Wide bandgap transistor amplifiers for improved performance microwave power and radar applications

Abstract

The generation of high RF output power, on the order of 100s to 1000s of watts necessary for transmitters for radars and wireless communications systems, remains a difficult challenge for semiconductor devices. RF power devices fabricated from standard semiconductors such as Si and GaAs are limited in the RF output capability by the inherent breakdown voltage of the semiconductor material. Recently, the development of wide bandgap semiconductors, such as SiC and GaN and related heterostructures, offers the potential to fabricate transistors with an order of magnitude improved RF output power compared to traditional devices. The wide bandgap semiconductor transistors offer the potential to fabricate high power transmitters for radars and communications systems, thereby permitting full semiconductor realization of advanced systems. However, the wide bandgap semiconductor devices currently suffer from several physical effects that are limiting the RF performance, and thereby, their application. These limitations are discussed and solutions presented.

Full text

56 March 2005
1527-3342/05/$20.00©2005 IEEE
R.J. Trew, G.L. Bilbro, W. Kuang, Y. Liu, and H. Yin
lthough solid-state
devices have
replaced vacu-
um electron-
ics in the
vast majority of microwave
electronic systems over the
past 30 years, the revolution is
not complete. In particular, the
areas of high RF power for
microwave and millimeter-
wave radar and communica-
tions transmitter applications,
the ability to produce ade-
quate RF power levels at fre-
quencies greater than 100
GHz, and the ability of
devices to operate at high tem-
peratures greater than about
250
◦
C remain dominated by
microwave tubes. Further solid-state material and device
development in these areas are among the last frontiers for
semiconductor electronics. In these areas, solid-state
devices have not been able to compete with vacuum tube
devices, and most systems that must deliver kilo- to
megawatt levels of power are designed using various
types of microwave tubes.
Current microwave power solid-state devices pro-
duce RF power levels less than about 100 W in S-band
and operate with reasonable
RF output power, on the
order of 1 W, to frequencies
in the range of 100 GHz.
Power-combining and
phased-array technology
permit the outputs of many
solid-state devices to be
combined, thereby produc-
ing significantly improved
RF output power and solid
state systems can, in prac-
tice, compete in terms of RF
output power with tube
based systems in certain
applications. There is much
interest in the development
of improved solid-state
devices with improved RF
output power. Increasing RF
power per device would greatly simplify power-com-
bining techniques and permit smaller and lower cost
transmitters for communications and radar applications.
Advances in the growth technology for wide
bandgap semiconductors offers the promise of produc-
ing microwave electronic devices with RF power capa-
bility an order of magnitude greater than that available
from comparable devices fabricated from standard
semiconductors, such as Si and GaAs. These devices
R.J. Trew (trew@ncsu.edu), G.L. Bilbro, W. Kuang, Y. Liu, and H. Yin are with ECE Department,
Box 7243, North Carolina State University, Raleigh, NC 27695-7243 USA.
©ARTVILLE
A

