Completed

TABLE II
ZTJ 1-PER-WAFER LIFE-CYCLE COUPONS
Orbit

Fig. 5 Partial view of ATKs NuStar solar array inside

Orbitals Pegasus Rocket
IV. QUALIFICATION AND ACCEPTANCE TESTS
The ZTJ has been qualified and characterized at the bare
cell, CIC, life-cycle coupon and flight panel acceptance levels.
The bare cell , CIC assembly and coupon qualifications were
performed according to the requirements of the AIAA-S1112005 and AIAA-S112-2005 standards, respectively.
Additional S-112 qualification and characterization tests were
performed on the 1-per-wafer ZTJ CICs and panels consisting
of life-cycle coupons, solar absorptance, angle of incidence,
bypass diode, emissivity, UV radiation and ESD
characterization at the CIC and coupon levels.
In addition, extensive life-cycle coupon testing has been
performed to various orbital environmental temperature
extreme conditions in the past several years. The coupons were
populated with both the 2-per-wafer and 1-per-wafer ZTJ
cells. These are summarized in Tables I and II, respectively,
and encompass the qualification requirements of the MMS,
CRS, AMOS-6 and NuStar programs.
The flight panels also undergo environmental acceptance
thermal cycle testing post manufacturing before they are
delivered for integration into solar arrays. The most common
tests include a high-temperature thermal vacuum bake-out,

978-1-4799-3299-3/13/$31.00 2013 IEEE

Number of Cycles
Temperature Extremes
(APTC)

Estimated
Completion

GEO

8 TVAC
4,000

-175C to +120C
-165C to +105C

Completed

GEO

2,800

-175C to +143C

Completed

GEO

2,000

-170C to +145C

June 2013

GEO

12 TVAC
2,000
7,500
5,000

-180C to +130C
-180C to +105C
-165C to +105C
-140C to +105C

2014

LEO

10 TVAC
2,400

-128C to +128C
-128C to +128C

Completed

LEO

14 TVAC
32,000

-70C to +130C
-70C to +130C

2014

LEO

9 TVAC
3 TVAC
700
600

-110C to +70C
-150C to +130C
-110C to +70C
-150C to +0C

Completed

TABLE III
ZTJ FLIGHT PANEL ACCEPTANCE TESTING
Orbit
LEO
LEO
LEO
LEO
LEO
LEO
LEO
LEO

LEO
GEO

Number of
Cycles (APTC)

Temperature Extremes

Vac. Bake-Out
8 TVAC
Vac. Bake-Out
8 TVAC
Vac. Bake-Out
8 TVAC
Vac. Bake-Out
14 APTC
Vac. Bake-Out
12 TVAC
Vac. Bake-Out
8 TVAC
Vac. Bake-Out
5 TVAC
Amb. Bake-Out
14 APTC
8 APTC
Vac. Bake-Out
10 TVAC

48 Hrs. +125C
-90C to +90C
24 Hrs. +100C
-101C to +117C
96 Hrs. +110C
-110C to +110C
48 Hrs. +125C
-103C to +103C
96 Hrs. +135C
-100C to +125C
24 Hrs. +135C
-100C to +130C
72 Hrs. +120C
-95C to +120C
24 Hrs. +125C
-97C to +95C
-90C to +90C
2 Hrs. +135C
-179C to +88C

Estimated
Completion

Luanch Date

Completed

June 2012*

Completed

2013

Completed

2014

Completed

2014

Completed

2014

2014

2015

2013

2015

2014

2015

Completed

2014

2013

2014

*Launched

2795

V. NUSTAR ON-ORBIT PERFORMANCE RESULTS

The NuStar spacecraft is orbiting at an altitude of ~600 km
at 6 inclination and represents the first flight of ZTJ solar
cells serving as primary power to the spacecraft. The
spacecrafts on-orbit solar array operation has been nominal
since its launch on June 13, 2012, including both a Summer
solstice and a Winter solstice.
The solar array consists of one deployable wing of 5 solar
panels (see Fig 6) that produces 750W of power at EOL and
operates at a load voltage of ~32.8V. The predicted output
current of the solar array is compared to the on-orbit measured
telemetry data provided by ATK and Orbital in Fig. 7.

Fig. 6 NuSTAR Spacecraft On-Orbit Configuration.

As shown in the figure, the on-orbit solar array performance

slightly exceeds predictions with power production varying
with earth sun distance and exhibiting minimal degradation as
expected.
This is consistent with other on-orbit results reported for
prior generations of Emcore triple-junction cells (i.e., TJ, ATJ
and BTJ), where the telemetry data showed on average about
1%-3% greater array output power than predicted.
VI. SUMMARY
The high-efficiency ZTJ triple-junction solar cells have been
in volume production at Emcore for the past 4 years. The ZTJ
is qualified and characterized at the bare cell and CIC/panel,
levels per the requirements of the AIAA-S111-2005 and
AIAA-S112-2005 standards, respectively. Several life-cycle
coupons have also been built and tested to the environmental
conditions of various orbits, with both the 1-per-wafer and 2per-wafer ZTJ form factors. The large-area 1-per-wafer ZTJ
cells were used in the manufacture of solar panels for the
MMS, CRS 4-8 and AMOS-6 missions. The on-orbit
performance results of NuStar solar array populated with ZTJ
cells were also presented and shown to be operating as
predicted. Finally, presently the ZTJ cell is baselined to power
more than 40 future satellites that are scheduled to launch in
the 2013-2018 time frame.
REFERENCES

[1] P. Sharps, et al., "Emcore high efficiency space solar cell

technology: 30% and beyond," in 38th European Space Power
Conference, 2011.
[2] M. Stan, et al., "Air Force mantech qualification of the 30% class
GaInP2/Ga(In)As/Ge solar cell to the AIAA S-111 standard:
results and recommendations," in 35th IEEE Photovoltaic
Specialist Conference, 2010, 657-S9
[3] B. Cho, M. Winter, I. Aeby, P. Patel, N. Fatemi and P. Sharps,
"The one-per-wafer ZTJ solar cell from Emcore confidence
testing and volume manufacturing," in 38th IEEE Photovoltaic
Specialist Conference, 2012, 925.

Fig. 7 NuStar solar array predicted and measured on-orbit

telemetry data

978-1-4799-3299-3/13/$31.00 2013 IEEE

2796

Recent Progress of High-Efficiency III-V Multijunction

Solar Cells
Lijie Sun, Mengyan Zhang, Xin Fang, Kaijian Chen, Wei Zhang,
Mengjiong Chen and Jianfeng Lu
Center for Photovoltaic Engineering, Shanghai Solar Energy Research Center, Shanghai, China

Abstract High-Efficiency III-V multijunction Solar

Cells are being developed at Shanghai Solar Energy
Research Center (SSRC) for use in space applications.
Recently, Inverted Metamorphic (IMM) 3-Junction cells
achieved a 1-sun, AM0 conversion efficiency of 31.88% with
an open-circuit-voltage of 2.978V, a short-circuit currentdensity of 16.94 mA/cm2, and a fill factor of 0.863. In
addition, GaInP/GaAs/InGaAsP/InGaAs 4-junction solar
cells have been prepared by direct wafer bonding (DWB)
technique after the epitaxial growth of GaInP/GaAs
heterojunction on GaAs substrate and InGaAsP/InGaAs
heterojunction on InP substrate, respectively. Moreover,
antireflection coating (ARC) has been studied over the range
of 300 nm to 1700 nm for 4J solar cells in order to reduce the
front surface reflectivity. The latest status of the
semiconductor-bonded 4J solar cells and the antireflection
coating will also be discussed.
Index Terms III-V semiconductor materials, Indium
gallium arsenide, Wafer bonding.

I. INTRODUCTION
InGaP/InGaAs/Ge triple-junction solar cells with over
30% efficiency, which consist of lattice-matching
epitaxial layers grown on a Ge substrate, have already
been put to practical use in power generators for space
satellites. However, this type of solar cell is not perfectly
matched to the solar spectrum [1, 2]. For the high
efficiency performance of monolithically fabricated III-V
multijunction solar cells, high quality epitaxial growth of
subcells with high internal quantum efficiency is a
requirement as well as an optimized combination of band
gap materials for efficient conversion of the solar
spectrum [1, 2]. The first approach is the invertedmetamorphic (IMM) structure [3], that is, the GaInP top
cell and the GaAs middle cell are grown lattice matched
and therefore virtually defect free on a GaAs substrate.
Then the lattice constant is changed by a metamorphic

978-1-4799-3299-3/13/$31.00 2013 IEEE

buffer layer to grow a GaInAs bottom junction with a 1

eV bandgap.
The second approach is to use direct wafer bonding
(DWB) to interconnect between subcells in multijunction
solar cells. This technique enables further flexibility in
achieving an optimal band gap sequence for solar energy
conversion and fabrication of dislocation-free active cells
by confining the lattice-mismatch accommodation to the
bonded interfaces. Recently, fabrication of a simple 2junction solar cell has been reported by Tanabe et al [4]
via direct bonding of GaAs and InP wafers.
In this study, high-efficiency III-V multijunction Solar
Cells are being developed using both of the IMM and
DWB structure. The IMM 3-Junction cells was grown by
Metalorganic Chemical Vapor Deposition and followed
by substrate removal process.
In order to monolithically interconnect between the
GaInP/GaAs and InGaAsP/InGaA subcells, the bonded
GaAs/InP heterojunction must be a highly conductive
Ohmic junction or a tunnel junction. In this study, the
bonding interfaces were designed by tuning the
conduction type and doping elements of GaAs and InP.
The electrical properties of p-GaAs/n-InP bonded
heterojunctions were analyzed from I-V characteristics.
Based on the above study, we finally obtained 4-junction
GaInP/GaAs/InGaAsP/InGaAs solar cells.
The wide bandwidth antireflection coating (ARC) is
another area that is critical to achieving higher
efficiencies of III-V multijunction solar cells. The ARC
over the range of 300 nm to 1700 nm has been studied for
4J solar cells in order to reduce the front surface
reflectivity.
II. EXPERIMENTAL
In this work, the typical structure of IMM-3J solar cell
in this study was shown in Fig. 1. All metamorphic buffer

2832

acoustic microscopy was used to determine the uniformity

and quality of the bonded interface. SIMS was measured
to investigate the diffusion of bonding interface atoms
during the wafer bond process. Post wafer bonding, the
GaAs substrate was removed and the now-upright cells on
InP substrate were processed with deposition of metal
contacts and anti-reflective coating.
III. RESULTS AND DISCUSSION

Fig. 2. shows the photovoltaic IV characteristics of

IMM-3J solar cells and the representative conversion
efficiency of 31.88% with an open-circuit-voltage of
2.978V, a short-circuit current-density of 16.94 mA/cm2,
and a fill factor of 0.863.
Fig. 3. show the internal quantum efficiency (IQE)
characteristics of IMM-3J solar cell. The good quality of
bottom cell can be confirmed form the high IQE values.

250
200

Current (mA)

layers in this work have been grown by Metalorganic

Chemical Vapor Deposition using an AIX2600-G3
reactor with 4-inch substrate at Shanghai Solar Energy
Research Center (SSRC). The n-GaAs substrates were
used with a 7 offset from the (001) to (111) B plane and
the carrier concentrations were (1~4)1018 cm-3. The
hydrid sources arsine and phosphine were used for the
group-V-growth. Trimethylgallium at 7.5 C and
ethyldimethylindium at 16 C were used as the group III
precursors. SiH4 was used as the dopant for n type doping
of III-V compound epitaxial layers, while diethylzinc and
CCl4 were used as the p type dopants. GaInAlAs buffer
structures were grown with a step-graded design (see Fig.
1.). The grade consisted of eight 250 nm steps and a 800
nm tuning layer which was varied in composition. On top
of the tuning layer, a final 200 nm GaInAs layer lattice
matched to the 1eV GaInAs Bottom Cell layer was grown.
Before direct wafer bonding, GaAs and InP lattice
matched subcells were grown by Metalorganic Chemical
Vapor Deposition (MOCVD) equipment. The MOCVD
growth conditions for each of these subcells were
optimized to yield a high quality material with minimum
defects and optimal electrical performance. Each of the
component wafers were cleaned and passivated to a high
quality of surface finish prior to subjecting the wafers to
bonding. First, all the epitaxial subcells were cleaned by
acetone, isopropanol and ethanol sequentially in an
ultrasonic bath, then rinsed by deionized (DI) water, and
finally dried with N2. The GaAs bonding surface was
etched in our homemade aqueous-based HCl:H2O2:H2O
(1:50:100) solutions and then dipped in hydrogen fluoride
solutions (HF) for surface oxide removal. The InP wafers
were first dipped in HF solution and then cleaned in
NH3OH:H2O2:H2O (1:4:20) solution to remove the
pollutions of carbon and metal oxide.

150
100
50
0
-50

Fig. 2.

1000

2000

Vbias (V)

3000

Photovoltaic IV characteristics of IMM-3J solar cell.

Contact layer
100

Step-graded buffer

80

Current (mA)

1eV GaInAs Bottom Cell

GaAs Middle Cell
GaInP Top Cell
substrate
Fig. 1.

The typical structure of IMM-3J solar cell.

The bonding was carried out using Karl Suss SB6e

wafer bonding equipment. The bonding parameters such
as temperature, pressure and bonding time were carefully
optimized to achieve high mechanical strength and low
electrical resistance across the bonded interface. Scanning

978-1-4799-3299-3/13/$31.00 2013 IEEE

60
40
Top
Middle
Bottom

20
0
400

600

800

1000

1200

Wavelength (nm)

Fig. 3.

IQE characteristics of IMM-3J solar cell.

2833

Integrated III-V multijunction solar cells were

fabricated by using direct wafer bonding technology to
join subcells grown on GaAs and InP substrates. For 4junction cells, two of the subcells were grown inverted on
GaAs substrate and two subcells were grown upright on
InP substrate. Appropriate current balance was achieved
between the top GaAs-based and bottom InP-based
subcells for 4-junction cells.
GaInP(1.89 eV)/GaAs(1.42 eV)/InGaAsP(1.05 eV)/
InGaAs (0.75 eV) 4-junction solar cell has been
fabricated by direct wafer bonding and the bonded
interface is that p-GaAs (Zn doped)/n-InP (Si doped) .

15

IV. CONCLUSION

Current (mA)

10
5
0
-5
-10
0

Fig. 4.

1000

2000

3000

Vbias (V)

Photovoltaic IV characteristics of DWB 4J solar cell.

50

4 layer + 1
experimental data
40

simulation data

R(%)

20

10

0
400

600

800

1000

1200

1400

Wavelength(nm)

1600

High-Efficiency III-V multijunction Solar Cells are

being developed at Shanghai Solar Energy Research
Center (SSRC) for use in space applications. Recently,
Inverted Metamorphic (IMM) 3-Junction cells achieved a
1-sun, AM0 conversion efficiency of 31.88% with an
open-circuit-voltage of 2.978V, a short-circuit currentdensity of 16.94 mA/cm2, and a fill factor of 0.863. In
addition, GaInP/GaAs/InGaAsP/InGaAs 4-junction solar
cells have been prepared by direct wafer bonding
technique after the epitaxial growth of GaInP/GaAs
heterojunction on GaAs substrate and InGaAsP/InGaAs
heterojunction on InP substrate, respectively. Moreover,
antireflection coating has been studied over the range of
300 nm to 1700 nm for 4J solar cells in order to reduce
the front surface reflectivity. The average reflectivity of
our ARC in the range of 300 nm to 1700 nm was less than
5%.
REFERENCES

30

200

The photovoltaic I-V characteristics of the 4-junction

solar cell are shown in Fig. 4. The opencircuit (Voc) for
this cell was elevated to 3.2 V, which is approximately
equal to the sum of the open-circuit voltages for the
GaInP, GaAs, InGaAsP and InGaAs subcells. This Voc
result indicates that the electrical properties of GaAs/InP
heterointerface are crucial to the solar cells parameter.
Fig. 5. show the preliminary results of reflectivity of
ARC for high efficiency III-V multijunction solar cells.
The average reflectivity of our ARC in the range of 300
nm to 1700 nm was less than 5%, which is helpful for
increasing the long wavelength absorption of the sun's
rays.

1800

[1] H. Huang, X. Ren, X. Wang, Q. Wang, and Y. Huang,

Low-temperature InP/GaAs wafer bonding using sulfidetreated surface, Appl. Phys. Lett. Vol. 88, pp. 06110410611043, 2006.
[2] K. Nakayama, K. Tanabe, and H. A. Atwater, Improved
electrical properties of wafer-bonded p-GaAs/n-InP
interfaces with sulfide passivation, Journal of Applied
Physics Vol. 103, pp. 0945031-0945035, 2008.
[3] D. Bhusari, D. Law, R. Woo, J. Boisvert, S. Mesropian, D.
Larrabee, W. Hong, and N. Karam, Direct Semiconductor
Bonding Technology (SBT) For High Efficiency III-V
Multi-Junction Solar Cells, in 37th IEEE Photovoltaic
Specialists Conference, Washington, Seattle, June, 2011.
[4] K. Tanabe, A. F. Morral, and H. A. Atwater, Direct-bonded
GaAs/InGaAs tandem solar cell, Appl. Phys. Lett. Vol. 89,
pp. 1021061-1021063, 2006.

The reflectivity of ARC (300 nm to 1700 nm) for

Fig.5.
DWB 4J solar cells.

978-1-4799-3299-3/13/$31.00 2013 IEEE

2834

Preliminary temperature Accelerated Life Test (ALT) on III-V

commercial concentrator triple-junction solar cells
P. Espinet-Gonzalez , C. Algora , V. Orlando , N. Nun ez , M. Vazquez , J. Bautista , H. Xiugang
L. Barrutia , I. Rey-Stolle and K. Araki

Instituto de Energa Solar, Universidad Politecnica de Madrid, Madrid, 28040, Spain

EUIT de Telecomunicacion, Universidad Politecnica de Madrid, Madrid, 28031, Spain
Daido Steel Co., Minami, Nagoya, Japan

AbstractA quantitative temperature accelerated life test on

sixty GaInP/GaInAs/Ge triple-junction commercial concentrator
solar cells is being carried out. The final objective of this
experiment is to evaluate the reliability, warranty period, and
failure mechanism of high concentration solar cells in a moderate
period of time. The acceleration of the degradation is realized
by subjecting the solar cells at temperatures markedly higher
than the nominal working temperature under a concentrator,
while the photo-current nominal conditions are emulated by injecting current in darkness. Three experiments at three different
temperatures are necessary in order to obtain the acceleration
factor which relates the time at the stress level with the time at
nominal working conditions. However, up to now only the test
at the highest temperature has finished. Therefore, we can not
provide complete reliability information but we have analyzed
the life data and the failure mode of the solar cells inside the
climatic chamber at the highest temperature. The failures have
been all of them catastrophic. In fact, the solar cells have turned
into short circuits. We have fitted the failure distribution to a two
parameters Weibull function. The failures are wear-out type. We
have observed that the busbar and the surrounding fingers are
completely deteriorated.
Index TermsAccelerated life test, ALT, reliability, concentrator, solar cells, III-V, CPV, Photovoltaics

to quantify them and to understand the failure mechanism [5]

to try to avoid them. Therefore, reliability tests go beyond
qualification. However, up to now there is not enough accumulated experience to evaluate the reliability of concentrator
III-V multi-junction solar cells because they have not been
long enough in the field. Silicon modules are reliable systems
which have performed very well in the field with less than 1
% degradation/yr for more than 20 years [5]. If concentrator
multi-junction solar cells are expected to have a useful life
similar to silicon solar cells, few concentrator solar cells would
fail or degrade importantly in a real time test of practical
length at nominal operation conditions. Therefore, quantitative
accelerated life tests (ALT) are required to provide reliability
information in a moderate period of time (weeks or months). In
this study we describe the phylosophy, design, set-up, progress
and the preliminary results of the temperature accelerated life
test which is being carried out on sixty commercial triplejunction solar cells GaInP/GaInAs/Ge lattice matched. To the
best of our knowledge, it is the first time that this kind of
experiment is carried out.
II. F UNDAMENTALS OF QUANTITATIVE ACCELERATED

I. I NTRODUCTION

LIFE TESTS

High concentration photovoltaic (HCPV) systems seem

to be one of the most promising ways to generate
electricity at competitive prices for terrestrial applications
[1]. The efficiency of multi-junction solar cells has increased considerably, reaching a 41.6 % efficiency in lattice matched GaInP/GaInAs/Ge solar cells and a 43.5 % in
GaInP/GaAs/GaInNAs solar cells [2]. Nevertheless, the CPV
community is aware of the fact that together with the efficiency
increase, high reliability is essential in reducing the cost of
solar electricity by extending system lifetime [3]. Aware of
this problem, the CPV community has developed the standard
(IEC-62108:2007) for qualifying CPV modules and assemblies
[4]. Qualification tests, also called qualitative accelerated test,
are designed to specify the minimum requirements that the
item under test should satisfy. However, the qualification tests
are not a good indicator of the item lifetime because their
duration is not long enough to cause wear-out degradation.
Therefore, in order to estimate the power degradation/year,
the projected returns and warranty costs of high concentration
solar cells, it is crucial to carry out reliability tests. Reliability
tests, also known as life tests, are designed to evaluate failures,

The purpose of quantitative accelerated life testing is to

find out how, when, and why failures occur in the product
more quickly than with data obtained under normal operating
conditions. For this purpose, one of the parameters of the
device under test is subjected to representative levels of
stress remaining the rest of the parameters constant working
under nominal operation conditions. The high levels of stress
used intentionally force failures by accelerating the effects
of natural aging. Assuming a physically reasonable statistical
model which relates the lifetime to the level of stress through
an acceleration factor, the life data from the ALT can be used
to estimate the reliability functions and parameters under
nominal stress levels.
It is important to point out some assumptions of ALT
models:
The failure-causing process at high stress is the same as
at the nominal stress.
A physical/chemical process causes a change in the
device under test and this change progresses over time
to eventually cause failure.

The applied stress accelerates reaction rates and this

acceleration can be described by a model that is adequate
over the range of testing.
In this study the factor which is going to be significantly
higher in order to accelerate the aging of the solar cells is
the temperature. In our test the ALT model that we are going
to use is the well-known and commonly used Arrhenius lifestress model which is defined with the expression 1. It has
been widely used when the acceleration variable is thermal.

EA

L(T ) = Ce kT

(1)

where L(T) is a temporal measurable characteristic of the

device under test life which depends on the temperature, k is
the Boltzmann constant and EA is the activation energy of
the mechanism which causes the failure. C is a parameter of
the Arrhenius model which depends on the L(T) used.
The activation energy (EA ) of the failure mechanism which
determines the acceleration factor (AF ) of the test between
two temperatures, is obtained from the linearized expression of
the Arrhenius life-stress model (expression 2). The activation
energy is obtained from the slope of the curve Ln(L(T))
versus 1/kT. Obviously, at least three different values of L(T)
obtained from three experiments at three different temperatures
are necessary in order to fit the curve Ln(L(T)) versus 1/kT
properly.
Ln(L(T )) = EA (1/kT ) + Ln(C)

(2)

Once the activation energy is calculated, obtaining the

acceleration factor between two temperatures (T1 and T2 ) is
direct:



EA 1
1
L(T1 )
= exp

(3)
AF =
L(T2 )
k
T1
T2
Therefore, once the parameters of the life distribution and
life-stress relationship have been estimated, the same information that is available from real time tests can be obtained with
the ALT.
III. E XPERIMENTAL APPROACH
The methodology followed in the temperature accelerated life test which is being carried out on commercial
GaInP/GaInAs/Ge triple-junction solar cells (lattice matched)
presented in this work, is an adaptation from the procedure
already carried out satisfactorily on GaAs concentrator solar
cells [6]. As we pointed out above, the parameter used to
accelerate the aging of the solar cells is the temperature
and at least three temperatures are necessary to calculate the
acceleration factor. Therefore, we have placed three groups
of twenty solar cells inside three climatic chambers at three
different temperatures. The working conditions are emulated
by injecting in darkness the equivalent current that the solar
cell would photo-generate in illumination at the nominal
concentration level. The dark I-V curve of the solar cells
is periodically monitored during the test inside the climatic

Fig. 1. Experimental measurements and simulation results of the dark I-V

curve of a triple-junction solar cells used in the ALT.

chambers in order to register the time interval in which each

solar cell fails. When all the solar cells have failed, the life
at the test temperature can be evaluated. Once the life at the
three temperatures is evaluated, the activation energy can be
calculated. From the activation energy, the acceleration factor
between two temperatures is obtained by means of equation
3. Finally, a reliability study is carried out to obtain the main
reliability functions and parameters, and a failure analysis is
accomplished in order to determine the most fragile part of the
device which causes the failure. We classify the steps followed
in the accelerated life tests into three parts: 1) design of the
test, 2) progress of the test, and 3) data analysis. Below, the
processes carried out in every part are described.
A. Design of the test
In order to start the experiment the following points have
to be carried out: 1) nominal working conditions under
concentration have to be defined. These particular solar cells
are expected to work at 820 X and 80 o C for an ambient
temperature of 25 o C. 2) The current which has to be
injected in darkness to emulate working conditions has to
be calculated. With this purpose, simulations with our 3D
distributed model described in [7] have been carried out.
In order to have reliable simulations, the parameters which
feed the model were obtained by fitting the experimental
dark I-V curve and the illuminated I-V curve under different
concentrations and spectral conditions. The excellent results
of the fitting are shown in figures 1 and 2.
Once we have a reliable model, the triple-junction solar
cell has been simulated under a uniform irradiance of 820 X.
In figure 3 false color maps of the photo-generated current
density through the top cell, middle cell and bottom cell when
the solar cell operates at 820 X at the maximum power point
are depicted. The current density in the active area is around

TABLE I
E QUIVALENT CONCENTRATION NEEDED TO PHOTO - GENERATE THE SAME
CURRENT DENSITY THAN THE CURRENT DENSITY WHICH FLOWS IN
DARKNESS DUE TO THE INJECTION OF 3.2 A.

