MODELING OF METAL-OXIDE-SEMICONDUCTOR CAPACITOR ON INDIUM GALLIUM

NITRIDE
1- CHANNEL MODEL
Tarik Menkad 1, student member, IEEE, Dimiter Alexandrov 1,2, and Kenneth Scott A. Butcher 1,2
1

Semiconductor Laboratory, Department of Electrical Engineering, Lakehead University, 955 Oliver Rd, Thunder Bay P7B 5E1, Ontario,
Canada.
2
Meaglow Ltd, 1294 Balmoral St, Suite 150, Thunder Bay P7B 5Z5, Ontario, Canada

2. MODEL

ABSTRACT
A new analytical model for a two terminal metal-oxideGallium Nitride/Indium Gallium Nitride heterojunction structure is
presented. This model characterizes the space charge layer created
by electron tunneling in the structures channel which is made of
intrinsic Gallium Nitride. A one dimensional (1-D) analysis is
adopted, and a set of hypotheses is stated to frame the present
work.

2.1. Nomenclature
,

Index Terms Gallium Nitride GaN, Indium Gallium

Nitride In0.5Ga0.5N, excitons.

,
.

1. INTRODUCTION
Excitons found in InxGa1-xN alloy are used as a basis for the
design of a high frequency field effect transistor [1]. This transistor
uses excitons as a quantum source of free electrons for its channel
made of intrinsic Gallium Nitride. Electrons, under the influence of
an external electric field, move in an intrinsic material and
experience reduced scattering. The high mobility in the channel
allows low-noise and high-speed performance. The metal- oxidesemiconductor on In0.5Ga0.5N capacitor, which forms the core of
this high-frequency field-effect transistor, is the subject of the
present work. Quasi-static C-V characteristic are obtained. For this
purpose, a set of hypotheses are stated in order to set the scope and
the domain for the validity of this model.
Channel conductivity is modulated by the value of the free
electric charge present in the intrinsic GaN material. This charge is
due to electrons of excitons origin, which tunnel through the
i-GaN/p-In0.5Ga0.5N heterojunction. The determination of the
charge density distribution, the electric field through the channel
thickness, and the potential drop across the channel are essential to
relate the electric charge present on the gate to the corresponding
applied gate voltage. For this purpose, a channel model is
presented. The reader will realize that such a model is a standalone
entity and can be introduced separately from the device model, we
chose to do it the other way to demonstrate the need for such a
channel model in the analysis of the MOS capacitor on In0.5Ga0.5N.

,
,
.

Front gate to back substrate voltage

Potential drop across In0.5Ga0.5N pn junction
and across the space-charge layer in p-type
In0.5Ga0.5N side of the i-GaN/p-In0.5Ga0.5N
heterojunction
Potential drop across i-GaN, oxide layers
Oxide and GaN layers thicknesses.
Charge per unit area on the gate, in GaN layer
Charge per unit area in space charge region
located
in
p-type
In0.5Ga0.5N
side of the i-GaN/p-In0.5Ga0.5N heterojunction.
Free electron concentration present in intrinsic
Gallium Nitride layer.
Electric field in GaN layer at the i-GaN/pIn0.5Ga0.5N heterojunction.
Thermal equilibrium hole and electron
concentrations in p-In0.5Ga0.5N.
Relative permittivity of oxide and GaN
Relative permittivity of In0.5Ga0.5N, and
permittivity of vacuum.
Thermal voltage
Elementary charge

2.2. Set of hypotheses

1) Consider a 1-D dimension model.
2) The oxide material is considered an ideal insulator, no trapped
charges exist.
3) The GaN surface at the interface oxide/GaN is considered to be
ideal, with no defects or traps or recombination centers near the
surface.
4) All semiconductors are considered to be non degenerate, so that
Boltzmann statistics can be used.
5) No temperature gradient is present. The device dimensions are
small enough, that we assume that heat distributes evenly
throughout the device.
6) Polarization is not considered; so that displacement and electric
vector fields are directly proportional.

