Вы находитесь на странице: 1из 6

Determination of thermal annealing eect in intrinsic

a-Si:H lm
T ulay Serin
*
, Necmi Serin, Celik Tarimci, Basri

Unal
Department of Engineering Physics, Science Faculty of Ankara University, 06100 Tandogan, Ankara, Turkey
Received 16 October 1999; received in revised form 10 January 2000
Abstract
In this study, the bias annealing eect on the density of state (DOS) in amorphous silicon lm on single crystalline of
silicon was investigated by means of Au/a-Si:H/c-Si/Al heterostructure. In order to fabricate the structure, at rst,
a-Si:H lm was coated on a single crystalline silicon substrate by means of a dc magnetron sputtering technique and
then gold and aluminum metal contacts were applied on the top of a-Si:H and c-Si, respectively. The samples were
annealed in the annealing temperature range 110175C with the negative end of the applied bias voltage kept at the
gold side and the capacitancevoltage characteristics heterostructure was measured. DOS around midgap of the i-re-
gion was determined by drive-level capacitance technique and the eect of thermal annealing on the DOS was indi-
cated. 2000 Elsevier Science B.V. All rights reserved.
1. Introduction
Photodetectors are important in optical-ber
communication systems operated in the near-
infrared region. They convert the optical variations
into electrical variations, which are subsequently
amplied and further processed. For such appli-
cations a good photodetector should have the
properties of a low current, a high response speed
and power linearity. The pin photodiode or
photodetector is one of the most common one
because the depletion-region (the intrinsic region)
thickness can be adjusted for optimum quantum
eciency and frequency response. A compromise
between quantum eciency and response speed
can be reached if the light is incident from the side,
parallel to the junction, thus increasing the eec-
tive absorption depth and at the same time keeping
the carrier transit distance small. The metal/semi-
conductor/metal type photodetector usually has
better performance and has simpler processing
requirements. To avoid large reection and ab-
sorption losses when the diode is illuminated
through the metal contact, the metal lm must be
very thin (~100
#
A) and an antireection coating
must be used. In this way a large fraction of the
incident radiation will be absorbed near the sur-
face of the semiconductor.
The investigations on the new metal/semicon-
ductor/metal photodetectors were mainly based on
the hydrogenated amorphous silicon (a-Si:H)
semiconductor materials in addition to IIIV
compounds [14]. The ambient temperature
changes aect the electrical characteristics of
metal/a-Si:H/metal structure extremely due to the
thermal annealing processes and for this reason it
Journal of Non-Crystalline Solids 276 (2000) 163168
www.elsevier.com/locate/jnoncrysol
*
Corresponding author.
E-mail address: nserin@eros.campus.ankara.edu.tr (T. Serin).
0022-3093/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 3 0 9 3 ( 0 0 ) 0 0 2 6 5 - 9
does not have a long-range periodicity [4]. Besides,
since its response is essentially within the visible
region it is not suitable for short-haul optical ber
communication centered at the optical signal
wavelength (of 830 nm). However, a-Si:H has the
following advantages: high dark resistance, high
absorption coecient, short carrier life-time, low-
temperature deposition process, low-cost and easy
integration with Si-based integrated circuits which
is very important if one remembers that metal/
semiconductor/metal photodetectors made of
IIIV group are dicult to couple to Si-based
integrated circuits.
Recently, it was shown that the metal/c-Si/metal
structure (operated at 830 wavelength) were com-
patible with Si-based integrated circuit technology.
However, since conventional metal/c-Si/metal
photodetector have a large current, low respon-
sivity and long transient tail, the performance of
c-Si device is needed to improve. The investiga-