March 2005 57
are expected to find wide application due to the high
reliability, small size, and potential low cost offered by
solid-state electronics. The most promising devices are
field-effect transistors (FETs) fabricated from the
AlGaN/GaN heterostructures. The two-dimensional
gas (2DEG) that is produced at the heterointerface has
both high sheet charge density (
n
ss
∼ 10
13
cm
−2
), high
electron mobility (
µ
n
∼ 1, 500 cm
2
/V-s), and high satu-
ration velocity (
v
s
∼ 2 × 10
7
cm/s). This permits a het-
erostructure field-effect transistor (HFET) device to
produce high RF current. Also, the nitride-based semi-
conductor materials have high critical electric field for
breakdown, permitting high-voltage operation. The
product of high current and high voltage results in high
RF power operation.
In this article, the microwave performance potential
and status of device development of AlGaN/GaN
HFETs are presented. Problem areas that are currently
limiting device performance are indicated, and the
potential RF performance of optimized devices is
shown. It is demonstrated that microwave power
amplifiers fabricated from these devices offer superior
RF power performance, particularly at elevated tem-
peratures, relative to comparable components fabricat-
ed from GaAs metal semiconductor FETs (MESFETs) or
Si transistors. These devices are expected to find appli-
cation in RF sources and power amplifiers for base sta-
tion transmitters for cellular telephone systems, satel-
lite transmitters, HDTV transmitters, power modules
for phased-array radars, surveillance and traffic control
radars, and other applications. The devices are particu-
larly attractive for applications that require operation at
elevated temperatures. However, before the potential
performance of these devices is available, current prob-
lems plaguing RF performance must be solved.
Background and State of the Art
The AlGaN/GaN HFET demonstrates excellent RF per-
formance. High sheet charge density resulting from high
Al incorporation in the AlGaN layer permits high chan-
nel current to be obtained [1] (see “Where Do the 2DEG
Electrons Come From?”). Initial HFETs were fabricated
on sapphire substrates, but recent work has focused
upon the use of semi-insulating or p-type SiC substrates
[2], [3]. Also, most AlGaN/GaN
HFETs are fabricated with unin-
tentionally doped AlGaN and
GaN epitaxial layers. However,
it is also possible to fabricate
AlGaN/GaN HFETs with good
RF performance using doped
channel designs [4], and 1.73-
W/mm RF output power with
good gain was obtained at 8.4
GHz. The small signal perfor-
mance of these devices demon-
strated gain bandwidth prod-
ucts of
f
T
= 39
GHz and
f
max
= 45
GHz. Small-signal
performance with intrinsic current gain-bandwidth
products up to
f
T
= 106
GHz for a
L
g
= 0.15 µ
m device
has been obtained [5]. These devices produced about 4
W/mm RF power and 41% power-added efficiency
(PAE) at 4 GHz. Very high RF power density has also
been obtained, and 9.8 W/mm RF power density with
47% PAE at 8 GHz has been obtained [6]. The devices
had gate widths of W
=
2 mm and the devices were
flip-chip mounted to AlN substrates for improved ther-
mal conductance. Other devices fabricated using SiC
substrates produced RF power as high as 10.7 W/mm
at 10 GHz with 40% PAE [7] when biased at
V
ds
= 45
v,
with further improvements yielding slightly over 11
W/mm. The use of field-plate technology (see “What’s
a Field Plate?”) results in much improved breakdown
voltages to be realized, thereby permitting much
greater bias voltage to be applied and a device biased
at
V
ds
= 120
v has produced RF output power density
greater than 30 W/mm [8]. Thermal management is a
key to obtaining high power amplifiers, and, as
improved technology is developed, the RF output
power of these amplifiers is rapidly improving,
approaching 100 W.
Basic High-Power Amplifier Operation
The basic configuration for an amplifier is shown in
Figure 1. The amplifier is a two-port network that con-
sists of a source that feeds the input with a load connect-
ed to the output. The network has gain and thereby
amplifies a signal passing through it from the source to
the load. RF power can only be generated from a real
source (i.e., resistance) and delivered through a network
to a real load (i.e., resistance). Electronic devices and
networks, as well as most microwave sources and loads,
also include reactance, which must be tuned in order to
optimally deliver energy from the source to the load.
The power delivered to the load from the network
can be written as
P
L
=
1
2
Re

V
L
I
∗
L

=
1
2
|
I
L
|
2
R
L
,(1)
where
V
L
and
I
L
are the voltage and current at the load
Figure 1. Basic amplifier configuration.
Network
P
in
V
s
Z
in
P
out
I
L
V
L
Z
L
=R
L
+jX
L
Z
out
Z
s
=R
s
+jX
s

58 March 2005
impedance and the star indicates the complex conju-
gate. The power delivered to the load can be written as
a function of the reflection coefficient at the load
P
L
= P
out

1 −
|

|
2

,(2)
and maximum RF power transfer occurs for the condi-
tion of no reflection from the load

