Вы находитесь на странице: 1из 44

SUMMER PRACTICAL TRAINING REPORT

AlGaN/GaN HEMTs and


Analysis of Transistors
K. Gagandeep G. Singh
3
rd
Year Undergraduate, Engineering Physics
Indian Institute of Technology, Hauz Khas, New Delhi -110016
Work from 12
th
May to 18
th
July at:
Solid State Physics Laboratory, Timarpur, Delhi -110054
1
SOLID STATE PHYSICS LABORATORY
DODO, MINISTRY OF DEFENCE
GOVERNMENT OF INDIA
21
st
July, 2014
CERTIFICATE
This is to certify that Mr. K. Gagandeep G. Singh, student of Engineering Physics (B. Tech)
3
rd
Year at IIT Delhi has undergone training under Mrs. Seema Vinayak of Solid State Physics
Laboratory (SSPL) for a period from 12
th
May to 18
th
July on AlGaN/GaN HEMTs and
Analysis of Transistors. He has completed their training successfully. This report does not contain
any secret information.
Signature of the Supervisor
(Dr. Seema Vinayak, Scientist G)
2
Contents
1 Introduction 6
2 Properties of GaN 7
2.1 Crystal Properties of GaN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Physical Properties of GaN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Figure of Merit (FOM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Substrate and Growth Techniques 12
3.1 Growth Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4 AlGaN/GaN Heterostructures 14
5 Transistors 15
5.1 Bipolar Junction Transistor (BJT) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.1.1 Common Emitter Conguration . . . . . . . . . . . . . . . . . . . . . . . . . . 16
5.2 Characteristics of MOSFET (Metal Oxide Semiconductor Field Emission Transistor) . 18
5.2.1 Saturation Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.2.2 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.2.3 Finding the Saturation Current . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.3 Characteristics of MESFET (MEtal Semiconductro Field Eect Transistor) . . . . . . 24
5.3.1 Saturation Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.3.2 Saturation Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6 Simulations for MOSFET Using COMSOL Multiphysics
R
28
6.1 Model Specications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.1.1 Parameters Used during the Simulation . . . . . . . . . . . . . . . . . . . . . . 29
6.2 Energy Level Diagram of MOSFET along the center of the device form the Gate Contact 29
6.3 Current Density of MOSFET and n-channel . . . . . . . . . . . . . . . . . . . . . . . 30
6.4 Electron Concentration for dierent Drain to Source Voltages . . . . . . . . . . . . . 31
6.5 Surface Charge Density of MOSFET . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
6.6 Current(I
D
) vs Voltage Characteristics(V
DS
) . . . . . . . . . . . . . . . . . . . . . . . 33
7 Semiconductor Heterostructures 33
7.1 Heterojunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.1.1 Band Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.1.2 Understanding the Notch Eect . . . . . . . . . . . . . . . . . . . . . . . . . . 35
8 HEMT Working Principle 36
9 Conclusion 40
10 References 41
11 Important Points 43
3
Acknowledgements
I sincerely express my sense of indebtedness to Dr. R. Muralidharan, Director SSPL, for grant-
ing me permission to undergo the summer training at SSPL.
I would like to sincerely thank my faculty coordinator Dr. Rajendra Singh, Professor IITD to provide
me this humble opportunity to work at SSPL. I warmly thank him for his co-operation and guidance.
I would like to express my sincere thanks to Mr. Amit and Mr. Chandan Sharma for their help
extended.
Finally I would like to sincerely express my thankfulness to our supervisor Mrs. Seema Vinayak,
Scientist G, for her guidance and for constantly motivating us to work harder.
I would like to state that the visits to the SSPL laboratory were very helpful in giving me an in-
sight into the practical work being done in the eld of my project.
4
Objectives of Training
The aim of the training program was to get familiarized with the components of fabrication of thin
lms at the laboratory and also to get the basic working principles behind the semiconductor devices
like MOSFETs, MESFETs and HEMTs. The objectives for the training period were:
To understand the basic operation principles and basic dierences between MOS and MES
FETs.
Calculation of Energy Band Diagrams in MESFETs and MOSFETs.
Understand the mechanism of current saturation in BJTs, MESFETs and MOSFETs.
DC Characteristics of MOSFETs simulated in COMSOL Multiphysics
R
.
5
1 Introduction
The continuously increasing need for electric power has led to development of new technology which
makes it possible to increase the eciency by miniaturization of device and by increasing power supply.
But now ecient use of the available energy has become the main concern of modern power electronics.
For the past several decades Si based semiconductor devices have been the most exhaustively used
technology for handling power circuitry. But due to its physical and performance short comes like
low power handling capacity, breakdown voltage, maximum operating temperature etc. no further
advancements in power electronics can be done any further. In particular, due to the band gap and
intrinsic carrier concentration of the material, Si devices are limited to work at a junction temperature
lower than 200 C which is less suitable for modern power applications. [1]
Where else, wide band semiconductors (WBG), such as silicon carbide (SiC), gallium nitride (GaN)
and related alloys, exhibit better physical properties which can serve better in satisfying the demand
of increased power, frequency and operating temperature of the devices.
Form these SiC technology is the most advanced technology amongst other wide band semiconduc-
tors. This is so because rst of all the size of SiC wafer has been continuously increasing and that too
defect free (especially micropipes). Furthermore, the device processing technology has reached a high
level of maturity and some SiC devices like Schottky diodes or MOSFETs operating in the range of
600-1200 V have already reached the market [2].
While SiC is the most technically advance WBG semiconductor, GaN and related alloys (like AlxGaN)
still suer from the several physical issues related to both surface and interfaces [3]. Furthermore,
the lack of good quality (and cheap) free standing GaN templates make also the material growth a
serious concern, since heteroepitaxy on dierent substrates (like Al2O31-x, SiC, or Si) is required.
Although for long time GaN has been mainly attractive because of the optoelectronics applications,
the recent improvement of the material quality of GaN-based heteroepitaxial layers provided the sci-
entic community with considerable incentive to investigate the potentialities of this material also for
applications in power electronics.
Although the low eld mobility of bulk GaN is much lower than that of other III- IV materials
like GaAs, GaN has a much higher saturation velocity and wide band gap, which makes it favorable
for high frequency power devices. Another good reason to look into advancing GaN is its ability to
form AlGaN/GaN heterostructures with a large band discontinuity which helps in the formation of
2DEG (2 Dimensional Electron Gas). 2DEG is formed due to the presence of both spontaneous and
piezoelectric polarization of the material.
The high polarizations and resulting electric elds in AlGaN/GaN heterostructures produce high
interface charge densities even for unintentionally doped materials. In particular, the 2DEG formed
is AlGaN/GaN heterostructures can have sheet carrier densities in the order of 1-3x10
13
cm
-2
, i.e.,
well in excess of those achievable in other IIIV systems like GaAs. Moreover, the possibility to use
undoped material results in a signicant improvement of the electron mobility in the 2DEG, due to
the reduction of Coulomb scattering with ionized impurities.
AlGaN/GaN heterostructures are particularly interesting for the fabrication of high electron mo-
bility transistors (HEMTs), based on the presence of the 2DEG.
6
Figure 1: Crystal structure of Zinc Blend(Left), Wurtzite(Middle) and Rock Salt(Right)
2 Properties of GaN
The research activity on GaN dates back to the rst 1930s. However, only in the 90s the material
started to attract interest for power and RF electronics, because of its superior properties such a high
band gap, a high breakdown eld and a high saturated electron velocity. In the next years, power
electronics will play an important role for the reduction of the energy consumption all over the world.
Many discrete power electronic devices are used in the power modules for the transmission and
the conversion of electric power. For these devices, a reduction of the static and dynamic losses can
directly result in the overall lowering of power consumption of the system. Also the next generation
of high-speed communication devices are becoming key technologies for network communication, re-
quiring increasing operating frequency associated with portability and convenience.
Si technology is approaching the theoretical limits imposed by the material properties, in terms of
maximum operation power, frequency and temperature. Therefore new materials have to be looked
forward to in order to overcome Si limitations. The use of wide bandgap (WBG) materials can be
considered as the best solution to meet the requirements of modern power electronics. In fact, WBG
semiconductors such as silicon carbide (SiC) and gallium nitride (GaN), have been known to exhibit
superior electrical characteristics compared to Si because of their inherent advantages such as high
electron mobility, higher breakdown eld strength and larger energy bandgap.
7
Figure 2: The wurtzite unit cell of GaN with lattice constants a0 and c0.
2.1 Crystal Properties of GaN
Any III-nitride material is expected to exist in one of the three forms namely rock salt, zinc-blend
and wurtzite.
Thermodynamically wurtzite is the most favorable structure for GaN at ideal temperature and pres-
sure. [4] The basis consists of four atoms two Nitrogen and two Gallium atoms. The unit cell consists
of six atoms and is characterized by two constants a
0
