CHAPTER

7
MESFET and Related Devices
7.1 METAL-SEMICONDUCTOR CONTACTS
7.2 MESFET
7.3 MODFET
SUMMARY

The metal-semiconductor field-effect transistor (MESFET) has current-voltage characteristics similar to those of
a metal-oxide-semiconductor field-effect transistor (MOSFET). However, it uses a metal-semiconductor rectifying
contact instead of a MOS structure for the gate electrode. In addition, the source and drain contacts of MESFET
are ohmic,* whereas in a MOSFET they are p–n junctions.
Like other field-effect devices, a MESFET has a negative temperature coefficient at high current levels; that is,
the current decreases as temperature increases. This characteristic leads to more uniform temperature distribution
and the device is therefore thermally stable, even when the active area is large or when many devices are connected
in parallel. Furthermore, because MESFETs can be made from compound semiconductors with high electron
mobilities, such as GaAs and InP, they have higher switching speeds and higher cutoff frequencies than do silicon
MOSFETs.
The basic building block of a MESFET is the metal-semiconductor contact. This contact is electrically similar
to a one-sided abrupt p–n junction, yet it can be operated as a majority carrier device with inherently fast response.
There are two types of metal-semiconductor contacts: the rectifying and the nonrectifying or ohmic types. In
this chapter, we begin with the two types of contacts and then consider the basic characteristics and microwave
performance of the MESFET. In the last section we discuss modulation-doped field-effect transistor (MODFET),
which is similar to the device configuration of a MESFET but can offer even higher-speed performance.

Specifically, we cover the following topics:

r The rectifying metal-semiconductor contact and its current-voltage characteristics.

r The ohmic metal-semiconductor contact and its specific contact resistance.
r The MESFET and its high-frequency performance.
r The MODFET and its two-dimensional electron gas.
r A comparison of three field-effect transistors—MOSFET, MESFET, and MODFET.

*
The concept of rectifying was discussed in Chapter 3; the concept of ohmic contact is presented in Section 7.1.
 MESFET and Related Devices 229

7.1 METAL-SEMICONDUCTOR CONTACTS

The first practical semiconductor device was the metal-semiconductor contact in the form of a point-contact
rectifier, that is, a metallic whisker pressed against a semiconductor. The device found many applications
beginning in 1904. In 1938, Schottky suggested that the rectifying behavior could arise from a potential barrier
as a result of the stable space charges in the semiconductor.1 The model arising from this concept is known as
the Schottky barrier. Metal-semiconductor contacts can also be nonrectifying; that is, the contact has negligible
resistance regardless of the polarity of the applied voltage. This type of contact is called an ohmic contact. All
semiconductor devices as well as integrated circuits need ohmic contact to make connections to other devices in
an electronic system. We consider the energy band diagram and the current-voltage characteristics of both the
rectifying and ohmic metal-semiconductor contacts.

7.1.1 Basic Characteristics

The characteristics of point-contact rectifiers were not reproducible from one device to another. The contact was
just a simple mechanical contact or formed by an electrical discharge process that could result in a small alloyed
p-n junction. The advantage of a point-contact rectifier is its small area, which can give very small capacitance,
a desirable feature for microwave application. The rectifiers are subject to wide variations such as the whisker
pressure, contact area, crystal structure, whisker composition, and heat or forming processes, and they have
been largely replaced by metal-semiconductor contacts fabricated by planar processes (see Chapters 11–15). A
schematic diagram of such a device is shown in Fig. 1a. To fabricate the device, a window is opened in an oxide
layer, and a metal layer is deposited in a vacuum system. The metal layer covering the window is subsequently
defined by a lithographic step. We consider the one-dimensional structure of the metal-semiconductor contact
shown in Fig. 1b, which corresponds to the central section in Fig. 1a, between the dashed lines.
Figure 2a shows the energy band diagram of an isolated metal adjacent to an isolated n-type semiconductor.
Note that the metal work function qϕm is generally different from the semiconductor work function qϕs. The work
function is defined as the energy difference between the Fermi level and the vacuum level. Also shown is the
electron affinity qχ, which is the energy difference between the conduction band edge and the vacuum

Fig. 1 (a) Perspective view of a metal-semiconductor contact fabricated by the planar process. (b) One-dimensional
structure of a metal-semiconductor contact.
 230 Semiconductors

Fig. 2 (a) Energy band diagram of an isolated metal adjacent to an isolated n-type semiconductor under thermal
nonequilibrium condition. (b) Energy band diagram of a metal-semiconductor contact in thermal equilibrium.

level in the semiconductor. When the metal makes intimate contact with the semiconductor, the Fermi levels in the
two materials must be equal at thermal equilibrium. In addition, the vacuum level must be continuous. These two
requirements determine a unique energy band diagram for the ideal metal-semiconductor contact, as shown in
Fig. 2b.
For this ideal case, the barrier height qϕBn is simply the difference between the metal work function and the
semiconductor electron affinity§:

qϕBn = qϕm– qχ. (1)

Similarly, for the case of an ideal contact between a metal and a p-type semiconductor, the barrier height qϕBp is
given by

qϕBp = Eg – (qϕm – qχ), (2)

where Eg is the bandgap of the semiconductor. Therefore, for a given semiconductor and for any metal, the sum of
the barrier heights on n-type and p-type substrates is expected to be equal to the bandgap:

q (φBn + φBp ) = Eg . (3)

On the semiconductor side in Fig. 2b, Vbi is the built-in potential that is seen by electrons in the conduction band
trying to move into the metal.

Vbi = ϕBn – Vn.. (4)

The qVn is the distance between the bottom of the conduction band and the Fermi level. Similar results can be
given for the p-type semiconductor.

§
Both qϕBn (in electron volts) and ϕBn (in volts) are referred to as the barrier height.
 MESFET and Related Devices 231

Fig. 3 Measured barrier height for metal-silicon and metal-gallium arsenide contacts.2,3

Figure 3 shows the measured barrier heights for n-type silicon2 and n-type gallium arsenide.3 Note that qϕBn
increases with increasing qϕm. However, the dependence is not as strong as predicted by Eq. 1. This is because
in practical Schottky diodes, the disruption of the crystal lattice at the semiconductor surface produces a large
number of surface energy states located in the forbidden bandgap. These surface states can act as donors or
acceptors that influence the final barrier height. For silicon and gallium arsenide, Eq. 1 generally underestimates
the n-type barrier height and Eq. 2 overestimates the p-type barrier height. The sum of qϕBn and qϕBp, however, is in
agreement with Eq. 3.
Figure 4 shows energy band diagrams for metals on both n-type and p-type semiconductors under different
biasing conditions. Consider the n-type semiconductor first. When the bias voltage is zero, as shown in the left side
of Fig. 4a, the band diagram is under a thermal equilibrium condition. The Fermi levels for both materials are
equal. If we apply a positive voltage to the metal with respect to the n-type semiconductor, the semiconductor-to-
metal built-in potential decreases as shown on the left side of Fig. 4b. This is a forward bias. When a forward bias
is applied, electrons can move easily from the semiconductor into the metal because the barrier has been reduced
by a voltage VF. For a reverse bias (i.e., a negative voltage is applied to the metal), the barrier is increased by a
voltage VR, as depicted on the left side of Fig. 4c. It is more difficult for electrons to flow from the semiconductor
into the metal. We have similar results for p-type semiconductor, however, the polarities must be reversed. In the
following derivations, we consider only the metal-n-type semiconductor contact. The results are equally applicable
to a p-type semiconductor with an appropriate change of polarities.
The charge and field distributions for a metal-semiconductor contact are shown in Fig. 5a and 5b,
respectively. The metal is assumed to be a perfect conductor; the charge transferred to it from the semiconductor
exists in a very narrow region at the metal surface. The extent of the space charge in the semiconductor is W, i.e.,
ρs = qND for x < W and ρs = 0 for x > W. Thus, the charge distribution is identical to that of a one-sided abrupt
p+-n junction.
The magnitude of the electric field decreases linearly with distance. The maximum electric field Em is located
at the interface. The electric field distribution is then given by
 232 Semiconductors