L
= 0.(3)
This condition occurs when the load impedance is
set to the conjugate of the network output impedance
Z
L
= Z
∗
out
.(4)
The amplifier PAE is
PAE =
P
L
− P
in
P
dc
× 100% = P
in
(G − 1)
P
dc
× 100%,(5)
where the network gain is
G =
P
L
P
in
.(6)
The dynamic characteristics for a transistor amplifier
are illustrated in Figure 2, which shows dynamic load
lines (i.e., I–V characteristics) for three conditions: lin-
ear operation,
−1
db in compression, and
−3
db in
compression. The dynamic load lines are superim-
posed upon the dc I–V characteristics for the active
device. For the situation shown in Figure 2, the transis-
tor is biased at
V
ds
= 8
v, and the network is tuned for
maximum PAE for each dynamic load line. For power
delivered from a real source to a real load, the dynam-
ic load line would be a straight line oscillating up and
down the dc load line for the network. However, since
HEMT devices have conducting channels formed by
a conducting layer of free electrons at the heteroint-
erface of a doped large bandgap semiconductor
(e.g., AlGaAs) and an undoped smaller bandgap
semiconductor (e.g., GaAs). Electrons from the edge
region of the doped AlGaAs transfer into the lower
bandgap semiconductor and gather in the quantum
well that forms in the GaAs at the interface between
the two materials. The electrons are confined in the
quantum well, which has a thickness on the order of
20–30 Å. The distribution of the electrons in the
quantum well is essentially two-dimensional due to
the much greater sizes of the channel length and
width and the very small thickness of the quantum
well. For this reason the charge density is termed a
two-dimensional electron gas (i.e., 2DEG) and is
characterized as a sheet charge density (n
ss
) with
units of cm
–2
. The sheet charge density for the
AlGaAs/GaAs heterointerface is on the order of n
ss
∼
2
×
10
12
cm
–2
. The advantage of the 2DEG is
that there are essentially no impurity atoms in the
undoped GaAs and the quantum well, and the elec-
trons in the conducting channel do not experience
significant impurity scattering, permitting them to
move with very high mobility, generally much greater
than can be obtained in bulk semiconductor materi-
al. Carrier velocity is high, and the conducting chan-
nel resistance is very low, permitting HEMTs to pro-
duce high-frequency and low-noise performance.
The first AlGaN/GaN heterostructures were fabri-
cated by growing thin AlGaN layers on thicker GaN
material as shown in Figure A. Originally, neither the
AlGaN nor the GaN layers were doped. However, it
was observed that, despite the lack of electrons from
intentionally doping, a 2DEG was, in fact, established
at the heterointerface. The 2DEG has very high sheet
charge density, on the order of n
ss
∼
10
13
cm
–2
,
which is factor of five greater than produced in the
AlGaAs/GaAs system.
The fundamental question is: since there are no
intentionally introduced impurity atoms to supply elec-
trons, what is the source of the electrons that form the
Figure A. Conduction band edge for an AlGaN/GaN
heterointerface showing quantum well and 2DEG.
Gate
Metal
AIGaN-
Undoped
GaN-
Undoped
E
C
E
F
qφ
b
2 DEG Electron Density (∼10
13
cm
−
2
)
+
+
+
+
+
+
+
+
+
Where Do the 2DEG Electrons Come From?