(3.18A) and c
0
(5.18A). The Ga and N atoms are
arranged in two interpenetrating hexagonal close packed lattices (HCP), each one with one type of
atoms, shifted 3/8 c0 each other. The covalent bonds allow that each atoms is tetrahedrally bonded
to four atoms of the other type. There is also a ionic contribution of the bound due to the large
dierence in electronegativity of Ga and N atoms. On a wurtzite structure there is no inversion
symmetry on the [0001] direction (c-axis). This latter means that it is possible to distinguish two
dierent orientation of GaN crystals, i.e., Ga-face and N-face, depending if the material is grown with
Ga or N on top and corresponding to the (0001) and (0001-) crystalline faces as is shown Figure 3.
Schematic drawing of the crystal structure of wurtzite Ga-face and N-face GaN. The spontaneous
polarization vector is also reported.
Now, because of the dierence in the electronegativity of N and Ga there exists an ionic bond along
with the already present covalent bond. This leads to a dipole moment in the (0001) direction. This
eect in known as spontaneous polarization, because it exists without the presence of any stress or
strain.
The strength of the spontaneous polarization depends on the non-ideal (asymmetric) structure of
the crystal. Not only covalent bond in the direction parallel to c0 plays an important role, but also
8
Figure 3: Ga-face and N-face structure of GaN
the other three covalent bonds of the tetrahedral structure. Their resultant polarization is aligned
with c0 but in a0 opposite direction, compensating the polarization in the (0001) direction. For this
reason in a wurtzite structure, the c/a ratio plays a fundamental role for the spontaneous polarization,
where the resultant Psp increases with reducing the asymmetry of the crystal, i.e. decreasing the c/a
ratio. For example a GaN crystal with a c/a ratio of 1.6259 will present a reduced Psp (-0.029 C/m2)
with respect to an AlN crystal (-0.081 C/m2) with a c/a ratio of 1.6010.
In this context, in the presence of factors that may change the ideality of the structure and the
c/a ratio, as stress or strain, the total polarization will be modied. The additional contribution to
the polarization, due to the presence of strain and stress in the crystal, is the so called piezoelectric
polarization Ppe. This contribution is particularly important in AlGaN/GaN heterostructures for the
generation of the two dimensional electron gas about which we will be seeing afterwards.
2.2 Physical Properties of GaN
Thanks to its superior physical properties, GaN is considered an outstanding materials for opto-
electronics, high power and high frequency devices. Properties like the wide band-gap, the high value
of critical electric eld and the saturation velocity can represent a big advantage in terms of electronic
devices applications. In Table 1, some of these properties, which are relevant for electronic devices
performances, are reported and compared to other semiconductors counterparts. [5]
The wide band-gap of GaN (3.39 eV) is responsible for the high critical electric eld (3.3 MV/cm),
which is one order of magnitude higher than that of Si. The high critical electric eld gives the possi-
bility to sustain the application of high bias values, thus making the material suitable for high-voltage
devices fabrication. A further implication of its wide band-gap is the low intrinsic electron concen-
tration ni. The value of ni in GaN at room temperature is in fact several orders of magnitude lower
9
Properties Si SiC (4-H) GaN GaAs Diamond
Bandgap Energy (E
g
) eV 1.12 3.26 3.39 1.42 5.45
Breakdown Field (E
c
) MV/cm 0.3 3.0 3.3 0.4 5.6
Intrinsic e
-
Concentration (n
i
) cm
-3
1.5x10
10
8.2x10
-9
1.9x10
-10
1.5x10
6
1.6x10
-27
e
-
Saturation Vel. (v
sat
) x10
7
cm/s 1.0 2.0 2.5 1.0 2.7
e
-
Mobiltiy cm
2
/V.s 1350 700 1200 8500 1900
Thermal Conductivity (k) W/cm.K 1.5 3.3-4.5 1.3 0.5 20
Relative Permittivity 11.8 10.1 9.0 13.1 5.5
Table 1: Properties of GaN compared with other conventional and wide band-gap semiconductors at
room temperature
with respect to that of Si or GaAs, and comparable with that of SiC. This characteristic enables to
increase the maximum operation temperature of the devices made of this material and have reduced
leakage currents.
Other parameters that describe the quality of the material are the relative permittivity (epsilon r) and
the thermal conductivity (k). The relatively high permittivity value (epsilon r) is a good indicator
of the capacitive loading of a transistor and passive components. On the other hand, the thermal
conductivity (k) describes the ease of heat conduction and, hence, the possibility to eciently extract
the dissipated power from the device. Materials with a lower thermal conductivity typically lead to a
device degradation at elevated temperatures. Although III-V semiconductors typically have a moder-
ate value of k, GaN has a thermal conductivity which is comparable to that of Si (but lower than SiC).
The amazing properties of GaN include also a high electrons saturation velocity (v sat), which in
turn is important for high current and high frequency operation of devices. Compared to other wide
band gap materials that show high v sat, GaN can also reach a high electron mobility (u) compara-
ble with Si. Undoubtedly, among wide band gap semiconductors, the unique feature of GaN is the
possibility to make band gap engineering considering the related AlxGa1-xN alloys. In particular, by
varying the Al content it is possible to tailor the band gap of the material. In this way, AlGaN/GaN
heterostructures can be fabricated, allowing to reach carrier mobility up to 2000 cm2/Vs in the two
dimensional electron gas (2DEG) formed at the interface.
Coming to an example of application of GaN, there are electric power converters which are inte-
grated practically in all the electronic systems to convert either DC or AC current. Their eciency is
also related to the possibility to have fast switching elements with increased power density. Typical
applications of ecient power converters are the energy conversion in solar systems, wind power sta-
tions and modern electric vehicles as well as for power supplies in mobile base stations and computer
systems. In all the aforementioned sectors, GaN represents today an attractive material. In fact, GaN
based switches have theoretically a better gure of merit with respect to Si and SiC. Figure shows
the comparison between the trade-o curves of the specic on resistance R ON vs breakdown voltage
for Si, SiC and GaN. [6]
10
Figure 4: On-Resistance vs Breakdown Voltage
As can be seen, at a given operation voltage, the on-state resistance of GaN devices can, in princi-
ple, outperform the competing Si or SiC devices. Since the specic on resistance is strictly related to
the power losses of the device, the use of GaN can signicantly lead to a reduction of the losses and to
an improvement of the eciency of the electronic systems. However, as can be seen, the experimental
data point are still far from the theoretical limits of the material.
2.3 Figure of Merit (FOM)
To better compare the potential power electronic performance for dierent semiconductors materials,
gures of merit (FOM) are commonly adopted. In particular, for high power and high frequency
devices three important FOM are considered, Johnson (JFOM), Baliga (BFOM) and Baliga high fre-
quency (BHFOM). JFOM = (v
sat
Ec)
2
is an indication of the maximum capability to energize carriers
by electric eld, BFOM=u.Es.Ec
3
measures the minimum conduction losses during DC operation
and BHFOM=uEc
2
give information about the minimum conduction losses during high frequency
Figure of Merit Si SiC GaN
JFOM (v
sat
Ec)
2
6x10
10
3.6x10
13
14.6x10
13
BFOM u.Es.Ec
3
248 20.9x10
4
79x10
4
BHFOM uEc
2
84 7200 20800
Table 2: Figure of Merit comparison
11
Figure 5: Typical operating frequencies and output power ranges of electron devices made using a
dierent semiconductor materials.
operation. All these gures of merit for GaN are reported in Table 2 and compared to Si and SiC,
clearly showing that GaN is potentially a superior material for the high power and high frequency
applications.
A comparison of the typical operating frequency and output power range for dierent semiconductor
materials is shown in Figure. In this case, it must be noted that with respect to SiC, GaN is more
suitable for higher frequencies but in a lower output power range.
3 Substrate and Growth Techniques
In spite of its outstanding material properties, the technological development of GaN has come later
than in other semiconductors. The reasons of this delay were mainly related to the diculty to have
high quality free-standing GaN substrates and, consequently, the diculty to fabricate vertical struc-
tures for power devices. In fact, for the growth of GaN other materials must be used as substrate.
Since the perfect substrate does not exist, an ideal candidate must have physical and crystallographic
properties, such as lattice parameters and thermal expansion coecients, close to those of GaN, in
order to avoid the formation of cracking of the lm, or defects formation during the growth of the
material. The lattice mismatch and the dierence in thermal expansion coecients (TEC) of the
common substrates used for GaN growth are reported in Table 3.
Out of the possible choices Sapphire is an interesting choice because it is insulating, it can with-
12
Substrate Latticce Mismatch Dierence in Thermal Expansion Cocient
Al2O3 (0001) +16 % -25.3 %
6H-SiC (0001) +3.5 % +33.3 %
3C-SiC (111) +3 % +24.4 %
Si (111) -17 % +55.5 %
AlN (0001) +2.5 % +33.3%
Table 3: Lattice mismatch and dierence in thermal expansion coecient of GaN with respect to the
most common substrates
stand the required high growth temperatures, and it is relatively cheap. Anyway the large lattice
mismatch (+16Silicon can be also a possible substrate for the growth of GaN layers. In the recent
years, GaN materials grown on Si is attracting a huge attention thanks to the low substrate cost,
the possibility of large substrate diameters and the potential integration with the well-developed Si