Fig. 4 Energy band diagrams of metal n-type and p-type semiconductors under different biasing conditions: (a) thermal
equilibrium; (b) forward bias; and (c) reverse bias.

qN D qN
E ( x) = (W − x ) = Em − D x, (5)
εs εs
qN DW
Em = , (6)
εs
where εs is the dielectric permittivity of the semiconductor. The voltage across the space-charge region, which is
represented by the area under the field curve in the Fig. 5b, is given by

EmW qN DW 2
Vbi − V = = . (7)
2 2ε s
The depletion-layer width W is expressed as

W = 2ε s (Vbi − V ) / qN D , (8)
 MESFET and Related Devices 233

Fig. 5 (a) Charge distribution and (b) electric-field distribution in a metal-semiconductor contact.

and the space-charge density, QSC, in the semiconductor is given as

QSC = qN DW = 2qε s N D (Vbi − V ) C/cm 2 , (9)

where the voltage V equal to +VF for forward bias and to –VR for reverse bias. The depletion-layer capacitance C
per unit area can be calculated by using Eq. 9:

∂Qsc qε s N D ε
C= = = s F/cm 2 (10)
∂V 2 (Vbi − V ) W
and
1 2 (Vbi − V )
C 2
=
qε s N D
( F/cm )
2 −2
. (11)

We can differentiate 1/C2 with respect to V. Rearranging terms we obtain:

2 ⎡ −1 ⎤
ND = ⎢ ⎥. (12)
qε s ⎢ d (1 / C ) / dV ⎥
2
⎣ ⎦
Thus, measurements of the capacitance C per unit area as a function of voltage can provide the impurity
distribution from Eq. 12. If ND is constant throughout the depletion region, we should obtain a straight line by
plotting 1/C2 versus V. Figure 6 is a plot of the measured capacitance versus voltage for tungsten-silicon and
tungsten-gallium arsenide Schottky diodes.4 From Eq. 11, the intercept at 1/C2 = 0 corresponds to the built-in
potential Vbi. Once Vbi is determined, the barrier height ϕBn can be calculated from Eq. 4.
 234 Semiconductors

Fig. 6 1/C2 versus applied voltage for W-Si and W-GaAs diodes.4

EXAMPLE 1
Find the donor concentration and the barrier height of the tungsten-silicon Schottky diode shown in Fig. 6.

SOLUTION The plot of 1/C2 versus V is a straight line, which implies that the donor concentration is constant
throughout the depletion region. We find

d (1 / C 2 ) ( cm / F)
2 2
6.2 × 1015 − 1.8 × 1015
= = −4.4 × 1015 .
dV −1 − 0 V
From Eq. 12,

⎡ 2 ⎤ ⎛ 1 ⎞
ND = ⎢ ⎥×⎜ 15 ⎟
= 2.7 × 1015 cm −3 ,
⎢⎣1.6 × 10 × (11.9 × 8.85 × 10 ) ⎥⎦ ⎝ 4.4 × 10 ⎠
−19 −14

⎛ 2.86 × 1019 ⎞
Vn = 0.0259 × ln ⎜ 15 ⎟ = 0.24 V.
⎝ 2.7 × 10 ⎠
Since the intercept Vbi is 0.42 V, then the barrier height is ϕBn = 0.42 + 0.24 = 0.66 V.

7.1.2 The Schottky Barrier

A Schottky barrier refers to a metal-semiconductor contact having a large barrier height (i.e., ϕBn or ϕBp >> kT) and
a low doping concentration that is less than the density of states in the conduction band or valence band.
The current transport in a Schottky barrier is due mainly to majority carriers, unlike a p-n junction where
current transport is due mainly to minority carriers. For Schottky diodes operated at moderate temperature (e.g.,
300 K), the dominate transport mechanism is thermionic emission of majority carriers from the semiconductor
over the potential barrier into the metal.
 MESFET and Related Devices 235

Fig. 7 Current transport by the thermionic emission process. (a) Thermal equilibrium; (b) forward bias; and (c)
reverse bias.5

Figure 7 illustrates the thermionic emission process.5 At thermal equilibrium (Fig. 7a), the current density
is balanced by two equal and opposite flows of carriers, and thus there is zero net current. Electrons in the
semiconductor tend to flow (or emit) into the metal, and there is an opposing balanced flow of electrons from
metal into the semiconductor. These current components are proportional to the density of electrons at the
boundary.
As discussed in Section 2.5 of Chapter 2, at the semiconductor surface an electron can be thermionically
emitted into the metal if its energy is above the barrier height. Here the semiconductor work function qϕs is
replaced by qϕBn, and

⎛ qφ ⎞
nth = N C exp ⎜ − Bn ⎟, (13)
⎝ kT ⎠
where NC is the density of states in the conduction band. At thermal equilibrium we have

| Jms|=| Jsm| ∝ nth (14)

or
⎛ qφ ⎞
J m → s = J s → m = C1 N C exp ⎜ − Bn ⎟ , (14a)
⎝ kT ⎠

where Jm→s is the current from the metal to the semiconductor, Js→ m is the current from the semiconductor to the
metal, and C1 is a proportionality constant.
When a forward bias VF is applied to the contact (Fig. 7b), the electrostatic potential difference across the
barrier is reduced, and the electron density at the surface increases to

⎡ q (φBn − VF ) ⎤
nth = N C exp ⎢ − ⎥. (15)
⎣ kT ⎦
The current Js→m that results from the electron flow out of the semiconductor is therefore altered by the same
factor (Fig. 7b). The flux of electrons from the metal to the semiconductor, however, remains the same because the
barrier ϕBn remains at its equilibrium value. The net current under forward bias is then
 236 Semiconductors

J = J s→m − J m→s
⎡ q (φBn − VF ) ⎤ ⎛ qφBn ⎞
= C1 N C exp ⎢ − ⎥ − C1 N C exp ⎜ − ⎟
⎣ kT ⎦ ⎝ kT ⎠
( )
= C1 N C e − qφBn / kT e qVF / kT − 1 .
(16)

With the same argument for the reverse-bias condition (see Fig. 7c), the expression for the net current is identical
to Eq. 16 except that VF is replaced by –VR.
The coefficient C1NC is found to be equal to A*T2, where A* is called the effective Richardson constant (in units
of A/K2-cm2) and T is the absolute temperature. The value of A* depends on the effective mass and is equal to 110
and 32 for n- and p-type silicon, respectively, and 8 and 74 for n- and p-type gallium arsenide, respectively.6
The current-voltage characteristic of a metal-semiconductor contact under thermionic emission condition is
then

J = J s ( e qV / kT − 1) , (17)

J s = A*T 2 e − qφBn / kT , (17a)

where Js is the saturation current density and the applied voltage V is positive for forward bias and negative
for reverse bias. Experimental forward I-V characteristics of two Schottky diodes4 are shown in Fig. 8. By
extrapolating the forward I-V curve to V = 0, we can find Js. From Js and Eq. 17a we can obtain the barrier height.
In addition to the majority carrier (electron) current, a minority-carrier (hole) current exists in a metal n-type
semiconductor contact. The electron-hole pairs can be easily created in the valence band (interband transition) in
the depletion region. Electrons in the valence band flow into the metal because there is no barrier there and holes
diffuse into the semiconductor to form the minority current under forward bias. The hole diffusion current is the
same as in a p-n junction, which is described in Chapter 3. The current density is given by

J p = J po ( e qV / kT − 1) , (18)
where
qD p ni 2
J po  . (18a)
Lp N D
Under normal operating conditions, the minority-carrier diffusion current is orders of magnitude smaller than
the majority-carrier current. Therefore, a Schottky diode is a unipolar device (i.e., predominately only one type of
carrier participates in the conduction process). The minimum minority-carrier storage makes the Schottky barrier
to operate at much higher frequencies (~100 GHz) compared to a p-n junction (~1 GHz).