tions show that intrinsic amorphous silicon
(a-Si:H) overlayer employed in c-Si will overcome
to some of these diculties and it is found that
a-Si:H formed a good hetereojunction with c-Si
[5,6] and improves the performance of solar cell
and hetereojunction bipolar transistor.
As is stated by many authors [719], the exis-
tence of density of states (DOSs) in the i-region
diminishes the electrical eld, increases the re-
combination and limits the collection eciency.
The thermal annealing highly aects the distribu-
tion of DOS. The changes of DOS with tempera-
ture highly aect the capacitancevoltage (CV)
and currentvoltage (IV) characteristics of metal/
a-Si:H/metal structure. The capacitance charac-
teristics measured at low frequency give some in-
formation about DOS, and hence this method has
been used widely for a-Si:H Schottky diodes re-
sulting in a large number of papers. The drive-level
capacitance method developed recently is a good
tool to determine the changes in DOS generated by
means of thermal annealing. Unlike deep level
transient spectroscopy (DLTS) and related tech-
niques which require doped samples with high
conductivities, this method is readily applicable to
the study of undoped lms and may be used to
map out the DOS distribution g(E) in such mate-
rials over a limited range of the mobility gap.
In this work, it was planned to fabricate
Au/a-Si:H/c-Si/Al heterostructures and to anneal
them in the annealing range 110175C and to
examine the changes occurred in their electrical
characteristics and to investigate the eect of bias
thermal annealing on the DOS by drive-level ca-
pacitance technique [15].
2. Theory
The frequency, temperature and band bending
in the sample determine the ac junction capaci-
tance. Only those states whose thermal emission
rate r is faster than 1/x can respond during a cycle
of the ac applied voltage and contribute to the
measured capacitance. These states lie between the
contour level E
E
and the conduction band mobility
edge E
C
. This level is determined by setting r = x
to give
E
x
= E
E
(x) E
C
(x) = k
B
T ln(max)Y (1)
where k
B
is the Boltzmann constant, T the abso-
lute temperature, and m is the escape frequency,
(~ 10
13
rad/s). Above this contour, the lled
states can equilibrate with the conduction band
within the ac period; below this contour the lled
states are frozen out and are unresponsive. Fig. 1
also shows a characteristic length x
E
, where
E
E
(x
E
) = E
F
.
The region 0 6x 6x
E
functions as a dielectric
with capacitance per unit area of C
d
= eax
E
and e
is the dielectric constant. The region x b x
E
con-
tributes a space charge capacitance C = eq
E
aF
E
,
where F
E
and q
E
are the electric eld and charge
density, respectively at x
E
. These two regions
constitute two capacitors in series, nally the
junction capacitance is given by
C
j
=
eq
E
eF
E
X
E
q
E
X (2)
The drive-level capacitance method [15] which
combines the temperature, frequency and dc bias
dependence of the deep state response in one
technique with the ac test signal amplitude (dV ) as
an additional independent variable is experimen-
tally the most dicult method. The parameter
(dV ) was xed in other admittance methods and
164 T. Serin et al. / Journal of Non-Crystalline Solids 276 (2000) 163168
the transfer function may be assumed to be linear
for such small test signals.
In the case, where the amplitude of the alter-
nating voltage (dV ) applied on a junction is neg-
ligible compared to the dc bias and diusion
voltage then the junction capacitance is given by
Eq. (2). If dV is not negligible relative to diusion
voltage, it gives higher order corrections to the
junction capacitance. Specically
C = C
0
C
1
dV C
2
(dV )
2
Y (3)
where C
0
is given by Eq. (2). The rst-order coef-
cient is
C
1
=
dC
dV
=
eq
2
E
2(eF
E
X
E
q
E
)
3
X (4)
The drive-level integrated defect state density N
DL
is dened in terms of C
0
and C
1
as
N
DL
=
C
3
0
2qeC
1
=
q
E
q
X (5)
The charge density which can respond at a given
E
x
is
q
E
= q