Subcell
TC
MC
BC

Busbar
450 X
440 X
430 X

Fingers
470-820 X
320 X
315 X

Active Area
250-340 X
330 X
320 X

for emulating 820 X is conservative. The maximum level

of current injection which fulfilled this criterion was 3.20
A. In figure 3 the current density distribution in darkness
through the different subcells when the solar cell is biased
with 3.20 A is shown. The external fingers in the top cell
are draining the highest current density (about 11.5 A/cm2).
It should be pointed out that we have also observed this
effect in experimental electroluminescence maps. Therefore,
3.2 A have been injected to the solar cells for emulating
the photo-generated current at 820 X. In table I we present
the equivalent concentrations which would be needed to
photo-generate the current density that each subcell manage
in the different regions of the solar cell when 3.2 A are
injected into the solar cells in darkness.
The next step in the design of the experiment is 3) to
determine the temperatures at which the tests will be carried
out. It was checked that the packaging was able to handle up
to 170 C. Therefore, the temperature accelerated life tests
have been carried out at the following solar cell temperatures:
113 2o C, 132 2o C and 164 1o C. These temperatures
are high enough to obtain a significant acceleration factor and
they are sufficiently separated from each other in order to
evaluate accurately the activation energy value. Finally, the
last step has been 4) to characterize the 60 solar cells in
order to carry out a failure analysis when the tests finish. The
following characterization techniques have been used: external
quantum efficiency (EQE), X-ray transmission imaging of the
back solder paste, I-V curve in darkness at 1X and at 400 X
and electroluminescence mappings.
Fig. 2. Experimental and simulation fitting of the main parameters of the
illuminated I-V curve of the same triple-junction solar cell presented in figure
1 under different concentrations and spectrums.

B. Progress of the test

11.5 A/cm2 and zero or negative in the bus bar and fingers.
Ideally this current density distribution should be emulated
in the ALT by forward biasing the solar cell in darkness
but in darkness the main of the current flows beneath the
busbar and fingers. We have used our 3D distributed model
to simulate the distribution of the current density through the
subcells when different levels of current were injected into
the solar cell. We have been increasing the current injected
into the solar cell so that the current density distribution
in darkness was as close as possible to the current density
distribution at 820 X, but without exceeding in darkness the
current density photo-generated at 820 X (11.5 A/cm2 ) in
any point on the solar cell. Therefore, the criterion followed

The solar cells have been divided into three groups of 20

solar cells. They have been introduced into three climatic
chambers at three different temperatures. In order to emulate
working conditions, 3.2 A 0.01 A are injected to the solar
cells heating them up to: 113 o C 2 o C, 132 o C 2
o
C and 164 o C 1 o C except for two solar cell in each
climatic chamber which are considered reference solar cells.
After the period of current injection, all the solar cells are
automatically disconnected from the current sources and there
is a period of temperature stabilization. After the stabilization
period, the dark I-V curve of each solar cell is measured. Once
all the solar cells have been measured, the current sources are
connected again to the solar cells and the cycle starts again.
This cycle is repeated until all the solar cells inside the climatic
chamber (except for the reference solar cells) fail. Figure 4

820 X, Maximum power point

0 X, 3.2 A

Fig. 3. False color maps representing the current density (A/cm2 ) through the top cell (top figures), middle cell (medium figures) and bottom cell (bottom
figures) for an illumination of 820 X at the maximum power point (left figures) and injecting 3.2 A in darkness (right figures). The vertical scale is A/cm2.

sketches the cycles of the solar cells (except the reference

solar cells) inside the climatic chambers.

information at nominal working condition and finally, 4) to

carry out a failure analysis to determine the part of the solar
cell responsible of the failure.

C. Data analysis
Once all the solar cells have failed, the analysis of
accelerated life test data consists of: 1) finding a life
distribution model which describes the solar cells failures
at different temperatures, 2) finding a life-stress model
that quantifies the manner in which the life distribution
changes across different temperatures, 3) to combine the life
distributions and the life-stress model to obtain reliability

Up to date, only all the solar cells of the test at the

highest temperature (164 o C) have failed. Therefore, we
can not calculate the acceleration factor (AF ) which relates
the time at certain stress level with the time at nominal
working condition (see expression 3). Consequently, complete
reliability information at nominal conditions can not be
obtained. Nevertheless, useful information can be derived

Fig. 4. Sketch of the thermal cycles of the solar cells inside the climatic
chamber.

from the existing data of the solar cells tested at 164 o C.

On finding a life distribution model which describes the
solar cell at 164 o C, we have decided to use the Weibull distribution, one of the most popular and widely used distribution
in failure data analysis. There are different estimation methods
to find the parameters of the Weibull distribution that are best
suited to a given set of data. Each method has a criterion which
yields estimates that are best in some situations. We have
chosen the MLE (Maximum Likelihood Estimation) method
which, from all possible values, it finds the parameter values
which maximize the likelihood of obtaining the data. Finally,
in order to account for the statistical uncertainty in estimates,
90 % confidence intervals have been used. In figure 5 the
Weibull plot is shown. A reasonably good fit of the failure
probability ditribution to the Weibull function is observed. The
parameters of the Weibull function which reproduce the failure
distribution obtained with the MLE method are:
The minimum life parameter () has been fixed to zero.
The shape parameter = 1.83.
The scale parameter or characteristic life = 32.04 hours.
In figure 6 the shape of the failure probability density
function (see expression 4) can be observed. It is slightly
positively skewed (has a right tail).

f (t) =

 1
t
t
e ( )

Fig. 5. Weibull plot of the failure probability versus time (hours) for the
solar cells at 164 o C.

(4)

In figure 7 the failure rate function (see expression 5)

versus time is shown. The failure rate function monotonically
increases corresponding with the wear-out failure part of the
bathtube curve.
 1
t
(t) =

Fig. 6. Probability density function (pdf) versus time (hours) for the solar
cells at 164 o C.

(5)

Regarding the failure analysis, the eighteen solar cells tested

have presented catastrophic failures. In fact, they have turned
into short circuits. However, neither of the two reference solar
cells have failed. It has to be pointed out that the reference
solar cells have not been subjected to current injection and
consequently they have not suffered thermal cycles either.
They have been at the constant temperature of the climatic

chamber (143 1 o C) through out the whole test. Therefore,

the high temperature of the climatic chamber (143 1 o C)
does not seem to be responsible for the failure. Due to the solar
cells turned into a short circuit, we have not been able to carry
out the same characterization techniques as prior to the test.
We only could make a visual inspection. We have observed
that the eighteen solar cells after the test present the busbar

generated current under a concentration of 820 X has been

emulated by injecting current in darkness. Only all the solar
cells inside the climatic chamber at the highest temperature
(164o C) have failed. All the failures have been catastrophic,
turning the solar cells into short circuits. By visual inspection
we have observed that the busbar and the surrounding fingers
of the solar cells tested are burned out. However, two reference
solar cells which were inside the same climatic chamber at
143o C have not failed and present the busbar and fingers intact.
In these reference solar cells current has not been injected
through them, thus they have not been subjected to thermal
cycles either. The failure probability distribution of the solar
cells at 164o C has been fitted to the two parameters Weibull
function and a good fit has been obtained with the shape
parameter = 1.83 and the scale parameter or characteristic
life = 32.04 hours. The failure probability density function is
slightly positively skewed (has a right tail) and the failure rate
function monotonically increases revealing that the failures are
wear-out type.
ACKNOWLEDGMENTS
Fig. 7. Failure rate function (failures/hour) versus time (hours) for the solar
cells at 164 o C.

This work was supported by the European Commission

under the project NGCPV A new generation of concentrator
photovoltaic solar cells, modules and systems with Grant
agreement no. 283798. The Spanish Ministerio of Economa
y Competitividad has contributed with the projects TEC201128639-C02- 01 and IPT-2011-1441-920000 (INNPACTO program) as well as the Comunidad de Madrid under the NUMANCIA II programme (S2009/ENE1477).
R EFERENCES

Fig. 8. Photograph of the metal deteriorated of the solar cells which were
injected 3.2 A in the ALT at 164 o C (left figure) and photograph of the metal
intact of the reference solar cells (right figure).

and the surrounded fingers completely deteriorated (burned

out) as shown in the left photograph of figure 8. However,
the reference solar cells present the metalization intact (see
the right photograph of figure 8).
IV. S UMMARY AND CONCLUSION
In this paper a quantitative temperature accelerated life test
on sixty commercial triple-junction concentrator solar cells has
been launched and the steps followed have been explained
in detail. In order to estimate the reliability and warranty
period of this solar cells under nominal working conditions,
the solar cells have been divided into three groups and they
have been placed into three climatic chamber at three different
temperatures. The aging of the solar cell have been accelerated
by stressing the solar cells at higher temperatures than nominal ones: 113 o C, 132 o C and 164o C. The nominal photo-

[1] M. Yamaguchi and A. Luque High efficiency and high concentration in

photovoltaics,IEEE Trans Electron Devices vol. 46, no. 20, pp. 21392144, 1999.
[2] Martin A. Green, Keith Emery, Yoshihiro Hishikawa, Wilhelm Warta and
Ewan D. Dunlop, Solar cell efficiency tables (version 39), Progress
in Photovoltaics: Research and Applications, vol. 20, pp. 12-20, no. 1,
2012.
[3] S. Kurtz, J. Granata and M. Quintana, Photovoltaic-Reliability R&D Toward a Solar-Powered World, in Society of Photographic Instrumentation
Engineers (SPIE) Solar Energy + Technology Conference, 2009
[4] IEC 62108 Ed 1.0: concentrator photovoltaic (CPV) modules and assembliesdesign qualification and type approval. 2007
[5] Manuel Vazquez and Ignacio Rey-Stolle, Photovoltaic module reliability
model based on field degradation studies,Prog. Photovolt: Res. Appl.,
vol. 16, pp. 419-433, no. 5, 2008.
[6] N. Nunez, J. R. Gonzalez, M. Vazquez,C. Algora, and P. Espinet,
Evaluation of the reliability of high concentrator GaAs solar cells by
means of temperature accelerated aging tests, Prog. Photovolt: Res.
Appl., 2012
[7] P. Espinet, I. Garca, I. Rey-Stolle, C. Algora and M. Baudrit, Extended
description of tunnel junctions for distributed modeling of concentrator
multi-junction solar cells, Solar Energy Materials and Solar Cells, vol.
95, pp. 2693 - 2697, no. 9, 2011.

International Conference on Electrical Engineering and Information & Communication Technology (ICEEICT) 2014

Challenges and Prospects of Cost-Effective Si-based

Solar Cells Fabrication in Bangladesh
Md. Shafiqul Islam, Mahbubul Hoq, Mohammod Abu Sayid Haque, Md.Abdur Rafiq Akand
Md. Rakibul Hasan and Mohd. Khairul Basher
Solar Cell Fabrication Laboratory
Atomic Energy Research Establishment (AERE)
Bangladesh Atomic Energy Commission (BAEC), Dhaka-1207, Dhaka, Bangladesh
Corresponding Authors Email: shafiq12@dhaka.net
AbstractSolar cells are the basic building blocks of Solar
Panels or Modules. There are presently nine companies
manufacturing solar modules in the country. The total demand
for solar cells in assembling solar modules of the nine companies
are about 80-90MWp annually. In assembling solar modules,
major raw materials include finished solar cells, bus bars, EVA,
Al frame, tempered glass etc. All of them are import dependent
for the case of Bangladesh. It is found that about 55-60% of total
panel/module cost depends on solar cells. Cost-effective power
generation using solar module solely depends on the quality of
solar cells. Now-a-days, cost of solar cells has been decreasing
rapidly over the last 3-4 years due to large scale production
plants, technological improvement, declining cost of raw
materials and competitiveness in the global markets. Solar cells
which were used to sell for USD2.5 per Wp can be bought on a
commercial scale for around USD1.2 per Wp. Mainly, cost of
solar cell depends on the Brand, type of crystal structure (Mono
or Multi) and the shape or size of wafer. For the first time in the
country, Bangladesh Atomic Energy Commission (BAEC) has
set up a laboratory to fabricate crystalline solar cells on a pilot
basis. The method used to fabricate solar cells is the low-cost
diffusion technique using POCl3 gas source. Several solar cells of
150150 mm2 sizes and 200 micrometer thick are produced and
characterized in the laboratory. The results have shown us some
challenges and good potentialities in getting the high-efficient and
cost effective solar cells. The paper addresses the challenges and
potentialities in fabricating the quality based crystalline solar
cells using slightly p-doped Si-wafers and chemicals in order to
adopt technology so as to reduce the cost of solar modules. If the
cost effective technology can be made familiar in Bangladesh
then it will help in solving the power crisis in the country a great
deal.
KeywordsSolar cell
effictiveness; energy crisis

I.

fabrication;

cell

efficiency;

cost

INTRODUCTION

The worlds increasing need for energy has made it necessary

that people rely on renewable energy sources rather than
conventional one. Solar energy often termed as the source of
nearly all energy available on earth, is one of the major
renewable energy sources and can give a solution to the
present day energy crisis. The solar energy captivated by
earths atmosphere, oceans and land is about 385000 EJ [1]. In
spite of this large energy captivation, only less than 1% useful

978-1-4799-4819-2/14/$31.00 2014 IEEE

energy comes from solar power [2]. Hence, this means the sun
shine can produce about 35000 times more power on the earth
than the daily power generation using solar energy. In
Bangladesh, the average sunshine hours are 6.69, 6.16 and
4.81 in winter, summer and monsoon, respectively.
Photovoltaic (PV) method which converts solar radiation
directly into electricity using solar cells is a simple and unique
method to use solar radiation as it is noiseless, reliable and has
a long life. Solar cells are made from different semiconductor
materials which affect the cost and efficiency. Solar cells can
be made of crystalline silicon, called Generation-I which
includes mono-crystalline silicon, polycrystalline silicon,
amorphous cadmium telluride, and copper indium
selenide/sulfide [7]. There are also thin film solar cells, called
Generation-II which includes cadmium telluride solar cell,
copper indium gallium selenide, gallium arsenide junction,
light-absorbing dyes (DSSC), quantum dot solar cells
(QDSCs), organic/polymer solar cells, silicon thin films [8].
The paper aims in analyzing the manufacturing processes that
is practiced in the laboratory with an aim to developing costeffective crystalline solar cell. Sect. II deals with the principle
how solar cell works.
II.

PRINCIPLE OF OPERATION OF SOLAR CELL

Solar cell is an electronic device that produces electricity by

directly converting energy from sunlight. This is the
photovoltaic phenomenon of semiconductor. The conversion of
energy from sunlight is based on the p-n junction method.
Crystalline type semiconductor solar cells are produced by
doping with acceptor atoms to create p-type region and
donors atoms to generate n-type region on semiconductor
crystal lattice. Due to the rule of diffusion, the holes are
diffused from the p type material and the electrons are
diffused from the n type material at the p-n junction as shown
in Fig 1. According to Fig. 1, atoms which release the
electrons become positively ionized and similarly the atoms
which release holes become negatively ionized. If the
electrons and holes were not charged due to built in electric
field, electrons and holes would continuously diffuse to the
opposite direction at the junction. But in real practice, after a
certain period, the next free electrons and holes coming from
the remote n type and p type regions respectively are opposed

by neutralized electron-hole pairs at the p-n junction.

Fig. 1: Formation of p-n Junction

Thus a junction barrier is formed at the p--n junction which is

known as the depletion region. In Fig.1, the depletion region is
indicated at the middle. Here a minor current and a very small
amount of voltage (generally 0.1-0.3V) are formed. In spite of
this, some electrons and holes pass the junction due to
diffusion force. Thus they form diffusion current. On the other
hand, due to the applied electric field, the electron and holes
are swept across the junction and form drift current. In an
open circuit solar cell, the diffusion and drift current are
balanced equally. Thus the solar cell becomes electrically
neutral. When the sunlight reaches the solar cell pp-n junction
a specific amount of incident light is reflected at the outer
atmosphere and the rest portion is being transmitted through
solar cell. Among these portions due to optical imperfections,
phonon and other quasi-particles a fraction of light is scattered
and reflected to the rear surface of the solar cell [9] that, an
anti-reflecting coating is given over the front surface of solar
cell which prevents reflection spectrum of the front surface. At
the rear surface, there is metal coating which totally reflects
back the incoming light inside the solar cell to increase the
amount of photon without any absorption. The photon hits the
valance electrons of the cell material atoms. These valance
electrons gain energy from the photon. If this energy is enough
to overcome the band gap (1.12 eV for silicon), the electrons
become free from the atom. By this process the light generated
current is being induced. But due to the recombination, the
light generated current and the voltage are being neutralized.
Thus solar cell cannot contribute to current flow. To eliminate
this problem, the emitter (n-type region) and base (p-type
region) are connected by wires to flow the electrons through
the external circuit by creating short circuit forward bias [6].
The electrons dissipate energy to the external load while
passing through the circuit and returns to the solar cell. Thus
the solar cell produces voltage and supplies electricity to the
external loads. The single solar cell can produce very small
amount of power. That is why, a group of solar cells are
combined inside a weather-tighten panel. And again a group
of panels by joining with wires are involved in electricity
supply projects. On the basis of this solar cell operation
theory, Sect. III describes the existing and cost effective
technology of crystalline solar cell fabrication process.
III.

LOW COST SI-BASED SOLAR CELL METHOD

A. Existing Technology
Solar cells, which has become a rapidly growing and

immensely important alternative to renewable energy, was

first practically demonstrated in 1950[6]. Since then by using
different technologies and different materials the efficiency of
solar cells (14-18%) has increased in a great deal. Today
different technologies are used for junction formation which is
basically a process of creating a new layer called emitter in
the substrate material. The technologies that are used for this
doping method are- phosphorus diffusion from Phosphorus
Oxy-Chloride (POCl3), Ortho-Phosphoric acid [10], Spin on
Dopant Process, Ion Implantation, Spraying method, Epitaxy
etc[11]. Adopting different diffusion techniques affect the
efficiency and cost consideration. Apart from these
techniques, there are also different chemical vapor deposition
methods that are widely used in fabricating solar cells. They
are Atmospheric Pressure Chemical Vapor Deposition
(APCVD), Low Pressure Chemical Vapor Deposition
(LPCVD), and Plasma Enhanced Chemical Vapor Deposition
(PECVD) etc [12]. Based on the cell fabrication technologies,
there are screen printed solar cells, buried contract solar cells,
high efficiency solar cells, and rear contract solar cells [6]. In
Sub Section B of Sect. III, the cost effective technology is
described in details.
B. Cost Effective Technology
The development and implementation of different
techniques and materials all these years was practiced to
increase the efficiency of solar cells and to make the
fabrication cost effective. Among all the methods that are
practiced now-a-days it is seen that Spin on Doping and
spraying method is cheaper than diffusion from a POCl3
source, Epitaxy etc. Spin on Doping is one of the most
conventional diffusion processes. In this process, liquid dopant
solution is applied to the wafer and the wafer is spun at high
speed to produce a thin film source [13]. This process is
relatively cost effective due to the simplicity of the whole
procedure. There is another diffusion process called POCl3
diffusion which is slightly expensive than the spin on doping
process. It is mainly the use of different advanced machinery,
chemicals and overall process that increase the fabrication cost.
It is widely used now-a-days by the solar cell manufacturers
[10]. Phosphorus is diffused from a liquid phosphorus OxyChloride (POCl3) source in a closed quartz tube. In this
procedure, double amount of samples can be processed at the
same time. This large processing capability is one of the main
advantages of POCl3 gas diffusion source. Another advantage
of this technology is the self-governing control of the predeposition and the drive-in, due to which the surface source can
more easily be made finite. Thus, a greater control of the
surface concentration of p diffused emitters is achieved. In
addition to that an extra parameter of freedom during process
optimization can be obtained [10]. The carrier lifetime of multi
crystalline wafers are increased tremendously in the POCl3
diffusion process as a result of impurity gathering [14]. Due to
all these advantages over the Spin on Doping process, POCl3
diffusion is industrially most wide-spread method for emitter
formation [8]. This is why BAEC has adopted this process in
their newly established Solar Cell Fabrication Libratory
(SCFL) under the project on Solar Energy Utilization and

Development of Related Technology. Fig.2 shows the actual

view of the overall laboratory set up of the SCFL.

Fig. 2: Solar Cell Fabrication Laboratory (SCFL)

Fig. 3: Flow chart of solar cell fabrication (Frond end process) processing
steps adopted in SCFL

According to the application of metal contacts, solar cell can

be divided as screen printed solar cells, buried contact, high
efficiency and rear contact solar cells. Among them, screen
printing is the simplest and the most cost effective method and
is going to be followed in BAEC. The details of solar cell
fabrication process using POCl3 diffusion technology in SCFL
is described in the following Sub Sect. C.
C. Solar Cell Fabrication Process using POCL3 Diffusion
Gas Source
Mono crystalline type solar cell fabrication deals with a set of
basic material, chemical components and equipment and
machinery. The specifications of the basic materials are stated
in Table 1. The basic material [6] used for solar cell
fabrication is mono crystalline type silicon wafers that are ptype doped initially. The usual square shape wafer is 150150
mm2 in size and 200m in thickness.

Fig. 4: Si wafer cleaning and texturing process performed at SCFL

Table 1: Basic Materials Involved in Solar Cell

Fabrication
Raw
material

Cell type

Doping

Shape & Size

Thickness

Silicon
wafer

Monocrystalline

P type

150150 mm2

200m

solar cell fabrication process which means the front end

process starts from the silicon wafer processing. Fig. 3 shows
the flow chart that briefly describes about the front-end solar
cell fabrication processing steps. At first the starting wafers
are p type doped with small amount of Boron. The starting
wafer remains uneven because of saw damage and it is coated
with cutting fluid. To remove the outer layer saw damage of
silicon, a strong alkaline (NaOH, KOH) etch is used. Figs. 4
and 5 show the wafer cleaning and texturing processes
performed during cell fabrication in wet benching machine.

Fig 5: Si wafer after cleaning and texturing

As phosphorus is diffused into both front and rear surfaces,

edge isolation is performed to separate the continuity by using
EDGE isolation paste around the cell. Fig. 6 shows the edge
isolated Si-wafer done mechanically by screen printer.
POCl3 diffusion is used for emitter formation. Phosphorus
is diffused from liquid Phosphorus Oxy-Chloride (POCl3)
source in a closed quartz tube.

Fig 8: Screen printed finished type Si wafer

Fig 6: Edge isolated Si wafer

Once the samples are loaded in the tube, the furnace is heated
up to the preferred temperature before any processing can take
place. During the initial stage or pre-deposition stage of the
diffusion process, carrier gas nitrogen is passed to feed liquid
POCl3 into the process chamber.

The cells are then fired at different higher temperatures in

three zones consecutively to form the Ohmic-contact between
conductor (Al & Ag) and semi-conductor (P type & N type).
Fig.9 shows the contact firing process of screen printed Si
solar cell.

Fig 9: Contact firing at RTC furnace

Fig 7: Phosphorous doped on P type Si wafer from diffusion chamber

There it evaporates and reacts with externally supplied O2

and forms P2O5 on the wafer surfaces, described in the
following Eq. (1).
4 POCl3 (g) + 3O2 (g) = 2 P2O5 (l) + 6Cl2(g)

(1)

After this, anti reflecting coating is provided to the cell

surface. The phosphorous doped on p type Si wafer from
diffusion chamber is shown in Fig. 7. Then front surface of
the cell is screen printed by applying silver paste. The wet
cells can be easily smudged and thus loaded into a drier to
evaporate off the organic binders in the paste. Fig.8 shows the
front surface screen printed Si solar cell. The cells are then
flipped to be screen printed on the rear. The rear surface is
printed using Al paste.

IV.

EFFICIENCY CALCULATION

After finishing a complete solar cell, some characterization

tests [6] are carried out. These tests show how efficient the
solar cell is. The major characterization equipment is LIV
tester which gives the reading of open circuit voltage (Voc),
short circuit current (Isc) and Fill factor (FF). These data are
required to calculate the efficiency of a solar cell from the ratio
of output maximum power to the input power.
Eq.2 shows the calculation of efficiency from the ratio of
output maximum power to the input power.

Voc I sc FF .. .(2)
=

Pin

where, Pmax = Voc I sc FF

V.

Fig. 10 LIV tester in SCFL

Fig.10 shows the LIV tester in SCFL. Solar cells are tested
under one-sun conditions using Xenon-arc lamps; a Xenon
spectrum is closest to the sunlight. Data acquisition based on
programmable current-voltage source power supplies capable
of handling currents up to 8A is used in conjunction with a
proprietary data acquisition system. Calibration of this LIV
measurements system is based on independently measured cSi solar cells at Sandia National Laboratories. The measured
LIV data are shown in Fig.11.

Fig. 11 Measured LIV data at SCFL

Fig. 11 shows the measured LIV data at SCFL. It gives the VI

curve and the efficiency of the fabricated solar cell. From Fig.
11 it is noted that the efficiency of the produced solar cells
were approximately 5.95% at the present condition.
Optimizing the processing techniques as well as the air
quality, water quality and other chemicals, it is very much
potential to raise the efficiency of the solar cell by more than
12%.

COST SCENARIO

The cost of solar cell mainly depends on the quality of cell

material, cell thickness, cell area, cell efficiency, life time, brand
and origin of the country. Now-a-days laboratory based Si solar
cell efficiencies has been reached up to 24.7% whereas at the end
of the 20th century it was struggling in the range of 1215%.
Laboratory based solar cells are basically research grade wafer
which are too expensive. At the same time, industrial scale solar
cells have low price having low efficiency level (12-14%). On a
positive note, the PV market, which grew at an average rate of
13% per year during 1982-1996, exploded with an average
annual growth rate of 3035% per year during 19962002 [25]
and its advantage is the significant cost reduction due to
increased production volume. Presently, there is no solar cell
fabrication company in Bangladesh but nine companies which
basically imports solar cells from foreign countries and assemble
them into panel. Almost all of the companies import low cost &
low efficiency graded solar cells and the average cost of those
solar cells is 41-57 BDT per Wp [26]. This shows that the cost
varies in the range of 3-14% from company to company in
Bangladesh. It is found that cost of the imported solar panels is
higher as well as lower quality rather than locally assembled
solar panels.
VI.

CONCLUSION

For the first time, a laboratory has been set up by BAEC

to fabricate the commercial grade solar cell locally to reduce
the cost from the imported one. To doing so, there are lots of
challenges. There are different recipes and techniques
available in processing solar cells. Of them, a relatively better
process is established to improve the efficiency of solar cells.
Efficiency varies significantly from one recipe to another and
performance indicators. The main challenges are to find out
the right recipe and technique which needs iterative methods.
Completion of a particular recipe and technique takes 3-4 days
and usages lots of consumables (raw materials, chemicals,
gases etc). In summary, significant progress has been made in
understanding and fabricating a wide range of silicon solar
cells with the highest device conversion efficiency of 5.95%.
For instance, cleaning and texturing process, optimum flow
rate for POCl3 and fixing the different temperature zones of
RTA process is achieved which can significantly enhance the
solar cell efficiency. Moreover the technology roadmap shows
that cost reduction can be achieved with better back surface
passivation, improved screen printing, improved surface
texturing process and improve the edge isolation technique.
Moreover, doping concentrations have to be further optimized
and carefully characterized to get more improvement. In the
near future, optimizing all the challenges the laboratory will
play a pioneer role to developing as well as promoting solar
cell fabrication technology in the country.
REFERENCES
[1]
[2]

Wikipedia- Solar Energy http://en.wikipedia.org/wiki/solar_energy.