electronics technology. Despite a large lattice (+17The residual stress is depending on the growth
condition and cool-down procedure. Moreover there exists a dependence of the stress on the impurity
concentrations that lead to an increase of the tensile stress with increasing the doping concentration.
[8, 9]
To relieve the tensile stress and achieve crack-free GaN heterostructures, several kind of transition
layers can be used, such as low temperature AlN [10], graded AlGaN buers [11] or AlGaN/GaN
superlattices [12]. It has been seen that the dislocation density in the material strongly depends on
the choice of the transition layer, and can be partially mitigated by using a high temperature AlGaN
intermediate layer that acts as a dislocation lter [13]. Moreover transition layers increase also the
series resistance in the GaN layer, reducing the crack density and providing a good electrical insula-
tion from the substrates.
3.1 Growth Technique
If the choice of a suitable substrate is an important issue for the development of GaN technology,
not less important are the growth techniques employed to obtain a high quality material, with a low
concentration of defects. The rst technique used to grow epitaxial GaN layers was the Hydride Vapor
Phase Epitaxy (HVPE).
Nowadays, the Metal Organic Chemical Vapor Deposition (MOCVD) has become the most used
method to grow GaN, owing its superior quality as high degree of composition control and uniformity,
reasonable growth rates (1-2 um/hr), the possibility to use high purity chemical sources and to grow
abrupt junctions. MOCVD uses the reaction of trimethylgallium (TEGa) and NH3 that occurs close
the substrate. To obtain high quality GaN lm, during the deposition the substrate must be kept
at a temperature of about 1000 C - 1100 C, to allow a sucient dissociation of the NH3 molecule,
at a pressure between 50 and 200 Torr. Moreover, another critical aspect is the control of the N/Ga
molar ratio that must be kept high in order to compensate N losses due to the high partial nitrogen
pressure at the elevated growth temperatures [14].
13
In fact the poor nucleation of GaN on Si at high temperatures results in a reaction of nitrogen
with Si and in a Ga-Si alloy formation which initiates a strong and fast etching reaction (melt back
etching) destroying the substrate and the epitaxial layer [15]. The most established method to prevent
the nitridation is starting the growth process with an AlN nucleation larger grown in the same reactor
with a few monolayers pre-deposition of Al. [16] The material doping can be tailored by the induction
of extra precursor on the reactor, as silane (SiH4) for Si doping (n-type) or biscyclopentadienylmag-
nesium (Cp2Mg) for Mg doping (p-type). Anyway the control of a low doping concentration (N
D
is less than 110
16
cm
-2
) is still a complex factor because the formation of nitrogen vacancies, which
act as donors leading to n-type doping of the material. Also the oxygen impurities present during
the growth process can act as donors, leading to an n-type material [17]. To improve the crystalline
quality of the grown GaN, pre-treatments can be required. For example the deposition of a thin low
temperature buer layer can be an advantage. The use of this layer, generally AlN or Si, can reduce
the lattice mismatch, providing a benet in terms of defects density (dislocations, oxygen impurity,
nitrogen and gallium vacancies, etc).
To reduce the defects density in the grown material, a dierent process called lateral epitaxial over-
growth (LEO) has been also developed [18]. It consists in the deposition of GaN on a patterned
dielectric substrates (like SiO) followed by the lateral expansion and coalescence of the grown ma-
terial. Although this technique can lead to a signicant reduction of the dislocation density (up to
610
7
cm
-2
) the extremely high cost of the process (which require the employment of lithographic steps
for the substrate preparation) has limited its practical application for GaN growth.
The Molecular Beam Epitaxy (MBE) is a slow (1 um/hr) but ecient technique for GaN growth,
that show comparable material quality to those grown by MOCVD. A problem is that the NH3 is
very stable at the low temperature (700-800 C) used in MBE. To solve this issues reactive species of
nitrogen, generated by electron cyclotron resonance (ECR) or radio frequency (RF) plasmas with low
energy, are generally used [19].
4 AlGaN/GaN Heterostructures
One of the most interesting aspects related to GaN materials is the possibility to grow AlGaN/GaN
heterostructures, in which a two dimensional electron gas (2DEG) is formed at the heterojunction.
The peculiarity of AlxGa1-xN alloys is the possibility to tailor the lattice constant and the energy
gap by varying the Al concentration x.
In particular, the in-plane lattice constant a of AlxGa1-xN alloys depends on the Al concentration x,
and is also related to the lattice constant of GaN and AlN by the relation [20]
a
AlGaN
(x) = xa
AlN
+ (1 x)a
GaN
(1)
On the other hand, also the band gap of a AlxGa1-xN alloy can be expressed as a function of the Al
mole fraction according to [21]
E
AlGaN
g
(x) = xE
AlN
g
+ (1 x)E
GaN
g
x(1 x) = [x6.13 + (1 x)3.42 x(1 x)]eV (2)
14
Figure 6: Dependence of the band gap energy (a) and of the lattice parameter a (b)on the Al mole
fraction for the AlxGa(1-x)N.
The dependence of the band gap energy and the lattice parameter with respect to the Al mole
fraction for the AlxGa1-xN is shown below.
The big advantage of AlGaN/GaN heterostructures consists in the formation of a two dimensional
electron gas (2DEG) at the interface, generated by the strain induced by the lattice mismatch between
GaN and AlGaN. The presence of the 2DEG allows the fabrication of an innovative device called High
Electron Mobility Transistor (HEMT).
5 Transistors
In order to understand more about working of High Electron Mobility Transistor(HEMT), the basic
knowledge of working principles of dierent transistor can give a better view about how the HEMTs
work.
5.1 Bipolar Junction Transistor (BJT)
When BJT is setup as shown in the following gure it is said to be in forward active mode. The
EB(Emitter-Base) junction is forward bias and the BC(Bas-Collector) junction is reverse bias.
15
The energy band diagram of this npn BJT under 0 bias and forward active mode is given as
When EB is forward bias the barrier height as seen by carrier reduces and the thermal excitation
energy kbt provide sucient energy to the electron permitting them to cross the region I and enter
region II. In region II there is an established gradient (approximately linear) of electron concentration
due to the majority carrier electron of region I which came into region II and are minority carrier here.
Due to gradient we get e- diusing towards the interface of region II and III where because of reverse
bias conguration of electric eld is easily able to drift away the e- at the interface. So, eventually we
can understand that the current through collector is dependent only on the barrier height on region
I and region II interface. Which in turn depends on the voltage between BE.
i
c
= eD
n
A
BE
dn(x)
dx
= eD
n
A
BE
_
n
B
(0) 0
0 x
B
_
(3)
i
c
=
eD
n
A
BE
x
B
.n
B0
.exp
_
V
BE
V
t
_
(4)
i
c
= I
s
.exp
_
V
BE
V
t
_
(5)
5.1.1 Common Emitter Conguration
16
As we have discussed earlier the transistor will work (i.e. generate current through collector) i
the CB junction is reverse bias. So keeping this in mind we write the KVL equation for loop 1.
V
CC
= IR + V
CB
+ V BE = IR + V
CE
(6)
Now if Vcc is large enough and Vr(=IR) is small enough then Vcb is greater than 0, that is BC
junction is reverse bias and thus the transistor is in the forward active region of operation.
Again as the forward bias BE voltage increases the collector current and hence Vr will also increase.
Increase in Vr results to decrease in the magnitude of Vcb. At some point Vr + Vcc may become
0 and at that point if the Ic is increased slightly further the setup will become forward biased and
ow of minority carrier through base collector barrier will be obstructed by the electric eld which
is opposing their ow. Thus we get a saturation current when the gradient of e- concentration is
not able to supply sucient force to the e- to move through the electric eld and contribute to the
current. The following thing can be understood using this Ic Vce graph.
The reverse bias is shown by linear region. The non-linear region before saturation shows that
some e- are able to pass through the negative electric eld and contribute to the current. And over
17
the saturation region even if we increase Vce we wont be able to increase current any further.
V
CE
= V
CC
I
C
R
C
(7)
I
c
=
V
CC
V
CE
R
C
(8)
[24,25]
5.2 Characteristics of MOSFET (Metal Oxide Semiconductor Field Emis-
sion Transistor)
Let us consider the region below the oxide layer under normal condition (No applied Voltage to the
gate) the energy band diagram is as follows
When a positive voltage is applied with respect to S the holes due to the eld move away and
leave behind some negative charge. The band diagram of which is as follows
When a positive voltage is applied whit respect to S the holes due to the eld move away and
leave behind some negative charge. The band diagram of which is as follows.
18
The excess of e- near the interface makes the band to bend such that Ef gets closer to Ec. When
even higher positive voltage is applied then there will be a situation where in the E will be crossed by
Ef. This is the situation when the region near the interface starts behaving as an n-type semiconductor
(as in n-type the Ef is abve E) and we get an inversion region. When seen this region in the following
diagram, the region appears as to be a channel of n-type semiconductor. The voltage above which
this happens is the threshold voltage (Vt).
So, under this situation n-channel is formed which can easily conduct electricity ids if small voltage
is applied over the drain. The n-channel has some conductance and for small values of Vds has its
conductance or resistance characteristics similar to a resistance. So,
I
D
= g
d
V
DS
(9)
g
d
=
W
L