E
x
E
F
g(E) dEX (6)
The coecients C
0
and C
1
are obtained by tting a
quadratic to C versus dV curve. The energy de-
pendence of N
DL
is determined by varying E
x
. The
slope of N
DL
versus E
x
directly gives the DOS
distribution g(E).
3. Experimental procedures
Hydrogenated intrinsic amorphous silicon
lms which have the following properties:
thickness 1.06 lm, dark electrical conductivity
r
d
= 7X9 10
13
X cm
1
, photoconductivity r
ph
=
7X5 10
7
X cm
1
, concentration 6X5 10
16
cm
3
,
Urbach energy E
rb
= 69 meV, C
H
= 14 were sep-
arately coated on single crystalline of silicon sub-
strate (c-Si), p-type, 10 X cm, in 1 0 0 orientation,
by means of dc magnetron sputtering technique. In
order to fabricate Au/a-Si:H/c-Si/Al heterostruc-
ture an aluminum metal contact and gold metal
contact were applied on top of c-Si surface and
a-Si:H lm, respectively under a vacuum of 110
7
Torr.
Subsequently, the samples were placed in an
oven at 110C for 30 min while maintaining a
0.5 V bias voltage (gold negative). After annealing
the samples were left to cool at room temperature
in the presence of bias.
The same annealing processes were repeated
at annealing temperatures of 130C, 150C and
175C for biases of 1.0, 1.5 and 2 V, respectively.
In this paper, the results for 2 V bias annealing are
presented since the largest changes have been
found for this case.
In the drive-level capacitance method, the ca-
pacitance was measured as dc voltage amplitude
V
A
was varied from )0.60 to )0.95 V (gold nega-
tive), and ac voltage amplitude dV was varied from
0.40 to 0.05 V at 200 Hz and at 23C . Thus the
maximum positive voltage applied to the device
(dV V
A
= V
max
) was maintained at )1.0 V for
all pairs of ac and dc biases. The capacitance
Fig. 1. Band diagram of Au/a-Si/c-Si/Al structure.
T. Serin et al. / Journal of Non-Crystalline Solids 276 (2000) 163168 165
measurements were carried out in the temperature
range 23110C. Drive-level capacitance mea-
surements were repeated after each annealing
temperature. The coecients C
0
and C
1
were ob-
tained as mentioned before by tting a quadratic
to C versus dV and N
DL
were calculated by means
of Eq. (5). Fig. 2 shows N
DL
versus E
x
for each
annealing temperature, N
DL
, increased with E
x
because progressively deeper states were probed.
The values of g
0
were determined from the slopes
of N
DL
versus E
x
plots for each annealing tem-
perature and the results are presented in Fig. 3.
4. Results
The midgap DOS found by drive-level capaci-
tance technique are shown in Figs. 2 and 3. It was
observed that the signicant changes in defect state
density (N
DL
) and midgap DOS were created by
means of bias thermal annealing in temperature
range of 110175C. The values of DOS around
the middle of i-region rst decreased with in-
creasing annealing temperature up to around
130C then increased with the increasing annealing
temperature as shown in Fig. 3.
5. Discussion
Field-induced changes so far have been report-
ed for doped lms with metastable increase of
shallow donor states and decrease of deep states
after reverse-bias annealing [1619]. These changes
are consistent with the auto-compensation mech-
anism proposed by Street et al. [17] to explain the
dependence of the doping-induced defect density
on the position of the Fermi level during lm
growth. Metastable solar cell improvements after
Fig. 2. The plot of charge density N
DL
versus emission energy
E
x
before annealing (r) and after annealing at 110C (j),
130C (N), 150C (M) and 175C (v).
Fig. 3. The plot of density of states versus emission energy E
x
before annealing (r) and after annealing at 110C (j), 130C
(N), 150C (M) and 175C (v).
166 T. Serin et al. / Journal of Non-Crystalline Solids 276 (2000) 163168
reverse-bias annealing have been found to origi-
nate in a quenched midgap defect density of the
undoped i-region [7,20]. The proposed quenching
mechanism was a lowered np product during re-
verse-bias annealing. Also in undoped a-Si:H
metastable changes of gap state proles with re-
verse-bias annealing have been investigated by
modulated photocurrent measurements and the
increase of the DOS above E
F
and the decrease of
the DOS below E
F
have been found [9,10]. The
bias-annealing eect has been investigated for
undoped a-Si:H photodiode by Ichinose et al. [14],
they have measured the ls-product, dark current
and transient photocurrent by evaluating the
properties of the photodiode. The ls-product and
transient photocurrent were improved by the bias-
annealing treatment at 125C and 150C. The re-
sults suggested that the defect density around
midgap decreases and that, above the Fermi level
it increases by the bias-annealing treatment. Sim-
ilar behavior was also observed in the case of pin
junctions [2123]. In our present case, the values of
DOS around the middle of i-region rst decreased
with increasing annealing temperature up to
around 130C, then increased with the increasing
annealing temperature as shown in Fig. 3.
Diusion of bonded hydrogen within the a-Si:H
matrix interconverts weak bonds and dangling
bonds. In undoped a-Si:H hydrogen does not be-
gin to diuse until temperatures as high as 470 K
[24]. DOS distribution in this material should not
change since our annealing temperature is below
this value. But there are the changes of the DOS
around the middle of i-region with bias annealing.
The explanation of these changes in DOS can be
attributed to the motion of bond centers hydrogen
(H