March 2005 59
the device has capacitance, the dynamic load line
shows elliptical behavior. While the device is operating
below saturation, the load line is confined within the dc
I–V characteristics. As the device is driven into satura-
tion, the dynamic load line shifts and extends outside
the dc I–V characteristics on both the high and low cur-
rent portions of the RF cycle. The average value of the
RF current also increases, indicating that the device dc
current increases as the device is driven into saturation.
The extension of the dynamic load line outside the dc
I–V characteristics is possible due to the complex
nature of the network. The total RF current consists of
conduction and displacement components, and,
although the conduction current is limited by the I–V
characteristics, the displacement current maintains cur-
rent continuity at the terminals. As the device is driven
into saturation, the conduction current is clipped by the
I–V characteristics for the transistor, but the total RF
current continuity is maintained by displacement cur-
rent. Network capacitance increases as the device is dri-
ven into saturation, and inductive tuning is necessary
to obtain optimum RF performance.
Optimized inductive tuning results in the reversal of
dynamic load line direction, as shown in Figure 2.
Under optimum tuning conditions, the network is
2DEG? Measured data for heterostructures fabricated
with a variety of growth conditions always produce a
high density 2DEG. It has been established that the
density of the 2DEG varies with Al concentration in the
AlGaN layer, with higher sheet charge density obtained
for higher Al concentration.
An argument for the formation of the AlGaN/GaN
2DEG can by explained by reference to the model
shown in the sketch in Figure B.
According to this model, the electrons that form the
2DEG result from the growth process. It is known that
the AlGaN semiconductor layer is both polar and piezo-
electric. During growth, the crystal atoms line up so that
the positive side of the atomic layers are aligned
towards the GaN layer. As the layer thickness increases
during growth, the atomic layers continue to align, cre-
ating an electric field internal to the AlGaN layer, with
the positive side of the dipole facing the GaN and the
negative side of the dipole facing the growth surface.
The magnitude of the electric field is very high, and on
the order of E
∼
10
6
V/cm. The magnitude of the elec-
tric field is sufficient to ionize some of the covalent
electrons, as well as any impurities that happen to be
present in the material. A large reservoir of electrons is
also available from loosely bond surface electrons. The
electric field will ionize electrons and cause them to
drift towards the heterointerface, where they fall into
the quantum well, thereby creating the 2DEG. As the
electrons move from the AlGaN into the GaN, the mag-
nitude of the electric field is reduced, thereby acting as
a feedback mechanism to quench the electron transfer
process. An equilibrium condition is established when
sufficient electrons are transferred into the quantum
well to reduce the magnitude of the electric field in the
AlGaN to the point where no further electrons are
transferred. The Al concentration in the AlGaN
becomes a control on the density of the 2DEG since it
produces stress at the AlGaN/GaN interface that
increases the polarization/piezoelectric charge density,
which defines the electric field in the AlGaN, according
to Gauss’ Law (shown in Figure 2).
Figure B. Model for formation of a 2DEG at the
AlGaN/GaN heterointerface.
Growth Front
EE
2DEG
Strain
Force
−
−
+
−
+
−
+
−
+
−
+
−
+
−
+
−
+
−
+
−
+
−
+
−
+
Al
x
Ga
1−x
N/GaN
E(y) = (−9.5x − 2.1x
2
) MV/cm
E
n
= n
ˆ
~10
6
V/cm
q
+
ρ
r
o
εε
→
ρ
+
Figure 2. Dynamic and dc load lines for a transistor
amplifier network.
Drain Voltage (V)
Drain Current (A)
−0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
051015 20 25
−1-db Compression
−3-db Compression
Linear