n
|Q

n
| (10)
I
C
=
V
CC
V
CE
R
C
(11)
So, form previous discussion we have
19
5.2.1 Saturation Current
When Vds increases to the point where the potential drop across the oxide ot the drain terminal is
equal to Vt, the induced inversion charge density is zero at the drain terminal. The following gures
show these eects
When Vds Becomes large than Vds(sat) value, the point in the channel at which the inversion
charge is just zero moves towards the source terminal. In this case, e- enter the channel at the source,
travel through the channel towards the drain and then, at the point where the charge goes to 0, the
e- injected into the space charge region where they are swept by the E-eld to the drain contact. The
region of Id vs Vds characteristic is referred to as the saturation region. When Vgs changes, the Id
vs Vds curve will increase so we have
20
5.2.2 Derivation
From Ohms law we have
J
x
= E
x
(12)
where
= e
n
n(y) (13)
The total current is found by
I
x
=
_
y
_
z
J
x
dydz (14)
So,
Q

n
=
_
en(y)dy (15)
where Qn is inversion layer charge per unit area. So,
I
X
= W
n
Q

n
E
x
(16)
From Charge neutrality
Q

m
+ Q

ss
+ Q

n
+ Q

sd
(max) = 0 (17)
From Gausss law _
s
E
n
dS = Q
T
(18)
21
Now, from surfaces 1 & 2, we assume that Ex is essentially a constant along the channel length. The
contributions of surfaces 1 & 2 cancel each other. Surface 3 is in the neutral p-region, so the electric
eld is zero at this surface. Only surface 4 contributes.
_
s
E
n
dS =
0x
E
ox
Wdx = Q
T
(19)
Where, epsillon ox is the permitting of the oxide. The total charge enclosed is
Q
T
= (Q

ss
+ Q

n
+ Q

SD
(max))Wdx (20)
So, from the previous two equations we have

0x
E
0x
= Q

ss
+ Q

n
+ Q

SD
(max) (21)
Now,
E
FP
E
FM
= e(V
GS
V
x
) (22)
Considering the potential barriers we have,
V
GS
V
x
= (V
0x
+

m
) (

E
g

st

fp
)
V
GS
V
x
= V
0x
+ 2
fp
+
mx
where, phi ms is the metal semiconductor work function dierence.

s
= 2
fp
for the inversion condition.
The electric eld in oxide is
E
0x
=
V
0x
t
0x
On combining the equations we have