2012 25th IEEE Canadian Conference on Electrical and Computer Engineering (CCECE)
978-1-4673-1433-6/12/$31.00 2012 IEEE

7) Both the oxide and the GaN have the same work function. No
built in potential exists at the oxide/GaN interface
8) Charge neutrality condition is assumed.
9) The existence of quasi-neutral regions in p-type and n-type
In0.5Ga0.5N layers is considered.
10) The i-GaN/p-In0.5Ga0.5N heterojunction is approximated by an
abrupt junction.
11) Complete ionization occurs in doped semiconductors. Device
is operating at room temperature.
12) The gate voltage varies slowly enough, so we can always
consider a zero-valued gate current.
13) The GaN material is considered to be depleted from free
charge carriers.

exp

(1)
.

(2)
(3)

(4)

(5)

i-GaN layer

p-In0.5Ga0.5N

n-In0.5Ga0.5N

(8)
"

Where
2

. ,

0 ,

exp

is the efficiency of the electron injection through the pn

1 ,
is a coefficient that depends on
homojunction 0
the lifetime of injected electrons in p-In0.5Ga0.5N 0
1 ,
is the
creation efficiency of exciton current
0
1 , " is the concentration of electrons occupying
"

the state in non equilibrium,

0 is the position of the
hydrogen like electron energy level
for
0, and is the
potential barrier width for level
[1]. Two equations of four
are obtained; first equation is given in
unknowns
, , ,
(1), and second equation is obtained by substituting the electric
field
and the average concentration
by theirs expressions
given in (7) and (8) respectively in (6).
is
, , ,
A system of two equations of four unknowns
not sufficient to solve for
given a value for the
, , ,
gate voltage
. Two more equations are needed. The proposed
GaN channel model will fulfill this need.

Gate

Oxide

The average value of free electron concentration expression was

obtained in terms of the potential drop across In0.5Ga0.5N
pn junction and the electric field in GaN layer at the
i-GaN/p-In0.5Ga0.5N heterojunction [1]. A good approximation of
this value is given by the expression (8)
.

A first set of relations, valid under any case of operation, is derived

from global equilibrium and Gausss Law:

1
(7)

exp

2.3. Analysis
Figure 1 illustrates a cross-section of the metal-oxidesemiconductor structure including the referenced potential drops
across regions of interest. The gate voltage breaks down to four
,
,
, and
.
potential drops

is given

at the semiconductor heterojunction. The electric field

as,

VGg

Figure 2: Block diagram of GaN Channel model

Figure 1: A cross-sectional schematics of MOS heterojunction

capacitor, with the corresponding referenced potential drops across
the regions of interest.
Relations (2)-(5) are grouped in one relation.

Figure 2 represents a block diagram of the proposed GaN channel

,
model. Average value of free electrons volume concentration
initial value of electric field , and GaN channel thickness
,
are inputs. Charge concentration, potential drop, and electric field
profiles are outputs. Application of the (1-D) Poisson equation on
the GaN channel, as indicated in Figure 3, results in the following
initial condition problem [2][3].

(6)
to potential drops, (1-D) Poisson
To relate the electric field
equation is solved for the space-charge region in p-type In0.5Ga0.5N

0
0

, where

exp

(9)

The value of volume concentration at

0, or at
0, is
referred to as , this value is generally different from the intrinsic
free carrier concentration . The solution of the differential equation
(9) gives the expression of the potential drop
at a point

(10),

x-axis

thGaN

(case1)
and
arctan

arctan

Figure 3: A cross-sectional schematics of the intrinsic GaN layer,

with the corresponding referenced potential drop, and electric field.

arctan

(11)

0,1
(case2)

2
2

The quantity

0,
1
2

0,

arctan

0,

arctan

The free electron concentration, the potential drop, and the electric
field
are
presented
as
functions
of
four
. The average concentration of free
variables ,
,
,
electrons is used to compute the free electron concentration initial
value. The initial value
is obtained by solving, for an
appropriate choice of the unknown, the following equation

2
1

0,

Where
1

and
1

represents a normalizing length, and

positive constant.
A closer look at the case corresponding to

0,

as case 2.