Need Earth, Solar power is our free source of energy,
http://www.needearth.com/ideas.jsp?s=8

[3]
[4]

[5]
[6]
[7]
[8]
[9]

[10]
[11]

[12]
[13]
[14]

www-inst.eecs.berkeley.edu/~ee143/fa10/lectu res/Lec_26.pdf
4. Rafique, S., Renewable Energy Scenario in Bangladesh: Estimation,
Expectation and Future Trend, presented and published as a Key-note
country paper in WREC (World Renewable Energy Congress) VIII,
Denver, Colorado, USA, Aug 29- Sept 3, 2004.
http://www.thedailystar.net/newDesign/news-details.php?nid=210387
Photovoltaic Education, http://www.pveducation.org
Mark Z. Jacobson (2009), Review of Solutions to Global Warming,
Air Pollution, and Energy Security, p. 4.
Wikipedia-Solar cell http://en.wikipedia.org/wiki/Solar_cell
Nigatu Teklu (2009) Nigatu Teklu 2009, Study the Effect of Ultra
Violate Radiation on Polycrystalline Silicon Solar Cell, Submitted in
Partial Fulfillment of the Requirements for the Degree of Master of
Science in Material Science, at ADDIS ABABA, ETHIOPIA March
(2010).
Andreas Bentzen, Phosphorus diffusion and gettering in Silicon solar
cells, University of Oslo.
Ferrada Martinez, Pablo, Diffusion through Oxide Barriers for Solar Cell
Applications, 2012-06-13,http://nbn resolving.de/urn:nbn:de:bsz:352196 190.
http://www.crystec.com/tridepe.htm
Spin-on Dopant Method, United States Patent by Bruce H. Justice,
Robert F. Aycock.
J. Lossen, L. C. Beneking. Making, Use of Silicon wafers with Low
Lifetimes by Adequate POCl3 Diffusion, 20th European Photovolatic
Solar Energy Conference and Exhibition, Barcelona, Spain, 6-10, 2005.

[15] www.festo.be/nl
[16] Martina A. Green, Crystalline Solar Cell, Photovoltaic Special
Research Centre, University of New South Wales, Sydney, N.S.W
Australia.
[17] Keith Emery, Measurement and Characterization of Solar Cell and
Module, National Renewable Energy Laboratory, Golden, CO, USA.
[18] Radiant Power, Bangladesh.
[19] Electro Solar Power Ltd, Bangladesh.www.electrosolarbd.com
[20] Rahimafrooz Renewable Energy Ltd (RRE), Bangladesh.
[21] AVA Renewable Energy Ltd. www.avasolarbd.com
[22] Greenfinity energy ltd. www.greenfinitybd.com
[23] How Its Made-Solar Panel by http://www.earth4energy.org on
www.youtube.com /watch?v=qYeynLy6pj8
[24] Wikipedia-Solar Panel. http://en.wikipedia.org/wiki/Solar_panel#Mod
ule_perfo rmance_and_lifetime.
[25] Energetics, Incorporated, Ed., Solar Electric Power: The U.S.
Photovoltaic Industry Roadmap.
[26] Md. Shafiqul Islam, Md. Rakibul Hasan, Fariba Mohammadi, Antara
Majumder and Ali Ahmed, Study of Cost-effective Crystalline Type
Solar Panels and Solar Cells Manufacturing Practices in Bangladesh,
Proceedings of the 6th Int. Mechanical Engineering Conf. & 14th
Annual Paper Meet (6IMEC&14APM), ET-12, Dhaka, Bangladesh, 2829 September (2012).

DEVELOPMENT OF LIGHTWEIGHT SPACE SOLAR CELLS

WITH 30% EFFICIENCY AT END-OF-LIFE
G.F.X.Strobl, L.Ebel, D.Fuhrmann,W.Guter, R.Kern,V.Khorenko, W.Kstler, M.Meusel
AZUR Space Solar Power GmbH Theresienstr. 2, 74072 Heilbronn, Germany

Abstract AZURs latest lattice-matched triple junction

space solar cell product has been designed for a high EOL
efficiency by focusing the development on radiation hardness
right from the beginning. These solar cells named as 3G30Advanced show an efficiency of 29.5% at BOL and of 28.1%
for EOL at 5E14 (1MeV) e-/cm2 (AM0, 1367 W/m2, 28C).
Meanwhile more than 125,000 cells of this cell type have been
delivered to customers worldwide. Cell thicknesses of 150 m,
80m or even as thin as 20m are available. Besides the
standard size of 8x4 cm2 (with cropped corners) larger sizes
such as 8x8cm2 and 12x6 cm2 are also available and provide
equal performance. AZURs next generation product will
comprise a metamorphic 4-junction device targeting at 30%
efficiency at EOL to be qualified in 2015/16. Furthermore, for
arriving at more than 30% EOL efficiency with 4 - 6 junction
solar cells in a subsequent step, appropriate lattice-matched
1eV dilute nitride materials, inversely grown solar cells and
semiconductor bonded cells are studied in cooperation with
external partners.
Index Terms end-of-life-efficiency, GaAs based space solar
cells, radiation damage, subcell design

I. INTRODUCTION
During the last years a large number of new satellite
orders was placed by the need in replacement of old
platforms, whereas in terms of applications, the major focus
currently is on TV broad-casting, high definition TV and in
the near future, 3-D TV. New payloads shall provide
significantly larger capacity with simultaneously higher
transmitting signal quality so that the demand in electrical
power to be provided by solar generators increases
correspondingly. This demand extremely sharpens the need
for lightweight, high-performing and cost effective
components and sub-systems for spacecrafts. In case of solar
cells this particularly means i) increase in solar array power
(W/m2), ii) reduction of specific mass (kg/m2), iii) reduction
of manufacturing costs (/W) and iv) increased operational
lifetime (>15y).
II. CURRENT SPACE TRIPLE CELLS 3G28
The 28%-class triple junction cell 3G28 was developed
mainly for applications in GEO. From 2009 up to now it was

978-1-4799-4398-2/14/$31.00 2014 IEEE

the workhorse for numerous applications in GEO and LEO.

More than 1,600,000 cells have been delivered to both the
main customers such as EADS-Astrium and ESA and to
further customers worldwide.
A. Standard cells 3G28C
The 3G28%-class solar cell was developed to the
production stage under ESA contract [1]. The 3G28 cell
structure consists of 3 p-n junctions based on InGaP,
InGaAs and Ge semiconductor layers. The In0.01 Ga0.99As
and In0.5 Ga0.5P layers are grown lattice-matched and with
high quality on a Ge substrate. The Ge sub-cell is formed by
diffusion of group V elements into the substrate during the
epitaxial layer growth in the MOCVD reactor. The main
focus during the development of the etriple-junction solar
cell stack at AZUR was set on EOL efficiency resulting in
the proprietary AZUR radiation hard cell design.
The solar cells are typically 4x8cm2 in size with 2 cropped
corners in order to utilize the 100mm epitaxial wafer
optimally. The total cell area of this geometry consists of
30.18 cm2. Successful qualification against ESA ECSS-E20-A has been performed on bare cell-, solar cell assembly-,
coupon- and panel-level [2]. Nominal cell data from the
ESA qualification program are given in Table I.
TABLE I
ELECTRICAL BOL- AND EOL-DATA (1MEV EL) OF 3G28%2
CLASS SOLAR CELLS (28C, AM0 1,367 W/M )
Average Voc
Average Isc
Average Vmp
Average Imp
Average Pmp
Average AM0

[mV]
[mA]
[mV]
[mA]
[mW]
[%]

BOL
2,667
506
2,371
487
1,155
28.0

5E+14
2,534
501
2,229
472
1,052
25.8

1E+15
2,480
486
2,205
458
1,010
24.7

On panel the cells can be protected against reverse

biasing either with an external silicon diode to be placed
into the cropped corners of the solar cell (Fig. 1) or with an
integral diode which is formed out of a part of the solar cell
structure (Fig. 2) and utilizes the bottom and middle sub-cell
junctions.

3595

Material quality,
design of layers + grid

GaInP
tunnel diode

Higher transparency

GaInAs
tunnel diode
buffer
Fig. 1.
scheme

nucleation

Interconnection of solar cells using external Si-Diode

Ge

Fig. 3.

Interconnection of solar cells using integral diode

B. Customized cells
For some ESA projects, special solar cells have been
derived from the basic 3G28 cell design. Based on
particular project specifications, various modifications in
cell dimensions, epitaxial structures, antireflective coatings,
grid designs, contact pads, as well as tunnel and bypass
diodes were realized. Examples are:
- Solar cell 3G28 with mesa groove and high density
grid and with a modified contact material system
allowing performance improvement with respect to
higher currents and higher temperature for near sun
missions such as BepiColombo and Sun Probe [3].
- 3G28 solar cell with LILT (low intensity, low
temperature) optimized design for deep space missions
such as Laplace demonstrating outstanding LILT
performance of about 35% with the absence of flat
spot effect on SCA and bare cell level [4].
For each special application a delta-qualification of the
modified cells has been carried out according to the very
special requirements of the individual mission.

III. SPACE TRIPLE CELL 3G30 - ADVANCED

The 3G30C - Advanced is a the result of detailed analyses
of the 3G28C cell design and systematic improvement of
identified cell structure elements (Fig. 3).

Fields of improvements with regard to the 3G28C

The most effective improvement could be achieved by

realization of a highly transparent upper tunnel diode by
using higher bandgap materials. Thus more photons are
available for absorption in the middle cell to increase the
middle sub-cell current. In order to keep the same current
mismatch between middle and top sub-cells as for the 3G28
solar cell, the InGaP top sub-cell current needed to be
increased. This has been achieved by optimizing the top subcell epitaxial layers. The InGaP top sub-cell is only about
500 nm thick leading to high radiation hardness. The total
thickness of solar cell epitaxy amounts to approximately
5m.
Finally average BOL efficiencies of 29.5% have been
reached. EOL current-matching has been designed for 5E14
1MeV electrons/cm2 as currently preferred by manufacturers
of telecommunication satellites. The efficiency degradation
under 1MeV-electron irradiation up to a fluence of
1E16/cm2 is shown in Fig. 4 [5], typical cell data from the
qualification program in Table II.
30

AM0 Efficiency [%]

Fig. 2.
scheme

Improvement of the
surface morphology
(substrate, nucleation,
buffer)

28
26
24
22
20
18

3G30-Advanced
1E14

1E15

-2

1E16

1 MeV Electron Fluence [cm ]

Fig. 4. Efficiency degradation of the 3G30-Advanced under

1MeV electron irradiation

978-1-4799-4398-2/14/$31.00 2014 IEEE

3596

TABLE II
ELECTRICAL BOL- AND EOL-DATA (1MEV EL) OF 3G30
2
CLASS SOLAR CELLS (28C, AM0 1,367 W/M )
Average Voc
Average Isc
Average Vmp
Average Imp
Average Pmp
Average AM0

[mV]
[mA]
[mV]
[mA]
[mW]
[%]

BOL
2,700
520
2,411
504
1,215
29.5

5E+14
2,564
514
2,290
501
1,147
28.1

1E+15
2,522
502
2,246
487
1,094
26.8

The 3G30-Advanced solar cell is available in both a size

of 48cm2 or 8x8cm2 produced on Ge wafers of 100mm
diameter and a larger size of 6x12cm2 produced on Ge
wafers of 150mm diameter (see Fig. 5). All cells have
cropped corners. The nominal cell thickness is 145m, the
cells have silver based weldable contacts with a gold flash
on top and a double layer antireflective coating.
These cells also are available as solar cell assemblies.
AZURs 3G30A-Advanced solar cell assembly (Fig. 6) is
composed of 50 to 150 m thick coverglass and welded
silver coated Kovar interconnectors. The coverglass
dimension provides a 100%-coverage of the cell.
Like the 3G28C the 3G30C-Advanced cells can be
protected either with an external silicon diode to be placed
into the cropped corners of the solar cell (Fig. 1) or an
integral diode which is part of the solar cell (Fig. 2). The
integral diode contact is placed on the cropped long side and
the silicon diode in one cropped corner of the cell (Fig. 6).

Fig. 6. Solar cell assemblies of type 3G30A-Advanced with

welded Kovar interconnectors and microsheet coverglass, bottom
SCA with external silicon by-pass diode, top SCA with integral
diode ID2*

A. Lightweight Cells
AZUR has developed a special proprietary thinning
process, combining mechanical as well as chemical removal
of the main portion of the Ge substrate. As the Ge substrate
also constitutes the bottom sub-cell, a remainder of Ge has
to be saved. Using this process the thickness of the Ge
substrate could be reduced from nominally 145 m to 80
m. Feasibility studies have shown that the cells can also be
thinned further to only 50 or even 20m. These cells have to
be stabilized by a coverglass to avoid coiling up and to
facilitate handling. This way, up to 70% of the bare cell
weight has been saved (Table III). Additionally, low-weight
and flexible cells allow for a redesign of the complete PVA
structure permitting further reduction in mass for several
elements and therefore also in cost.
TABLE III
MASSES OF VARIOUS SCA-CONFIGURATIONS

Fig. 5. Bare cells of type 3G30C-Advanced with size of

48cm2 , 8x8 cm2 and 612 cm2 .

978-1-4799-4398-2/14/$31.00 2014 IEEE

Cell Thickness [m]

145
145
80
80
50
50
20
20
20

Cell Mass
[mg/cm2]
84
84
49
49
33
33
17
17
17

Cover Thickness [m]

100
80
100
80
100
50
150
100
50

SCA Mass
[mg/cm2]
114
108
79
74
63
50
60
47
34

3597

The efficiency distribution for delivery lots of 130,000

3G30 cells in total shows a tight symmetric Gaussian
efficiency distribution peaking at 29.8% (Fig. 7). The
quantity of low performers lies below 1%. These lots well
reproduce the values of the 700 cell qualification lot [6].
AZURs BOL and EOL performance data of the 3G30CAdvanced have recently been reconfirmed by an
independent ESA verification program with average EOL
efficiencies of 27.9% at 5E14 1 MeV e /cm2 and 26.6% at
1E15 1 MeV e/cm2. This still represents best-in-class values.

Fig. 8. AZUR TET-1 payload consisting of three 4cmx8cm

3G30 advanced cells and three 2cmx2cm experimental cells of
future design

15

Production Share [%]

particular case, even the thinnest cell shows the highest

efficiency.

10
110

<28
22.0

28

29

Efficiency [%]

30

31

Fig. 7. Efficiency distribution at AM0 (1367 W/m, 28C) for

several delivery lots of 130k 3G30-Advanced cells [10]

Pmp,nom[%]

150 m

80 m

20 m

105

100

95
07/2012 11/2012 03/2013 07/2013 11/2013

Date

B. Flight Data
Cells of the 3G30-Advanced type were flown on the
German technology satellite TET-1. TET-1 was launched on
July 22, 2012 from Baikonur with a Sojus Fregat Launch
vehicle into a circular sunsynchronous orbit at 506 km
altitude. The AZUR payload consisted of a CFC-panel with
6 advanced solar cells three of 3G30-Advanced type (Fig.
8). The cells had a size of 4cmx8cm each, were of different
thickness ranging from 150m, 80m to 20m and were
covered with a CMX100 coverglass from Qioptiq (UK). The
IV-characteristics of each cell as well as the substrate
temperature were measured regularly several times a week.
The first data transmission occurred on July 26, 2012 and
was continued until mission end on October 31, 2013. The
sun orientation was first at 30 due to thermal constraints
and was adjusted to 0 in May 2013. The transmitted IVdata were normalized to AM0 standard conditions at 28C.
This allowed comparison with the calibration measurements
prior to integration.
The accuracy of each measurement was 2%. By the high
number of measurement points the standard deviation could
be kept below 1%. In Figure 9 the evolution of maximum
power over operation time of the satellite is depicted. The
majority of the data lie well above the calibration points.
Thinning has hardly any impact on cell performance. In this

978-1-4799-4398-2/14/$31.00 2014 IEEE

Fig. 9. In-orbit Pmax data of 3G30-advanced cells with various

thicknesses on board of DLR satellite TET1 normalized to AM0,
28C standard conditions. Dashed lines indicate lab-calibration
values.

IV. UPRIGHT METAMORPHIC 4J SOLAR CELLS

The main weakness of lattice-matched designs such as the
3G30-Advanced solar cells is the waist of excess current in
the bottom sub-cell [7]. Thus in order to substiantially
increase the solar cell efficiency a 4J upright metamorphic
cell design has been introduced which has an efficiency
potential of 33% at BOL [7]. This design requires a
metamorphic buffer layer with a high lattice-mismatch of
around 1.5% to Ge as well as the introduction of larger
bandgap Al-compound materials for the upper junctions. A
straight-forward approach uses AlInGaP (1.9 eV) and
AlInGaAs (1.4 eV) for the upper sub-cells, combined with
InGaAs (1.1 eV) and Ge (0.7 eV) for the lower ones.
In Fig. 10 the external quantum efficiency (EQE) of the
sub-cells of a realized UMM 4J solar cell structure is shown
both BOL and EOL (after 1 MeV electron irradiation with a
fluence of both 5E14 e-/cm2 or 1E15 e-/cm2). Within this
cell AZURs proprietary EOL design feature has already

3598

been included. First studies reveal a similar extraordinary

radiation hardness for voltage and current as already
achieved for the 3G30-Advanced solar cell. Nevertheless the
current matching for all subcells was not yet perfect in this
cell structure (Fig. 10).

30

lot [%]

100

80

20

EQE

60

10
40
BOL
-2
5E14 cm
-2
1E15 cm

20

400

600

800

1000 1200
[nm]

1400

1600

0
1800

Fig. 10. External quantum efficiency (EQE) of subcells of an

UMM 4J solar cell structure both BOL and EOL [5].

The achieved open-circuit-voltage so-far was 3.3V.

Anyhow the strong metamorphic growth of these 4J solar
cell materials still yields small losses of around 30mV
compared to the lattice-matched growth which has in depth
been studied by appropriate isotype cells [6] [8].
Already in 2009 a terrestrial CPV world record efficiency
of 41.1% at 454 suns has been achieved with a metamorphic
triple junction solar cell on Ge substrate [9]. AZURs
current 42% terrestrial CPV product 3C42 is based on this
metamorphic technology [10].
AZUR SPACE adapted the cell structure of this CPV
3C42 for the operation under AM0 conditions and in a space
radiation environment, resulting in a UMM 3J cell with a
lattice-mismatch of about 0.3%. An average efficiency of
31.3% could be achieved for standard AM0 conditions as
shown in Figure 11.
Compared to the 3G30C-Advanced, the open circuit
voltage Voc of 2.6 V is about 100 mV lower (3.7%), on the
other hand the short-circuit-current Isc of 565 mA is about 45
mA higher (+8.6%), whereas the fill factor is hardly
affected.
After 1MeV electron irradiation approximately the same
remaining factors could be achieved as for the latticematched 3G30-Advanced solar cell.

978-1-4799-4398-2/14/$31.00 2014 IEEE

30.0

30.5

31.0

31.5

eta [%]
Fig. 11. Efficiency distribution at AM0 (1367 W/m2, 28C) for
UMM 3J bare solar cells in standard thickness of 145m [11].

In the next development steps for the 4J-UMM solar cell

the proper current matching for all 4 sub-cells has to be
achieved and the material quality of the Al-containing layers
must be improved. Then the target efficiency shown in
Table IV will be achievable for AZUR SPACEs new 4J
upright metamorphic cell. Qualification of this cell type is
planned for 2015/16.
TABLE IV
TARGET PERFORMANCE FOR 4G33 UPRIGHT METAMORPHIC
2
SOLAR CELLS (28C, AM0 1367 W/M ) BOTH BOL AND EOL
(1MEV E-)
Voc
Isc
FF
AM0

[mV]
[mA/cm2]
[%]

BOL
3370
14.9
0.87
32.0

EOL(5e14) EOL(1e15)
3200
3150
14.7
14.3
0.87
0.86
30.0
28.5

V. COMPLEMENTARY TECHNOLOGY
DEVELOPMENTS
In order to surpass the 32% efficiency for the next
generation solar cell further cell concepts are also
investigated by AZUR and partners [6]. Essentially all other
4J approaches target about the same bandgap combination
as the 4J UMM cell structure, but implement it with
different means.
Lattice-matched 4J cells incorporating InGaNAs still
require intensive material research on basic R&D level.
InGaNAs material can be grown either by MBE or

3599

MOCVD. In case of MOCVD growth this would be the

preferable method due to its manufacturability and costeffectiveness, anyhow the MOCVD grown GaInNAs
material still is of much lower quality compared to the MBE
grown one. Inverted metamorphic 4J cells are also
considered as back-up for the upright metamorphic approach
although they need extremely strong lattice-mismatch and
long epitaxial growth time.
Semiconductor bonding technology is considered by
AZUR only for the next but one generation solar cell
product, it may require epitaxial growth on other substrates
than Ge or GaAs. This intricate technology may pave the
way for highest achievable cell efficiency conceivable at the
moment.
VI. CONCLUSION
AZURs latest qualified product, 3G30-Advanced solar
cell with its outstanding EOL performance is now in mass
production. Both bare cells in several sizes and solar cell
assemblies with Kovar interconnectors are available. For
mass critical missions low weight versions with cell
thicknesses of 20-50m can be delivered.
Currently a 4J upright metamorphic solar cell is
developed to target the 33% efficiency class, qualification is
planned for 2015/16.
Complementary approaches such as inverted metamorphic
growth or diluted nitrides are investigated in cooperation
with partners such as ISE. Semiconductor
bonding
technology is considered to bear AZURs next but one
generation cell product.
ACKNOWLEDGEMENT
The authors like to thank our colleagues from AZUR
SPACE for their support and our coworkers from
Fraunhofer/ISE for their contribution. Special thanks go to
Gnther LaRoche for managing the TET1/DLR project and
for his general support with paperwork. Part of the R&D
work presented in this paper was supported by the European
Space Agency (ESTEC contracts No. 22969/09/NL/GLC,
4000100600 and 4000104852) and the German Aerospace
Agency DLR/BMWI (Contract No. 50RN1002).

[3] C.G.Zimmermann,
C.Nmayr,
W.Kstler,
A.Caon,
E.Fernandez, C.Baur and H.K.Fiebrich, Photovoltaic
Technology Development for the Bepi Colombo Mission,
Proc. of 9th European Space power Conference ESPC 2011,
Saint Raphael, France
[4] V.W.Khorenko, G.F.X.Strobl, R.Hoheisel, F.Dimroth,
R.Campesato, M.Casale and C.Baur, Triple-Junction Solar
Cell design for Low Intensity, Low Temperature Space
Applications, Proc. of 9th European Space power Conference
ESPC 2011, Saint Raphael, France
[5] W. Guter, L. Ebel, D. Fuhrmann, W. Kstler and M.Meusel,
Development of Space Solar Cells with 30% EOL
Efficiency, Space Power Workshop 2014, Manhattan Beach
(USA, CA), 2014.
[6] G.F.X.Strobl, D.Fuhrmann, W.Guter, V.Khorenko, W.Kstler
and M.Meusel, About AZURs 3G30-advanced Space Solar
Cell and next Generation Product, Proc. of the 27th
EUPVSEC, Frankfurt, p. 1BO10.3, 2012.
[7]

W. Guter, W. Bensch, R. Kern, W. Kstler, M. Meusel, G.

Strobl, S. v. Riesen, T. Gerstmaier, A. Gombert, F. Dimroth
and A. Bett, Development, Cost Reduction and Customized
Design of Industrial Concentrator Solar Cells with
Efficiencies Approaching 40% and Above, Proceedings 35th
IEEE PVSC, Honolulu (USA), pp. 196, 2010.

[8] W. Guter, R. Kern, W. Kstler, T. Kubera, R. Lckenhoff, M.

Meusel, M. Shirnow and G. Strobl, III-V Multijunction
Solar Cells New Lattice-Matched Products And
Development Of Upright Metamorphic 3J Cells,
Proceedings of the CPV-7, LasVegas (USA), pp. pp. 5-8,
2011.
[9] W. Guter, J. Schne, S. Philipps, M. Steiner, G. Siefer, A.
Wekkeli, E. Welser, E. Oliva, A. Bett and F. Dimroth,
Current-matched triple-junction solar cell reaching 41.1%
conversion efficiency under concentrated sunlight, Appl.
Phys. Lett. 94, (2009) 223504, vol. 94, p. 223504, 2009.
[10] D. Fuhrmann, M. Meusel, L. Ebel, W. Guter, T. Kubera W.
Kstler and G.Strobl, About 42%-Class CPV Cells and
Pathways Beyond, Proc. of the CPV9, Myazaki, 2013
[11] W. Guter, L. Ebel, D. Fuhrmann, W. Kstler and M.Meusel,
Development, Qualification and Production of Space Solar
Cells with 30% EOL Efficiency, Proc. of the 10th ESPC
2014, Noordwijkerhout (NL), to be published 2014.

REFERENCES
[1] AZUR Final Summary Report HNR 0002274-01
Development of advanced GaAs-based triple-junction solar
cells under ESA Contract No. 15134/01/NL/FM
[2] AZUR document HNR 0002922-00, Executive Summary
Report Space Qualification of European Triple-Junction
Solar
Cell
AZUR-3G28
under
ESA-Contract
18118/04/NL/US

978-1-4799-4398-2/14/$31.00 2014 IEEE

3600

Potential and Present Status of 111-VlSi Tandem Solar Cells

Masafumi Yamaguchi
Toyota Technological Institute, 2-12-1 Hisakata, Tempaku, Nagoya 468-8511, Japan
Abstract

expected

to

III-V compound and Si tandem solar cells are

have

great

potential

of

space

and

terrestrial

application because of high efficiency, light weight and low cost

potential. However, problems to be solved are high density of
dislocations and large residual strain in III-V compound on Si in
the case of hetero epitaxy of III-V compound materials on Si
substrates. There are other approaches such as epitaxial lift-off
and

direct

bonding

technologies

in

order

to

prevent

such

problems. This paper presents efficiency potential of III-V/Si

are other approaches such as epitaxial lift-off and direct bonding

technologies.
This paper presents effects of dislocation upon solar cell
properties of III-V single-junction solar cells and III-V/Si tandem
solar cells, approaches on reduction in dislocations in III-V
compound films on Si and improvements in efficiencies of III-V
compound single-junction solar cells on Si substrates and III-V/Si
tandem solar cells

tandem solar cells and effects of dislocation upon solar cell

properties of III-V-on-Si single-junction solar cells and III-V/Si
tandem

solar

cells.