EXAMPLE 2
For a tungsten-silicon Schottky diode with ND = 1016 cm–3, find the barrier height and depletion-layer
width from Fig. 8. Compare the saturation current Js with Jpo, assuming that the minority-carrier lifetime in
Si is 10–6 s.

SOLUTION From Fig. 8, we have Js = 6.5 × 10–5 A/cm2. The barrier height can be obtained from Eq. 17a:

⎛ 110 × 3002 ⎞
φBn = 0.0259 × ln ⎜ −5 ⎟
= 0.67 V.
⎝ 6.5 × 10 ⎠
 MESFET and Related Devices 237

Fig. 8 Forward current density versus applied voltage of W-Si and W-GaAs diodes.4

This result is in close agreement with the C-V measurement (see Fig. 6 and Ex. 1). The built-in potential is given by
ϕBn – Vn, where

⎛N ⎞ ⎛ 2.86 × 1019 ⎞
Vn = 0.0259 × ln ⎜ C ⎟ = 0.0259 ln ⎜ 16 ⎟ = 0.17 V..
⎝ ND ⎠ ⎝ 1× 10 ⎠
Therefore
Vbi = 0.67 – 0.17 = 0.50 V.

The depletion-layer width at thermal equilibrium is given by Eq. 8 with V = 0:

2ε sVbi
W= = 2.6 × 10−5 cm.
qN D
To calculate the minority-carrier current density Jpo, we need to know Dp, which is 10 cm2/s for ND = 1016 cm–3, and Lp
which is D pτ p = 10 × 10−6 = 3.1× 10−3 cm . Therefore,

1.6 × 10−19 × 10 × ( 9.65 × 109 )

2
qD p ni 2
J po = = = 4.8 × 10−12 A/cm 2 .
Lp N D ( 3.1×10−3 ) ×1016
 238 Semiconductors

The ratio of the two current densities is

Js 6.5 × 10−5
= = 1.3 × 107.
J po 4.8 × 10−12

From the comparison, we see that the majority-carrier current is over seven orders of magnitude greater
than the minority-carrier current.

7.1.3 The Ohmic Contact

An ohmic contact is defined as a metal-semiconductor contact that has a negligible contact resistance relative to
the bulk or series resistance of the semiconductor. A satisfactory ohmic contact should not significantly degrade
device performance and can pass the required current with a voltage drop that is small compared with the drop
across the active region of the device.
A figure of merit for ohmic contacts is the specific contact resistance RC, defined as

−1
⎛ ∂J ⎞
RC ≡ ⎜ ⎟ Ω − cm 2 . (19)
⎝ ∂V ⎠V 0

For metal-semiconductor contacts with low doping concentrations, the thermionic-emission current dominates the
current transport, as given by Eq. 17. Therefore,

k ⎛ qφ ⎞
RC = *
exp ⎜ Bn ⎟. (20)
qA T ⎝ kT ⎠

Equation 20 shows that a metal-semiconductor contact with a low barrier height should be used to obtain a small
RC.
For contacts with high doping concentration, the barrier width becomes very narrow, and the tunneling
current becomes dominant. The tunneling current, as described in the upper inset of Fig. 9, is proportional to the
tunneling probability, which is given in Section 2.6 of Chapter 2:

I ~ exp ⎡ −2W 2mn ( qφBn − qV ) / ? 2 ⎤ , (21)

⎣ ⎦

where W is the depletion-layer width, which can be approximated as ( 2ε s / qN D ) (φBn − V ) , mn is the effective mass,
and h- is the reduced Planck constant. Substituting W into Eq. 21, we obtain

⎡ C (φ − V ) ⎤
I ~ exp ⎢ − 2 Bn ⎥, (22)
⎢⎣ ND ⎥⎦
where C2 = 4 mn s / ? . The specific contact resistance for high doping is thus

⎛C φ ⎞ ⎛ 4 mnε s φBn ⎞
RC ~ exp ⎜ 2 Bn ⎟ = exp ⎜ ⎟. (23)
⎜ N ⎟ ⎜ N ? ⎟⎠
⎝ D ⎠ ⎝ D
 MESFET and Related Devices 239

Equation 23 shows that in the tunneling range the specific contact resistance depends strongly on doping
concentration and varies exponentially with the factor Bn / N D .
The calculated values of RC are plotted6 in Fig. 9 as a function of 1/ N D . For ND ≥ 1019 cm–3, RC is dominated
by the tunneling process and decreases rapidly with increased doping. On the other hand, for ND ≤ 1017 cm–3,
the current is due to thermionic emission, and RC is essentially independent of doping. Also shown in Fig. 9 are
experimental data for platinum silicide-silicon (PtSi-Si) and aluminum-silicon (Al-Si) diodes. They are in close
agreement with the calculated values. Figure 9 shows that a high doping concentration, a low barrier height, or
both must be used to obtain a low value of RC. These two approaches are used for all practical ohmic contacts.

EXAMPLE 3
An ohmic contact has an area of 10–5 cm2 and a specific contact resistance of 10–6 Ω –cm2. The ohmic contact is
formed in an n-type silicon. If ND = 5 × 1019 cm–3, ϕBn = 0.8 V, and the electron effective mass is 0.26 m0, find the
voltage drop across the contact when a forward current of 1A flows through it.

Fig. 9 Calculated and measured values of specific contact resistance. Upper inset shows the tunneling process, lower
inset shows thermionic emission over the low barrier.6
 240 Semiconductors

SOLUTION The contact resistance for the ohmic contact is

RC
= 10 6 Ω − cm 2 / 10 5 cm 2 = 10 1 Ω,
A
4 0.26 × 9.1× 10 31
× (1.05 × 10 10
)
C2 = 4 mnε s / ? = 34
1.05 × 10
= 1.9 × 1014 ( m 3/ 2
/ V ).
From Eq. 22,
⎡ C (φ − V ) ⎤
I = I 0 exp ⎢ − 2 Bn ⎥,
⎢⎣ ND ⎥⎦
∂I A ⎛ C ⎞ ⎛ −C φ ⎞
= = I0 ⎜ 2 ⎟ exp ⎜ 2 Bn ⎟
∂V RC ⎜ N ⎟ ⎜ N ⎟
V 0 ⎝ D ⎠ ⎝ D ⎠

A⎛ ND ⎞ ⎛C φ ⎞
or I0 = ⎜ ⎟ exp ⎜ 2 Bn ⎟
RC⎜ C2 ⎟ ⎜ N ⎟
⎝ ⎠ ⎝ D ⎠

⎛ 5 × 10 × 10 ⎞
19 6
⎛ 1.9 × 1014 × 0.8 ⎞
= 10 × ⎜ ⎟ exp ⎜⎜ ⎟
⎜ 1.9 × 10 14 ⎟ 19 6 ⎟
⎝ ⎠ ⎝ 5 × 10 × 10 ⎠
= 8.13 × 108 A.