(12)

by its expression,
Substituting the free electron concentration
got from first equation in set (10) or set (11) according to the case
considered, in (12) we obtain;
1

(case1)

(13)

(case2)

(14)

is a dimensionless
and

1 shows that the

Using the two bijections

1

for case1,
1,

0,

and

as case 1, and to:

0

initial value of electron concentration is entirely determined by the

electric field initial value. In other words, regardless of the value of
, the initial value
is pinned. This
the average concentration
is incorrect since, two different values of the average concentration
lead to two different initial values under the same conditions. This
case is, thus, discarded.
, voltage , and electric field
at
The electron concentration
any point in the GaN layer are derived. From now on, we refer to:

is the normalized abscissa, it is defined as

0,

0,1
arctan

0, ,

for case2,

the following variable change is justified

Start

(case1)

Get (F0, nch,ave, thGaN)

(case2)

Computation of
parameters a and b

Expressions (13) and (14) can be put under the forms

,

0,

(Case1)

Yes

(15)

Case 1
Yes

(Case2)
0, arctan

No
No

a(b+1)=2

(16)
Cannot provide a solution
for given (F0, nch,ave, thGaN)

0,

a>=2

Yes

No

a(b+1)>2

Case 1

Case 2
Case
1(2)

End

Equation (12), put under form (15) or (16), is solved numerically

for the unknown using an iterative method. One can prove that a
solution, different from the trivial one
0, exists and is unique if
constants ,
satisfy one of the mutually exclusive following
conditions.
Condition1:

(Case 1)

Condition2:

(Case 2)

Constants ,
are expressed in terms of the electric field initial
value, the average value of free electron concentration, and the
Gallium Nitride layer thickness. The product
1 is directly
obtained from the input
,
,
, and indicates which case
(1 or 2) is taken.
Suppose that Equation (12), put under form (15) or (16), has a
solution , then the initial value of free electron concentration
is computed using reciprocals of bijections introduced earlier to
is given as,
justify the variable change. Initial value
1

(Case 1)

(17)

(Case 2)

(18)

that satisfies condition 1(2),

For an input
, ,
Equation (12), put under form (15)((16)) is solved for the unknown
. The electron concentration initial value
is evaluated using
(17)((18)). The electron concentration, potential drop, and electric
field profiles can be computed using Equation set (10)((11)). The
steps followed to calculate the free electron concentration, the
, ,
starting from the
potential drop, and the electric field
, are shown by the flowchart in Figure 4.
input
, ,

Solve equation
and compute value n0
Compute parameters
d, d2, and Ld

Yes

thGaN within range

No

Compute and Plot (n,v,F)

Cannot compute (n, v, F)

for this thickness value
thGaN

End

End

Figure 4: Solution flowchart for proposed intrinsic GaN Channel

model

3. DISCUSSION
For a given GaN layer thickness, here we considered
thGaN=50nm as in [1]. Constants a and b depend on the electric
field initial value F0, and on the free electron concentration average
value
. To emphasize the difference between case 1 and case 2,
consider the following practical situations. First, the GaN layer is
heavily charged, and under the influence of an electric field whose
initial value is low. Second, the GaN layer is lightly charged, and
under the influence of an electric field whose initial value is high.
The first situation satisfies condition 2 and thus falls under case2,
while the second satisfies condition 1 and thus falls under case1.
Condition1 is always satisfied if constant
2. This corresponds
to situations where
1.04 10 V/cm at room temperature for
a GaN layer thickness of 50nm, regardless of the charge density
value.