0x
E
0x
=

0x
t
0x
[(V
GS
V
x
) (
ms
+ 2
fp
)]
So,
I
x
= W
n
C
0x
dV
x
dx
[(V
GS
V
x
) V
T
] (23)
22
Where Ex = -dVx/dx and Vt is the threshold Voltage.
So, on integrating
L
_
0
I
x
= W
n
C
0x
Vx(L)
_
Vx(0)
[(V
GS
V
x
) V
T
]dV
x
(24)
Assuming mu n to be a constant above.
For the n-channel device the drain current enters the drain terminal and is a constant along the
entire channel length. Letting Id = -Ix the equation becomes
I
D
=
W
n
C
0x
2L
[2(V
GS
V
T
)V
DS
V
2
DS
] (25)
which is valid for V
GS
V
T
and V
DS
(sat) V
DS
.
5.2.3 Finding the Saturation Current
The above equation of current through the drain can be plotted as follows
Since Id is valid below Id(sat)
Id sat can be found by taking the maxima of the above plots. So, from
dI
D
dV
DS
= 0
we get
V
DS
= V
GS
V
T
(26)
The value of Vds is just Vds(sat). For Vds is graVds(sat) the ideal drain current is a constant
and is equal to
I
D
(sat) =
W
n
C
0x
2L
(V
GS
V
T
)
2
(27)
So, nally the plots we obtain are
23
[24,25]
5.3 Characteristics of MESFET (MEtal Semiconductro Field Eect Tran-
sistor)
When negative voltage is applied to the gate the e- below the gate feels repulsion and positive charge
is left near the schottky junction. This region thus forms a depletion region where no free charge
is present. As the negative voltage magnitude increases the width of depletion layer increases and
nally covers completely the n channel. Thus making the gate to close the conduction. Such gates
are active low gates which are activated only at zero magnitude of applied potential.
Now when n-channel is not depleted then even on the application of small positive voltage at the
drain the majority carrier negative move from the source through n-channel to the drain and thus
conduct electricity and cause current ow. Now, if Vd(Drain Voltage) is increased then the region
near the drain will become reverse biased and thus the depletion region width near this region will
increase.
24
Now if Vd is increased further then a voltage will come at which the depletion region near the
drain will completely cover the n channel. At this condition it is presumed that current will halt at
once, but due to strong electric ied the e will be pulled. Under this condition we reach the saturation
current value and even on any further increase of Vd we wont get any increase in value of Id. Thus
we get the saturation region.
5.3.1 Saturation Voltage
The depletion region width will vary with distance h(x) throughout the channel. hi is function of
Vbi(built In Voltage) and Vgs(Gate Voltage) and hm(max. depletion region width) is given by
h
m
=

2
s
(V
bi
+ V
DS
V
GS
)
eN
d
(28)
Pincho occurs when hm = a. At this point we reach the Saturation condition. So,
a =

2
s
(V
bi
+ V
DS
(sat) V
GS
)
eN
d
(29)
Or,
V
bi
+ V
DS
(sat) V
GS
=
ea
2
N
d
2
s
= V
p0
(30)
So,
V
DS
(sat) = V
p0
(V
bi
V
GS
) (31)
25
5.3.2 Saturation Current
From Ohms Law, the dierential resistance of the channel at a point x in the channel is
dR =
dx
A(x)
(32)
where p is the resistivity and A(x) is cross sectional area.
Also,
=
1
e
n
N
d
A(x) = (a h(x))W
So,
dR =
dx
e
n
N
d
(a h(x))W
(33)
also
dV = I
D1
dR(x)
where Id1 is the constant current throughout the channel.
So,
dV (x) =
I
D1
dx
e
n
N
d
W(a h(x))
(34)
Idx = e
n
N
d
W(a h(x))dV (x) (35)
Using,
h(x) =
_
2
s
(V (x) + V
bi
V
GS
)
eN
d
_
1/2
Where V(x) is the potential in the chennel due to the drain-to-source voltage. Solving for V(x)
and taking dierential we have,
dV (x) =
eN
d
h(x)dh(x)

s
(36)
and using this in Id1 equation we have
I
D1
=

n
(eN
d
)
2
W

s
[ah(x)dh(x) h(x)
2
dh(x)]
On Integration, we have,
I
D1
=

n
(eN
d
)
2
W

s
_
a
2
(h
2
m
h
2
i
)
1
3
(h
2
m
h
2
i
)
_
(37)
Using
h
2
m
=
2
s
(V
DS
+ V
bi
V
GS
)
eN
d
h
2
i
=
2
s
(V
bi
V
GS
)
eN
d
26
and
V
p0
=
ea
2
N
d
2
s
We can rewrite the above Idq equation as,
I
D1
=

n
(eN
d
)
2
Wa
3
2
s
L
_
_
_
V
DS
V
P0

2
3
_
V
DS
+ V
bi
V
GS
V
p0
_
3/2
+
2
3
_
V
bi
V
GS
V
p0
_
3/2
_
_
_
(38)
we say,
I
P1


n
(eN
d
)
2
Wa
3
6
s
L
as the pinch o current.
And thus we have
I
D1
= I
P1
_
_
_
V
DS
V
P0

2
3
_
V
DS
+ V
bi
V
GS
V
p0
_
3/2
+
2
3
_
V
bi
V
GS
V
p0
_
3/2
_
_
_
(39)
The above equation is valid for
0 |V
GS
| |V

|
and
0 V
DS
V
DS
(sat)
Now, as we have shown earlier that the drain becomes pinched o, for the n-channel MESFET
when
V
DS
= V
DS
(sat) = V
P0
(V
bi
V
GS
)
So, in the saturation region the saturation drain current is determined by using Vds=Vds(sat)
I
D1
= I
D1
(sat) = I
P1
_
1 3
_
V
bi
V
GS
V
p0
__
1
2
3

V
bi
V
GS
V
p0
__
(40)
27
For MESFETs the pincho voltages are known as threshold voltages so we have
V
T
= V
bi
V
P0
So,
V
bi
= V
T
+ V
P0
Using this value in the previous equation we have
I
D1
(sat) = I
P1
_
_
_
1 3
_
1
_
V
GS
V
T
V
p0
__
+ 2
_
1
_
V
GS
V
T
V
p0
__
3/2
_
_
_
(41)
The above equation is valid for Vgs is greater than Vt.
When transistor turns on, we have (Vgs-Vt) is less than Vp0. So, the above equation can be ex-
panded using Taylor Series and we obtain
I
D1
(sat) I
P1
_
3
4
_
V
GS
V
T
V
P0
__
2
(42)
Substituting Ip1 and Vp0 the above equation becomes
I
D1
(sat) =

n
W
2aL
(V
GS
V
T
)
2
(43)
This can further be written as
I
D1
(sat) = k
n
(V
GS
V
T
)
2
(44)
k
n
=