At I = 1 A, we have
ND ⎛I ⎞
φBn − V = ln ⎜ 0 ⎟ = 0.763 V
C2 ⎝ I ⎠
or
V = 0.8 – 0.763 = 0.037 V = 37 mV.

Therefore, there is a negligibly small voltage drop across the ohmic contact. However, the voltage drop may
become significant when the contact area is reduced to 10–8 cm2 or smaller.

7.2 MESFET

7.2.1 Basic Device Structures

The metal-semiconductor field-effect transistor (MESFET) was proposed7 in 1966. The MESFET has three metal-
semiconductor contacts—one Schottky barrier for the gate electrode and two ohmic contacts for the source and
drain electrodes. A perspective view of a MESFET is illustrated in Fig. 10a. The basic device parameters include
L, the gate length, Z, the gate width, and a, the thickness of the epitaxial layer. Most MESFETs are made of n-
type III-V compound semiconductors, such as gallium arsenide, because of their high electron mobilities, which
help to minimize series resistances, and because of their high saturation velocities, which result in the increase in
the cutoff frequency.
Practical MESFETs are fabricated by using epitaxial layers on semiinsulating substrates to minimize
parasitic capacitances. In Fig. 10a, the ohmic contacts are labeled source and drain, and the Schottky barrier is
labeled gate. A MESFET is often described in terms of the gate dimensions. If the gate length (L) is 0.5 μm and
the gate width (Z) is 300 μm, the device is referred to as a 0.5 × 300 μm device. A microwave- or millimeter-wave
device typically has a gate length in the range 0.1–1.0 μm. The thickness a of the epitaxial layer is typically one-
third to one-fifth of the gate length. The spacing between the electrodes is one to four times that of the gate length.
The current handling capability of a MESFET is directly proportional to the gate width Z because the cross-
sectional area available for channel current is proportional to Z.
 MESFET and Related Devices 241

Fig. 10 (a) Perspective view of a metal-semiconductor field-effect transistor (MESFET). (b) Cross section of the gate
region of a MESFET.

7.2.2 Principles of Operation

To understand the operation of a MESFET, we consider the section under the gate, Fig. 10b. The source is
grounded, and the gate and drain voltage are measured with respect to the source. Under normal operating
conditions, the gate voltage is zero or reverse biased and the drain voltage is zero or forward biased; that is,
VG ≤ 0 and VD ≥ 0. Since the channel is n-type material, the device is referred to as an n-channel MESFET. Most
applications use the n-channel MESFET rather than the p-channel MESFET because of higher carrier mobility
in n-channel devices.
The resistance of the channel is given by

L L L
R=ρ = = , (24)
A q μn N D A q μn N D Z ( a − W )

where ND is the donor concentration, A is the cross-section area for current flow and equals Z(a – W), and W is the
width of the depletion region of the Schottky barrier.
When no gate voltage is applied and VD is small, as shown in Fig. 11a, a small drain current ID flows in
the channel. The magnitude of the current is given by VD/R, where R is the channel resistance given in Eq. 24.
Therefore, the current varies linearly with the drain voltage. Of course, for any given drain voltage, the voltage
along the channel increases from zero at the source to VD at the drain. Thus, the Schottky barrier becomes
increasingly reverse biased as we proceed from the source to the drain. As VD is increased, W increases, and the
average cross-sectional area for current flow is reduced. The channel resistance R also increases. As a result, the
current increases at a slower rate.
 242 Semiconductors

Fig. 11 Variation of the depletion-layer width and output characteristics of a MESFET under various biasing
conditions. (a) VG = 0 and a small VD. (b) VG = 0 and at pinch-off. (c) VG = 0 at post pinch-off (VD > VDsat).
(d) VG = –1V and a small VD.

As the drain voltage is further increased, eventually the depletion region touches the semiinsulating substrate,
as shown in Fig. 11b. This happens when W = a at the drain. We can obtain the corresponding value of the drain
voltage, called the saturation voltage, VDsat, from Eq. 7 where V = –VDsat:

qN D a 2
VDsat = − Vbi for VG = 0. (25)
2ε s
At this drain voltage, the source and the drain are pinched off or completely separated by a reverse-biased
depletion region. The location P in Fig. 11b is called the pinch-off point. At this point, a large drain current called
the saturation current IDsat can flow across the depletion region. This is similar to the situation caused by injecting
carriers into a reverse-biased depletion region such as the collector-base depletion region of a bipolar transistor.
Beyond the pinch-off point, as VD is increased further, the depletion region near the drain will expand and
point P will move toward the source, as indicated in Fig. 11c. However, the voltage at point P remains the same,
VDsat. Thus, the number of electrons per unit time arriving from the source to point P, and hence the current
flowing in the channel, remain the same because the potential drop in the channel from source to point P does
not change. Therefore, for drain voltages larger than VDsat, the current remains essentially at the value IDsat and is
independent of VD.
When a gate voltage is applied to reverse bias the gate contact, the depletion-layer width W increases. For a
small VD , the channel again acts as a resistor but its resistance is higher because the cross-sectional area available
for current flow is decreased. As indicated in Fig. 11d, the initial current is smaller for VG = –1 V than for VG = 0.
When VD is increased to a critical value, the depletion region again touches the semiinsulating substrate. The value
of this VD is given by
 MESFET and Related Devices 243

qN D a 2
VDsat = − Vbi − VG . (26)
2ε s

For an n-channel MESFET, the gate voltage is negative with respect to the source, so we use the absolute value of
VG in Eq. 26 and in subsequent equations. We see from Eq. 26 that the application of a gate voltage VG reduces the
drain voltage required for the onset of pinch-off by an amount equal to VG.

7.2.3 Current-Voltage Characteristics

We now consider a MESFET before the onset of pinch-off, as shown in Fig. 12a. The drain voltage variation
along the channel is shown in Fig. 12b. The voltage drop across an elemental section dy of the channel is given by

I D dy
dV = I D dR = , (27)
q μn N D Z ⎡⎣ a − W ( y ) ⎤⎦

where we used Eq. 24 for dR and replaced L by dy. The depletion-layer width at distance y from the source is given
by

2ε s ⎡⎣V ( y ) + VG + Vbi ⎤⎦
W ( y) = . (28)
qN D

Fig. 12 (a) Expanded view of the channel region. (b) Drain voltage variation along the channel.
 244 Semiconductors

The drain current ID is a constant, independent of y. We can rewrite Eq. 27 as

I D dy = q μ n N D Z ⎡⎣ a − W ( y ) ⎤⎦ dV . (29)

The differentiation of the drain voltage dV is obtained from Eq. 28:

qN D
dV = WdW . (30)
εs
Substituting dV into Eq. 29 and integrating from y = 0 to y = L yields

1 W2 qN
ID = ∫ q μn N D Z ( a − W ) D WdW
L 1W εs

Z μ n q 2 N D2 ⎡ 2 3 3 ⎤
⎢ a (W2 − W1 ) − 3 (W2 − W1 ) ⎥
2 2
=
2ε s L ⎣ ⎦
or

⎡V 2 ⎛V + V + V ⎞ 3/2 2 ⎛V + V ⎞ 3/2 ⎡
I = IP ⎢ VD – 3 ⎜ D V G bi⎟ + 3 ⎜ G V bi ⎟ ⎢ , (31)
⎣ P ⎝ P ⎠ ⎝ P ⎠ ⎣
where