F in V/cm

0
0.2
0.4 0.6
0.8
1
Electric field for nch ave=1e17cm 3, F0=1e-2,1e-1V/cm, and thGaN=50nm
5
10
10

10

-5

v in mV

0
0.2
0.4 0.6
0.8
1
Voltage diff. for nch ave=1e17cm 3, F0=1e-2,1e-1V/cm, and thGaN=50nm
100
50
0
0

0.2

0.4
0.6
Normalized abscissa

0.8

Figure 5: Free electron concentration, electric field, and potential

drop are plotted as functions of normalized abscissa for inputs
= (1e17,1e-2,50), and
= (1e17,1e, ,
, ,
1,50).
-

F in V/cm

n in cm-3

Free elec. conc. n for nch ave=1e18cm 3, F0=1e3,1e4 V/cm, and thGaN=50nm
20
10
18
10
16
10
0
0.2
0.4
0.6
0.8
1
Electric field F for nch ave=1e18cm-3 and F0=1e3,1e4V/cm, and thgan=50nm
6
10
4

10

F0=1e3 V/cm
F0=1e4 V/cm

10

v in mV

0
0.2
0.4
0.6
0.8
1
Voltage diff. v for nch ave=1e18cm-3,F0=1e3,1e4V/cm, and thGan=50nm
400
200
0
0

0.2

0.4
0.6
Normalized abscissa

0.8

Figure 6:. Free electron concentration, electric field, and potential

drop are plotted as functions of normalized abscissa for inputs
=(1e18,1e3,50),and
=(1e18,1e4,50)
, ,
, ,

n in cm-3

10

20
10

0
0
0.2
0.4
0.6
0.8
1
Voltage diff. for nch ave=1e18cm-3, F0=5e4,1e5V/cm, and thGaN=50nm
1000

v in mV

F in V/cm

10
0
0.2
0.4
0.6
0.8
1
Electric field F for nch ave=1e18cm-3, F0=5e4,1e5V/cm, and thGaN=50nm
6
x 10
2
F0=5e4
1
F0=1e5

500
0
0

0.2

0.4
0.6
Normalized abscissa

0.8

Figure 7: Free electron concentration, electric field, and potential

drop are plotted as functions of normalized abscissa for inputs
=(1e18,5e4,50),and
=(1e18,1e5,50)
, ,
, ,
Free elec. conc. for nch ave=1e14-15-16-17cm-3,F0=5e2V/cm,thGaN=50nm
20
10
nch ave=1e17
nch ave=1e16
15
10
nch ave=1e15
nch ave=1e14
10
10
0
0.2
0.4
0.6
0.8
1
Electric field for nch ave=1e14-15-16-17cm-3,F0=5e2V/cm,thGaN=50nm
6
10
n in cm-3

n in cm-3

10

16

Free elec. conc. n for nch ave=1e18cm-3,F0=5e4,1e5V/cm, and thGaN=50nm

30
10

F in V/cm

Free elec.conc n for nch ave=1e17cm 3, F0=1e-2,1e-1V/cm, and thGaN=50nm

18
10
F0=1e-2 V/cm
17
10
F0=1e-1 V/cm

Another trend is noticed; for a given value of the initial electric

field, the charge density distribution gets steeper as the average
concentration value increases when the GaN layer thickness is kept
constant. The concentration density varies from 9.49*1013cm-3 to
1.06*1014cm-3 for
=1014cm-3 and from 2.09*1016cm-3 to
9.50*1017cm-3 for
=1017cm-3 under the same condition of
F0=5*102V/cm (Figure 8). Figure 9 illustrates the effects of
varying the Initial value of electric field together with the average
concentration on charge density distribution, electric field, and
potential drop profiles for the same GaN thickness of 50nm.