The

paper

also

reviews

approaches

on

reduction in dislocation density in III-V compound films on Si

and improvements in efficiencies of 111-V compound single

junction solar cells on Si substrates and III-V/Si tandem solar

cells. We have demonstrated 20% efficiency (under I-sun of
AM1.5G) with hetero-epitaxially grown GaAs single-junction
solar cells on Si and Ohio State University has also demonstrated
17.1% (AMO) efficiency with GaAs single-junction solar cells by
using Ge buffer layer. Nagoya Inst. Tech. has achieved 22.1%
with GaAs/Si tandem solar cell by hetero epitaxy. Most recently,
Fraunhofer ISE has demonstrated 27.9 % efficiency (under 48.3suns of AM1.5D) InGaP/GaAs/Si 3-junction solar cells by using
epitaxial lift-off and direct bonding. Future prospects are also
discussed.

Index Terms

III-V/Si tandem solar cells, hetero-epitaxy of

III-V compounds on Si, dislocation behavior, minority carrier

lifetime, solar cell properties.

'.50

------,,---=TW::-:O:::-::T2::-:Eft:7:-::IN AL

, 25 ___-c-.lit=-'l:iI'1I"'_A_M_' _
.

:;o

"'

CJ

0.75

-,,-'5'-0 ---'-==--=2.0-="": --=0 2.2'0:5:-:

: 2"'
.5'"0

t,GO

THREE CEl.1.S

$eRIU'ONH(CTEO

lonOM,CUl BANDGAP . 1.00 ,v

..

': .-

,, ------c.:.....-

-n9'fff;,fT-----:-:::-

I.$O ____

;-

TOP-CELL BANDGAP leV)

(a ) 2-junction cell

SINGLE AND III-V/SI TADEM SOLAR CELLS

Because III-VlSi system has large misfit and difference in
thermal expansion coefficient, generation of misfit and thermal stress
induced dislocations affect on solar cell properties. Effects of
dislocations upon minority-carrier lifetime are derived from
considering one dimensional transport of minority carriers to
dislocations [2]. One dimensional continuous equation is expressed
by
d!1nI dt = Dd2!1n I dx 2 11,
(1

where, I1n is excess minority-carrier density, and D is minority

carrier diffusion coefficient.
The following boundary conditions have been assumed.

n = 0,

at

(2)

at

(3)

where, Xc is radius of unit cell affected by one dislocation, and N d

III-V compound and Si tandem solar cells are expected to have

great potential of space and terrestrial application because of high
efficiency, light weight and low cost potential. Figure 1 shows high
efficiency potential [1] of III-VlSi 2-junction and 3-junction solar
cells.

g.

ON SI UPON SOLAR CELL PROPERTIES OF III-V- ON-SI

d!1nldx = 0,

I. INTRODUCTION

II. EFFECTS OF DISLOCATIONS IN III-V COMPOUND FILMS

( b) 3-junction cell

is dislocation density.
The resulting dislocation-limited minority-carrier lifetime is given by

IITd = ,,3NdDI4,(4)

Therefore, the effective minority-carrier lifetime, eff is expressed

by the following equation [2]:

lITeff =lITrad +lITd =BN +,,3NdDI4,(5)

where, 'rad is minority-carrier lifetime for radiative recombination,

B is the radiative recombination probability, and N is the carrier
concentration.
Figure 2 shows effects of dislocation density upon III-V-on-Si
single-junction, III-VlSi 2-junction and concentrator 2-junction solar
cells. In order to realize high efficiency solar cells with similar
efficiency by homo-epitaxially grown solar cells, low density
dislocation of less than 3x105 cm,2 is necessary.

Fig.l. High efficiency potential of monolithic cascade III-V/Si multi

junction solar cells.
However, problems to be solved are large density of dislocations
and large residual strain in III-V compound on Si in the case of
hetero-epitaxy of III-V compound materials on Si substrates. There

978-1-4799-4398-2/14/$31.00 2014 IEEE

0821

50
,-.

40

....
u
s::

30

:::...
.ilI

<.i
<=
'..
s::

.!:

'"
""
..
..
c
0

Coocentrator

Tandem

20

Because dislocation coalescence is thought to be the main

mechanism on dislocation annihilation in the thennally cycle
annealed materials, K,D KzDz were assumed.
The resulting solution is given by

D=Do 1[1+DoK2t ]

Cells

(12)

In the analysis, reaction constant

proportional to dislocation velocity.

Single-junction

f3 -dislocations

K2

was also assumed to be

In particular, the velocity of

in GaAs bulk crystals [5] under a stress of a was

used as follows:

10

v=9.86xlO-9(j'L6exp(-1.35e VlkT)
K =vlh=av
(14)

0
103

10

IO

10

Oisloc atioo den sity

10

(an -2 )

Fig. 2. Effects of dislocation density upon III-V-on-Si single-junction,

III-VlSi 2-junction and concentrator 2-junction solar cells.
III. APPROACHES OF REDUCTION IN DISLOCATION

[emI5]

2
where V is dislocation velocity and, a is a constant, determined to
8
be 7.2xlO cm from experimental results for GaAs-on-Si [6]. The
stress as a driving force of dislocation motion is thought to be
derived from misfit stress at around annealing temperature. In this
case, stress value of 107 dyne/cm2 was used as the stress value.

Hetero-epitaxial Film
Prop'gation

Deflection

Combination
+Re-emlsslon

+Propgation

DENSITY IN III-V COMPOUND FILMS ON SI

Hetero epitaxy of III-V compound materials on Si has great
potential of electronic, optical devices including solar cells and
optoelectronic integration. There are, however, problems to be
solved are high density of dislocations and large residual strain in
III-V compound on Si in the case of hetero epitaxy of III-V
compound materials on Si substrates. Based on our previous studies,
some approaches for reducing dislocations in III-V compound
materials on Si substrate are shown.
3.1.

ThermalCycle Anneali ng (TCA)

In order to reduce dislocation density in III-V compound film

on Si, effectiveness of thermal cycle annealing [3] and insertion of
strained layer super-lattices [4] have been demonstrated. Figure 3
shows mechanisms of dislocation motion in III-V film hetero
epitaxially grown on Si substrate [3].
Dislocation annihilations, such as movement toward edges and
interfaces, dislocation coalescence, and dislocation re-emission due
to thermal annealing, were considered in the analysis as follows:

K
D-' a nnihilat ion

( 6)

Kz
D+D- Dz

(7)

(13)

Fig. 3. Mechanisms of dislocation motion in III-V film hetero

epitaxially grown on Si substrate.
Figure 4 compares experimental results with calculated results
for changes in dislocation density for GaAs hetero-epitaxially grown
on Si using thermal cycling annealing at the annealing temperature
of 700, 800 and 900c [3]. Good agreement between experimental
values and calculated results validates the analytical model.
However, it is necessary to optimize the thermal cycle annealing
process in order to obtain excellent quality of III-V compound film
on Si.

where D,

D are densities of dislocations and coalescenced

2
dislocations, respectively, and K"
K are the corresponding rate
2

constants.
The reaction equations for dislocations are given by

dDldt =-KID- KzD

z
dDzldt =K2D l2

(8)
(9)

The boundary conditions are a s follows:

D(t = 0) =Do,

D2 (T = 0) = 0

The resulting solutions are given by

D=1/[ (11 Do + K2 1 K, )exp(K, t ) - K2 1 K, ]

(10)
2
z
Dz = (Do KzIK, )[I- (exp(-K,t ]
(11)

978-1-4799-4398-2/14/$31.00 2014 IEEE

10e
o

________

=5

________

________

10

Cycle number N

16

Fig. 4. Comparison between experimental and calculated results for

changes in dislocation (etch-pit) density for GaAs hetero-epitaxially
grown on Si

using thennal cycling annealing at the annealing

temperature of 700, 800 and 900C.

0822

3.2.

St ra i ned -Layer Super-latt ice (SLS) I nsert ion

Figure 5shows changes in etch-pit density (EPD) in GaAs films

grown by insertion of Ino I Gao9AsIGaAs or Ino I Gao9AsiIGaAso sP0.2

SLSs (Strained Super-Lattices) after TCA (thermal cycle annealing)

on Si substrates, as a function of individual SLS layer thickness.
High quality GaAs hetero-epitaxial films with dislocation densities
of about 1x106 cm-2, on Si substrates, haven obtained by combining

Ino lGao9As/GaAs SLS insertion and TCA [4,7]. The reduction of the
dislocation density is found to become more remarkable as the

individual SLS thickness increases. However, for films thicker that

the critical thickness (10-20 nm) of the SLSs, the dislocation density
in the GaAs/Si seems to increase with an increase in SLS thickness.
These results suggest that further reduction of dislocation density in
III-V compound films on Si substrates will be realized by optimizing
SLS insertion and TCA.
o ilSilrown

108 r-------.
10 period

;:;'

0 InO.1GaO.9As/GaAs

(TEM)
(EPD)
TCA
(PO)
(900'C, N=4)

.....

lIS'grQwn

Ino., Gao.9As/GaAso.sP 0.2

E
.3

10
Dista nee from Si

(.um)

Fig. 6. Changes in the dislocation density, determined by KOH

etching and TEM observation, for as-grown and thermal cycle
0
annealed (TCA) GaAs/Si (Ta =900 C, N=4) as a function of film
thickness

101
1

10

Individual SLS thickness (A)

tot

Fig. 5. Changes in etch-pit density (EPD) in GaAs films on Si due to

10 periods of SLS insertion after TCA as a function of individual
SLS layer thickness.
3.3.

Thicker Layer Growth

Figure 6 shows changes in the dislocation density, determined by

KOH etching and TEM observation, for as-grown and thermal cycle
0
annealed (TCA) GaAs/Si (Ta =900 C, N=4) as a function of film
thickness [7]. The result shown in Fig. 5 suggests that a dislocation
coalescence mechanism [8, 9] is the main mechanism of dislocation
annihilation in the as-grown and thermal cycle annealed GaAs/Si.
This result also suggests that dislocation density in III-V compound
films on Si substrates can be decreased because dislocation density
depends on the inverse of film thickness.

>-

t:

and

open

and

closed

circles

were

determined

by

density

decreases

as

the

film

thickness

increases.

dependence seems to separate into two different regions.

Its

8 107
I-

,
"

b,

...... '"

<..>

-- '"

(/)

6
0 10

10 5

TEM

results show the same dependence. It is clear that dislocation

7 [10]. Closed

observation and molten KOH experiments, respectively. Both

109

(/) 10
z
UJ
Q

The dislocation density dependence on film thickness of

GaAs films on Si substrates is shown in Fig.

GoAs/SI

Fig.

201

100
GoAs FILM THICKNESS

(I'm)

7. Changes in dislocation density in GaAs films on Si as a

function of GaAs film thickness. Closed and open and closed

circles were determined by TEM observation and molten
KOH

experiments,

respectively.

The

lines

represent

the

calculated values.

978-1-4799-4398-2/14/$31.00 2014 IEEE

0823

Dislocation density

Nd depends on the inverse of film

thickness in the region of film thickness < 10 Jim . This

dependence is the same as that in Ref. 8.

Nd(X)

C1 X

x < IO,um,

for

(15)

where x(cm) is the GaAs film thickness, and C is constant and

I
S.6x104 cm- for this experimental result.

However, in the film thickness region > 50 Jim (and/or

dislocation density < 107 cm-2), a lower dislocation density is
obtained as expected from eq. (1S), and dislocation density depends
exponentially on film thickness.

for
x>50,um,
(16)
2
7
C 5xlO-3 cm for these
1.9xl0 cm- ,

NAx)= NdOexP(-xlxo)
where constants NdO

experimental results.
Based on one-dislocation reaction such as annihilation and
deflection of dislocation, and two-dislocation reaction such as
coalescence of dislocations, the equation of dislocation density
GaAs film thickness x is given

dD(x) 1 dx

-aD - bD2

D for

(17)

where D,

D are densities of one dislocations and coalescenced

2
dislocations, respectively, and a, b are the corresponding rate

constants.
Therefore,

dislocation

density

Nd

dependence

on film

thickness x is expressed by

Nd(x) = 1/[(11 NdOO + b 1 a)exp(ax) - b 1 a],

(18)

where NdOO is the dislocation density at the GaAs-Si interface.

From eq. (18), in the region of film thickness <IOm,
dislocation density is thought to depend on the inverse of film
thickness x:

NAx)11 bx

(19)

However, in the larger film thickness region >SOm, dislocation

density is thought to depend exponentially on film thickness:

Nd(x)(alb)exp(-ax).

(20)

From these experiments and obtained equations from eq. (17) to

eq. (20), dislocation coalescence reaction mainly occurs in high
dislocation density crystals, and additional annihilation reaction,
deflection of dislocation, occurs in low dislocation density crystals.
Although low dislocation density <4x105 cm-2 has been achieved
with a GaAs film thickness of 180m on Si substrate, improvements
for increase rate constants a and b is very important for obtaining
high quality III-V materials on Si...
3.4.

Select ive Area Growth

High quality GaAs film with an EPD of less than lx106 cm-2 on
Si substrates have been obtained by selective area growth [11, 12].
The selective area growth is thought to aid stress relief and thus
prevent dislocations from generating. The major component of the
9
large residual stress of about I.Sx10 dyne/cm2 in the as-grown GaAs
film on Si substrates is thermally induced stress after growth; stress
relief in the films fabricated by selective area growth has been
analyzed using the bi-metal model [13] as follows:
0"=

where

O"o[l-exp{-k(w- x)}].

a,

(21)

0'0 are residual stress for selective area growth and

distance from the center of the patterned film, and kis the interfacial
compliance parameter related to the elastic constant, young's
modulus of film and substrate, and film thickness,
Numerical analysis [12] suggests that high quality III-V on Si
with a dislocation density less than 105 cm-2 will be obtained by
optimizing selective area growth and TCA, but selective area
effective for low dislocation density is thought to be limited with
selective area less than 20m..
3.5. Low Temperature Growth

Because dislocations in III-V compound materials grown Si

substrates are mainly induced during cooling process after growth
[14], low dislocation density will be realized by low temperature
growth with growth temperature of less than 400 C. High quality
GaAs films with dislocation density of 7x104 cm-2 have been
fabricated on Si substrates by combining migration-enhanced
epitaxial growth at 300C and InGaAs/GaAs SLS insertion by NTT
labs. [IS]. Similar high quality GaAs films with dislocation density
of less than lx105 cm-2 have also been fabricated on Si substrates
using atomic hydrogen assisted molecular beam epitaxial growth at
330 C and high minority-carrier lifetime of 8 ns has been achieved
by Univ. Tsukuba [16]. However, high performance GaAs-on=Si
solar cells have not yet been fabricated using low temperature growth.
High temperature device processing is thought to degrade properties
of solar cells fabricated using low temperature growth.
IV. APPROACHES OF HIGH EFFICIENCY GAAS-ON-SI

SINGLE JUNCTION SOLAR CELLS AND III-V/SI TANDEM

SOLAR CELLS
Figure 8 shows a structure of a high-efficiency GaAs thin
film solar cell fabricated on a Si substrate by the metal-organic
chemical vapor deposition (MOCVD) method [3, 17, 18]. A
low-pressure MOCVD was used for GaAs heteroepitaxial
o
growth on Si substrates 0.S -2 tilted toward [Oil] using
trimetyl-gallium (TMG) and arsine (AsH 3). To remove native
oxides from the Si substrates, the substrates were preheated at
0
about 1000 C in a H 2 flow for lOmin. The two-step growth
method [19] was applied. After growing a GaAs layer with a
film thickness of I.S-2m, thermal cycle annealing (TCA) [3,
20] was carried out. After the TCA process, a GaAs layer was
grown on the annealed GaAs/Si at a temperature of Tg' The
In
addition,
total
film
thickness
was
3.S-4/lm.
Ino I Gao 9As/GaAso sP 0 2.
or
Ino I Ga0 9As/GaAs,
Al0 6Ga04AI/GaAs strained-layer supperlattices (SLSs) [4, 20,
21] were inserted after TCA (annealing temperature: 800C,
cycle number: 3) to reduce the dislocation density in the GaAs
films on Si. As an intermediate layer, S to 10 pairs of SLS
layer were used, with each SLS layer thickness ranging from S
to 100 nm. After the TCA and SLS insertion process, a GaAs
layer was grown on the annealed GaAs/Si and SLS layer at a
temperature of Tg The total film thickness was also 3.S-4 /lm.
As a result of combination of TCA process and SLS
insertion, high quality GaAs films with a dislocation density of
1-2xl0 6cm- 2 have been realized.
Reduction in dislocation
density of GaAs-on-Si layers due to TCA and SLS insertion
have been explained [4] by coalescence of dislocations under
high temperature and high stress condition and dislocation
movement towards edge and interface by mismatch stress,
respectively.

normal growth, respectively, wis half the patterned width, x is the

978-1-4799-4398-2/14/$31.00 2014 IEEE

0824

a combination, high efficiency GaAs-on-Si solar cells have

been realized using high quality GaAs films with a dislocation
density of 3.7xI06cm- 2 .
Further improvements in conversion efficiency of GaAs-on-Si
solar cells must be attained because higher minority-carrier lifetime
of 1O.Sns that is due to low density of threading dislocations as low
as 8xlOScm-2 has been obtained using SiGe buffer layer by the Ohio
State Univ. [22]. 17.1% (AMO) efficiency (Voc=1.023 V,
Jsc=28.3mA/cm2, FF=80.S%) has been realized with GaAs single
junction solar cell with 0.04 cm 2 area on Si by using SiGe buffer
layer [22].
Nagoya Inst. Tech. group has demonstrated 21.2% (AMO)
efficiency (Voc=1.S7 V, Jsc=23.6 mAlcm2, FF=77.2 %) has been
realized with A0 1SGao8SAs/Si 2-junction solar cell by TCA [23].

n+/n+---AGIGaAsaAs
G
aAs
n+
n+/n-i-lnGGaAaAss IIl1

SL

SLS

Direct bonding technique of thin-film III-V compound solar cell

made by using epitaxial-lift-off technology and crystalline Si solar

Fig.8. A schematic cross section of a high-efficiency GaAs-on

cell is thought to be one of more realistic approaches in order to

Si solar cell.

overcome difficult hetero-epitaxy technologies. Fraunhofer ISE

By employing high quality GaAs films on Si, heteroface

GaAs thin films solar cells with an Alo8Gao.2As window layer
were fabricated on Si substrates. The thickness of Si substrates
was 200l1m. The n-GaAs base layer (carrier concentration of
Sxl01 7 cm-3) with a thickness of 1.3-l.Sl1m was grown on an
Alo 6Gao4As/GaAs super lattice (SL) layer (20nm1l00nm x
Spairs), In0 1 Ga0 9 As/GaAs SLS (IOnml10nm x S pairs), and
o
thermal cycle annealed GaAs buffer layer (TCA:900 C x S
times, 211m thick) which were grown on Si substrates. The
thickness and carrier concentration of p-GaAs emitter layer
8
were O.S 11m and 2xl01 cm-3, respectively. The total epitaxial
layer thickness was about 3.S-S 11m. A Si 3N 4 layer was used as
an anti-reflection film. Ti/Ag was used as electrode metal.
The cell area was I cm 2 .

mAlcm2, FF=82.6 %) under 48.3 suns with InGaP/GaAs/Si 3junction solar cells with 0.OS4 cm2 area by using direct bonding

40
c:t

...
c:

G)
...
...
::s

SUN,AMO

T=28C

Pmax

lOr-

0
0

technique [24].
V. SPACE FLIGHT DEMONSTATE OF GAAS-ON-SI SOLAR
CELLS
In order to apply GaAs-on-Si solar cells in space, they have
undergone a space flight test [2S]. Forty-eight 2 cm x 2 cm GaAs-on
Si solar cells with an average AMO total-area efficiency of 16/9%
were evaluated in the Engineering Test Satellite No. 6 (ETS-VI),
launched on 28th August 1994. Figure 10 shows AMO efficiency
distribution of the solar cells with SO and 100 11m cover glass. The
beginning-of-life (BOL) specific power for the GaAs-on-Si cells
(with SOl1m thick) studied was evaluated to be 68SW/kg and was
found to be higher than lS8W/kg for the GaAs-on-GaAs cells and
SS8W/kg for the SOl1m thick Si cells.

30 r20-

group has demonstrated 27.9% efficiency (Voc=3.33S V, Jsc=483

Isc

Voe

FF
7]'0'01
1) active
i

33.2mA
0.940V
0.791
18.3%
19.2%

10
...
Q)
.0

::J

0.5
Voltage (V)

high
AMO

Figure 9 shows current-voltage characteristics of a GaAs

on-Si solar cell, measured under an AMO illumination
condition (13S.3mW Icm 2 ).
The total-area conversion
efficiency of 18.3% at AMO and 20.0% at AMl.S were the
highest ever reported for GaAs-on-Si single-junction solar cells.
Such high-efficiency GaAs-on-Si cells have been attained by
combining TCA, SLS and SL buffer layers. As a result of such

978-1-4799-4398-2/14/$31.00 2014 IEEE

Average

AMO

Efficiency

16. 86%

4
2
0

, I
1.0

Fig. 9. Light-illuminated I-V characteristics of a

efficiency GaAs-on-Si solar cell measured under
illumination (13S.3mW/cm 2 ).

16.0

16.4

16.8

AMO Efficiency

17.2

(%)

17.6

Fig. 10. AMO efficiency distribution of 48 2cmx2cm GaAs-on-Si

cells with SOl1m and 100l1m cover glass.
Figure II shows remaining factor of maximum output power P
for the 200l1m thick GaAs-on-Si solar cells after 94 days from
launching in comparison with those for liquid-epitaxy grown 280l1m
thick GaAs-on-GaAs cells and SOl1m, 100l1m and 200l1m thick Si
cells. As a result of an apogee engine failure, the ETS-VI has been
placed in an elliptical orbit that passes through the Van Allen
radiation belts. Therefore, SOl1m thick Si space cells used for main
power source have been heavily damaged and various solar cells after
l
only 94 days are thought to be equivalently exposed to 2.8xlO s cm-2

max

0825

I-MeV electrons. The GaAs-on-Si solar cells have been

demonstrated to be more radiation-resistant in space than GaAs-on
GaAs cells and 50llm, 100llm and 200llm thick Si cells.

.s

0.9
0.8

0.7

0.6
0.5

E
Q)
0:: 0.4

GaAs on Si
(50um Cover

GaAs-on-Si
(1 OOum Cover

Cell
Glass)

e
1----I
--i -=---------------1
Q
t '""c:"'; "::'=:':
-'
-i ()U--5""'':"'"-":
---1
1----__

fruitful collaboration and discussion.

REFERENCES

Cell
Glass)
0

Dr. M. AI-Jassim and Dr. R. Ahrienkiel, NREL for their

----

2 OOum-thick Si Cell
...
I--------.-=---____;.;;;_='__l
( um Cover Glass)
.. 5
...

GaAs-on-GaAs

Bulk Cell

G '''=S)_--7100O''u=m-oCth' =ick''Cs:c-rn=ec'-'
ov=ec-"=
1-----"(5=O=um",C=
( 0um Cover Glass)
5

---1

--

--"---___--'---___---'
L-___-"---___
Various space solar cells evaluated using ETS-VI

for the
Fig. II. Remaining factor of maximum output power P
GaAs-on-Si solar cells after 94 days from launching in comparison
with those for liquid-phase epitaxy grown GaAs-on-GaAs cells and
50llm, 100llm and 200llm thick Si cells.
max

The end-of-life EOL specific power for the GaAs-on-Si cells

(with 50llm thick) studied was evaluated to be 564W/kg and was
found to be higher than 116W/kg for the GaAs-on-GaAs cells and
442W/kg for the 50llm thick Si cells.
These results show that the GaAs-on-Si solar cells have great
potential space applications. Further improvements in conversion
efficiency of GaAs-on-Si solar cells must be attained because
recently, higher minority-carrier lifetime of 10.5ns has been obtained
using SiGe buffer layer by the Ohio State Univ. [22].
VI. SUMMARY

III-V compound and Si tandem solar cells are expected to have

great potential of space and terrestrial application because of high
efficiency, light weight and low cost potential. We have
demonstrated 20% efficiency (under I-sun of AMI.5G) with hetero
epitaxially grown GaAs single-junction solar cells on Si. The first
space flight of forty-eight 2 cm x 2 cm GaAs-on-Si solar cells with
an average AMO total-area efficiency of 16/9% have been carried out
using the Engineering Test Satellite No. 6 (ETS-VI). The GaAs-on-Si
solar cells have been demonstrated to be more radiation-resistant on
space than GaAs-on-GaAs solar cells and 50, 100 and 200-m-thick Si
space solar cells. These results show that the III-VlSi-based multi
junction solar cells have great potential for space applications.
Fraunhofer ISE has demonstrated 27.9 % efficiency (under 48.3suns of AMI.5D) InGaP/GaAs/Si 3-junction solar cells by using
epitaxial lift-off and direct bonding.
Further improvement in efficiency for GaAs-on-Si single
junction solar cells and III-VlSi multi-junction solar cells can
be obtained by reducing dislocation densities to less than
s
2

3xlO cm- .

ACKNOWLEDGEMENT

The author would like to express sincere thanks to Prof. A.