Z μ n q 2 N D2 a 3 (31a)
IP ≡
2ε s L
and

qN D a 2
VP ≡ . (31b)
2ε s

The voltage VP is called the pinch-off voltage, that is, the total voltage (VD + VG + Vbi) at which W2= a.
In Fig. 13 we show the I-V characteristics of a MESFET having a pinch-off voltage of 3.2 V. The curves
shown are calculated for 0 ≤ VD ≤ VDsat using Eq. 31. Beyond VDsat the current is taken to be constant in accordance
with our previous discussion. Note that there are three different regions in the current-voltage characteristics.
When VD is small, the cross-section area of the channel is essentially independent of VD and the I-V characteristics
are ohmic or linear. We refer to this region of operation as the linear region. For VD ≥ VDsat, the current saturates
at IDsat. We refer to this region of operation as the saturation region. As the drain voltage is further increased,
avalanche breakdown of the gate-to-channel diode occurs and the drain current suddenly increases. This is the
breakdown region.
In the linear region where VD <<VG + Vbi, Eq. 31 can be expanded to give

IP ⎡ ⎛ V + Vbi ⎞⎤
ID ≅ ⎢1 − ⎜ G ⎟ ⎥ VD . (32)
VP ⎢ ⎝ VP ⎠ ⎥⎦
⎣
 MESFET and Related Devices 245

Fig. 13 Normalized ideal current-voltage characteristics of a MESFET with VP = 3.2 V.

An important parameter of a MESFET is the transconductance gm, which represents the change of drain current
at a given drain voltage on a change in gate voltage. From Eq. 32, we obtain

∂I D IP VP
gm = = 2
VD . (33)
∂VG VD
2VP VG + Vbi
In the saturation region, the drain current can be calculated from Eq. 31 by evaluating the current at the
pinch-off point; that is, by setting VP = VD + VG + Vbi :

⎡ 1 ⎛ V + V ⎞ 2 ⎛ V + V ⎞3 / 2 ⎤
I Dsat = IP ⎢ − ⎜ G bi
⎟+ ⎜
G bi
⎟ ⎥. (34)
⎢⎣ 3 ⎝ VP ⎠ 3 ⎝ VP ⎠ ⎥⎦

The corresponding saturation voltage is given by

VDsat=VP – VG – Vbi. (35)

The transconductance in the saturation region can be obtained from Eq. 34:

⎛
IP V + Vbi ⎞ Z μn qN D a ⎛ V + Vbi ⎞
gm = ⎜⎜1 − G ⎟⎟ = ⎜⎜1 − G ⎟. (36)
VP
⎝ VP ⎠ L ⎝ VP ⎟⎠
In the breakdown region, the breakdown voltage occurs at the drain end of the channel, where the reverse
voltage is the highest:

VB (breakdown voltage) = VD + |VG| (37)

For example, in Fig. 13 the breakdown voltage is 12 V for VG = 0. At |VG | = 1, the breakdown voltage is still 12 V
and the drain voltage at breakdown is (VB – |VG|) or 11 V.
 246 Semiconductors

EXAMPLE 4
Consider an n-channel GaAs MESFET at T = 300 K with a gold contact. Assume the barrier height is 0.89 V. The
n-channel doping is 2 × 1015 cm–3 and the channel thickness is 0.6 μm. Calculate the pinch-off voltage and the built-
in potential. The dielectric constant of GaAs is 12.4.

SOLUTION The pinch-off voltage is

qN D 2 (1.6 × 10 ) ( 2 × 10 )
−19 15

× ( 0.6 × 10−4 ) = 0.53 V.

2
VP = a =
2ε s 2 × 12.4 × ( 8.85 × 10 )
−14

The difference between the conduction band and the Fermi level is given by

kT ⎛ N C ⎞ ⎛ 4.7 × 1017 ⎞
Vn = ln ⎜ ⎟ = 0.026 ln ⎜ 15 ⎟
= 0.14 V.
q ⎝ ND ⎠ ⎝ 2 × 10 ⎠

The built-in potential is

Vbi = fBn - Vn = 0.89 - 0.14 = 0.75 V.

So far we have considered only a normally on (or depletion-mode) device; that is, the device has a conductive
channel at VG = 0. For high-speed, low-power applications, the normally off device is preferred. This device does
not have a conductive channel at VG = 0; that is, the built-in potential Vbi of the gate contact is sufficient to deplete
the channel region. This is possible, for example, in a gallium arsenide MESFET with a very thin epitaxial layer on
a semiinsulating substrate. For a normally off MESFET, a positive bias must be applied to the gate before channel
current begins to flow. The required voltage, called the threshold voltage VT, is given by

VT = Vbi – VP (38a)
or
Vbi = VT + VP, (38b)

where VP is the pinch-off voltage defined in Eq. 31b. Near the threshold voltage, the drain current in the saturation
region can be obtained by substituting Vbi of Eq. 38b in Eq. 34 and by using the Taylor series expansion assuming
(VG – VT)/Vp  1. We obtain

⎧⎪ 1 ⎡ ⎛ V − V ⎞ ⎤ 2 ⎡ ⎛ V − V ⎞ ⎤ 3/ 2 ⎫⎪
I Dsat = I P ⎨ − ⎢1 − ⎜ G T ⎟ ⎥ + ⎢1 − ⎜ G T ⎟ ⎥ ⎬
⎪⎩ 3 ⎢⎣ ⎝ VP ⎠ ⎥⎦ 3 ⎢⎣ ⎝ VP ⎠ ⎦⎥ ⎪⎭
or

Z μnε s
(VG − VT ) .
2
I Dsat ≈ (39)
2aL
In deriving Eq. 39 we used a negative sign for VG to take into account its polarity.
The basic current-voltage characteristics of normally on and normally off devices are similar. Figure 14
compares these two modes of operation. The main difference is the shift of threshold voltage along the VG axis.
The normally off device (Fig. 14b) has no current conduction at VG = 0, and the current varies as in Eq. 39 when
VG > VT. Since the built-in potential of the gate is less than about 1 V, the forward bias on the gate is limited to
about 0.5 V to avoid excessive gate current.
 MESFET and Related Devices 247

Fig. 14 Comparison of I-V characteristics. (a) Normally on MESFET. (b) Normally off MESFET.

The transconductance for a normally off device can be obtained from Eq. 39:

dI Dsat Z μn ε s
gm = = (VG − VT ) . (40)
dVG aL

7.2.4 High-Frequency Performance

For high-frequency applications of MESFETs, an important figure of merit is the cutoff frequency ƒT , which is
the frequency at which the MESFET can no longer amplify the input signal. The small-signal input current is the
product of the gate admittance and the small-signal gate voltage, assuming that the device has negligibly small
series resistance:

iin = 2π fCGυg , (41)

where CG is the gate capacitance equal to ZL( s / W ) and W is the average depletion-layer width under the gate
electrode. The small-signal output current is obtainable from the definition of the transconductance:
 248 Semiconductors

∂I D iout
gm = =
∂VG υg
(42)
or
iout = g mυg .