n

s
W
2aL
The factor kn is called conduction parameter. [24,25]
6 Simulations for MOSFET Using COMSOL Multiphysics
R
In order to get a better picture of how the device is going to perform for a given set of device param-
eters, various simulation softwares are available which can help us get a glimpse of how the device
may react on application of certain parameters. This enables us to drive away many of the possible
errors which might come due to petty human error. Such errors can cause huge waste of both human
labour and money. Thus they can increase the manufacturing time signicantly. To overthrow these
sorts of errors, we can rely on good simulation softwares like Comsole Multiphysics
R
.
Now in the following semiconductor device simulation, we will be simulating the DC characteris-
tics of a MOSFET.
6.1 Model Specications
This model calculates the DC characteristics of a MOS (metal-oxide semiconductor) transistor using
standard semiconductor physics. In normal operation, a system turns on a MOS transistor by applying
a voltage to the gate electrode. When the voltage on the drain increases, the drain current also
increases until it reaches saturation. The saturation current depends on the gate voltage.
28
6.1.1 Parameters Used during the Simulation
Vds = 0[V] is the Drain-to-source voltage
Vbs = 0[V] is the Base-to-source voltage
Vgs = 2[V] is the Gate-to-source voltage
phim = 5.0535[V] the Metal work function
Na = 1E17[1/cm3] is the Background doping
Nd = 1E18[1/cm3] is the Maximum donor doping concentration
W mos = 1e-6[m] is the width of MOSFET
h mos = 0.2[um] is the height of MOSFET
L mos = 1[um] is MOSFET length
Lg = 0.24e-6[m] is Gate length
L s = 0.32[um] is Source length
L d = 0.32[um] is Drain length
eps ins = 4.2 is Insulator relative permittivity
d ins = 5E-9[m] is Insulator thickness
6.2 Energy Level Diagram of MOSFET along the center of the device
form the Gate Contact
Form the above gure obtained we can see the energy diagram along the center of the device
from the gate contact for Vgs = 2 V and Vds = 3 V. This gure shows the separation of the quasi-
Fermi levels (size of the depletion region) as well as the inversion region where the intrinsic energy
level crosses the electron quasi-Fermi level (Efn).From this gure one can see the length of inverted
region starting from the surface (y = 0) to the crossing of the intrinsic energy level with the electron
quasi-Fermi level (y approx - 0.03 um).
29
6.3 Current Density of MOSFET and n-channel
The above gure displays the logarithm of the norm of the current density in the device under
the specied conditions. In the gure, one can notice the inverted region that allows the current to
pass between the n-doped regions (drain and source). Here we can clearly see the inverted n-channel
between the two n-doped regions.
30
6.4 Electron Concentration for dierent Drain to Source Voltages
The above gure shows the logarithm of the electron concentration along the inverted channel for
dierent drain-to-source voltages. We can clearly see that a reduction of the electron concentration
near the drain creating the saturation of the drain current. The gure shows the electrons pinched-o
as the charge near the drain end is reduced by the channel potential, i.e. as the drain-to-source voltage
increases.
31
6.5 Surface Charge Density of MOSFET
The gure shows the inverted channel (in blue) as well as the depletion region (orange). The red
regions show the close-to neutral areas. The drain-side of the device shows a larger depletion region
to compensate the vanishing space charge at the pinch-o point. We can see a larger depletion width
on the drain-side of the MOSFET compared to the source side. This is a consequence of the electrons
drawn from the drain-side of the inverted channel.
32
6.6 Current(I
D
) vs Voltage Characteristics(V
DS
)
The above gure shows the I
D
vs V
DS
curve of the device where a the current rises linearly at
low drain-to-source voltage to slowly saturates as the electron concentration vanishes at the pinch-o
point
7 Semiconductor Heterostructures
The previous sections helped us get a good understanding of how the basic transistors work. For
us to understand this we must be clear with what is the physics associated with Semiconductor
Heterostructures.
7.1 Heterojunctions
Heterojunctions are two types
Iso type Conductivity types are same bot n or both p.
Aniso type Conductivity types are dierent. One is p and other is n.
33
For Heterostructures to form, both the joining materials must have similar thermal properties. So
that while operation one might not crack over the operation of other.
For Heterojunction the most important requirement is that their lattice should match. If not so,
we will have defective regions. And if the interface is defected we have mobility eects which is not
permissible. Similar to what happens during high doping AlGaN/GaN have perfect match of lattice.
Homojunctions are formed from same material doped dierently. Heterojunctions are formed by
dierent materials which may or may not be dierently doped.
7.1.1 Band Diagram
Let us consider two semiconductors with X1 is greater than X2; Eg1 is smaller than Eg2; then
34
So, here we see a discontinuity in the energy band diagram. This 5 or notch is caused due to
Notch Eect which is explained below.
7.1.2 Understanding the Notch Eect
Let us rst take the example of homojunction. The diagram below describes p-n homojunction (The
conduction band only)
The e- will keep transferring from the n side to p side till the maximum energy of e- concentration
on both sides are equal. And that is when we get the built in potential.
Now in Heterojunction
If notch was not to be considered then only Vbi would have been the built in potential but due
to the notch the built in potential is actually increased.
So, band bending in case of heterojunction is more that that of homojunction. Now, as were the
case in MOSFET here too due to the extra band bending the notch might cross the fermi level and
35
so the portion lower to the fermi level will be accumulated of e-. So, basically we try to maximize the
delta Ec. So that e- concentration in unbiased condition increases.
Due to the discontinuity delta Ec in the conduction band of AlGaAs and GaAs, the band bend-
ing in the undoped GaAs is more than in the GaAs homojunctions of similar doping levels.
Due to this eect, large concentration of e- are present at the GaAs surface adjacent to AlGaAs and
they remain there due to the notch in the cconduction band.
The e- have been supplied form the doped layer to undoped region (or where doping concentration is
low) as a result ionied impurity scattering eect is absent.
8 HEMT Working Principle
The HEMT is a peculiar device, since it can oer optimal characteristics in terms of both high
voltage, high-power and high frequency operation. Its operation principle is founded on the presence
of the 2DEG at interface of an heterostructure, like for example an AlGaN/GaN system. It is a
three terminal device where the current between the two Ohmic contacts of source and drain, owing
through the 2DEG, is controlled by the electrode of gate (typically a Schottky contact). Practically,
the bias applied to the gate controls the ow of electrons through the channel. The gure shows a
schematic of an HEMT device. To conne the electron ow in the 2DEG and isolate HEMT devices,
deep trenches (cutting the 2DEG) or ion implantation are typically used.
36
The below gure illustrates, in a schematic band diagram of an AlGaN/GaN HEMT structure,
how the 2DEG is inuenced by the dierent gate bias conditions. This schematic is reported for the
case of a n-type doped AlGaN barrier layer. At Vg = 0V there are allowed levels below the Fermi level
in the subbands of the quantum well, resulting in the presence of a high sheet carrier concentration
and in the on-state of the device. By increasing the gate bias (Vg is greater than 0 V), the Fermi
level rises, increasing the density of allowed states below the Fermi level in the conductive band,
and therefore increasing the sheet carrier concentration of the 2DEG. By decreasing the gate bias
V towards negative values (V is less than 0 V) the Fermi level drops depleting the 2DEG, until the
position of the Fermi level lies below the quantum well Under this condition, the level in the energy
subbands are completely empty and the device is in the o-state.
In the following gure we can see the Ids vs Vds characteristics of a HEMT. In the Ids-Vds
characteristics by applying a positive potential dierence between source and drain (Vds ), the current
will start to ow in the 2DEG. By increasing the drain bias, the current ow in the channel will increase
linearly up to certain value. After this value the current through the channel starts to saturate. The
maximum saturation value Idss depends on the sheet carrier concentration n of the channel. Looking
at the trans characteristics, for a xed Vds the drain current I rises with a parabolic behaviour with
increasing gate bias.
37
The drain current(Ids) can be controlled by the bias applied to the gate electrode. In particular,
Ids decreases with increasing the negative value of the gate bias (Vg), since the region of the channel
under the gate is depleted. The value of Vg which determines the pinch-o of the channel (where the
sheet carrier concentration in the channel becomes zero) is called threshold voltage (Vth) of the device.
In a AlGaN/GaN HEMT at any point x along the channel, neglecting the extrinsic series resistance
of source and drain, the sheet carrier concentration depend by the applied Vg
n
s
(x) =