Yamamoto, Dr. M. Sugo, Dr. M. Tachikawa, Dr. T. Nishioka,

[1] J.C.C. Fan, B.Y. Tsaur and RJ. Palm, Proceedings of the 16th
IEEE Photovoltaic Specialists Conference, (IEEE, New York,
1982) p.692.
[2] M. Yamaguchi and C. Amano, J. Appl. Phys., 58, 3601 (1985).
[3] M. Yamaguchi, A. Yamamoto, M. Tachikawa, Y. Itoh and M.
Sugo, Appl. Phys. Lett., 53, 2293 (1988).
[4] M. Yamaguchi, T. Nishioka and M. Sugo, Appl. Phys. Lett. 54,
24 (1989).
[5] S.K. Choi, M. Mihara and T. Ninomiy Jpn. J. Appl. Phys. 16,
737 (1977).
[6] M. Yamaguchi, M. Tachikawa, Y. Itoh, M. Sugo and S. Kondo, J.
Appl. Phys. 68, 4518 (1990).
[7] M. Yamaguchi, J. Mater. Res., 6. 376 (1991).
[8] P. Sheldom, K.M. Jones and M.M. AI-Jassim, J. Appl. Phys. 63,
5609 (1988).
[9] H. Kroemer, T.Y. Liu and P.M. Petroff, J. Crystal Growth 65, 96
(1989).
[10] M. Tachikawa and M. Yamaguchi, Appl. Phys. Lett. 56, 484
(1990).
[II] S. Onozawa, T. Ued M. akiyama and M. Sakut Annual
Meeting of the Jpn. Soc. Appl. Phys., Toyam 7p-Y-7, 1988.
[12] M. Yamaguchi, M. Tachikaw M. Sugo, S. Kondo and Y. Itoh,
Appl. Phys. Lett. 56, 27 (1990).
[13] E. Suhir, J. Appl. Mech. 53, 657 (1986).
[14] M. Tachikawa and H. Mori, Appl. Phys. Lett. 56, 2225 (1990).
[15] K. Nozawa and Y. Horikoshi, J. Electron. Mater. 21, 641 (1992).
[16] Y. Okada, H. Shimomura and M. Kawabe, J. Appl. Phys. 73,
7376 (1993)..
[17] Y. Itoh, M. Yamaguchi, T. Nishioka, and A. Yamamoto, J. Appl.
Phys. 61, 762 (1987).
[18] Y. Kadot M. Yamaguchi, and Y. Ohmachi, Proceedings of the
4th International Photovoltaic Science and Engineering
Conference, Sydney, (The Institute of Radio Electronics and
Engineers Australi Edgecliff), 1989, pp. 873-880.
[19] M. Akiyama, Y. Kawarada, and K. Kaminishi, Jpn. J. Appl.
Phys. 23, L843 (1984)..
[20] H. Okamoto, Y. Watanabe, Y. Kadot and Y. Ohmachi, Jpn. J.
Appl. Phys. 26, L1950 (1987).
[21] Y. Watanabe, Y. Kadot H. Okamoto, M. Seki, and Y.
Ohmachi, 1. Cryst. Growth 93, 1988, pp.459.
[22] SA Ringel, R.M. Sieg, J.A. Carlin, S.M. Ting, EA Fitzgerald,
M. Bulsara, and B.M. Keyes, Proceedings of the 2nd World
Conference and Exhibition on Photovoltaic Solar Energy
Conversion, Vienn (European Communities, Ispra, Italy, 1998),
pp.3594.
[23] T. Jimbo et aI., Science and Technology of Advanced Materials,
6, 27 (2005).
[24] K. Derendorf et aI., IEEE Journal of Photovoltaics, 3, 1423
(2013).
[25] M. Yamaguchi, Y. Ohmachi, T. Oh'har Y. Kadota, M.
Imaizumi and S. Matsuda, Progress in Photovoltaics 9, 191
(2001).

Dr. Y. Itoh, Dr. S. Kondo, Dr. N. Uchida, Dr. H. Mori, Dr. Y.

Kadota, Prof. Y. Ohmachi, Dr. T. Oh'hara, previous NTT
members, Dr. M. Imaizumi and Mr. S. Matsuda, JAXA, and

978-1-4799-4398-2/14/$31.00 2014 IEEE

0826

Digital Noise Spectroscopy System and Its Application to Solar Cells

2
1
l
l
3
Armin Mehrabian , Raymond Hoheisel , David J. Nagel , Scott R. Messenger and Sergey Maximenko
I
2

Abstract

The George Washington University, Washington DC 20052, USA

University of Maryland Baltimore County, Baltimore MD 21250, USA

3
Naval Research Laboratory, Washington DC 20375, USA

A creative combination of hardware and software

has produced a new digital noise spectroscopy system. It provides

a low cost, user-friendly, rapid, non-destructive and non-invasive
capability to acquire noise spectra. Thorough testing showed that
17 2
the system has a very low noise floor (3xl0- V /Hz), as well as a
5
broad frequency range (1 to 10 Hz).
The system has been
applied to survey solar cells with a variety of materials and
geometries.

In particular, we acquired noise spectra from (a)

single junction Si and GaAs solar cells and (b) triple-junction

solar cells that were heavily irradiated with 3 MeV protons. The
results should make it possible to determine characteristics of
solar cells, for example, trap densities, from the spectra.

Later

work will compare noise spectroscopy with normal means of

characterizing solar cells.

Index Terms

data analysis, digital data acquisition, GaAs

solar cells, mUlti-junction solar cells, noise spectra, silicon solar

cells, solar cell noise spectra interpretation.

I.

INTRODUCTION

Noise has a dark side, a bad reputation, since it interferes

with many types of communications, ranging from ordinary
verbal discussions to high speed wireless connectivity.
However, noise also has a bright side. The information
embedded within electronic noise can be used to learn
interesting and practical aspects of devices and systems. The
fundamental point is that the same quantities and mechanisms
that can influence the performance of diverse chips and
circuits are active in producing noise. For example, scattering
of electrons by impurities and traps determines both mobilities
in devices and the noise that appears within them. So,
measurements of voltage noise can sometimes be used to infer
information about the quality and reliability of electronic
devices and systems, including solar cells.
There are two basic types of voltage noise measurements,
either total power (V2/R) or the spectral (frequency)
distribution of noise power. The fIrst of these has signifIcant
value.
However, noise spectra contain much more
information, so their measurement can be quite valuable. A
basic question regarding noise spectroscopy of solar cells is
the value of the technique in comparison to commonly used
methods of characterizing solar cells. These include basic
Current-Voltage and Dark Current measurements, Capacitance
and Quantum EffIciency (QE) measurements, Deep-Level
Trap Spectroscopy (DLTS), Photo-, Electro-, Cathode
Luminescence measurements, Electron Beam Induced Current
(EBIC), lifetime spectroscopy, thermography and Hall Effect
measurements. This paper reports the fIrst steps toward

978-1-4799-4398-2/14/$31.00 2014 IEEE

answering the question on the relative value of noise

spectroscopy for characterization of solar cells [ 1,2,3,4].
Four steps are necessary to determine the value of noise
measurements of photovoltaic materials and devices. The fIrst
is the development of a modern system for such
measurements. The second is to use the system to measure
noise spectra from a wide variety of solar cells as a function of
illumination, load, bias, temperature, and radiation damage.
The third step is to extract information from the spectra that
can be used to learn details of the structure and behavior of
solar cells. Finally, it will be necessary to directly compare
what can be obtained from noise spectra with the results of
analyzing data from other characterization tools, once data on
the same devices are obtained by alternative methods. This
paper deals with the fIrst two of these steps. We report on the
conception, design, fabrication and testing of a facile double
channel system for obtaining quantitative noise spectra and its
initial use for measurements of solar cells.
The next section provides a brief review of earlier studies of
noise from various electronic devices, in particular solar cells.
The following section discusses the hardware, software and
testing of the new digital system for noise spectroscopy. The
attractive features of the new system are also summarized in
that section. Then, in Section IV, we exhibit spectra obtained
from solar cells with different materials and structures, both
single- and multi-junction cells. A sununary of what was
learned from these cells is in the concluding section. That
section also looks ahead to additional applications of the new
digital system to solar cells, both to learn more about what
information can be extracted from solar cell noise spectra and
to determine the advantages of noise spectroscopy relative to
other common techniques for characterizing solar cells.

II.

EARLIER WORK ON NOISE SPECTROSCOPY OF SOLAR CELLS

The use of noise power measurements for characterization

of electronic devices started over four decades ago.
A
thorough review of noise as a diagnostic tool for determining
the quality and reliability of electronic devices was published
in 1994 with 80 references [ 5]. Since then, there have been
two major trends. The fIrst is to measure more noise spectra,
initially by use of analog instrumentation [ 6]. The second and
more recent advance is the use of digital tools to obtain noise
spectra [ 7]. In addition, there have been increasingly better
studies on the extraction of information from noise spectra [ 8].

1884

III. DIGITAL NOISE SPECTROSCOPY SYSTEM

C. Testing

A. Hardware

The hardware for the new noise spectroscopy system

consists of four main subsystems. One is a shielded copper
enclosure containing the device being tested, a probe station
for contacting the device and a blocking capacitor to pass
frequencies of interest, but not DC voltages. The second major
component is a pair of Stanford Research SR560 Low Noise
Amplifiers (LNA). Then, another copper shielded box
contains the 200 kSample/second, 16 bit analog-to-digital
(ADC) converter 9223 from National Instruments in a
Compact DAQ chassis. Finally, a laptop computer is
employed to store the twenty million data points obtained in
each 100 second run, and to convert such time series into
spectra. Fig. 1 shows the relationship of these components.

Fig.! Schematic of the layout of the components of the digital

noise spectroscopy system.

Shielding enclosures around the

device-under-test (DUT) and the ADCs are not shown. The DUT
enclosure also contains the probe station.

The housing for the DUT is provided with a white LED and a
optical diffuser made of three-layers of tissue paper.
B. Software

MatLab was used for the bulk of the computations, and

LabVIEW was employed for some testing operations. Fast
Fourier Transform algorithms in MatLab gave single channel
spectra, or cross correlated spectra from the dual channels,
each spectrum containing ten million points in the 1 to
100 kHz frequency range. The raw time series could be
turned into quantitative noise spectra in less than two minutes.
Fig 2. is the flow diagram for the system software.

Fig. 2

Functions executed sequentially by the MatLab program.

The Power Spectral Densities (PSD) can be obtained either from a

single time series or the cross correlation of two such time series.
Averaging

and

fitting

are

optional

means

smoothed spectra prior to display.

978-1-4799-4398-2/14/$31.00 2014 IEEE

of

handling

the

The shielding for the DUT and for the ADC was adequately
effective. The performance of the two LNAs, compared over
a gain range from 1 to 10,000, was nearly identical. Spectra
were obtained using ADC windows of 1 and 10 V without
noticeable differences. Noise floor spectra were reproducible
to within a factor of 2. Spectra taken with our new digital
system compared well with spectra from a recently
recalibrated HP 3588 analog spectrum analyzer. The single
channel noise floor was competitively small at 3 x 10.17
V2/Hz. The dynamic range was measured to be over one
million. Sharp lines of unknown origin seen in some spectra
had widths of less than 0.04 Hz, so they are ignorable. The
hardware, software and results of testing the new digital
system will be described fully in another paper [9].
D. Features and Advantages

The measurements made with the new system showed it to

have several very desirable characteristics. The system is
relatively low cost. It is quite easy to set up. Sample insertion
is fast, as are acquisition, analysis and display of spectra. The
system is both non-destructive and non-invasive.
Being
digital, many signal processing options are available.
Quantization noise and ambient noise are not problems. A
second channel can be used to improve the noise floor by
almost a factor of ten, although that is not needed for most
work. Besides the low noise floor, the system has a dynamic
range of about a factor of one million. And, it spans a broad
frequency range from 1 Hz to 100 kHz. The system is
accurate in both frequency and amplitude. The overall high
performance enables applications to diverse physical,
chemical and biological systems, in addition to solar cells.
The system will find uses in several engineering disciplines, as
well as in medical research and clinical diagnostics.
IV.

SAMPLE SOLAR CELL NorSE SPECTRUM

We obtained spectra from several silicon solar cells and

some GaAs cells, each under four conditions: illuminated or
dark with either no load or a 100 Q resistor. The spectra were
significantly different for the different devices and conditions.
Sample spectra will be exhibited in the remainder of this
section, beginning with spectra that show the differences
between the single- and two-channel spectra.
Fig. 3 gives the noise spectra from a GaAs solar cell as
measured with one and both of the channels in our system.
The cell was illuminated and had no load when the data were
acquired. It is seen that the spectra are essentially identical
below about 20 kHz. There is a distinct region with a slope of
1If2 due to generation-recombination (G-R) noise, with the
usual lIf spectrum at frequencies less than about 1 kHz. The
other two spectra are the noise floor, again from single and
cross-correlated methods. They were obtained by shorting the

1885

system at the position normally occupied by the device being

tested. It is seen that cross correlation gives something less
than one order of magnitude improvement in the noise floor.
Hence, we do not routinely use cross-correlation. The rest of
the spectra presented in this paper are from single channel
time series.

Fig. 3.

Spectra from single-channel and two-channel cross

correlated methods from the noise floor (bottom two) and a GaAs
solar cell (top two). Noise on the noise spectra is unimportant.

A. Single Junction Cells

We measured single junction cells of Silicon and GaAs with

the four conditions, illuminated or dark with the cell terminals
either open or connected through a 100 n resistor. First, we
compare spectra from three types of single-junction silicon
cells, and then show how the spectra for one of those cells
vary with illumination and load. The three cells are (a) a
small array of two 50 by 20 mm cells from RadioShack, (b) a
single 22 by 35 mm cell with a surface absorber labelled
IXYS (SLMD480HI2L) from DigiKey, and (c) a single 13 by
35 mm cell labelled TRONY from the company EnOcean
(ECS3000).
The noise spectra from the trio of cells under conditions of
illumination and no load are shown in Fig. 4. It is seen that
the noise levels from the cells are as much as a factor of one
million higher than the noise floor of our system at low
frequencies. Cell-to-cell differences as high as two orders of

magnitude are seen at frequencies near 10 kHz. The spectra

from all three cells have general similar shapes, with close to
lIf behavior at the lowest frequencies and lIf slopes from that
region to the white noise floor. The regions in which G-R
noise is dominant vary with the type of cell. Those regions,
and the corner frequencies at which the G-R noise meets the
white noise, are given in Table 1. It is seen that both the G-R
region and the corner frequencies vary significantly for the
three types of Silicon cells.
TABLE I
PARAMETERS FROM SPECTRA OF SI CELLS
Cell

G-R Range (kHz)

Corner Freq. (kHz)

RadioShack

0.6-14

14

IXYS

2.4-80

80

TRONY

6-40

40

The next set of spectra show variations in the noise levels

for different conditions of illumination and load. Fig. 5
exhibits spectra from the RadioShack cells. The spectrum for
the conditions of illumination and no load is the same as that
in Fig. 4. It is seen that connecting the cell to a 100 n load
reduces the noise be as much as four orders of magnitude.
Turning off the light in the no-load condition produces a
similarly large reduction in noise. But, the spectrum for the
dark and no-load conditions has a region from about 10 to 200
The
Hz where there is little variation with frequency.
spectrum for the conditions of dark and the 100 n load is
within a factor of 2 of the noise floor for all frequencies. Lack
of illumination or use of the load eliminates the G-R region,
which is seen for the illuminated and no-load conditions.

101
I,o"
10"t

:::
I

Fig. 5.

10

100

f'Dqurenc:y(

It

.. ... .. ..
"

ICM!

loot

Spectra from the RadioShack Silicon cells under four

conditions of illumination and load.

Fig. 4. Noise spectra from three Silicon solar cells in comparison

to the system noise floor.

The feature near 10 Hz and the sharp

lines at higher frequencies are of unknown origin, but do not

interfere with extracting parameters from the noise spectra.

978-1-4799-4398-2/14/$31.00 2014 IEEE

The last of the single junction cells was a GaAs substrate

[ 10], again measured under the four conditions of illumination
and load. The spectra are in Fig. 6. They show that either
eliminating the light or introducing a 100 n load reduces the
level of the noise by four orders of magnitude at frequencies
below 1 kHz.
The conditions of dark and no-load again
produce a flat region in the spectrum, but at frequencies below
about 40 Hz, that is, lower than is the case for the Si cell dark
and no-load spectrum in Figure 5. For the GaAs cell, as for
the RadioShack Si cell, the G-R region disappears without
illumination and with the load.

1886

Fig. 6. Spectra from a GaAs solar cell under four conditions of

illumination and load. The G-R region is from 0.6 to 20 kHz.

B . Triple Junction Cells

The multi-junction devices used in this study are

monolithically grown GalnP/GaAs/Ge advanced triple
junction (ATJ) solar cells from Emcore Photovoltaics with a
device area of 2 cm x 2 cm. To assess the effects of radiation
on the characteristics of these cells, 3 MeV proton irradiations
were performed previously with two steps per decade up to a
total fluence of 1015 protons/cm2. Earlier work on this set of
cells has already been published, including IV and QE data
[ 1 1] and CL and EBIC data [ 12].
Noise spectra for one of the triple-junction devices, which
had been irradiated with 1015 protons/cm2, are given in Fig. 7.
The overall behavior is qualitatively similar to those already
shown for the Si and GaAs cells. However, there is no flat
(frequency independent) region for the conditions of dark and
no-load. The G-R region for this cell is from about 0. 1 to 8
kHz, significantly lower than the G-R regions for the Si and
GaAs cells.

range. Spectra can be obtained within a few minutes. We

recorded hundreds of spectra in the course of testing and using
the system.
The second goal was to measure noise spectra from a
variety of solar cells under various conditions. Many of the
results were exhibited in this paper. They show that the noise
spectra vary widely with the material and type of solar cell
and, especially, with the conditions of illumination and load.
Details of the noise spectra, including the range over which G
R noise was dominant, vary significantly for different solar
cells. It is expected that similar variations will be found with
cell bias, temperature and irradiation conditions.
The third phase of our program will be to wring additional
information out of the noise spectra and their parametric
variations. What can be said about, for example, the densities
of traps, and their variation with irradiation or annealing,
remains to be determined. The final phase of this program
will be to do head-to-head comparisons of information from
noise spectra with results from the more familiar means of
characterizing solar cells. Possibly, noise spectroscopy will
become a routine tool for characterization of solar cells.
REFERENCES
[I]

S. M. Sze. Physics of Semiconductor Devices. John Wiley and

Sons, 1981.

[3]

Fundamentals o/Solar
Cells: Photovoltaic Solar Energy Conversion, Ac. Press, 1983.
H. J. Hovel. Semiconductors and Semimetals. Volume II : Solar

[4]

S. Rein. "Lifetime Spectroscopy, A Method of Defect

[2]

A. L. Fahrenbruch and R. H. Bube.

Cells. Academic Press, 1976.

Characterization in Silicon for Photovoltaic Applications",
Springer, 2005.
[5]

L. K. J. Vandamme, "Noise as a diagnostic tool for quality and

reliability of electronic devices," Electron Devices,

IEEE Trans.

on EI. Devices, vo1.41, no.11, pp. 2176,2187, Nov 1994

[6]

Z. Chobola and A. Ibrahim, "Noise and scanning by local

illumination as reliability estimation for silicon solar cells."

Fluctuation and Noise Letters,

[7]

01 :01, L21-L26, 2001

O. Alink, "RF spectrum sensing in CMOS exploiting cross

correlation." University of Twente,

CTIT Ph.D. Thesis Series

No. 12-236 ISBN 978-90-365-3497-0, 2013

[8]

R. Macku and P. Koktavy, "Analysis of fluctuation processes in

forward-biased solar cells using noise spectroscopy", Physics
Status Solidi, A 207, pp. 2387-2394, 2010

[9]

A. Mehrabian, D. 1. Nagel and C. E. Korman, "Digital Noise

Spectroscopy System", to be published, 2014.

Fig. 7.

Spectra from a heavily irradiated triple-junction cell

under four conditions of illumination and load.

[10] S. R. Messenger, 1. H. Warner, R. Uribe, and R. 1. Walters,

"Monte Carlo Analysis of the NEO Beam Electron Beam
Facility for Space Solar Cell Radiation Qualification", IEEE
Trans. Nuclear Sci., 57, pp. 3740-6, 2010

V.

CONCLUSION

As noted in the introduction, this paper reports on two parts

of a four phase program to determine the utility of noise
spectroscopy for characterizing solar cells. We have achieved
both of the initial goals. The first was the development and
thorough testing of a digital system for measuring noise
spectra, which has a low noise floor and broad frequency

978-1-4799-4398-2/14/$31.00 2014 IEEE

[11] S.R. Messenger, R. Hoheisel, 1. Lorentzen, D. Scheiman,

J.H.Warner, M. Gonzalez, P.P. Jenkins, "High fluence
th
irradiations on triple junction solar cells", Proc. 39 IEEE
PVSC, pp. 3121-3127, 2013
[12] S. I. Maximenko, S. R. Messenger, R. Hoheisel, D. Scheiman,
M. Gonzalez, J. Lorentzen, P. P. Jenkins, and R. J. Walters,
"Characterization of high fluence irradiations on advanced triple
th
IEEE PVSC, pp. 2797-2800,

junction solar cells", Proc. 39

2013

1887

IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 40, NO. 6, DECEMBER 1993

1505

Results from the High Efficiency Solar Panel Experiment Flown on CRRES
K.P. Ray and E.G. Mullen
Phillips Laboratory, PL/GPSP, 29 Randolph Rd.
Hanscom AFB, MA 01731-3010

T.M. Trumble
Wright Laboratory, WLIPOOC-2
Wright-Patterson AFB, OH 45433-6563

solar cells, and simple static models cannot accurately predict

degradation.

Abstract
This paper presents results from the High Efficiency Solar
Panel Experiment (HESP) flown on the Combined Release and
Radiation Effects Satellite (CRRES). The on-orbit solar cell
degradation is correlated with the proton and electron environments. Comparisons between gallium arsenide germanium
(GaAs/Ge) and silicon (Si) solar cells are presented, and
results from three different annealing methods of like GaAs
solar cells are compared.

I. INTRODUCTION
Damage to solar cells in space can reduce overall spacecraft power which can affect operations and curtail mission
lifetimes. In recent years, the standard silicon solar cell has
been replaced by thinner junction silicon cells and different
materials like GaAs and InP. Silicon solar cell performance in
space has been well documented [l-21.Recently, GaAs and
InP solar cell results have been published [3-41.
Much of the analysis done on the space performance of
solar cells is based on the common practice of reducing the
fluences produced by the NASA models, both AP8 (protons)
and AE8 (electrons), to a 1 MeV equivalent electron fluence
and then looking up a damage versus fluence curve in a solar
cell radiation handbook [5]. However, when discrepancies
arise, there is no way of knowing whether the estimations of
the space environment are in error or if the solar cell is not
degrading as expected.
The data from CRRES combines HESP solar cell degradation data from advanced technologies like GaAs and GaAs/Ge
with proton and electron differential fluence data in units of
particles/cm2/sr. From this data set, we have been able to
compare the performance of GaAs and silicon solar cells as
well as correlate the degradation of maximum power to
changes occurring in the space radiation environment. Finally,
we have compared three different annealing methods.
We do not attempt to reduce the radiation environment to
a 1 MeV equivalent electron fluence and compare the results
with the NASA models. Instead, our goal is to demonstrate
that the very dynamic nature of both the proton and electron
radiation belts and the changes brought about by solar flares
and geomagnetic storms directly affect the performance of

11. SPACECRAFT
& INSTRUMENTATION
The CRRES mission lasted 1067 orbits from July 25, 1990
to October 12, 1991. The satellite had an highly elliptical orbit
(350 km X 35,000km)and an inclination of 18.2". The nine
hour and fifty minute orbit took the satellite through the heart
of the outer (electron) and inner (mostly proton) radiation
belts.
HESP was designed to study the degradation of both
gallium arsenide germanium (GaAs/Ge) and silicon (Si) solar
cells in the space radiation environment and to study the
effectiveness of different types of annealing methods in GaAs
solar cells. Five different solar cell parameters were measured:
maximum power, short-circuit current, open-circuit voltage,
and voltage and current at maximum power. Other parameters
studied included solar cell efficiency, GaAs/Ge cells versus Si
cells, coverglass thickness, coverglass material, and soldered
versus welded interconnects. Two different solar panels were
built: an Ambient panel and an Annealing panel.
The Ambient panel was made up of a total of 12 cell
strings, 10 GaAs/Ge strings with four cells per string and 2
silicon strings with five cells per string. All solar cells were 2
cm X 2 cm in area. The two Si solar cell strings, both built by
Spectrolab, were the K4-314 solar cell with a solar cell
efficiency of 12.3 % and the thin K7-314solar cell with a solar
cell efficiency of 14.8%. The K4-314cell string served as the
reference string for the entire experiment. The K4-314 N/P
cell has a 10 n-cm base resistivity, 0.15 pm junction depth, a
dual layer Ti0,A1203 AR coating, PdTiAg contacts, no back
surface field, a back reflector, and is 200 pm thick. The thin
K7-3/4N/P cell has a 10 Q-cm base resistivity, a 0.15 pm
junction depth, a dual layer Ti0,A1203 AR coating, PdTiAg
contacts, a back surface field, a back reflector, and is 62 pm
thick [6].
The ten GaAs/Ge PIN cell strings were built by Applied
Solar Energy Corp (ASEC) using a Metal Organic Chemical
Deposition (MOCVD) process. The cells had a GaAs buffer
region of 6-9 pm thick, a 0.47 pm junction depth, a 0.06 pm
AlGaAs window, an AuZnAg front contact, a AuGeNiAg rear
contact, a Ti0,Al,03 AR coating, and were 200 pm thick. The
efficiencies of the ten GaAs/Ge cells ranged from 17.5 to
19.2%. Table I summarizes the 12 cell strings in the Ambient

U.S.Government work not protected by U.S. copyright

1506

Table 11. Annealing Panel Solar Cell Strings

panel. Two types of materials were used as coverglass: Quartz

(fused silica) and ceria-doped microsheet (CMX). All covers
were bonded to the solar cells with a DC93500 silicone
adhesive [6].
Table I. Ambient Panel Solar Cell Strings
String # &
(Efficiency)

Coverglass &
Thickness (mils)

Cell Type

1 (18.0%)

Quartz 3.5

GaAslGe

2 (19.2%)

Quartz 6.5

GaAslGe

3 (18.5%)

Quartz 4.5

GaAslGe

4 (14.8 %)

Quartz 12

Thin Silicon

String # &

K7-314
5 (17.7%)

Quartz 6.5

GaAslGe

6 (17.5W )

Microsheet 4

GaAslGe

7 (18.0%)

Microsheet 12

GaAslGe

8 (18.2%)

Quartz 12

GaAslGe

9 (12.3%)

Microsheet 12

Si(Reference)

K4-314
10 (18.7 %)

Quartz 20

GaAslGe

11 (18.4%)

Quartz 20

GaAslGe

Cell Type

(Efficiency)

Coverglass &
Thickness (mils)

1 (16.1%)

Quartz 12

GaAs

2 (15.1 %)

Quartz 6

GaAs

3 (16.5 %)

Quartz 12

GaAs

4 (16.7%)

Quartz 6

GaAs

5 (15.4%)

Quartz 12

GaAs

6 (14.7%)

Quartz 6

GaAs

7 (16.4%)

Quartz 12

GaAs

8 (16.2%)

Microsheet 6

GaAs

logarithmically spaced from 1 to 100 MeV. The instrument

provided proton fluences in p+/cm2/sr [7]. Electron data was
provided by the High Energy Electron Fluxmeter (HEEF)
which measured electron fluence in e-/cm2/sr with energies in
the range of 1 to 8.3 MeV [8].