Equating Eqs. 41 and 42a, we obtain the cutoff frequency

gm I P / VP μ qN a 2
fT = < ≈ n D2 , (43)
(
2π CG 2π ZL ε s / W 2πε s L )
where we used Eq. 36 for gm. From Eq. 43 we see that to improve high-frequency performance, we should use a
MESFET having high carrier mobility and short channel length. This is the reason that the n-channel MESFET,
which has higher electron mobility, is preferred.
These derivations are based on the assumption that the carrier mobility in the channel is a constant
independent of the applied field. However, for very-high-frequency operations, the longitudinal field, i.e., the
electric field directed from the source to the drain, is sufficiently high that the carriers travel at their saturation
velocity.
Under these conditions, the saturation channel current is given by

I Dsat = ( area for carrier transport ) × qnυ s ,

(44)
= Z ( a − W ) qN Dυ s .
The transconductance is then

∂I Dsat ∂I Dsat ∂W ⎛ 1 ⎞
gm = = ⋅ = ⎡⎣ qN Dυ s Z ( −1) ⎤⎦ ⎜ ⎟ (45)
∂VG ∂W ∂VG ⎝ −qN DW / ε s ⎠
or
gm = Zυsεs/W. (45a)

In Eq. 45, we obtain ∂W/∂VG from Eq. 28.

From Eq. 45a, we can obtain the cutoff frequency under saturation-velocity condition:

gm Zυ s ε s / W υ
fT = = = s . (46)
2π CG 2π ZL ( ε s / W ) 2π L

Therefore, to increase ƒT, we must reduce the gate length L and employ a semiconductor with a high saturation
velocity. Figure 15 shows the electron drift velocity versus electric field for five semiconductors.8 Note that GaAs
has an average velocity§ of 1.2 × 107 cm/s and a peak velocity of 2 × 107 cm/s, which are 20%–100% higher than the
saturation velocity of silicon. Also note that Ga0.47In0.53As and InP have higher average and peak velocities than
GaAs. Consequently, the cutoff frequencies of these semiconductors will be higher than that from GaAs.
1
⎡ 1 L dx ⎤
§
The average velocity is defined as υ ≡ ⎢ ∫ 0 ⎥ . If ' ( x) is a constant υ0, then ' = υ0.
⎢⎣ L υ ( x ) ⎥⎦
 MESFET and Related Devices 249

Fig. 15 The drift velocity versus the electric field for electrons in various semiconductor materials.8

7.3 MODFET

7.3.1 MODFET Fundamentals

The modulation-doped field-effect transistor (MODFET) is a heterostructure field-effect device. The MODFET
has been an alternative to MESFETs in high-speed circuits. Other names commonly applied to the device include
high electron mobility transistor (HEMT), two-dimensional electron gas field-effect transistor (TEGFET),
and selectively doped heterostructure transistor (SDHT). Frequently, it is referred to by a general name of
heterojunction field-effect transistor (HFET).
The most-common heterojunctions for the MODFETs are the AlGaAs/GaAs, AlGaAs/InGaAs, and InAlAs/
InGaAs heterointerfaces. Figure 16 shows a perspective view of a conventional AlGaAs/GaAs MODFET. The
special features of a MODFET are its heterojunction structure under the gate and its modulation doped layers.
For the device in Fig. 16, AlGaAs is the wide bandgap semiconductor, whereas GaAs is the narrow bandgap
semiconductor. The two semiconductors are modulation doped, i.e., the AlGaAs is doped (~1018 cm-3), except for
a narrow region do, which is undoped, whereas the GaAs is undoped. Electrons in the AlGaAs will diffuse to the
undoped GaAs, where a conduction channel can be formed at the surface of the GaAs. The net result of
this modulation doping is that channel carriers have high mobilities because there is no impurity scattering. The
undoped AlGaAs spacer layer is used to reduce the Coulomb scattering from ionized donors in doped AlGaAs
and leads to enhanced carrier mobility in the channel.
A comparison of low-field electron mobility of the modulation-doped 2-D channel to bulk GaAs at
different doping levels is shown in Fig. 17. In a MESFET the channel has to be doped to a reasonably high level
(> 1017 cm-3), and thus the electron suffers from impurity scattering. However, the modulation-doped channel
has much higher mobilities at all temperatures. It is also interesting to compare the modulation-doped channel,
which usually has an unintentional doping below 1014 cm-3, to lowly doped bulk samples (with a similar impurity
concentration of 4 × 1013 cm-3). The bulk mobility as a function of temperature shows a peak but drops at both
 250 Semiconductors

Fig. 16 Perspective view of a conventional modulation-doped field-effect transistor (MODFET) structure.

Fig.17 Comparison of low-field electron mobility of modulation-doped 2-D channel to bulk GaAs at different
doping levels.9

high temperature and low temperature. The decrease of bulk mobility with an increase of temperature is due to
phonon scattering. At low temperatures, the bulk mobility is limited by impurity scattering. It depends on the
doping level and also decreases with a decrease in temperature. In the modulation-doped channel, its mobility at
temperatures above ~ 80 K is comparable to the value of a lowly doped bulk sample. However, mobility is much
enhanced at lower temperatures. The modulation-doped channel does not suffer from impurity scattering,
 MESFET and Related Devices 251

Fig. 18 Energy band diagrams for a normally-off MODFET at (a) thermal equilibrium, and (b) the onset of threshold.
d1 and d0 are the doped and undoped regions, respectively.6

which dominates at low temperatures. This benefit stems from the screening effect of a two-dimensional electron
gas (2 DEG), where its conduction path is confined to a small cross-section, smaller than 10 nm, with high volume
density.
Figure 18a shows the band diagram of a MODFET in a thermal equilibrium condition. Similar to a
standard Schottky barrier, qϕBn is the barrier height of the metal on the wide-bandgap semiconductor.10 ΔEC is the
conduction band discontinuity for the heterojunction structure, and VP is the built-in potential given by

q qN D d12
N D ( x ) xdx =
d1
VP =
εs ∫ 0 2ε s
, (47)

where d1 is the thickness of the doped region in AlGaAs and εs is the dielectric permittivity.
A key parameter for the operation of a MODFET is the threshold voltage VT , which is the gate bias at
which the channel starts to form between the source and drain. With reference to Fig. 18b, VT corresponds to the
situation where the bottom of the conduction band at the GaAs surface coincides with the Fermi level:
 252 Semiconductors

ΔEC
VT = φBn − − VP . (48)
q

The threshold voltage VT can be adjusted by using different values for ϕBn and Vp. However, ΔEC is fixed for a given
set of semiconductors. Figure 18b has a positive VT, and the MODFET is an enhancement-mode device (normally
off), as opposed to a depletion-mode device with a negative VT (normally on).
When the gate voltage is larger than VT, a charge sheet ns(y) is capacitively induced by the gate at the
heterojunction interface. The charge sheet is similar to the charge Qn /q in the inversion layer of a MOSFET (see
Section 5.1):

Ci ⎡⎣VG − VT − V ( y ) ⎤⎦
ns ( y ) = , (49)
q
where
εs
Ci = , (49a)
d1 + d 0 + Δd

d1 and d0 are the doped and undoped AlGaAs thickness (Fig. 16) and Δd is the channel thickness or the thickness
of the inversion layer, estimated to be about 8 nm. V(y) is the channel potential with respect to the source. The
channel potential varies along the channel from 0 to the drain bias VD, similar to that shown in Fig. 12b. The
charge sheet is also called a two-dimensional electron gas. This is because that the electrons in the inversion layer
are confined in the x-direction by ΔEC on the left side and by the potential distribution of the conduction band on
the right side (Fig. 18b). However, these electrons can make two-dimensional movements: in the y-direction from
the source to the drain and in the z-direction parallel to the channel width (Fig. 16).
Equation 49 shows that a negative gate bias will reduce the two-dimensional electron gas. If, on the other
hand, a positive VG is applied, ns will increase.