0

AlGaN
qd
AlGaN
[V g V th V (x)] (45)
where d
AlGaN
is the distance of the gate to the 2DEG channel, corresponding to the AlGaN
thickness. The gate-to-channel capacitance (per unit of area) can be approximately assumed as
independent of n
s
using the expression C
2DEG
=
0

AlGaN
/qd
AlGaN
.
It is now possible to dene the threshold voltage of the device, as the gate bias necessary to turn-o
the 2DEG, resulting in a n
s
=0. Looking at the AlGaN/GaN schematic band diagram showed in above
gure, it is clear that the threshold voltage depends on dierent parameters like the Schottky barrier
height
B
, the conduction band oset at the AlGaN/GaN interface delta Ec, the concentration of
38
donor atoms in the AlGaN layer N
D
, the relative dielectric constant
AlGaN
, the thickness d
AlGaN
and
the Al concentration of the AlGaN. Besides these parameters, in order to have a complete expression
of the threshold voltage the contribution of the polarization induced charge density must be taken
into account. Thus simply the threshold voltage can be expressed as
V
th
=
B
E
C

qN
D
AlGaN
d
2
AlGaN
2
0

AlGaN

d
AlGaN
(46)
Assuming a constant mobility and remembering the Ohmic law, for a two-dimensional electron
gas the conductivity of the channel will be directly proportional to the sheet carrier concentration
n
s
and to the electrons mobility in the channel
= q.n
s
. (47)
It is possible to write the drain current as:
I
D
= .W.Q(x)
dV (x)
dx
(48)
where Q(x) is the charge considered in the channel. Integrating both sides in the all length of the
channel and considering the expression of Q(x) we have
I
D
= .
W
L
.C
2DEG
_
V
g
V
th

V
DS
2
_
V
DS
(49)
The drain current of a HEMT in linear region is often expressed in a form similar to that used for
a MOSFET, i.e., :
I
D
= .
W
L
.C
2DEG
_
V
g
V
th

V
DS
2
_
V
DS
(50)
Increasing V
DS
upto certain value called V
DSsat
, the drain current I
D
is constant and so the derivate
of I
D
will be zero.
dI
D
dV
DS
= q
W
L
C
2DEG
(V
g
V
th
V
DS
) = 0 (51)
and V
DSsat
is given by
V
DSsat
= V
g
V
th
(52)
At bias condition of V
DSsat
the I
D
will be expressed as
I
DSS
=
1
2