111. DATARESULTS

Our data analysis efforts consisted of two parts. The first

part comprised of quantifying the space radiation environment
The Annealing panel was made up of a total of eight GaAs and correlating changes in solar cell maximum power with
cell strings built by Hughes Research Laboratory (HRL). Each changes in the environment. This helped explain several
string was built using a liquid-phase epitaxy (LPE) process unusual results seen in the data. The second dealt with direct
and consisted of four cells, each with an area of 2 cm X 2 cm. comparisons between cell materials, coverglass materials and
The GaAs P/N cells had a 1 X lok8cm-3Te doped n-type thicknesses, and different biasing methods.
The majority of particulate radiation the solar cells were
substrate, a Sn doped buffer region 10 pm thick with a carrier
exposed
to came from the inner (mostly proton) and outer
concentration of 1 X 10" ~ m -a~0.5-0.6
,
pm junction depth,
(electron) radiation belts. Particles from solar flares were only
a 0.05 pm AlGaAs window doped with Be ( > 2 X 10" ~ m . ~ ) ,
an AuZnAg front contact, an AuGeNiAg back contact, a a minor contribution. As an illustrative example of the proton
T%O, AR coating, and were 300 pm thick. Cell numbers 1-7 environment, figure 1 shows three channels from the Proton
were covered with quartz coverglass and cell number 8 was Telescope, channel 9 (6.0 - 7.7 MeV with a 6.8 MeV center
covered with CMX coverglass. All coverglasses were bonded energy), channel 12 (9.9 11.5 MeV with a 10.7 MeV center
to the cells usifig DC93500 adhesive. The solar cell efficien- energy) and channel 16 (18.0 - 20.8 MeV with a 19.4 MeV
cies ranged from 14.7 to 16.7%. String numbers 1 and 2, center energy). The figure displays proton fluence in p+/cm2/constant heater strings, were heated continuously to 170" C sdorbit versus orbit number. The protons measured in
using an external heater. String numbers 3 and 4, pulsed Channel 9 (6.8 MeV) would have penetrated 12 mil quartz
heater strings, were heated to 250" C for 0.5 hour per week coverglass. The protons measured in channel 12 (10.7 MeV)
using an external heater. String numbers 5 and 6, forward bias would have penetrated all quartz coverglass except for the 30
strings, were heated to 190" C for 0.5 hour per week using a mils over Ambient String #12. Finally, channel 16 protons
forward biasing technique. The last two strings, numbers 7 (19.4 MeV) would have penetrated through all quartz coverand 8, were not heated and served as a reference for the glass on the experiment.
The period from the beginning of the mission to orbit 587
Annealing panel. Table I1 summarizes the 8 cell strings in the
was a relatively quiet period, devoid of solar flares or geoAnnealing panel [6].
Proton environmental data was provided by the Proton magnetic stonns. The proton fluence per orbit peaks around
Telescope (PROTEL) on CRRES. PROTEL measured the orbit 200 and steadily declines until orbit 587. The difference
differential energy spectrum of protons in 24 energy channels in proton fluence from orbit to orbit was a result of the satell12 (18.4%)

Quartz 30

GaAslGe

1507
1.OEt08

>b 1 .OE+07
c

E
2

MeV

8 l.OEt06
z,
3
M ~ V
1.OEt05

200

400

600

800

1000

1200

ORMT NUMBER

Figure 1. Proton fluenceper orbit versus orbit number for 6.8,10.7,

and 19.4 MeV protons.

results from simply integrating figure 1. The 10.7 and 19.4

MeV protons have been multiplied by a factor of 10 to
facilitate comparisons. The fluence data has been plotted on a
linear scale to clearly demonstrate the changes in overall
proton fluence. As the figure shows, the rate of increase in
total proton fluence begins to decline somewhat during orbits
400-587.The effect of the solar flare is clearly seen by the
sudden jump in the data at orbit 587. The total fluence begins
to increase rapidly from orbits 600 - 900 before beginning to
level off at the end of the mission.
The electron environment was extremely dynamic especially when compared to the proton environment. Figure 3 shows
channel numbers 1 (0.8 - 1.3 MeV, 1.1 MeV center energy)
and 10 (6.7 - 8.3 MeV, 7.5 MeV center energy) of the High
Energy Electron Fluxmeter. The figure shows electron fluence
in e-/cm2/sr/orbit versus orbit number. The absence of data
during orbits 1 - 49 and 383 - 410 was the result of dropouts
in the data.
1.OEt12

ite's 18.2" inclination which resulted in the satellite cutting

through the radiation belts on different tracks each orbit. On
22 March 1991 (orbit # 584) a large solar flare occurred at
22:47 Universal Time (UT). Following the flare, a solar
particle event was measured on CRRES beginning at about
7:30 UT on 23 March. While the solar particle event was in
progress, a sudden storm commencement occurred at 3:42 UT
(due to the interaction of a large shock with the magnetosphere), and created a second proton radiation belt [SI.The
creation of this new belt led to a significant increase in proton
fluence per orbit over the fluence preceding the event. The
proton fluence per orbit steadily increases up to orbit 900 and
then decreases as the second inner radiation belt begins to
decay. During orbits 600-900,there were also several other
particle events during orbits 765 to 793, 829 to 834,and 843
to 847.

ORBIT NUMBER

Figure 2. Total proton fluence versus orbit number for 6.8, 10.7, and
19.4 MeV protons. The 10.7 and 19.4 MeV proton fluences have
been multiplied by a factor of 10 to facilitate comparisons.

Figure 2 shows the total proton fluence in p+/cm2/srversus

orbit number for 6.8, 10.7, and 19.4 MeV protons. Figure 2

\
N

1.OEt08

>

1.0L+07

1.OEt06

1.OEt05

v
W
W

1.OEt04
1.OEt03

200

400

600
800
ORBIT NUMBER

1000

1200

Figure 3. Electron fluence per orbit versus orbit number for 1.1 and
7.5 MeV electrons.

The beginning of the mission to orbit 587 was also a quiet

period in the electron environment. The electron fluences per
orbit peak around orbit 200 and steadily decline until orbit
600. The sudden storm commencement at orbit 587 and
subsequent storms during orbits 600-900increased the electron
fluence per orbit several orders of magnitude. The 1.1 MeV
electron fluence peaked at orbit 600 at 1.85 X 10" e-/cm2/sr
compared to 1.86 X 108 e-/cm2/sr 16 orbits before (approximately one week).
Figure 4 is a similar plot of the total electron fluence in e/cm2/sr versus orbit number. Here the dynamic aspect of the
electron belts is clearly visible. The increase in electron
fluence from orbits 300 - 600 is relatively small. However,
after the geomagnetic storm at 600, there is a sharp increase
in fluence followed by an even larger increase during orbits
800 - 1000.
All the HESP data were reduced and corrected to 1 sun
AM0 at 28' C using temperature coefficients generated from
the same batch of solar cells tested at the Jet Propulsion
Laboratory. Operating temperatures for both the Annealing

1508
1 .OEt 14

O.OEtO0

1.OE+13
tl

1.OEt12

-1.OE+09

l.OE+l 1

5 l.OE+lO

0.08

-1.5E+09

0.075

-2.OE+09

1.OEt09

-I
3

1.OEt07

2
E
$

.,
0

1.OE+08

20.085
-

-5 .OEt08

(Y

3 0.09

I
.
.

1 0.07

-2.5Et09

9 0.065

-3.OE+O 9

;
2

1 .OEt06

0.06
1.OEt05

200

400

600

800

1000

1200

4
0

200

400

600
800
ORBIT NUMBER

1000

1;

-3.5Et09
IO

ORBIT hUMeER

Figure 4. Total electron fluence versus orbit number for 1 . 1 and 7.5
MeV electrons.

Figure 5 . Maximum power per cell versus orbit number for the thin
silicon K7-314 cell under 12 mils quartz coverglass and the GaAslGe
cell under 4 mils CMX coverglass. The total 10.7 MeV proton
fluence is inverted and plotted on the secondary y-axis.

and Ambient panel were provided by sensors on the two

panels. The solar cells on the Ambient panel operated between
85 and 90" C except during eclipses. In the annealing panel,
the pulse heater cells (unheated), the forward bias cells
3 0.09
-2.OEt12
(unheated), and the reference cells operated between 94 and
o
c
100" C except during eclipses.
5
- 0.085
-4.OE+12
0
The various solar cell parameters were plotted versus orbit
& 0.08
number to see if any irregularities appeared in the data. We
a
-6.OE+12
found that the degradation in a number of strings seemed to
s
0.075
have leveled off between approximately orbits 400-600. The
a
9
-8.OEt12 i;
-0
degradation in three different cell parameters, maximum
0.07
-x
I-0
power per cell, short circuit current, and open circuit voltage,
- 1 .OEt13
0.065
seemed to have stopped or even slightly recovered during this
.
period.
0.06 4
1-1.2Et13
0
200
400
600
800
1000
1200
We investigated the possibility that the measurement
ORBIT NUMBER
electronics could have become faulty during the period. No
evidence for erroneous behavior was uncovered. We believe
that the cell recovery was due in fact to the low particle Figure 6. Maximum power per cell versus orbit number for the thin
fluences observed during that period. Figure 5 shows the silicon K7-3/4 cell under 12 mils quartz coverglass and the GaAs/Ge
maximum power per cell for the thin silicon K7-314 cell with cell under 4 mils CMX coverglass. The total 1.1 MeV electron
fluence is inverted and plotted on the secondary y-axis.
12 mils quartz coverglass and for the GaAs/Ge cell with 4
mils CMX coverglass versus orbit number. As the figure
As further evidence that the leveling off of the solar cell
shows, the degradation in maximum power per cell levels off
during orbits 400-600. The total 10.7 MeV proton fluence is degradation was due to the space environment and not from
inverted and plotted on the secondary y-axis for comparison. the supporting electronics, we looked at the solar cell data
The rate df increase in the total proton fluence is less during from the Hipparcos satellite. The Hipparcos satellite was
orbits 400-600. At this time, it appears that the cells are launched 8 August 1989 with an inclination of 7" and a
annealing at the same rate as they are being damaged. After transfer orbit of 651 km X 35,835 km. Like the CRRES
orbit 587, the rate of proton fluence increases as well as the satellite, Hipparcos traveled through the radiation belts four
times per day. The satellite has both GaAs/Ge and silicon
degradation rate of the cells.
Figure 6 is a similar plot except that the total 1.1 MeV solar cells with CMX coverglass. Results from these solar
electron fluence is inverted and plotted on the secondary y- cells, both silicon and GaAs/Ge, show that the degradation in
axis. It appears that the electrons were not responsible for the cell parameters such as short-circuit current and open-circuit
damage. The electron fluence per orbit is relatively level from voltage did indeed level off during similar periods seen by
orbits 100-400 while the maximum power steadily declines. CRRES [lo].
Because of the large volume of data generated by the
Also, the degradation in solar cells after orbit 600 is relatively
consistent and does not correlate with the large increases in experiment, the data set had to be summarized, and only
maximum power per cell plots are shown. Figure 7 compares
electron fluence at orbit 600 and again at orbits 800-900.
L
t

1509

orbit. The shallow junction in the GaAs/Ge (0.5 micron) solar

cells is one of the main keys to its better performance in a
radiation environment. Cells with deeper junctions ;1 micron)
normally have a better BOL efficieucy; however, their
radiation performance is only slightly better than that of silicon
t61.
0.084
Figure 8 shows the same three cells compared cI7 a
normalized
plot. At the end of fourteen and a half months, the
Thin Si K7-3/4
0.07
GaAs/Ge cell was producing 87% of its BOL power, the
silicon reference cell 84% of BOL power, and the thin silicon
5 0.065
solar cell 78% of its BOL power. All spacecraft are designed
for a specific end-of-life (EOL) power after a fixed number of
years in orbit. Neglecting GaAs annealing, the higher BOL
0.05
200
400
600
BOO
1000
1200
efficiency of the GaAs/Ge solar cells coupled with lower
ORBIT NWBER
losses makes GaAs/Ge solar cells prime candidate to replace
silicon solar cells for high radiation orbits. The thin silicon
Figure 7.Maximum power per cell versus orbit number for the thin solar cell is usually selected for spinner spacecraft where the
silicon K7-3/4cell, the GaAs/Ge cell, and the Si K4-3/4(reference) colder cell operation (25' C to 45" C) is an advantage for
cell. All three cells were under 12 mils of coverglass.
silicon over GaAs/Ge. At high temperatures, 85" C or above,
the temperature coefficient for GaAs/Ge results in substantially
the maximum power per cell versus orbit number for the more power than can be provided by +hedegraded silicon solar
GaAs/Ge, the K4-3/4Si (reference) and the K7-314 (thin cell.
silicon) solar cell strings. It is important to note that all the
'
0
5
k
data presented are from cell strings (5 cells per string for
silicon, and 4 cells per string for GaAs/Ge and GaAs/GaAs).
This provides an average value of power per cell and improves the statistical baseline. All three cell strings were under
12 mils coverglass with the Si K4-3/4(reference) cell string
under microsheet and the GaAs/Ge and thin silicon K7-314cell
strings under quartz. The beginning-of-life (BOL)efficiencies
for the GaAs/Ge, Si reference and thin silicon solar cells were
18.2596, 12.3%, and 14.8%, respectively.
0.1

100

70

200

400

600

800

1000

1200

ORBIT NUMBER

Figure 9. Percentage of BOL maximum power per cell versus orbit

number for six GaAslGe solar cells under 3.5,4.5,6.5,12,20,and
30 mils quartz coverglass.

90

a5

75

200

400

600

800

1000

1200

ORBIT NUMBER

Figure 8. Percentage of BOL maximum power per cell versus orbit

number for the thin silicon K7-314cell, the GaAslGe cell, and the Si
K4-3/4(reference) cell. All three cells were under 12 mils coverglass.

The initial rapid drop in the thin silicon solar cell performance is due to the fact that it has a back surface field. The
back surface field is particularly susceptible to radiation
damage [ll] and usually becomes ineffective after a year in

Figure 9 is a normalized curve for output power of

GaAs/Ge solar cells with quartz coverglass thicknesses of 3.5,
4.5, 6.5, 12, 20, and 30 mils. The three thin coverglasses,
3.5,4.5 and 6.5 mils indicate a cell power loss of over 25%.
The three thicker coverglasses, 12, 20, and 30 mils show a
loss of less than 13%. The distribution of power loss is
reasonably predictable based upon laboratory studies of proton
damage protection of different quartz covergiass thicknesses to
the GaAs/Ge solar cells. The 12 mil coverglass thickness is
about optimum for maximum solar cell protection with
minimum weight for the orbit selected. The percentage of
lifetime gain derived from the use of the thicker 20 mils and
30 mils thick covers is negated by almost double and triple the
weight penalty when compared to the 12 mils coverglas?. For
example, the net cell improvement by using a 30 mil: cover-

1510

glass over the 12 mils coverglass is only 4 % .

102
E

100

98

b
L

96

94

50
170

Refsrence 100

82

200

400

600

800

1000

1200

200

400

600

800

1000

1200

ORBIT NUMBER

ORBIT NWBER

Figure 10. Percentage of BOL maximum power per cell versus orbit
number for four annealing GaAs solar cells under 6 mils coverglass.

Figure 10 displays the percentage of maximum power lost

by GaAs solar cells with 6 mils coverglass as a function of
time as shown by orbit number under different active annealing conditions. Figure 11 is a similar plot comparing 12 mils
coverglass GaAs cells. The first seven out of eight annealing
strings were covered with quartz, while cell string number
eight, the 6 mils reference cell, was covered with microsheet.
Before orbit 400, the pulse heater 6 mils cell string closely
tracks the reference cell string, indicating very little, if any
improvement over the reference cell string. During the
quiescent period there appears to be abrupt annealing in which
the pulse heater cell surpasses both the reference cell and
constant heater cell. Comparing this information with figure
11, the 12 mil coverglass pulse heater performance is noticeably better than both the constant heater and reference cell
strings for the entire flight.
The 6 mils pulse heater cell and reference cell strings had
solar efficiencies of 16.7 and 16.2% compared to 14.7 and
15.1% efficiencies for the forward bias and constant heater
cells. The efficiency advantage of the pulse heater and
reference cell is eliminated at approximately orbit 375, at
which time the pulse heater annealing begins to become more
effective. The 12 mils cells had a much narrower range of
efficiencies: 16.5, 16.1, 16.4, and 15.4 for the pulsed,
constant, reference, and forward biased cells. The 12 mils
pulsed and constant heater cells track each other closely until
orbit 375 where the pulsed heater also becomes more effective
at annealing out damage. It should also be mentioned that 6
mils quartz stops -4.5 MeV protons while 12 mils quartz
coverglass stops -6.5 MeV protons. Examination of the
measured proton spectra reveals that the fluence in p+/cm2/sr
drops approximately an order of magnitude in going from 4.5
MeV to 6.5 MeV resulting in a damage rate in the 6 mils cells
that was greater than the annealing rate until the quiescent
period.
Looking at figure 10, the forward biased string has
retained 82% of its BOL power compared to 79% BOL for

Figure 11. Percentage of BOL maximum power per cell versus orbit
number for four annealing GaAs solar cells under 12 mils Quartz
coverglass.

the pulse heater string, 77% BOL for the constant heater
string, and 75% BOL for the reference string. Figure 11
reveals the forward biased string has retained 88 % of its BOL
maximum power compared to 87%, 86%, and 84% BOL
maximum power for the pulse heater string, the constant
heater string, and the reference string respectively. Passing
current in the forward direction through the cells results in a
different voltage drop across each cell. Localized low resistance paths in a solar cell will result in very high current
densities at that point which will locally anneal out damage
sites. The forward bias and the pulse annealing experiment
were designed to operate at 190" C and 250" C, respectively,
for a half an hour each week. The pulse heater experiment
uses a heater under the cells to achieve annealing and thus is
not as effective in annealing out localized dislocations or
trapping centers.
A comparison between quartz and microsheet coverglass
shielding revealed essentially no difference between the two
materials in terms of degradation rates. Figure 12 shows the
percentage of BOL maximum power per cell versus orbit
number for four GaAs/Ge cells behind 4.5 mils (string #3) and
12 mils (string #8) of quartz and 4 mils (string #6) and 12
mils (string #7) of microsheet. The cells under 12 mils
essentially tracked one another with the CMX ending with
87 % of its BOL maximum power versus 85 % BOL maximum
power for the quartz. The cell under 4 mils CMX finished
with 77% of BOL maximum power versus 75% of BOL
maximum power for the 4.5 mils of quartz.
We also saw no difference between soldered and welded
electrical interconnects. During each eclipse, the temperature
in the ambient panel dropped approximately 140" C (90" C to
-50" C). There was no indication that the thermal stress
created during each eclipse over the 1067 orbit mission
degraded either the welded or soldered electrical interconnections.

1511
We found essentially no difference in the degradation of
maximum power per cell for GaAslGe solar cells under
similar thicknesses of quartz and microsheet coverglass. There
were also no differences observed between soldered and
welded interconnects.

pm 0.95
6

a 0.9

s
5
s

V. ACKNOWLEDGEMENT

0.85

0.8

We would like to thank Joe Powe of Hughes Aircraft Company for

writing the software used to analyze the solar cell data.

0.75

VI. REFERENCES

0.7

200

400

600

800

1000

1200

ORBIT NUMBER

Figure 12. Percentage of BOL maximum power per cell versus orbit
number for four GaAs/Ge solar cells under 4.5 and 12 mils quartz
coverglass and 4 and 12 mils microsheet coverglass.

IV. DISCUSSION
& CONCLUSIONS
We have shown that the space radiation environment is
very dynamic, and that extreme care must be used in predicting solar cell degradation using constant particle fluences per
day derived from static models. Solar cell degradation showed
good correlation with changes in measured proton fluences.
Electron fluence levels did not appear to effect solar cell
performance, even after the multitude of geomagnetic storms
during orbits 600 to 900 increased the 1.1 to 7.5 MeV
electron fluence per orbit by several orders of magnitude.
However, it should be noted that during an extended mission,
these high energy electron fluences could eventually degrade
performance.
GaAslGe solar cells performed better than two different
types of silicon solar cells under equivalent coverglass
shielding. The thin silicon K7-3/4 solar cell degraded much
faster in the beginning of the mission due to its back surface
field. However, after fourteen and a half months, the K7-3/4
still had a higher maximum power per cell than the silicon K43/4 reference solar cell.
A comparison of like GaAs/Ge solar cells under different
quartz coverglass thicknesses showed that increasing the
coverglass thickness from 6.5 mils to 12 mils resulted in a
12 % increase (75 % to 87 %) in percentage of BOL maximum
power at the end of the mission. Increasing the quartz thickness from 12 mils to 30 mils resulted in only a 4 % increase
in percentage of BOL maximum power at the end of the
mission at almost triple the weight penalty.
Results from the Annealing panel showed that forward bias
annealing outperformed both constant heating and pulsed
heating biasing techniques for cells under 6 and 12 mils
coverglass. From an economic and practicality standpoint, the
forward biasing annealing method is much simpler and less
costly than methods using external heaters in providing high
annealing temperatures.

[l] Gelb, S.W. and Goldhammer, L.J., In-Orbit Performance of

Hughes HS 393 Solar Arrays, Proceedines of the 22nd Photovoltaic
Specialists Conference, 1429-1433, 1991.

[2] Gelb, S.W., Goldhammer, L.J., and Kerola, D.X., In-Orbit

Performance of Hughes HS 376 Solar Arrays, Proceedings of the
18th Photovoltaic Specialists Conference, 362-367, 1985.
[3] Takata, N., et al., Space Proven GaAs Solar Cells - Main Power
Generation for CS-3, Proceedings of the 21st PhotovoltaicSuecialists
Conference, 1219-1225, 1990.
[4] Waters, R.J., et al., Space Radiation Effects in InP Solar Cells,
IEEE Transactions on Nuclear Science, NS-38, 1153-1158, Dec,
1991.
[SI Anspaugh, B.E. and Downing, R.G., Radiation Effects in Si and
GaAs Solar Cells Using Isotropic and Normally Incident Radiation,
1 Sept. 1984, JPL Publication 84-61.
[6] Trumble, T.M., AFAPL-801 Flight ExperimentResults, WL-TR92-2070, June, 1993.

[A Violet, M., et al., The Proton Telescope (PROTEL) on the

CRRES Spacecraft, IEEE Transactions on Nuclear Science, NS-40,
242-245, April, 1993.
[8] Dichter, B.K., et al., High Energy Electron Fluxmeter,
Transactions on Nuclear Science, NS-40, 252-255, April, 1993.
[9] Mullen, E.G., et al., A Double-Peaked Inner Radiation Belt:
Cause and Effect as Seen on CRRES, IEEE Transactions on Nuclear
Science, NS-38, 1713-1718, Dec, 1991.
[lo] Crabb, R.L. and Robben, A.P., In-Flight Hipparcos Solar Array
Performance Degradation After Three and a Half Years, to be
published by ESAIESTEC.

[ll] Anspaugh, B.E. et al, Solar Cell Radiation Handbook, 1 Nov.

1982, JPL Publication 82-69, p. 1-34.

ULTRATHIN, DIRECTLY GLASSED, GaAs SOLAR CELLS

C Hardingham, C. Huggins, T Cross
EEV Ltd,
Chelmsford, Essex, England.
A Gray, K. Mullaney, C. Kitchen
Pilkington Space Technology
Bodelwyddan, Clywd, England.
the first 5 microns of material. Resultant cells would have
superior power output per unit mass compared with all other
cell technologies. Mass specific power values as high as 673
W/kg [ref 71 should be achievable. For space solar arrays
this could offer very significant power to mass, inertia and
cost benefits, particularly for satellites in Geostationary orbit.

ABSTRACT
Due to its direct bandgap and high optical absorption
coefficient, GaAs offers the potential for ultrathin (< 10
micron) solar cells. Such cells have a very high output power
per unit mass, and applications both as single junction cells
and as top cells to mechanically stacked multijunction cells.
This paper reports the development of ultrathin GaAs solar
cells, with results up to 20% AM0 1 sun, based on
conventional coverglass and interconnect techniques.

ULTRATHIN GaAs SOLAR CELLS

EEV have previously reported progress [ref 71 in developing
the technologies necessary for manufacturing a high efficiency
GaAs ultrathin cell (UTC). These technologies have been
used to manufacture devices.

Direct (glueless) bonding of coverglass to GaAs

materials and cells will also be reported. To achieve a direct
bond, an electric field is placed across the cell and coverglass,
whilst they are held in compression at elevated an
temperature. Progress towards developing a directly glassed,
ultrathin GaAs solar cell will be reported.

Bevice Structure
The epitaxy (Figure 1) is similar to that for conventional
GaAs or GaAs/Ge cells, but includes a rear AlGaAs etch stop
layer, under a bifacial-style structure. A highly doped
(> 1E18,n-type) rear GaAs contact layer is grown under a thin - rear window/carrier reflector of AIGaAs, for which the
Aluminum mole fraction is chosen to give strong reflection
and good window properties, combined with good growth. A
conventional GaAs cell structure is grown thereafter, with a
2-4 micron, moderately doped (1E17<n< 1E18) base under
a thinner (0.5-1 micron), higher doped (2E18<p< 1E19)
emitter, a thin (optimally a quarter-wave thickness), high
Aluminum mole-fraction, AIGaAs window and a GaAs
contact cap layer.

INTRODUCTION
It has long been known that, for power- or area-limited
applicationsin space, GaAs solar cells are very attractive, due
to their high beginning of life efficiencies and superior
radiation resistance and temperature coefficients. GaAs cell
technologies have been developed for many years, both LPE
[refs 1-41, and the more advanced MOVPE [ref 5). However,
the advantages of early GaAs cells were somewhat offset by
their higher unit cost, increased handling requirements(due to
size limitations and high fragility compared to conventional
silicon cells), and higher optical absorptance - and hence
operating temperatures. One way of addressing these issues
has been growth of GaAs on a foreign substrate such as Ge which is now a commercially competitive reality [ref 61.