10
10

4
2

0
0.2
0.4
0.6
0.8
1
Voltage diff. for nch ave=1e14-15-16-17cm-3,F0=5e2V/cm,thGaN=50nm
100
v in mV

Figure 5 shows the charge density distribution, electric field,

and potential drop across a GaN layer of thickness 50nm for an
average carrier concentration of 1017cm-3. Although the electric
field initial value increased by one order of magnitude from
10-2V/cm to 10-1V/cm, no noticeable difference in the charge
density and potential drop was observed. This comes from the fact
that the electric field initial values are too small to have noticeable
distinct effects on the charge density distribution which varies from
2.18*1016 cm-3 to a final value of 9.42*1017cm-3. When the electric
field initial value is high, it pushes free charge carriers away from
the corresponding GaN layer side, increasing the charge depletion.
On the other side, an accumulation region is created. For strong
initial fields, a change by one order of magnitude in the value of
initial electric field can result in a dramatic difference in the charge
distribution profile, as well as for the potential drop across the GaN
layer. Figures 6 and 7 illustrate this case; initial concentration
density changes by almost six orders of magnitude from
1.09*1016cm-3 for F0=104V/cm to 1.35*1010cm-3 for F0=105V/cm.
The potential drop across the GaN layer varies from 235mV for
F0=104V/cm to 600mV for F0=105V/cm.

-100
0

0.2

0.4
0.6
Normalized abscissa

0.8

Figure 8: Free electron concentration, electric field, and potential

drop are plotted as functions of normalized abscissa for following
= (1e14-15-16-17, 5e2, 50).
, ,
inputs
To this point, the Gallium Nitride layer thickness was kept
constant, however the analysis allows us to consider different
thickness values. Changing its value allows us to find the
maximum value before breakdown for predetermined conditions of
=1016cm-3 and an
operation. For an average concentration
4
electric field initial value F0=10 V/cm, the charge density
distribution varies from 5.52*1015cm-3 to a final value of
1.75*1016cm-3 for a thickness of 25nm and from 1.83*1014cm-3 to

Free elec. conc. for nch ave=1e14-15-16cm-3, F0=5e4,1e5V/cm, thGaN=50nm

20
10

elec. conc. for nch ave=1e16cm-3,F0=1e4V/cm,thGaN=25-50-75-100-125nm

18
10

10
n in cm-3

1.06*1017cm-3 for a thickness of 125nm (Figure 10a). The potential

drop across the GaN layer varies from 29.82mV for 25nm to
164.60mV for 125nm (Figure 10c). The electric field final value is
around 1.48*104V/cm for 25nm and 3.38*104V/cm for 125nm
(Figure 10b).

10

Concentration n in cm-3

10
10

10

16

thGaN=25
thGaN=50
thGaN=75
thGaN=100
thGaN=125

15

15

10

10

17

(nch ave,F0)=(1e14,5e4)
(nch ave,F0)=(1e14,1e5)
(nch ave,F0)=(1e15,5e4)
(nch ave,F0)=(1e15,1e5)
(nch ave,F0)=(1e16,5e4)
(nch ave,F0)=(1e16,1e5)

10

0.2

0.4
0.6
Normalized abscissa

0.8

14

0.2

0.8

(a)
Elec.field for nch ave=1e16cm-3,F0=1e4V/cm,thGaN=25-50-75-100-125nm
4
x 10
3.5
3

(a)

thGaN=25
thGaN=50
thGaN=75
thGaN=100
thGaN=125

F in V/cm

2.5

Electric field for nch ave=1e14-15-16 cm-3, F0=5e4,1e5V/cm, thGaN=50nm

4
x 10
11

0.4
0.6
Normalized abscissa

2
1.5

10

F in V/cm

(nch ave,F0)=(1e14,5e4)
(nch ave,F0)=(1e14,1e5)
(nch ave,F0)=(1e15,5e4)
(nch ave,F0)=(1e15,1e5)
(nch ave,F0)=(1e16,5e4)
(nch ave,F0)=(1e16,1e5)

9
8
7

0.5
0

0.2

0.4
0.6
Normalized abscissa

0.8

150

200

v in mV

Volt diff. for nch ave=1.0e14-15-16cm-3, F0=5e4,1e5V/cm, and thGaN=50nm

600
(nch ave,F0)=(1e14,5e4)
(nch ave,F0)=(1e14,1e5)
500
(nch ave,F0)=(1e15,5e4)
(nch ave,F0)=(1e15,1e5)
400
(nch ave,F0)=(1e16,5e4)
(nch ave,F0)=(1e16.1e5)
300
v in mV

0.8

Volt. diff. for nch ave=1.0e16cm-3,F0=1e4V/cm,thGaN=25-50-75-100-125nm

200

(b)

thGaN=25
thGaN=50
thGaN=75
thGaN=100
thGaN=125

100

50

0
0

0.2

0.4
0.6
Normalized abscissa

0.8

(c)
Figure10: (a) Free electron concentration, (b) electric field, and (c)
=(1e16,1e4,25-50-75-100, ,
potential drop for inputs
125).