W
L
C
2DEG
(V
g
V
th
)
2
(53)
The above equation is approximation valid for long channel devices. However, for HEMTs with
a short gate length (l is less than 10 um) the electron transport occurs under high electric elds and
the expression of the saturation current is dierent. If the electric elds exceeds a certain critical
value, the speed of the electrons in the 2DEG begins to saturate. Taking into account the eects of
the saturation velocity model the saturation current is expressed as
I
DSS
= q.n
s
.v
sat
(54)
Considering the expression of the drain current, it is also possible to dene the transconductance
of the device as the change in drain current I
D
resulting from a variation of gate voltage V
g
for a xed
V
DS
:
g
m
=
I
D
V
g
(55)
39
at constant V
DS
.
Similarly the output conductance of the device is dened as the I
D
response to a V
DS
variation
for a xer gate bias V
g
g
d
=
I
D
V
DS
(56)
at constant V
g
. [26]
9 Conclusion
As the need for power is ever growing, present technology based on Si is not able to provide sucient
power handling and high frequency operation capabilities. Therefore better alternative to the domi-
nant Si based technology has to be looked forward to. For this purpose we explored some alternatives
and came to the conclusion that WBG semiconductor like GaN can help us in this deed.
We saw that 2DEG which is formed at the interface of AlGaN/GaN heterostructure can be used
to make high power, high frequency transistors, based on HEMT principle. In order to understand
about the working principle of HEMTs we went to the basics and started with BJT then proceeding
to MOSFETs, MESFETs and nally came to HEMT. On understanding the working principles of all
the devices we came to know that in both MOSFET and HEMT 2DEG are formed, but because in
MOSFET the 2DEG is formed in the doped region, the scattering there is high, which results in lower
saturation velocity. Where as in HEMTs the 2DEG is formed in the undoped region which leads to
lack of ion scattering and thus the mobility can be raised highly. Although the low eld mobility
of GaN is lower as compared to other III-IV materials, but the high saturation velocity and higher
bandgap makes it ideal candidate for high power and high frequency device.
We compared the various gure of merits for dierent capable contenders and again came to the
conclusion that our choice for GaN was the best. Further the possibility increasing the polariza-
tion (piezoelectric polarization) by inducing stress or strain on the material makes route for further
possibilities which can help in tweaking or enhancing the properties of AlGaN/GaN HEMTs.
40
10 References
[1] J. Millman and C.C. Halkias in Electronic Devices and Circuits, Tata McGrawHill
Publishing Company Ltd., New Delhi, 1991.
[2] http://www.cree.com/LED-Chips-and-Materials
[3] Wide Energy Bandgap Electronic Devices Fan Ren and John C. Zolper (Pg. 13 15)
(Book Source: books.google.co.in/books?isbn=9812382461)
[4] http://www.ioffe.ru/SVA/NSM/Semicond/GaN/bandstr.html
[5] GaN-Based RF Power Devices and Amplifiers By Umesh K. Mishra, Fellow IEEE,
Likun Shen, Thomas E. Kazior, and Yi-Feng Wu. (Source: http://ieeexplore.ieee.org/
stamp/stamp.jsp?tp=&arnumber=4414367)
[6] Power Electronic Europe Issue 4, 28 (2010), Alberto Guerra and Jason Zhang,
International Rectifier, El Segundo, USA (Source: http://www.power-mag.com/pdf/
feature_pdf/1283339996_IR_Feature_Layout_1.pdf)
[7] Growth of GaN on SiC(0001) by Molecular Beam Epitaxy
(Source: http://onlinelibrary.wiley.com/doi/
10.1002/1521-396X(200112)188:2%3C595::AID-PSSA595%3E3.0.CO;2-S/pdf)
[8] High quality AIN and GaN epilayers grown on (00-1) sapphire (100) and (111)
silicon substrates (Source: http://scitation.aip.org/content/aip/journal/apl/
66/22/10.1063/1.114242)
[9] Effect of Si doping on strain, cracking, and microstructure in GaN thin films
grown by metalorganic chemical vapor deposition (Source: http://scitation.aip.org/
content/aip/journal/jap/87/11/10.1063/1.373529)
[10] Metalorganic Chemical Vapor Phase Epitaxy of Crack-Free GaN on Si (111)
Exceeding 1 m in Thickness (Source: http://iopscience.iop.org/1347-4065/39/11B/
L1183)
[11] Metalorganic chemical vapor deposition of GaN on Si111: Stress control and
application to field-effect transistors (Source: http://scitation.aip.org/content/
aip/journal/jap/89/12/10.1063/1.1372160)
[12] The nature of arsenic incorporation in GaN (Source: http://scitation.aip.org/
content/aip/journal/apl/79/20/10.1063/1.1418030)
[13] AlGaN/GaN High Electron Mobility Transistors Grown on 150 mm Si(111)
Substrates with High Uniformity (Source: http://iopscience.iop.org/
1347-4065/47/3R/1553)
41
[14] Influence of sapphire nitridation on properties of gallium nitride grown by
metalorganic chemical vapor deposition (Source: http://scitation.aip.org/content/
aip/journal/apl/68/11/10.1063/1.115687)
[15] GaN-Based Devices on Si (Source: http://onlinelibrary.wiley.com/doi/
10.1002/1521-396X(200212)194:2%3C361::AID-PSSA361%3E3.0.CO;2-R/pdf)
[16] Effect of the N/Al ratio of AlN buffer on the crystal properties and stress
state of GaN film grown on Si(111) substrate (Source: http://
www.sciencedirect.com/science/article/pii/S0022024803017433)
[17] Activation energies of Si donors in GaN (Source: http://scitation.aip.org/
content/aip/journal/apl/68/22/10.1063/1.115805)
[18] Thick GaN Epitaxial Growth with Low Dislocation Density by Hydride Vapor Phase
Epitaxy (Source: http://iopscience.iop.org/1347-4065/36/7B/L899)
[19] Wide Energy Bandgap Electronic Devices Fan Ren and John C. Zolper (Book
Source: books.google.co.in/books?isbn=9812382461)
[20] Two-dimensional electron gases induced by spontaneous and piezoelectric
polarization charges in N- and Ga-face AlGaN/GaN heterostructures (http://
scitation.aip.org/content/aip/journal/jap/85/6/10.1063/1.369664)
[21] Optical constants of epitaxial AlGaN films and their temperature dependence
(Source: http://scitation.aip.org/content/aip/journal/jap/82/10/10.1063/1.366309)
[22] Polarization-induced electron populations (Source: http://scitation.aip.org/
content/aip/journal/apl/77/7/10.1063/1.1288817)
[23] Spontaneous polarization and piezoelectric constants of III-V nitrides
(Source: http://journals.aps.org/prb/pdf/10.1103/PhysRevB.56.R10024)
[24] Semiconductor Physics and Devices: Basic Principles - Donald A. Neaman (Book)
[25] NPTEL - Electronics and Communication Engineering - High Speed Devices and
Circuits (Video Lectures form IIT Madras) (Source: http://nptel.ac.in/video.php?
subjectId=117106089)
[26] AlGaN/GaN heterostructures for enhancement mode transistors (Source: http://
dspace.unict.it/bitstream/10761/1347/1/GRCGPP82S30C35MA-Giuseppe%20Greco-%
20AlGaN-GaN%20heterostructures%20for%20enhancement%20mode%20transistors.pdf
42
11 Important Points
Si based technology has lived through its glorious period and because of its low power handling
capacity, its low band gap and intrinsic carrier concentration of the material it has a to have a
junction temperature less than 200 C to work properly.
WBG semiconductors like SiC, GaN etc. overcome the above short comes of Si and can copeup
with increased power, frequency and operating temperatures. SiC technology is the most ad-
vance amongst all the WBG SCs but it too suers from micropipes crystal defect.
GaN and its alloys have not been able to advance as much as SiC because of various physical
issues realted to their surface and interface. Also, because of the lack of free standing GaN
substrate heteroepitaxy has to be performed on substrates like Al2O3, SiC or Si.
Although the bulk mobility of GaN at low elds is much lower than that of other III-IV materials,
but the high saturation velocity and higher band gap makes it an ideal contender for high
frequency power devices. Also ability to form heterostructures of AlGaN/GaN leads to the
formation of 2DEG at the interface which can be further used in the fabrication of HEMT
devices.
2DEG is due to the formation of both spontaneous and piezoelectric polarization. Which can
be used to make HEMTs.
There are three possibilities for III-Nitrides namely zinc-blend, wurtzite and rock salt, of which
GaN exists in the wurtzite form which is thermodynamically most stable at room temperature
and pressure.
Because of the absence of inversion symmetry along the c-axis it is easily possible to distinguish
between dierent orientations of GaN namely G-faced and N-faced, depending if the material
is grown with Ga on top or N on top corresponding to (0001) and (0001-) crystalline faces.
There exists a polarization in the GaN crystal due to the dierence in the electronegativity
of the atoms. This leads to a polarization known as spontaneous polarization which in turn
depends on the c/a ratio. The less the ratio the more the polarization.
There exists another polarization because of the induced stress and strain which leads to change
in the c/a ratio. This becomes considerably important in AlGaN/GaN heterostructres for 2DEG.
The wide band gap of GaN (3.30 eV) is responsible for the high value of critical electric eld
(3.3 MV/cm) which is order of magnitude higher than that of Si. The high critical electric eld
makes it suitable for high voltage devices.
Also, because of low intrinsic electron concentration the maximum operation temperature can
be made to rise without leading to rise in leakage current.
The relatively good value of relative permittivity makes it a good contender for capacitive loading
of transistor and passive components. Also, the thermal conductivity value being almost equal
to that of Si makes it capable to better transfer the heat produced and thus not making it to
degrade at high temperature working conditions.
Also, its capability to form AlGaN alloys makes the band gap alteration easy which comes handy
for the formation of heterostructures needed for the formation of 2DEG.
43
In principle the on state resistance at a given voltage of GaN can outperform competing Si and
SiC devices. Which will reduce the power losses.
In order to better compare the power electronic performance for dierent semiconductor ma-
terials gure of merit are employed. In particular for high power and high frequency JFOM is
an indication of the maximum capability to energize carriers by electric eld, BFOM measures
the minimum conduction losses during DC operation and BHFOM give information about the
minimum conduction losses during high frequency operation. On comparing these values with
Si and SiC we can easily come to the conclusion that GaN is a better choice.
The inability to form free standing GaN substrate, other materials must be used as substrate
for the growth of GaN.
Hexagonal silicon carbide (6H-SiC) can be used as a substrate as the lattice mismatch for (0001)
oriented GaN lms is small and the thermal conductivity is higher. But the catch here is the
high defect density (10
7
cm
-2
for screw dislocations and 10
9
cm
-2
for edge dislocation density).
Si can also be used as a substrate. But owing to large lattice and thermal expansion coecient
mismatch can lead to defects and cracks on the material.
To relieve the tensile stress and achieve crack-free GaN heterostructures, several kind of tran-
sition layers can be used, such as low temperature AlN, graded AlGaN buers or AlGaN/GaN
superlattices.
MOCVD is now the most popular method to grow GaN owing to its good quality and reasonable
growth rates.
By the formation of 2DEG at the junction interface we have a pool of electrons captured in the
quantum well which is free to ow in the plane parallel to the junction interface.
The benet of 2DEG is that because of the absence of ionic counterparts in the well the scattering
is reduced signicantly which contributes to better eciency.
It is because of the piezoelectric polarization the formation of high sheet charge density is
possible even without using undoped layers.
Saturation of current in BJT (NPN) happens when base is higher than emitter, but collector is
not higher than base.
Saturation in case of MOSFET happens when the width of the channel below the oxide layer
near the collector becomes zero, this width is related to Vds and Vgs.
Saturation current in case of MESFET is because of the depletion of n-channel below the Gate,
where on the application of required voltage Vgs the width of n-channel can be controlled. Also
the potential dierence Vds causes the n-channel to pinch o near the drain.
It is the formation of Notch near at the interface of heterojunction which acts as the place for
formation of 2DEG.
44