BC
) in electrical eld [25].
The changes of the DOS around the middle of
i-region with bias annealing can be explained by
defect pool method [26]. Bias annealing intercon-
verts weak-bonds and dangling bonds by the fol-
lowing reactions:
WB e D

Y (7)
WB h D

X (8)
The equilibrium distribution depends on the for-
mation energy E
Form
and the time to reach equi-
librium depends on the barrier to form U, which is
the barrier for defect formation. If the formation
energy of the neutral defect with one electron is
identied as energy dierence between the valence
band states (the weak bonds) and the dangling
bond energy [27], then the expressions for the
formation energies of charged defects are
E
Form
(D

) = E
F
E
t
Y (9)
E
Form
(D

) = 2E
+
E
F
U E
t
Y (10)
where E
+
is the defect energy, E
t
the valence band
energy and U is the correlation energy. The dis-
tribution of defects D(E) which are actually
formed depends on the product of P(E) (a pool of
potential defects) and the probability of defect
formation at an energy E
D(E) = P(E) exp(E
Form
akT
+
)Y (11)
where T
+
is the equilibrium temperature while the
energy position of the states does not change.
Lowering E
F
leads to an exponential decrease in
the density of D

states and an exponential in-


crease in D

states.
With the help of comments given above, the
change of DOS around midgap of i-region due to
bias annealing can be explained as follows: the bias
applied to sample during the annealing and the
cooling causes the lowering of Fermi level, and so
the defect formation energy for the defects below
the Fermi level increases and the defect DOS
decreases exponentially. For this reason DOS
around midgap decreases while the annealing
temperature increases up to 130C. The defect
formation energy at larger annealing temperatures
than 130C is small, because there is a raising of
Fermi level with the increasing annealing temper-
ature in addition to the lowering of Fermi level due
to reverse bias. Therefore, the defect DOS around
midgap increases at larger annealing temperatures
than 130C.
During the bias annealing treatments, we ap-
plied bias to the sample, annealed it to 110175C
range and then cooled to room temperature with
the bias applied. In our opinion, this causes re-
equilibration to a new Fermi energy and a new
DOS, which was frozen in cooling.
T. Serin et al. / Journal of Non-Crystalline Solids 276 (2000) 163168 167
6. Conclusion
Au/a-Si:H/c-Si/Al heterostructure was annealed
by bias annealing. DOS around midgap of i-region
was determined by ``drive-level-capacitance'' method
after each annealing processes maintained in the