EXAMPLE 5
Consider an AlGaAs/GaAs heterojunction with n-AlGaAs doped to 2 ×1018 cm–3 and a thickness of 40 nm.
ΔEC
Assume the undoped spacer layer is 3 nm and the Schottky barrier height is 0.85 V and q = 0.23 V. The dielectric
constant of the AlGaAs is 12.3. Calculate the two-dimensional electron gas concentration for such heterojunction
at VG = 0.

SOLUTION

qN D d12 1.6 × 10 × 2 × 10 × ( 40 × 10 )
−19 18 −7 2

VP = = = 2.35 V.
2ε s 2 × 12.3 × 8.85 × 10−14
The threshold voltage is

ΔEC
VT = φBn − − VP = 0.85 − 0.23 − 2.35 = −1.73 V.
q
Therefore, the device is a normally on MODFET.
The two-dimensional electron gas at the source for VG = 0 is

12.3 × 8.85 × 10−14

ns = × ⎡0 − ( −1.73) ⎤⎦ = 2.29 × 1012 cm −2 .
1.6 × 10−19 × ( 40 + 3 + 8 ) × 10−7 ⎣
 MESFET and Related Devices 253

7.3.2 Current-Voltage Characteristics

The current-voltage characteristics of a MODFET can be obtained by using the gradual channel approximation
similar to that of a MOSFET. The current at any point along the channel is given by

I = Zq μ n nsEy

dV ( y )
= Z μ n Ci ⎡⎣VG − VT − V ( y ) ⎤⎦ . (50)
dy
Since the current is constant along the channel, integrating Eq. 50 from source to drain (y = 0 to y = L) gives

Z ⎡ V2 ⎤
I= μn Ci ⎢(VG − VT )VD − D ⎥ . (51)
L ⎣ 2 ⎦
The output characteristics for an enhancement-mode MODFET are similar to those shown in Fig. 14b. In the
linear region where VD << (VG – VT), Eq. 51 can be reduced to

Z
I= μn Ci (VG − VT )VD . (52)
L

For a large drain voltage, the charge sheet n(y) at the drain is reduced to zero. This is the pinch-off condition
previously discussed and shown in Fig. 11b. From Eq. 49, we obtain the saturation voltage VDsat, at which ns (y = L)
= 0,

VDsat = VG ,VT, (53)

and the saturation current can be obtained from Eqs. 51 and 53,

Z μ n Ci Z μnε s
(VG − VT ) = (VG − VT ) .
2 2
I= (54)
2L 2 L ( d1 + d 0 + Δd )

Note that this equation is very similar to Eq. 39. A similar expression can be obtained for the transconductance
given in Eq. 40.
For high-speed operations, the longitudinal field (i.e., the field along the channel) is sufficiently high to cause
carrier velocity saturation. The current in the velocity-saturation region is

I sat = Zυ s qns
(55)
≅ Zυ s Ci (VG − VT ).
The transconductance becomes

∂I sat
gm = = Zυ s Ci . (56)
∂VG
Note that Isat is independent of gate length and gm is independent of both gate length and gate voltage in the
velocity-saturation regime.
 254 Semiconductors

Fig. 19 Cutoff frequency versus channel or gate length for five different field-effect transistors.8,11

7.3.3 Cutoff Frequency

The speed of a MODFET is measured by the cutoff frequency

gm Zυ s Ci
fT = =
2π ( total capacitance ) 2π ( ZLCi + CP )
υs (57)
= .
2π ( L + CP / ZCi )

where Cp is the parasitic capacitance. To improve ƒT, we should consider a semiconductor with a large ' s , a gate
structure with an ultrashort gate length, and a device configuration with minimum parasitic capacitance.
A comparison of the cutoff frequencies of various FETs is shown in Fig. 19. The cutoff frequency ƒT is
plotted against the channel or gate length.8,11 Note that for a given length, a silicon n-type MOSFET has the lowest
ƒT because of the relatively low mobility and a low average velocity of electron in silicon. GaAs MESFET has an
ƒT about three times higher than does a silicon MOSFET.
Also shown are three MODFETs. The conventional GaAs MODFET (i.e., AlGaAs-GaAs structure) has
an ƒT about 30% higher than that of the GaAs MESFET. The pseudomorphic SiGe MODFET (i.e., Si-SiGe
structure where the SiGe lattice is shrunk slightly to match the silicon lattice) has an ƒT comparable to the GaAs
MODFET. SiGe MODFETs are attractive because they can be processed in a silicon fabrication facility. For
even higher cutoff frequencies, we have the Al0.48In0.52As-Ga0.47In0.53As MODFET formed on an InP substrate. The
superior performance is mainly due to the high electron mobility in Ga0.47In0.53As and its high average and peak
velocities. It is estimated that at a gate length of 50 nm ƒT can be as high as 600 GHz.
 MESFET and Related Devices 255

SUMMARY

When a metal makes intimate contact with a semiconductor, it forms a metal-semiconductor contact. There are
two types of contacts. The first type is the rectifying contact, also called the Schottky barrier contact, which has a
relatively large barrier height and is formed on a semiconductor with a relatively low doping concentration. The
potential and field distribution in a Schottky barrier are identical to those of a one-sided abrupt p–n junction.
However, the current transport in a Schottky barrier is by thermionic emission and, therefore, has an inherent fast
response.
The second type is the ohmic contact, which is formed on a degenerate semiconductor in which carrier
transport is by the tunneling process. An ohmic contact can pass the required current with a very small voltage
drop across it. All semiconductor devices and integrated circuits need ohmic contacts to make connections to
other devices in an electronic system.
Metal-semiconductor contacts are building blocks for MESFET and MODFET devices. By employing
a Schottky barrier as the gate electrode and two ohmic contacts as the source and drain electrodes, we form a
MESFET. This three-terminal device is important for high-frequency applications, especially for monolithic
microwave integrated circuits (MMIC). Most MESFETs are made with n-type III-V compound semiconductors
because of their high electron mobilities and high average drift velocities. Of particular importance is GaAs,
because of its relatively mature technology and the availability of high-quality GaAs wafers.
The MODFET is a device with enhanced high-frequency performance. This device structure is similar to that
of a MESFET except there is a heterojunction under the gate. A two-dimensional electron gas, i.e., a conductive
channel, is formed at the heterojunction interface, and electrons with high mobility and high average drift velocity
can be transported from the source through the channel to the drain.
The output characteristics of all field-effect transistors (FETs) are similar. They all have a linear region at
low-drain biases. As the bias increases, the output current eventually saturates, and at a sufficiently high voltage,
avalanche breakdown occurs at the drain. Depending on whether it requires a positive- or negative-threshold
voltage, FET can be either normally off (enhancement mode) or normally on (depletion mode).
The cutoff frequency ƒT is a figure of merit for the high-frequency performance of an FET. For a given
length, the silicon MOSFET (n-type) has the lowest ƒT and the GaAs MESFET has an ƒT about three times higher
than that of silicon. The conventional GaAs MODFET and the pseudomorphic SiGe MODFET have ƒT about
30% higher than that of the GaAs MESFET. For even higher cutoff frequencies, we have the GaInAs MODFET,
which has a projected ƒT of 600 GHz at a gate length of 50 nm.