P+c.P

p N G a U Window

n GaAs 8.-

A more fundamental approach to overcoming the limitations

of GaAs is to remove the substrate completely. This is
electrically feasible, due to the strong optical absorption in
GaAs, characteristic of a direct band-gap material, which
means that essentially all the carrier photogeneration occurs in

0-7803-1220-1/93 $3.00 0 1993 IEEE

Fig. 1. Schematic of epitaxial structure of an Ultra-Thin Cell.

1399

&ice

F-

.
SOLAR CELL PHOTOVOLTAIC CURRENT-VOLTAGE PLOT
A W , 1 sun. 25 'C
Cell E2011 3 EEV 8 ym UIw.lhln G U S

Figure 2 illustrates the basic device fabrication sequence.

Front contact is made to the device by standard methods,
using e-beam evaporated metal contact layers, thickened with
electroplated gold. The GaAs cap layer is selectively etched
between the metal grid, to expose the AlGaAs front window.
A quarter-waveantireflection coating is applied, which on the
earlier UTCs presented here has been a PECVD SIN SLAR
(Single Layer Antireflection)coating. For the direct glassing
trials, a better quarter-wave optical match is provided using
TiO, layer, as part of a multilayer structure AR along with
A1,0,.

For conventionally interconnected UTCs, an interconnect is

ultrasonically bonded to the front contact grid, and a
coverglass applied to the cell with space-grade coverglass
adhesive.

01

02

03

04

05

00

07

08

OQ

ID

MLTS

Fig.3. Photovoltaic 1V measurement of an Ultra-Thin Cell.

Using these techniques, along with standard space grade

coverglass adhesive, and standard EEV interconnects,
nominal 4cm sq. devices of up to 19.7% AM0 1 Sun
efficiency have been manufactured (Figure 3). To date, best
results have been obtained using continuousrear contacts, due
to the complexities of forming a grid with conventional
(highly flexible) interconnects attached.
DIRECT GLASSING OF GaAS SOLAR CELLS
Etpansion Matched Glass and Potentl'al for -A
Bonding
Fig.2. Fabrication schematic of an Ultra-Thin Cell.

Once the adhesive is cured, the cell is potted on a rigid

supporting superstrate, which further acts as to protect the
coverglass and interconnect during subsequent etching. The
cell is etched using a fast but highly uniform (see ref 7) nonselective etch, followed by a slower, highly selective etch, to
completely remove the GaAs or Ge substrate, and stop at the
AlGaAs etch-stop layer. The etch stop layer is then removed
using an etchant selective against the GaAs buffedcap, which
itself is lightly etched, to improve the subsequent contacting.
After etching, the cell is dewaxed, and a continuous rear
metal contact e-beam evaporated onto the rear bufferkap.
Sintering of this contact is severely restricted due to the very
low thermal stability of the adhesive layer. For bifacial
operation, a grid pattern may be photolithographicallydefined
in the rear contact, and the rear cap etched between the grid.
Approximate alignment of the rear grid to the front grid IS
straight-forward, without resorting to complex infrared
imaging, by means of the strain fields induced by the front
contact grid in the thin semiconductor material.

Adhesiveless bonding requires that the coverglass and solar

cell have similar thermal expansion coefficients, both to
minimise the stress induced in the assembly by the bonding
process, and to reduce the cyclical stresses induced by orbital
temperature variations.
I
I
30
~~

10

(1

-5
-200

-100

100

200

300

400

Temperature IC

Fig.5. Thermal Expansion of CMG, compared with GaAs.

Following the successful development of a glass (CMZ) with

good expansion match for Si, a glass (CMG) has been
developed, which is suitable for space use and expansion
matched to GaAs [ref lo]. It is a cerium daped borosilicate
glass incorporating a UV absorbing filter. (The UV-blocking
properties of CMZ were retained such that the glass could
also be of benefit to conventional adhesively bonded solar cell
assemblies.) It has thermal expansion closely matched to
GaAs over a wide temperature range (see Figure 3, and is
suitzble for electrostatic bonding (Direct Glassing).
Cell M

Silicon Nitride (SiNJ, and Titania (Ti03 are two prime

candidates, neither of which have a high enough refractive
index to make an optimum AR coating, but both of which are
widely used in commercial devices. From previous work (ref
8), it is known that bonding directly to Tiox is possible, but
success depends on the stoichiometry of the film and
particularly its oxygen content.
However, its optical
properties show similar dependence. To partially deconvolve
the direct glassing optimisation from the ARC optimisation,
a TiO, layer with optimum antireflective properties has been
used as part of a two layer system, along with a directglassing compatible bonding layer.

A good candidate for the bonding layer is Alumina (A1203),

In order to successfully achieve an adhesive-free bond

between the cell and coverglass, electro-static glass bonding
has been carried out. The technique is wellestablished in the
manufacture of glasdmetal seals, in thermionic valves, lightbulbs, etc. In this technique, a substrate material is placed in
intimate contact with the glass at moderate temperatures, and
a voltage imposed across the interface, which results in a
strong bond between the two materials.

which is known to form good bonds with TiO,. Alumina has

a refractive index (1.61) close to CMG, so has little effect
opticall , and so samples have been bonded with thin
(< IOOi) Al,O, layers over the TiO,, as have samples with
the conventional quarter-wave thickness. These layers have
proved stable to the direct-glassingprocess, at temperatures up
to 4oooC, with no special deposition conditions necessary for
the TiO,. It has been found that the thinner Al,O, layer
facilitates lower bonding temperatures.

Despite its name, the technique is more accurately described

as electrolytic than as electrostatic, the probably mechanism

being as follows (seealso ref 9): At room temperature, glass

Alternatives to the A1,0, bonding layer, such as magnesia

(MgO), and single layer SiN., have been used, but these have
shown evidence of delamination from the TiO. (despite good
bonding to the glass), and optical degradation respectively.

is a good insulator. However, as it is heated towards its glass

transition temperature, cations within the glass become more
mobile. The glass then can act as an electrolyte, conducting
electricity through ionic (cation) flow.

Direct BondinP To GaAs Ce11s.

If glass, at such temperatures, is placed in intimate contact

with a surface and a sufficiently high (negative) bias applied
to the glass, the positive cations in the glass will be repelled
from the (positive) interface, and will diffuse towards the
opposite (negative) glass face. This will continue, until the
electrostatic field created by the displacement of the positive
ions balances the applied field. Beneficially, if local areas of
the sample are not in intimate contact with the glass, the field
will tend to draw the surfaces together. Initially, current
flows round the circuit, by the electrons associated with the
vacant cation sites being liberated by bonds formed between
oxygen atoms in the glass network and suitable substrate
atoms. This cannot continue indefinitely as the cation sites
are fmed in the glass network, so only those near the interface
can form bonds. This limits the charge transfer. (This
assumes that there are no mobile cations available from the
substrate - if there are any of suitable size, current will
continue to flow until the supply of these is exhausted.)

Bonding layer evaluation has been carried out using both LPE
and MOCVD grown GaAs solar cell material (ie AR coated
wafers with no metal contacts). No discernable difference in
bond characteristics has been observed between the two types
of structure. In order to bond to actual solar cells, grooves
have been provided within the coverglass, to accommodate the
collection grid pattern. Initial trials have used development
cells, and coverglasses (without any front surface ARC) with
recesses manufactured by diamond sawing; the resultant
coverglasses had wide grooves, with over I2 % of the cell area
in non-contact (compared to the 2-396 grid obscuration). First
results indicate that good bonding has been achieved, with
little degradation. Figure 6 shows an early result, and Table
1 shows the photovoltaic parameters before and after bonding
for this cell. The drop in current is attributable to: (i) the
relatively large (12%) non-contact area, which is very lossy
since these initial coverglasses have a rough, sawn finish, with
no light-bending; and (ii) non-optimised anti-reflective
coating. The initial measurement was made with a standard
SiN, coating which was stripped, before the DLAR applied
prior to bonding. The reflection plot shown in Figure 7
demonstrates that there is little optical change from bonding.

It follows from the above that the surface layer of the

substrate is of prime importance. Where the substrate is a
solar cell, this layer is part of the anti-reflective coating.
Due to the high refractive index of GaAs, the number of
candidate materials, which have suitable refractive index and
are transparent over the required spectrum, is very limited.

1401

__ -

0 Covsrglass has no front AR cwPng

0 AR cosbng not optlmised lor Direct Glassing

Glycerol natched Sample

a 2000

D i r e c t Glassed ( 3 2 0 C )

Table 1. Comparison of IV Parameters of a Directly

Glassed GaAs Solar Cell Before and After Bonding

Fig.7. Reflection trace from Directly-GlassedGaAs Solar Cell

material.

In order to use ultrathin, bifacial, GaAs solar cells in flight,

further development will be required on integration of cells
onto coupons and panels. This should be facilitated, using
interconnection techniques being developed along side the
direct glassing development.

Fig.6. Photovoltaic IV measurement of a Directly- Glassed

GaAs Solar Cell.

CONCLUSIONS.
High efficiency, ultrathin GaAs solar cells have been
demonstrated. Directly bonding of CMG coverglass to GaAs
solar cell materials, with good adhesion has been
demonstrated. Initial bonding trials of GaAs solar cells have
proved successful, and a first stage optimisation, to reduce
obscuration losses, is in progress. Directly glassed ultrathin
GaAs solar cells show promising potential for mass-critical
flights, and as components into mechanically stacked multijunction cells.

Interconnection

No special consideration need be given to the interconnection

technique in manufacture of standard thickness DC cells, such
as those reported above, since the cell itself provides some
level of rigidity and support for the interconnect. However,
it is clear that for commercial application of Ultrathin cells,
novel interconnection techniques will be required, since the
thin (< 10 micron) cell cannot give any support to the
interconnect, and would be under significant stress during
orbital thermal excursions. To this end, development work is
in progress to provide a metallic track in the coverglass,
making contact from the cell to this, and using it to take the
current from the cell. Figure 4 shows a schematic of the
arrangement.

ACKNOWLEDGEMENTS
This work has heen carried out with the support of the
procurement executive, Ministry of Defence.
The development of CMG was carried out by PST in
conjunction with Pilkington Group Research, under contract
to DRA Farnborough.

REFERENCES
1.
\ /

COVEAGUSS

2.
Fig 4. Interconnection schematic for an Ultra-Thin Cell.

I402

N. Mardesich et al, "Production and Characterization

of High Efficiency LPE AlGaAs Space Solar Cells.",
Proc 18th IEEE PVSC, Las Vegas, Oct. '85.
J Burrage et al, "Gallium Arsenide Solar Cell
Development.", Proc 5rh Europeun Symposium:
Photogenerurors in Spuce, Scheveningen, Sept. '86
p37.

3.
4.

5.

6.

7.
8.
9.

10.

L. Bertotti et al "Assessment and Improvement of the

GaAs Solar Cell Technology for a Production Pilot
Line. " ibid p43.
S Yoshida et al, "Liquid Phase Epitaxy Technology of
Large Area AlGaAslGaAs wafers of GaAs Solar Cells
for Space Applications.", Proc. 17th IEEE PVSC,
Kissimmee, '84 p42.
Y C M Yeh et al, "Large Scale OMCVD Growth of
GaAs Solar Cells", ibid p36.
G C Datum & S A Billets, "Gallium Arsenide Solar
Arrays - A Mature Technology." h o c 22nd IEEE
PVSC, Las Vegas, Oct. '91 p1422.
C Huggins et al, "Ultrathin GaAs Solar Cells Using
Germanium." ibid p 3 18.
P A White "Electrostatic bonding of Silicon Solar
Cells." Proc 2nd ESPC, Florence, Sept. '91 p575
M J Nowlan, S Tobin, & G Darkazelli, "Direct
Coverglass Bonding to GaAs and GaAs/Ge Solar
Cells." Proc 22nd IEEE PVSC, Las Vegas, Oct. '91
p1480
P White, "CMG - Coverglass for GaAs Cells", Proc
2nd ESPC, Florence, Sept. '91 p547

1403

STABILITY STUDIES OF COMPONENT CELLS FOR MULTIJUNCTION

AMORPHOUS SILICON ALLOY SOLAR CELLS
Xixiang Xu, Jeffrey C. Yang, and Subhendu Guha
United Solar Systems Corp.
1100 West Maple Road, Troy, MI 48084

ABSTRACT
We have studied the light-induced degradation and
saturation behavior of component cells for double- and triplejunction a-Si alloy solar cells. Top a-Si alloy cells reach
saturation after 300 hours of one-sun light soaking with 1012% degradation. Both middle and bottom a-SiGe alloy cells
show signs of saturation at around 600 hours under simulated
light illumination appropriate for these cells; the degradation
after lo00 hours of exposure is 16-18 %. Based on this study,
it is estimated that double-junction cells will degrade by 1218 %. and triple-junction cells will degrade by 10-18 %.
Experimental results also reveal that the degradation of top
cells, particularly in open-circuit voltage, is determined by
the bulk properties instead of the interface between p and i
layers. Further improvement of bulk intrinsic layer quality of
top cells can reduce the degradation of multijunction solar
cells.
INTRODUCTION
Amorphous silicon (a-Si) alloy solar cells have
received a great deal of attention for their potential as lowcost energy source. It is well known that prolonged lightexposure lowers the efficiency of these cells and extensive
studies have been carried out [l] to investigate light-induced
degradation in both a-Si alloy materials and solar cells.

obtaining high efficiency double- and triple-junction cells

[4]. The bottom cells in this study used a-SiGe alloys for the
intrinsic layers. For middle cells, both a-Si and a-SiGe alloys
were used for the intrinsic layers. The thicknesses of the
middle cells were adjusted to obtain short-circuit current
density similar to that produced by the corresponding cell in
the multijunction configuration. The Ge concentration in aSiGe middle and bottom cells was profiled to achieve the best
cell performance [5]. The top and middle cells were deposited
on textured substrates without any back reflector, since in the
multijunction configuration these cells do not see much
reflected light. In addition, one top cell was deposited on a
specular substrate for comparison purpose. The bottom cells
were deposited on A m 0 back reflector.
The characteristics of the component cells in the
initial state are listed in Table I. For both light soaking and
cell measurement, the top cells were exposed to white light,
while for both the middle and the bottom cells, different color
filters were used to adjust the light spectrum and obtain the
desired short-circuit current density. The middle cells were
illuminated with filtered light using 530 nm cut-on filter, and
the bottom cells for the double and triple junction devices
were illuminated with filtered light using 590 run and 610 nm
cut-on filters, respectively.
The temperature during
degradation was kept at 50 "C.
RESULTS AND DISCUSSION

An attractive way to reduce light-induced

degradation in solar cells is to use the multijunction
approach. In this case, the individual cell thicknesses are
small and hence the light-induced degradation is much lower
than that in a single-junction cell of comparable initial
efficiency [2]. We have recently reported that multijunction
a-Si alloy solar cells degrade by 12 to 16 % under prolonged
one-sun light-soaking [3]. In order to understand where the
degradation takes place with a view to designing more stable
multijunction a-Si alloy solar cells, we have studied the lightinduced degradation and saturation of component cells.
EXPERIMENTAL
The component cells were made using deposition
parameters that have been used in our laboratory for

0-7803-1220-1/93 $3.00 0 1993 IEEE

cells for

devica

Figures 1 (a)-(d) show normalized cell efficiency

(q), short-circuit current density (Jgc), open-circuit voltage

(Vw), and fill factor (FF) as a function of light soaking time

for both the top and bottom cells of a double-junction device.
As presented in these figures, the top cell reaches saturation
in degradation around 300 hours, while the bottom cell shows
signs of saturation around 600 hours. After 1000 hours of
light soaking, the degradation of the top cell is 12% and the
bottom cell 18%. We should mention that in a separate
experiment we have degraded the bottom cell with one-sun
illumination and have observed saturation beyond 800
hours.

97 1

For single junction a-Si alloy solar cells [6-81, the

largest degradation is in the fill factor. For a 2000 thick
top cell, J,, barely degrades, but the FF degrades by 7% and
V,, by 40 mV. In the case of the bottom cell, the fill factor
and short-circuit current density degrade by 12% and 5%,
respectively, and Voc degrades by 20 mV.

devices, the top cells degrade by about 11 %, the middle and

bottom cells degrade by about 17-18 %; in the doublejunction case, the top cells degrade by 12 % and the bottom
cells degrade by 18 %. Based on these data, one can estimate
that the double-junction cells will degrade between 12-18 %.
and the triple-junction cells will degrade by 11-18 %.
Since the top cells have the highest fill factor and
the least degradation in both the initial and degraded states, it
is desirable to design a multijunction stmcture to be top-cell
limited. Using this procedure, we have achieved an 11.2 %
stabilized active-area conversion efficiency for dual-bandgap
double-junction a-Si alloy solar cell [3]. The optimization of
triple-junction a-Si based solar cells is in progress.

devicu

Light-induced degradation data of the component

cells for triple configuration are presented in Fig. 2. As
expected, the top cell shows the least degradation, but it still
degrades by about 11% in cell efficiency after IO00 hours of
one-sun light soaking. It is interesting to note that both aSiGe middle and bottom cells show very similar degradation
behavior although the middle cell has less germanium content
than the bottom cell. Both cells show signs of saturation in
degradation around 600 hours, but did not fully saturate even
at 1000 hours. They also show a similar degradation of about
17-18% after lo00 hours of the simulated light soaking.

TODcells
It is surprising to observe that the top cells,
although very thin, degrade by 10-12 %, of which the
degradation in V,w is about 5%. Note that the V,, of thin top
cells on the specular and textured substrates in this work are
0.98 and 0.94 V, respectively. Obviously, reducing the
degradation of WO, of top cells can improve multijunction
solar cell stability.

The top cell deposited on a specular substrate shows

similar degradation and saturation behavior as the one on a
textured substrate (see Table II). However, the cell deposited
on the specular substrate has higher V, and FF. The V, and
FF are 0.98 V and 0.76 for the initial state and 0.93 V and
0.71 after IO00 hours of one-sun light soaking, respectively.
As pointed out earlier, the top cells for the triple-junction
configuration in this study show J,, between 6.4 to 7
mA/cm2. J,, for similar cells when deposited on standard
textured back reflector is 10-1 1 mA/cm2.

In order to investigate whether the degradation of

V,, of the top cells is determined by the interface defects or
bulk defects introduced by light soaking, the following two
experiments have been carried out. The light generation
profile in the intrinsic region adjacent to the p/i interface was
modified by inserting a "buffer" layer between p and i layers.
A 100 thick wide bandga (-1.85 eV) intrinsic a-Si layer
deposited at high rate (-20 $sec) was grown between p and i
interface to spread the generation profile away from the p/i
interface. In order to observe the degradation behavior within

From the degradation data of the component cells

shown in Figs. 1 and 2. we can see that in the triple-junction

Table I. Characteristics of the component cells in the initial state. The top cells are measured under global
AM1.5 illumination, and the middle cell with global AM1.5/530 nm cut-on filter. The bottom cells for
double- and triple-junction cells are measured with global AM1.5/590 nm and global AM13610 nm cut-on
filters, respectively.
Sample

Designation

Pmax
(mW/cm*)

Jsc
(mNcm2)

voc

FF

(VI

5344 (Si)

top/double

6.84

10.3

0.94

0.71

5342 (Si)

top/triple

4.39

6.50

0.94

0.72

5343 (SiGe)

middldtriple

3.51

7.02.

0.77

0.65

5351 (Si)

middle/mple

4.06

6.44

0.90

0.70

5346 (SiGe)

I;ottom/double

4.16

9.80

0.68

0.63

5345 (SiGe)

bottom/triple

3.65

8.59

0.68

0.63

972

1.1

-*a:---.---

*.:-.

0
h

0.9

.C(

a
.
:
:
*
.

mm

ia

.r(

0.7

. = - 0 - 5 3 4 4/top
=

0.5

- - 0 - 5 3 4 4 /top
= - * = 5 3 4 6 /bottom

5 34 6 /bottom

fi

.0.5
1o-2

1 oo

1 o2

1 o4

Light-soaking Time (hr)

1.1

10.7

-- 0-Q-

0.5

1 o-2

5344ltop
5 34 6 /bottom

0.5
1 oo

1 o2

1 o-2

1 o4

1 oo

1 o2

1 o4

Light-soaking Time (hr)

Light-soaking Time (hr)

Fig. 1. (a) Normalized efficiency , (b) short-circuit cwent density, (c) open-circuit voltage. and (d) fill factor of
the top and bottom cells for the double-junction configuration as a function of light-soaking time.

973

1.1

1.1

BN
?j

0.7

-*

--*-5342/top
--W=5343/middle
-I -5345/bottom

*-be 5 3 4 3 /middle

0.5
lo-*

-5345bottom
1

oo

o2

Light-soaking Time (hr)

0.5

1 o4

1 oo

1 o2

Light-soaking Time (hr)

1.1 7
' ,.dq-"l

a
.
g0.7
e
d

i?
0.5

:;: 5343/middle
53421top
- * -5345/bottom

--e5 34 2 4 0 p

5343/m idd Ie
--- --5345/bottom
H

1 o2

Light-soaking Time (hr)

Fig. 2. (a) Normalized efficiency , (b) short-circuit current density, (c) open-circuit voltage, and (d) fill factor of
the top. middle and bottom cells for the triple-junction configuration as a function of light-soaking time.

974

a short time, intense light-soaking was used to

accelerate the degradation of the cells. After a 5-minute. 30
sun light soaking, no appreciable difference was observed
between the cell with or without the buffer. Both show about
50 mV &gradation in V,.

8 suns. Degradation in V, of 34-39 mV was observed in both

cases. These experimental results indicate that the
degradation of a-Si alloy top cells is mainly determined by
the properties of bulk intrinsic layer rather than the thin
interface layer between p and i layers.

The second experiment is to provide different

generation profiles in the cells by changing the spectrum the
of light source. Two groups of devices selected from the same
sample were illuminated under blue light (with a 550 nm cutoff filter) and red light (with a 610 nm cut-on filter),
respectively, for 30 minutes at 34C. The light intensity in
both cases was adjusted so that J,, is equal to that obtained for

We have also investigated the degradation of V, of

top cells in which the intrinsic layer was deposited with
increased hydrogen dilution. The initial performance of the
cell on a textured substrate without back reflector is as
follows: V,, = 1.00 V, Jsc= 6.22 mA/cm*. FF = 0.76. After
30-sun. 5-minute, 48 T degradation, V,, went down by 40
mV. Studies on one-sun degradation are in progress.

Table 11. Characteristics of the top cells in both initial and saturated states. The top cells are measured under
global AM1.5 illumination.
Sample

5340 (Si)

5342 (Si)

SubstJate
specular

textured

State

Pmax
(mW/cm2)

Jsc

voc

(cm2)

(VI

FF

initial

4.92

6.67

0.98

0.76

saturated

4.37

6.62

0.93

0.7 1

degradauon(%) 11.2

0.8

5.1

6.6

initial

4.39

6.50

0.94

0.72

saturated

3.90

6.44

0.89

0.68

degradation(%) 11.2

0.9

5.3

5.6

Table 111. Characteristicsof a-Si and a-SiGe alloy middle cells in both initial and degraded stam. The middle
cells are degraded and measured under the illumination of global AM13530 nm cut-on filter.
Sample
5351 (Si)

5343 (SiGe)

State

Jsc

Pmax
(mW/cm2)

voc

FT

(mA/cm2)

cv)

4.02

6.44

0.90

0.70

3.29

6.22

0.87

0.61

degradation (%)

18.2

3.4

3.3

12.9

initial

3.51

7.02

0.77

0.65

2.89

6.85

0.74

0.57

degradation(%) 17.7

2.4

3.9

12.3

initial

975

ACKNOWLEDGEMENTS

Middle celh
We have also assessed the applicability of using aSi alloy for the intrinsic layer of the middle cell in the triplejunction configuration. The comparison between a-Si and aSiGe middle cells is listed in Table III for both initial and
degraded states. Although the a-Si middle cell reaches
saturation sooner than the a-SiGe cell, both cells show 1718% degradation after lo00 hours of light soaking. Another
difference between these two cells in the degraded state is that
a-Si cell has a higher overall efficiency of 3.29% than the aSiGe, 2.89%. However, the J,, of a-Si cell is only 6.22
mA/cmZ, which is much lower than that of the a-SiGe, 6.85
mA/cm2. Higher ,J for an a-Si middle cell requires a thicker
intrinsic layer which will cause more degradation.
From a stability study of another set of a-Si middle
cells with different thicknesses (3400 A - 4800 A), we found
that under no case could the middle cells produce more than 7
mA/cm2 of current density (under 530 nm cut-on filter) in the
degraded state after 1000 hours of one-sun light soaking.
Note, however, that this condition of light soaking is more
severe than that for the middle cells in the triple-junction
configuration, which may overestimate the degradation of aSi middle cells. A more systematic study of degradation of aSi cells with different thicknesses and preparation conditions
is required to further study this issue. We believe that 7
mA/cm2 of J, in the degraded state is necessary to obtain
stable efficiencies higher than 12% for triple-junction cells.

CONCLUSIONS
From a study of the stability of the component cells
of multijunction a-Si alloy solar cells. we show that it is
possible to reduce light-induced degradation to approach 1 1 %
by making the current top-cell limited. We also show that it
is necessary to incorporate germanium in the middle cell in
order to achieve a stabilized multijunction cell efficiency
above 12 %. Finally. we have demonstrated a stabilized
small-area efficiency of 11.2 % using a dual bandgap doublejunction structure.

The authors are grateful to E. Chen, G. Hammond,

M. Hopson, and N. Jackett for sample preparation and

measurements, and to K. Arao, A. Banerjee. T. Glatfelter and
K. Hoffman for interesting discussions. The work was
supported in part by the National Renewable Energy
Laboratory under Subcontract No. ZM-1-19033-2.