100
0
-100
0

0.4
0.6
Normalized abscissa

(b)

6
5
0

0.2

0.2

0.4
0.6
Normalized abscissa

0.8

(c)
Figure 9: (a) Free electron concentration, (b) electric field, and (c)
potential drop are plotted as functions of normalized abscissa for
= (1e14-15-16,5e4-1e5,50).
, ,
following inputs

An increase of the electric field final value is expected; for

constant average concentration and electric field initial values, an
increase of the GaN layer thickness results in an increase of the
total net charge and thus of the electric field final value according
and F0 are kept constant, then the electric
to Gausss law. If
field final values F1 and F2 corresponding to the GaN layer

thicknesses th1, th2 respectively, satisfy

proportionality (derived from Gausss law).

the

following

(19)
Values F0=104V/cm, F1=1.48*104V/cm, F2=3.38*104V/cm,
th1=25nm, and th2=125nm satisfy proportionality (19).
As stated before, our primary motivation to develop a model for
the intrinsic GaN channel of the MOS capacitor is the need for two
to get a consistent
, , ,
more equations of unknowns
system of four equations. If the initial value of the electric field in
the GaN layer is taken at the i-GaN/p-In0.5Ga0.5N heterojunction,
the third equation is obtained by finding the potential drop across
the GaN channel. Its expression is that of the second equality in set
(10) for
1 if case1 applies, or that of the second equality in set
(11) for
1 if case 2 applies. The fourth equation is obtained
first, by finding the expression of the electric field final value and
second, by using the flux continuity property at the oxide/GaN
interface. The final value of the electric field in the GaN layer is
obtained from the third equality in set (10) for
1 if case1
applies, or from the third equality in set (11) for
1 if case 2
applies. The two equations are;

(20)

(case1)

arctan

arctan

(21)

(case2)
In fact, instead of having one system of four equations, two
systems were obtained. We will demonstrate in a coming paper (in
progress) that in the case where there is a presence of free charge
carriers in the GaN channel, only case 1 is taken and thus equations
(20) together with equations (1) and (6) form the needed four
equations system. The case corresponding to an initial zero-valued
electric field was not discussed in this work because no charge
tunneling occurs for this value [1], however this case will be
considered in the discussion of the flat band condition of the MOS
capacitor. Theoretical quasi-static C-V characteristics will be
compared against experimental measurements on a test MOS
capacitor which is being fabricated in the Semiconductor
Laboratory at Lakehead University.

4. CONCLUSION
A quasi-static analysis was performed for a MOS capacitor on
In0.5Ga0.5N. In order to obtain C-V characteristics, the need to
analyze charge density distribution, electric field and voltage drop
across the channel of this 2-terminal device was demonstrated.
A channel model was presented. Future work (paper in progress) is
defined.

5. REFERENCES
[1] Dimiter Alexandrov, "Design of High Frequency Field Effect
Transistor on InxGa1-xN," in 21st Biennial Symposiun on
Communications, Kingston ON, Canada, 2002.
[2] G. C. Dousmanis and R. C. Duncun Jr, "Calculation on the
Shape and Extent of Space Charge Regions in Semiconductor
Surfaces," Journal of Applied Physics, vol. 29, no. 12, pp.
1267-1269, December 1958.
[3] J. R. Hauser and M. A. littlejohn, "Approximations for
Accumulation and Inversion Space Charge Layers in
semiconductors," Solid State Electronics, vol. 11, no. 7, pp.
667-674, July 1968.