REFERENCES

1. W. Schottky, “Halbleitertheorie der Sperrschicht,” Naturwissenschaften, 26, 843 (1938).

2. A. M. Cowley and S. M. Sze, “Surface States and Barrier Height of Metal Semiconductor System,”
J. Appl. Phys., 36, 3212 (1965).
3. G. Myburg et al., “Summary of Schottky Barrier Height Data on Epitaxially Grown n- and p-
GaAs,” Thin Solid Films, 325, 181 (1998).
4. C. R. Crowell, J. C. Sarace, and S. M. Sze, “Tungsten-Semiconductor Schottky-Barrier Diodes,”
Trans. Met. Soc. AIME, 23, 478 (1965).
5. V. L. Rideout, “A Review of the Theory, Technology and Applications of Metal-Semiconductor
Rectifiers,” Thin Solid Films, 48, 261 (1978).
6. S. M. Sze and K. K. Ng, Physics of Semiconductor Devices, 3rd Ed., Wiley Interscience, Hoboken,
2007.
7. C. A. Mead, “Schottky Barrier Gate Field-Effect Transistor,” Proc. IEEE, 54, 307 (1966).
 256 Semiconductors

8. S. M. Sze, Ed., High Speed Semiconductor Devices, Wiley, New York, 1992.
9. P. H. Ladbrooke, “GaAs MESFETs and High Mobility Transistors (HEMT),” in H. Thomas, D.
V. Morgan, B. Thomas, J. E. Aubrey, and G. B. Morgan, Eds., Gallium Arsenide for Devices and
Integrated Circuits, Peregrinus, London,1986.
10. K. K. Ng, Complete Guide to Semiconductor Devices, McGraw Hill, New York, 1995.
11. S. Luryi, J. Xu, and A. Zaslavsky, Eds., Future Trends in Microelectronics, Wiley, New York, 1999.

PROBLEMS (* DENOTES DIFFICULT PROBLEMS)

FOR SECTION 7.1 METAL-SEMICONDUCTOR CONTACTS

1. Calculate the theoretical barrier height and built-in potential in a metal-semiconductor diode for zero applied bias.
Assume the metal work function is 4.55 eV, the electron affinity is 4.01 eV, and ND = 2 × 1016 cm–3 at 300 K.
2. (a) Find the donor concentration and barrier height of the W-GaAs Schottky barrier diode shown in Fig. 6. (b)
Compare the barrier height with that obtained from the saturation current density of 5 × 10–7 A/cm2 shown in Fig. 8.
(c) For a reverse bias of –1 V, calculate the depletion-layer width W, the maximum field, and the capacitance.
3. Copper is deposited on a carefully prepared n-type silicon substrate to form an ideal Schottky diode. ϕm = 4.65 eV, the
electron affinity is 4.01 eV, ND = 3 × 1016 /cm3, and T = 300 K. Calculate the barrier height, the built-in potential, the
depletion-layer width, and the maximum field at a zero bias.
*4. The capacitance of a Au-n-type GaAs Schottky barrier diode is given by the relation 1/C2 = 1.57 × 105 – 2.12 × 105 Va,
where C is expressed in μF and Va is in volts. Taking the diode area to be 10–1 cm2, calculate the built-in potential, the
barrier height, the dopant concentration, and the work function.
5 Calculate the value of Vbi and ϕm in an ideal metal-Si Schottky barrier contact. Assume the barrier height is 0.8 eV, ND =
1.5 × 1016 cm–3, and qχ = 4.01 eV.
6. In a metal-Si Schottky barrier contact, the barrier height is 0.75 eV and A* = 110 A/cm2 – K2. Calculate the ratio of the
injected hole current to the electron current at 300 K, assuming Dp =12 cm2 s–1, Lp = 1 × 10–3 cm, and ND = 1.5 × 1016
cm–3.

FOR SECTION 7.2 MESFET

7. Given ϕBn = 0.9 eV and ND = 1017 cm–3, find the minimum value of the thickness of the epitaxial layer for a GaAs
MESFET to be a depletion mode device (i.e., VT < 0).
8. Assume the doping in a GaAs MESFET is ND = 7 × 1016 cm–3 and the dimensions are a = 0.3 μm, L = 1.5 μm, Z = 5
μm, μn= 4500 cm2/V-s, and ϕBn = 0.89 V. Calculate the ideal value of gm for VG = 0, and VD = 1 V.
9. The n-channel GaAs MESFET shown in Fig. 10 has a barrier height ϕBn = 0.9 V, ND =1017 cm–3, a = 0.2 μm, L = 1 μm,
and Z = 10 μm. (a) Is this an enhancement or depletion mode device? (b) Find the threshold voltage (the enhancement
mode indicates VT > 0; the depletion mode indicates VT < 0).
10. An n-channel GaAs MESFET has a channel doping ND = 2 × 1015 cm–3, ϕBn = 0.8 V, a = 0.5 μm, L = 1 μm, μn= 4500
cm2/V-s, and Z = 50 μm. Find the pinch-off potential, threshold voltage, and the saturation current at VG = 0.
 MESFET and Related Devices 257

11. The barrier height ϕBn of two GaAs n-channel MESFETs are the same and equal to 0.85 V. The channel doping in
device 1 is ND = 4.7 × 1016 cm–3, and that in device 2 is ND = 4.7 × 1017 cm–3. Determine the channel thickness required in
each device so that the threshold voltage is zero for each device.

FOR SECTION 7.3 MODFET

12. For an abrupt AlGaAs/GaAs heterojunction with the n-AlGaAs layer doped to 3 × 1018 cm–3, the Schottky barrier is
0.89 V and the heterojunction conduction-band edge discontinuity ΔEC is 0.23 eV. Calculate the thickness of the doped
AlGaAs layer d1 so that the threshold voltage is –0.5 V. Assume the permittivity of the AlGaAs is 12.3.
*13. Find the thickness of the undoped spacer layer d0 so that the two-dimensional electron gas concentration of an
AlGaAs/GaAs heterojunction is 1.25 × 1012 cm–2 at zero gate bias. Assume that the n-AlGaAs is doped to 1 × 1018 cm–3
and has a thickness d1 of 50 nm, the Schottky barrier height is 0.89 V, and ΔEC/q = 0.23 V. The permittivity of the
AlGaAs is 12.3.
14. Consider an AlGaAs/GaAs HFET with a 50 nm n-AlGaAs and a 10 nm undoped AlGaAs spacer. Assume the
threshold voltage is –1.3 V, ND is 5 × 1017cm–3, ΔEC = 0.25 eV, the channel width is 8 nm, and the permittivity of
AlGaAs is 12.3. Calculate the Schottky barrier height and the two-dimensional electron gas concentration at VG = 0.
15. The AlGaAs/GaAs has a two-dimensional electron gas concentration of 1 × 1012 cm–2; the spacer is 5 nm, the channel
width is 8 nm, the pinch-off voltage is 1.5 V, ΔEC/q = 0.23 V, and the doping concentration of AlGaAs is 1018 cm–3. The
Schottky barrier height is 0.8 V. Find the thickness of the doped AlGaAs and the threshold voltage.
16. Consider an n-AlGaAs–intrinsic GaAs abrupt heterojunction. Assume that the AlGaAs is doped to ND = 3 × 1018cm–3
and has a thickness of 35 nm (there is no spacer). Let ϕBn = 0.89 V, and assume that ΔEC = 0.24 eV and the dielectric
constant is 12.3. Calculate (a) VP and (b) ns for VG = 0.