REFERENCES
[l]See, for example, "Amorphous Silicon Materials and Solar
Cells", AIP COP$ Proc. No. 234, edited by B. Stafford
(American Institute of Physics, New York, 1991).
[2]J. Yang. R. Ross, R. Mohr, and J. Foumier. "Status of
fluorinated amorphous silicon alloy multi-junction solar
cells", Materials Research Society Symposia Proc. 70, p.475
(1986).
[3] J. Yang and. S. Guha, "Double-junction amorphous
silicon-based solar cells with 11% stable efficiency". Appl.
Phys. Lett. 61, 2917(1992).
[4] S. Guha, Final Report, SERInP-211-3918,
the National
Renewable Energy Laboratory, Golden. Colorado, 1990.
[5]S . Guha, J. Yang. A. Pawlikiewicz, T. Glatfelter. R. Ross,
and S. Ovshinsky, "Band gap profiling for improving the
efficiency of amorphous silicon alloy solar cells", A p p l .
Phys. Lett. 54, 2330 (1989).
[6]A. Banerjee, S. Guha. A. Pawlikiewicz. D. Wolf, and J.
Yang, "Light-induced degradation in a-Si:H alloy solar cells
at intense illumination", AIP Conf. Proc. No. 234. 186
(1 99 1).
[7] L. Yang. L. Chen, and A. Catalano, "Intensity and
temperature dependence of photodegradation of amorphous
silicon solar cells under intense illumination", Appl. Phys.
Lett. 59, 840 (1991).
[8]S. Wagner, X. Xu, X. R. Li, D.S. Shen, M. Isomura. M.
Bennett, A. E. Delahoy. X. Li. J. K. Arch, J.-L. Nicque, and S.
J. Fonash, "Performance and modeling of amorphous silicon
solar cells soaked at high light intensity". Proc. of 22nd IEEE
Phofovolfuic Specialists Conference, p.1307 (1991).

976

A UNIFIED MODEL FOR RADIATION-RESISTANCE

OF ADVANCED SPACE SOLAR CELLS
Masafumi Yamaguchi, Shingo Katsumoto* and Chikara Amano'*
Toyota Tech. Inst., 2-12-1 Hisakata, Tempaku, Nagoya 468, Japan
*Univ. of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113, Japan
**NIT Opto-Electr. Labs., Atsugi, Kanagawa 243-01, Japan

ABSTRACT
1-MeV electron irradiation effects on MBE-grown
InGaAs and AlGaAs solar cells have been examined in
comparison with our previous results for radiation
damage of InP and GaAs solar cells in order to clarify
radiation -resistance of advanced space solar cells.
Moreover, 1-MeV electron irradiation results of several
space solar cells such as InP, InGaP, InGaAsP, GaAs,
AIGaAs, InGaAs, Si, Ge, and CulnSe2 cells have also
been analyzed by considering their damage constants,
bandgap energies and optical absorption coefficients.
We believe that this study will provide a unified model for
radiation-resistance of advanced space solar cells.
INTRODUCTION
One of the key points for developing space solar cells
is to improve their radiation-resistance. Previously, the
InP and CulnSe2 solar cells were found to be superior
radiation-resistant than Si and GaAs solar cells [1,2].
These results show that new radiation-resistant materials
are expected to be found and advanced space solar cells
with hig h-efficiency and radiation-resistance are
expected to be developed. Material-oriented studies
should be carried out in order to design superior radiationresistant solar cells.
In this study, 1-MeV irradiation effects on advanced
space solar cells have been examined and their radiationresistance has been analyzed by a unified model in which
their damage constants, bandgap energies and optical
absorption coefficients are considered.

0.72V and 0.76, respectively. 1-MeV electron irradiation

effects on MBE-grown AlGaAs and InGaAs solar cells
have been examined in comparison with our previous
results for radiation damage of InP and GaAs solar cells
[3] in order to clarify radiation-resistance of advanced
space solar cells.
In order to clarify properties of radiation-induced
defects in various solar cells, deep-level transient
spectroscopy (DLTS) and electron-beam induced current
(EBIC) measurements for solar cells have also been
carried out.
1-MeV ELECTRON IRRADIATION EFFECTS
OF AlGaAs AND InGaAs SOLAR CELLS

Figure 1 shows 1-MeV electron irradiation effects on

MBE-grown A10.35Ga0.65As and Ino.15Ga0.85As solar
cells in comparison with our previous results for radiation
damage of InP and GaAs solar cells [3].
It can be seen in Fig.1 that radiation-resistance of
AlGaAs solar cells is better than that of GaAs cells but
that of InGaAs cells is worse than that of GaAs cells.
1
x

0.8

c
L

0.6

InGaAs

EX PERIMENTAL
The samples used in this study were heteroface p-on-n
AlxGal -xAs (x= 0.2-0.35) and homo-junction p-on-n
In G a l - y A s (y = 0.15) solar cells grown by MBE
(dlecular Beam Epitaxy) on n-GaAs substrates. Their
junction depths were about 0.3pm.
Typical AM1.5
efficiency 7 , short-circuit current density Jsc, opencircuit voltage Voc and fill factor FF of A10.3Ga0.7As
solar cells were 14.6%, 16.2mA/cm2, 1.28V and 0.707,
respectively.
Typical 7 , Jsc, Voc and FF of
ln0.1gGa0.85As solar cells were 13%, 23.55mAkm2,

1014

id5

ld6

1-MeV electron fluence (cm2 )

Fig.1. 1-MeV electron irradiation effects of AlGaAs and
InGaAs solar cells in comparison with those of GaAs
and InP cells.
However, damage constants for minority-carrier
diffusion length, determined by the EBIC (Electron-Beam
Induced Current) method, of AlGaAs and InGaAs due to

2149
CH3365-4494/0000-2149 $4.00 0 1994 IEEE

First WCPEC; Dec. 5-9, 1994; Hawaii

1-MeV electron irradiation have been found to be the

almost same as that of GaAs as shown in Fig. 2. These
results suggest that differences of radiation-resistance
for GaAs-based cells must be affected by bandgap
energy effects on solar cell property degradation as
described below.

semiconductor solar cells is explained by their directtransition band structures, a unified model should be
proposed in order to clarify variety of radiation-resistance
for solar cells.
A UNIFIED MODEL O F RADlATlO
RESISTANCE OF SPACE CELL5

I?

Changes in solar cell properties due to 1-MeV electron

irradiations were calculated by our analytical method
previously reported [3] based on Hovel expressions [9].
In this study, especially, changes in solar cell properties
with irradiations were assumed to be dependent on
bandgap energy Eg, optical absorption coefficient (I
and minority-carrier diffusion length L.
And damage
constant KL for minoritycarrier diffusion length is given
by
A(1/L2) = 1 b 2 - 1/b2
=KL~,

'

!2
6
n 10

Id4

, , , ,,,, 1

, ,

where suffixes 0 and 0 show before and after irradiation,

and 0 is electron fluence.
In this study, n+-p junction solar cell structure has
been assumed for simplification. Changes in properties
of the solar cells fabricated with various materials due to
1-MeV electron irradiations were calculated using the
following theoretical equations [9]for the solar cell
properties and Eq. (1).
Photogenerated hole density Jp in the n+-layer is given
by

,,,,,I
1oI6

10

1-MeV electron fluence

(11

(cm-2 )

Fig.2. Changes in minority-carrier diffusion length

for AIGaAs, InGaAs, GaAs and InP with 1-MeV
electron irradiations as a function of electron
fluence.

- 1111x [ { ( ~ p ~ +d ~~ p- )~ x P ( -xj)
~p = [qF LP/{( 2 ~ p 2
((SpLp/Dp)COSh(Xj/Lp)+ sinh(xj/Lp))}/{(SpLdDp)
sinh(xj/Lp) + msh(xj/Lp)} - pLpexp(- xi)],
(2)
where suffixes p and n show carriers (hole and electron),
Xj is the junction depth, F the photon flux, (Y the
absorption coefficient, S the surface recombination
velocity and D the diffusion coefficient for the minority
carrier. P hotogenerated current density Jd, collected
from the depletion region, is
Jd = qF[1 - eXp(- W)]eXp(- Xj),

where W is depletion layer width.

Photogenerated
electron current density Jn, in the p-layer, is given by

a-Si (i-layer hckne

Jn = [qF Ln/(l + Ln)]exp(- Xj - W).

1d4

ldS
1-MeV electron fluence (cm-2 )

(3)

(4)

ld6 Total current density, delivered by the cell, is expressed

by the following equation and is dependent upon bandgap
and
energy Eg, optical absorption coefficient
minority-carrier diffusion length L as expressed in the
above equations.
(Y

Fig. 3. 1-MeV electron irradiation effects on output

power degradation of various solar cells [2-81.
Figure 3 demonstrates 1-MeV electron irradiation
effects on output power degradation of various solar cells
[2-81. CulnSe2 (CIS) and InP solar cells in particular
have much better radiation-resistance than Si cells.
Although superior radiation-resistance of compound

2150

Jx = f(Eg,

, L) Jp + Jd + Jn.

(5)

The open-circuit voltage V, is mainly dependent upon

bandgap energy Eg, and is related to the diode saturation

current density Jo that is dependent upon minority-carrier

diffusion length L as expressed by

where ri is the diode ideality factor and q is the electronic

charge.
Maximum current lm at the maximum power point can be
determined as a solution for the following equation

where Rs is total series resistance for the solar cell.

Power conversion efficiency 7 at the maximum power
point is

and fill factor FF is

Optical absorption coefficient effects

Moreover, radiation-resistance of several space solar
cells such as InP, GaAs, InGaAs, GaSb, Si, Ge, and
C u l n S e 2 cells studied by the authors and other
researchers [2-81 has also been analyzed by considering
their damage constants, bandgap energies and optical
absorption coefficients. Figure 5 compares experimental
results for changes in conversion efficiencies of various
cells after I O1 cm-2 irradiation of 1-MeV electrons and
analytical results for radiation degradation of them
calculated by considering their bandgap energies Eg,
damage constants KL, and optical absorption
coefficients a . Good correlations between them imply
the adequacy of analytical model and pertinent
assumptions. Superior radiation-resistance of CulnSe2
cells [2] is found to be explained by higher optical
absorption coefficientofCulnSe2 compared to other
materials.

where Pin is input light power.

1

Bandgap energy e f f e c t s

0.8

Figure 4 compares experimental results for changes in

conversion efficiencies of AIGaAs, InGaAs and GaAs
solar cells after I OI5 cmw2irradiation of 1-MeV electrons
and analytical results for solar cell property degradation
of them calculated by considering bandgap energy Eg
and damage constant KL. Good agreements between
them were obtained. Differences of radiation-resistance
for AIGaAs, GaAs and InGaAs cells are found to be
explained by considering bandgap energy effects on
solar cell property degradation.

0.6

0.4

1-MeV electron
0 = 1015c-2
Xj = 03ym

Calculated

0.2

1000

'

'

'

"

'"I

10000

'

'

'

'

".'I
100000

'

'

' ' "'

1000000

Optical absorption coefficient a (cm-I )

Fig. 5. Analytical results calculated by considering

1their bandgap energies Eg, damage constants KL,

and optical absorption coefficients a on radiation

damage for various space solar cells after IOl5 cm-2
irradiation of 1-MeV electrons in comparison with
experimental results.

Calculated
I

Defect i n t r o d u c t i o n rate effects

Better radiation-resistance of InP-based solar cells
such as quaternary compound InGaAsP and ternary
compound InGaP cells has also been found to be
explained by lower damage constants KL of InP-based
materials.
Figure 6 illustrates radiation-resistance of solar cells
made with InP-related materials such as InP [3], InGaP
[ l o ] and InGaAsP [I I], measured using I-MeV electron
irradiations, in comparison with that of GaAs cells. It can
be seen in this figure that solar cells fabricated with InPrelated materials show superior radiation-resistance
compared to GaAs as composition of InP increases.

Bandgap energy (eV)

Fig. 4. Analytical results calculated by considering
bandgap energy Eg and damage constant KL effects
on radiation damage for AIGaAs, InGaAs and GaAs
solar cells after 10'5 cm-* irradiation of 1-MeV
electrons in comparison with experimental results.

2151

SUMMARY
1-MeV electron irradiation effects on MBE-grown
InGaAs and AlGaAs solar cells were examined in
comparison with our previous results for radiation
damage of InP and GaAs solar cells in order to clarify
radiation-resistance of advanced space solar cells.
Moreover, 1-MeV electron irradiation results of several
space solar cells such as InP, InGaP, InGaAsP, GaAs,
AIGaAs, InGaAs, Si, Ge, and CulnSe2 cells were also
analyzed by considering their damage constants,
bandgap energies and optical absorption coefficients. It
is found in this study that 1) Differences of radiationresistance of GaAs-based solar cells are explained by
considering bandgap energy effect on solar cell property
degradation; 2) Superior radiation-resistance of CulnSe2
cells is understood by higher optical absorption
coefficient of CulnSe2; 3) Better radiation-resistance of
InP-based solar cells is explained by lower damage
constants of InP-related materials.
We believe that this study will provide a unified model
for radiation-resistance of advanced space solar cells.

InP

0.8

1
0.4

'

10'~

10 IS

1-MeV electron thence

1Ol6
)

Fig.6. Changes in conversion efficiency for various

solar cells made with InP, InGaP and InGaAsP,
measured using 1-MeV electron irradiations,
in comparison with that of GaAs solar cells.

REFERENCES

Figure 7 shows changes in damage constant KL for

minority-carrier diffusion length, measured by the EBlC
method, for InP-related materials and GaAs, determined
by 1-MeV electron irradiations, as a function of InP
composition. The better radiation-resistance of InPrelated materials such as InGaP and InGaAsP and their
solar cells shown in Figs. 6 is explained by lower damage
constant KL of InP-related materials compared to that of
GaAs. Moreover, the radiation-resistance Qf solar cells
fabricated with InP-related materials is found to be further
improved as composition of InP increases.

x 10-8
100

0.2

0.6
InP Composition
0.4

0.8

[l] M. Yamaguchi, C. Uemura and A. Yamamoto, J. Appl.

Phys. 55, 1429 (1984).
[2] B.J. Stanbery , J.E. Avery, R.M. Burgess,
W.S. Chen, W.E. Devaney, D.H. Doyle, R.A. Mickelsen,
R.W. McClelland, B.D. King, R.P. Gale and J.C.C. Fan,
Proc. 19th IEEE PVSC, (IEEE, New York, 1987) p. 280.
[3] M.Yamaguchi, C. Uemura, A. Yamamoto and
A. Shibukawa, Jpn. J. Appl. Phys. 23, 302 (1984).
[4] H.B. Curtis, C.K. Swartz and R.E. Hart, Jr., Proc.
19th IEEE PVSC, (IEEE, New York, 1987) p. 727.
[5] R.Y. Loo, G.S. Kamath and S.S. Li, IEEETrans.
Electron Devices, ED-37, 485 (1990).
[6] R.D. Krut, B.T. Cavicchi and D.R. Lillington, Proc.
22nd IEEE PVSC, (IEEE, New York, 1991) p. 90.
[7] J.R. Woodyard and G.A. Landis, Solar Cells 31, 297
(1991).
[8] P.E. Gruenbaum, J.E. Avery and L.M. Fraas, Proc.
22nd IEEE PVSC, (IEEE, New York, 1991) p. 1440.
[9] H.J. Hovel, Semiconductors and Semimetals V01.2,
(Academic Press, New York, 1966) p.49.
[ l o ] P.K. Chiang, P.S. Vijayakumar and B.T. Cavicchi,
Proc. 23rd IEEE PVSC, (IEEE, New York, 1993) p. 659.
[ l 11 P.R.Sharps, M.L.Timmons, R.Venkatasubramanian,
R. Pickett, J.S. Hills, J. Hancock, J. Hutchby, P. Illes,
C.L. Chu, M. Wanlass and J.S. Ward, Proc. 23rd IEEE
PVSC, (IEEE, New York, 1993) p. 633.

Fig. 7. Changes in damage constant KL for minoritycarrier diffusion length for InP-related materials
and GaAs as a function of InP composition.

2152

THE SOLAR CELL EXPERIMENT OF THE FIRST BRAZILIAN SATELLITE:

500 DAYS OF DATA ANALYSIS
Paulo Nubile, Nelson Veissid and Antonio F. Beloto
Instituto Nacional de Pesquisas Espaciais
Caixa Postal 515, CEP 12201-001
S2o Jose dos Campos, ,720 Paulo, BRAZIL

ABSTRACT

The experiments conducted aboard satellites show

the importance of the study of joint effects of thermal cycles,
ionizing radiation damage, ultra-violet light and vacuum
degradation on solar cells [ 1,2].
The first satellite of the Brazilian Complete Space
Mission, launched on February 9, 1993, hosts the Solar Cell
Experiment (SCE). The objective of this experiment is to
qualify in mission and real time the space use silicon solar
cells developed and manufactured in Brazil. The SCE is a
small solar module composed by a single array of six silicon
solar cells connected in series. The solar cells have the
following characteristics: 2x2 cm2 of area, base resistivity of
10 ohm.cm and efficiency over 11% under AM0 condition.
This module was laid down in an aluminum mechanical
structure. The signal generated by the array is sent to a peak
detector circuit located inside the mechanical structure. Three
signals are available for the SCE behavior analysis: the peak
detector output, the temperature sensor output, located inside
the SCE mechanical structure under the solar array, and the
attitude sensor signal, which gives the sunlight incidence angle
with respect to the SCE plane. A computationalmethod, which
permits the recovering of the entire solar cell's current versus
voltage characteristics, was developed to the solar cells'
parameter extraction using these three signals. During the
first 500 days of the satellite operation, the SCE solar cells'
degradation is smaller than expected.

satellite is illuminated during 65 minutes, the rest of the time

being in the shadow cone of the earth.
The satellite mechanical structure supports nine
solar panels: one octagonal top panel and eight rectangular
lateral panels. The SCE is located at the middle of a
rectangular panel. This experiment has been designed with the
following constraints: no electrical energy was available to
supplji the electrical conditioning unit, the maximum allowed
weight was 300 grams, and only two data channels were
allocated to the experiment. Despite these severe conditions,
the SCE could be designed to qualify Brazilian solar cells in a
real mission.
In January, 1984, a similar experiment was launched
with a Chinese scientific satellite [2]. The Chinese experiment
measured the solar cell short circuit current. and the received
data were used to calibrate solar simulators. The SCE was
designed in such a way that all that the whole solar cells
current versus voltage characteristic could be recovered from
the SCE output signal and temperature channels.
The aim of this work is to discuss 500 days of
received data.

INTRODUCTION
The Brazilian Complete Space Mission satellite is
the first space engine designed and constructed in Brazil. The
first data collection satellite (SCD1) is a real time repeater of
environmental data collected by ground platforms. The SCDl
is in orbit since 9 February, 1993 in a circular orbit 750 Km
high, with a period of 100 minutes, at an inclination angle of
250 with respect to the equation plane. In this orbit the

SOLAR CELL EXPERIMENT DESCRIPTION

The SCE is composed by an array of six solar cells in
series, connected by home made silver electrical
interconnectors, 50 Fm thick, with a geometric pattern
designed to relieve thermal stress. The solar array is laid
down over the external surface of an aluminum mechanical
structure.
A lot of one hundred 4 00> 10 ohm.cm boron doped
silicon substrates, grown by float-zone, resulted in 80 devices.
The solar cells were manufactured by the following
procedure: boron difhsion to make the back surface field;
phosphorus diffusion to make the emitter with junction depth
of 0.3 pm; front contact deposition by sputtering of titanium,
palladium and silver layers; antireflective coating deposition
by sputtering of tantalum pentoxide and final cut of the
substrate, resulting in a solar cell with 2x2cm size.

2232

CH3365-4/94/0000-2232$4.00 0 1994 IEEE

First WCPEC; Dec. 5-9, 1994; Hawaii

The conditioning electrical circuit, shown in figure 1,

is accommodated inside the mechanical structure. A 30 !2 load
resistartce is used to polarize the string in near the maximum
power point in the k V characteristics.The conditioning circuit
consists of a peak detector, formed by the diode D1 and the
300 pF capacitance. All other components are used to preserve
the solar array and the telemetry input circuit.

Csl and Cs2 are the saturation current constants. Eq.

(1) and (2) depends on the energy gap, which also varies with
temperature [5], as shown by eq. (8).

E,[T]= E,[O]+XT'/(T+ p )
Figure 1. Electrical circuit of the solar cell
experiment.

THEORETICAL FUNDAMENTS
Two problems had to be solved to analyze the SCE
behavior: how to extract the current versus voltage
characteristics from the one orbit SCE output voltage versus
temperature curve and how to obtain all the electrical
parameters used to adjust the current versus voltage curve.
The characteristics current versus voltage of a solar
cell can be represented by the double exponential model [3,4],
see eq. (1).

where IL is the photo generated current, R, and % are

respectively the series and parallel resistances, A is the diode
perfection factor, Isl and Is2 are the saturation currents and
all other symbols follows the usual notation. All these
parameters vary with temperature [4], and these variations
had to be taken into account. Eq. ( 2 ) to (10) shows the
expressions used to calculate the main parameters as a
function of temperature.

I,,

= Cs,T3exp(--)

E,

kT

I,,, = C,,T"* e x p ( -Em, )

(2)
(3)

2233

(8)

where Eg[0]=1.17 eV, a = 0.473 meVKand = 636 K.

CE, CA, C b and C,, are temperature independent
proportionality constants and Tref =25 OC is the reference
temperature used to obtain and compare the solar cell
performance data.
Eq. (2) to (8) combined with eq. (1) are used to
calculate from the current versus voltage curve all the
electrical parameters necessary to follow the solar cells
behavior. The following step is to recover the current versus
voltage curve from the telemetry data.
Supposing that the six solar cells are identical, we
can use eq. (1) to obtain the expression of the ECS voltage
output. Eq. (9) gives the output voltage as a function of
temperature. In this expression, the parallel resistance effect is
neglected and A=2, for the second exponential term of eq. (1).

A numerical method has been developed to extract

the desired parameters. Utilizing the cosine law, the
temperature correction and the attenuation factor as suggested
by Heinmaki and Guekos [6], the current versus voltage is
obtained. The method used to fit the telemetry curve is the
Down-Hill technique [7]. After several iterative calculations,
the minimum parabolic that satisfies a curve fitting criterion is
achieved. The criterion chosen to obtain this minimum is the
chi-square minimization, using the uncertainty in the SCE
telemetry voltage of 20 m'v, since the ACDC converter uses 8
bits for a 5V full scale.

LABORATORY AND IN ORBIT EXPERIMENTAL

DATA

DISCUSSION

Before launching, the SCE was submitted to an

extensive test battery. Besides the tests which verifies if the
subsystem will support the launching and environmental
conditions, as temperature cycling, humidity, spin and
vibration tests, all components and devices of the satellite has
to tested to qualify them for space use. The test plan for solar
cells was based in then norms MIL-STD-883and ESA-PSSPSS-25.
We have devoted special attention to the SCE's
temperature behavior, since the experiment is exposed to
temperaturc variations from -30 OC, when the satellite leaves
the shadow cone, to +30 O C at the end of the illumination
period. We measured the IxV curves for 4 solar cells in
temperatures varying from 5 OC to 75 OC, with 10 OC steps.
The IxV curves were obtained in the dark and under
illumination, with a solar simulator calibrated for the AM0
condition (135.3 mW/cm2). The results of this test is found in
ref. [SI.
Solar cell's degradation was expected after few
month of SCE operation. Therefore, radiation damage tests
were performed in a set of four solar cells. They were
irradiated with 5x1014 e/cmz, the equivalent fluence for 62.5
years in orbit. The irradiations were performed at the Instituto
de Pesquisas Energtticas e Nucleares (IPEN-USP), using a
Dynametron accelerator.
The constants of eq. (2) to (7) were determined
using the current versus voltage measurement at different
temperatures for both irradiated and non irradiated solar cells.
Since the first orbit, the satellite is sending
telemetry data for the SCE behavior analysis. Figs. 2 and 3
show the telemetry data for the output voltage and
temperature obtained in September 20, 1994. The data are
sent every half second and received by earth stations located
in S o Jost dos Campos and Alcgntara. In this day, the sun
aspect angle was 6 8 O and the spin was 63.8 rpm.
The voltage drop, observed in fig. 2 , at t=0.5 hour,
when the satellite enters the shadow cone, is damped by the
peak detector capacitor. This effect and the voltage drop in
the Shottcky diode, D1, were considered to correct data for
parameters extraction.
The sun aspect angle is important for the generated
current correction. The Heinamaky and Guekos attenuation
factor used in this work is 0.92, considering an aspect angle
of 6 8 O . Small variations around this angle do not affect
significantly the attenuation factor.

Daily telemetry data as illustrated in fig. 2 and 3 are

fitted to Equations 1 to 9 to obtain single solar cells IxV
curves and their parameters. Fig. 4 to 6 shows the results of
these calculations for the open circuit voltage, short circuit
current and efficiency. Each point represents the parameter
monthly mean value, for the AM0 condition at T = 25'C.

a
m
m
TIME (niinute)

Figure 2. SCE output voltage versus time for one

orbit.

im

TIME (iiimute)

Figure 3 . Temperature versus time for one orbit.

Figure 4. Open circuit voltage as a function of SCE

time in orbit.

2234

Open circuit voltage, short circuit current and

efficiency behavior, as shown in fig. 4 to 6, show that there is
no evidence of solar cell radiation damage after 500 days in
orbit. A short circuit current degradation of 0.5% was
expected for the first year, according to conventional space
use solar cell data [9]. However, previous radiation damage
study for our solar cells showed an expected degradation of
0.2% in the first year of operation.

REFERENCES
[I] D.H. Walker and R. L. Statler, "A satellite experiment to
study the effects of space radiation on solar cell power
generation." Solar Cells, 23, 1988, p. 245-250.
[2] C Huang et al., "Solar cell calibration experiment on
Chinese scientific satellite." Eighteenth IEEE PVSC, 1985,
pp. 634-637.
[3] H.J. Hovel, Solar Cell in: R. K. Willardson and A.C. Beer,
Semiconductor and Semimetals. Academic Press Inc., New
York, 11, 1976.
[4] H.S. Rauschenbach, Solar cell array design handbook,
Van Nostrand Reinhold Company, New York, 1980.

[5] S. M. Sze, Physics of SemiconductorDevices. Wiley, New

York (1981).

Figure 5. Short circuit current as a function of SCE

time in orbit.
The results show that the processes adopted in the
solar cell and SCE fabrication proved to be effective, since
the telemetry signals remain practically unchanged since
launching.

[6] A. Heinamaki and G. Guekos, "Solar cell short circuit

current dependence on the angle of the incident radiation,"
Solar Cells, 1977,p. 75.

[7] W.H. Press et al., Numerical Recipes in Pascal,

Cambridge University Press (1992).

[8] N.Veissid and A.M. Andrade, "The I-V silicon solar cell
characteristic parameters temperature dependence: an
experimental study using the standar deviation method."
Tenth E.C. Photovoltaic Soalr EnergV Conference, 1991, pp.
43-47.
[9] H.Y. Tada et al. Solar Cell Radiation Handbook, Thrid
Edition, JPL Publication 82-69 (1982).

Figure 6. Efficiency as a function of SCE time in

orbit.
The results also show that the method adopted to
the I-V characteristic recovering in real time is satisfactory.

2235