annealing temperature range 110175C. It was
observed that DOS was signicantly changed by
means of thermal annealing. The changes were
explained by `defect pool model'.
Acknowledgements
Authors would like to thank Dr B. Schr oder
(Fachbereich Physik, Universitat Kaiserslautern)
for the sample preparation facilities and ``The
Scientic and Technical Research Council of
Turkey'' (TUBITAK) for the project TBAG-AY/
190 and Ankara University Research Fund.
References
[1] L. Laih, J. Wang, Y. Chen, T. Jen, W. Tsay, J. Hong, Solid
State Electron 41 (1997) 1693.
[2] C.S. Harder, B. Van Zeghbroeck, H. Meier, W. Patrick,
P.J. Vettiger, Lightwave Technol. 9 (1989) 171.
[3] H. Schumaacher, H.P. Leblanc, J. Soole, R. Bhat, IEEE
Electron Device Lett. 11 (1988) 607.
[4] N.F. Mott, E.A. Davis, Electronic Processes in Non-
Crystalline Materials, 2nd Ed., Oxford University, Oxford,
1979 (Chapter 6).
[5] H. Mimura, Y. Hatamaka, J. Appl. Phys. 61 (1987) 7.
[6] K. Okuda, H. Okamoto, Y. Hamakawa, Jpn. J. Appl.
Phys. 22 (1983) L603.
[7] G.A. Swartz, Appl. Phys. Lett. 44 (1984) 697.
[8] R.A. Street, J. Kakalios, Philos. Mag. Lett. B 54 (1986)
L21.
[9] G. Schumm, G.H. Bauer, J. Non-Cryst. Solids 114 (1989)
660.
[10] G. Schumm, G.H. Bauer, Phys. Rev. B 39 (1989) 5311.
[11] M.J. Powell, C. Van Berkel, S.C. Deane, J. Non-Cryst.
Solids 137&138 (1991) 1215.
[12] R. Miyagawa, T. Yamaguchi, A. Furukawa, J. Non-Cryst.
Solids 137&138 (1991) 1317.
[13] Z. Chen, W. Yu, Thin Solids Films 251 (1994) 96.
[14] H. Ichinose, Y. Ishizuka, H. Nozaki, A. Furukawa,
H. Tango, O. Yoshida, J. Non-Cryst. Solids 198200
(1996) 322.
[15] C.E. Michelson, A.V. Gelatos, J.D. Cohen, Appl. Phys.
Lett. 47 (4) (1985) 412.
[16] D.V. Lang, J.D. Cohen, J.P. Harbison, Phys. Rev. Lett. 48
(1982) 421.
[17] R.A. Street, J. Kakalios, C.C. Tsai, T.M. Hayes, Phys.
Rev. B 35 (3) (1987) 1316.
[18] E.Z. Liu, W. Spear, J. Non-Cryst. Solids 137&138 (1991)
167.
[19] G. Kr otz, G. M uller, J. Non-Cryst. Solids 137&138 (1991)
163.
[20] Z. Smith, S. Aljishi, V.C. Solobodin, S. Wagner, P.M.
Lenohan, R.R. Arya, M.S. Bennett, Phys. Rev. Lett. 57
(1986) 2450.
[21] T. Serin, N. Serin, Semicond. Sci. Technol. 6 (1991) 679.
[22] T. Serin, N. Serin, Semicond. Sci. Technol. 9 (1994) 2097.
[23] T. Serin, Semicond. Sci. Technol. 12 (1997) 1.
[24] R.A. Street, C.C. Tsai, J. Kakalios, W.B. Jackson 56 (1987)
305.
[25] N.M. Johnson, J. Non-Cryst. Solids 137&138 (1991) 11.
[26] K. Winer, Phys. Rev. B 41 (1990) 12150.
[27] M. Stutzmann, Philos. Mag. B 56 (1987) 63.
168 T. Serin et al. / Journal of Non-Crystalline Solids 276 (2000) 163168

Вам также может понравиться