60 March 2005
essentially a resonant circuit with the reactive energy
shifting between the capacitive and inductive fields. As
the network is driven further into saturation, the cur-
rent clipping behavior increases, with a net increase in
both dc current and device capacitance.
The dynamic behavior of the amplifier network defines
the factors that determine the RF performance limits of the
device and the materials from which it is fabricated. The
power delivered to the load is a product of the RF voltage
and RF current that can be established at the load, and this
Standard power GaAs FET designs are generally limit-
ed to drain bias voltages in the range of 8–12 V,
which limits the RF voltage and RF output power that
can be developed [1]. It has been shown that the use
of field-plate technology suppresses gate breakdown
and permits significantly higher drain bias voltages to
be applied [2]. Field-plate power GaAs FET’s biased
with drain voltage of 35 V have produced RF power
density of 1.7 W/mm of gate periphery, and a 230-W
amplifier when the FET was biased at V
ds
=
24 V [3].
Wide bandgap semiconductors such as those based
upon the III-N materials system have much improved
critical electric fields for breakdown compared to GaAs
and HFETs fabricated from these materials can sustain
significantly improved bias voltages, with V
ds
>
40 V
before breakdown is observed. Field-plate technology
is also being widely used with nitride-based HFETs [4]
to permit even greater drain voltage to be applied,
and a field-plate HFET when biased at a V
ds
=
120 V
has produced over 30 W/mm RF power density at S-
band [5] and over 5 W/mm at 30 GHz with a drain
bias of V
ds
=
30 V [6].
The concept for a field plate dates back to work on
the development of guard rings for high-voltage diodes.
Basically, the principle is to provide a conducting plane
near junctions and other locations where high electric
fields exist. The conducting plane provides a means to
terminate and smooth high electric fields, thereby
reducing the high electric field peaks that result in elec-
tronic breakdown. In this manner, it is possible to engi-
neer devices for increased breakdown voltage by
spreading the electric field over a larger region.
The field plate is applied to FETs as shown in the fig-
ure above for an AlGaN/GaN HFET. The field plate con-
sists of a conducting electrode placed over the gate and
extending into the gate-drain region of the FET structure.
The field plate is generally designed so that it is electri-
cally connected to the gate but is held at a short dis-
tance above the surface of the semiconductor.
Normally, the field-plate metal is located on top of a
surface passivation dielectric, which is designed for a
specific thickness. Both the extension of the field plate
into the gate-drain region and the thickness of the
dielectric are critical design parameters for optimum per-
formance. The field plate functions by reducing the elec-
tric field at the edge of the gate on the drain side. Due
to the geometry of the FET, the electric field at this posi-
tion is very high (normally on the order of E
∼
10
6
V/cm), which is sufficient to produce a tunneling current
from the gate metal to the semiconductor. It has been
shown that the primary breakdown mechanism for FETs
is tunnel leakage of electrons from the gate to the sur-
face region of the semiconductor [7]. Under high-volt-
age operation, the electrons can tunnel into the semi-
conductor with sufficient energy to produce avalanche
ionization. Field plates produce sufficient reduction of
the electric field at the gate edge to permit high volt-
ages to be applied, and this, in turn, permits high RF
power to be generated.
References
[1] T.A. Winslow, D. Fan, and R.J. Trew, “Gate-drain breakdown effects
upon the large signal RF performance of GaAs MESFETs,” in 1990
IEEE MTT-S Int. Microwave Symp. Dig., pp. 315–317.
[2] A. Asano, Y. Miyoshi, K. Ishikura, Y. Nashimoto, M. Kuzuhara, and M.
Mizuta, “Novel high power AlGaAs/GaAs HFET with a field-modulating
plate operated at 35 V drain voltage,” in 1998 IEDM Dig., pp. 59–62.
[3] A. Wakejima, K. Ota, K. Matsunaga, and M. Kuzuhara, “A GaAs-based
field-modulating plate HFET with improved WCDMA peak-output-
power characteristics,” IEEE Trans. Electron Dev., vol. 50, pp.
1983–1987, Sept. 2003.
[4] S. Karmalkar and U.K. Mishra, “Enhancement of breakdown voltage
in AlGaN/GaN high electron mobility transistors using a field plate,”
IEEE Trans. Electron Dev., vol. 48, pp. 1515–1521, Aug. 2001.
[5] Y.F. Wu, A. Saxler, M. Moore, R.P. Smith, S. Sheppard, P.M. Chavarkar, T.
Wisleder, U.K. Mishra, and P. Parikh,
“30-W/mm GaN HEMTs by field plate
optimization,” IEEE Electron Dev. Lett.,
vol. 25, pp. 117–119, Nov. 2004.
[6] C. Lee, P. Saunier, J. Yang, and M.A.
Khan, “AlGaN-GaN HFEMT’s or SiC
with CW power performance
>
4
W/mm and 23% PAE at 35 GHz,”
IEEE Electron Dev. Lett., vol. 24,
pp. 616–618, Oct. 2003.
[7] R.J. Trew and U.K. Mishra, “Gate
breakdown in MESFETs and
HEMTs,” IEEE Electron Dev. Lett.,
vol. 12, pp. 524–526, Oct. 1991.
A field-plate AlGaN/GaN HFET design.
Reduced Tunneling
Under High Field
Fixed Positive Charge
2DEG Electrons
E Field
Channel Depleted of Electrons as Gate
Voltage Nears Pinch-Off
Depletion Oscillates with RF Voltage
+++++++++++++++++++++
What’s a Field Plate?

March 2005 61
is determined by the active device. Semiconductors are
limited in the bias voltage that can be applied by the criti-
cal electric field for breakdown of the semiconductor mate-
rial. Therefore, semiconductors
that have high critical electric
fields for breakdown are desirable
for power device applications. The
critical field for breakdown is a
function of bandgap energy, and
wide bandgap semiconductors are
desirable for power applications.
Large current capability
requires semiconductor materi-
als that have high electron
velocity. In general, both high
mobility and high saturation
velocity semiconductors are
desirable to generate high RF
current. Traditional semicon-
ductors, such as Si and GaAs,
have electron saturation veloci-
ties that are limited to about
v
s
= 1 × 10
7
cm/s, and this lim-
its both the power that can be
generated and the frequency
response of the device. The electron transport charac-
teristics of AlGaN/ GaN are significantly greater than
for standard semiconductors (
v
s
 2 × 10
7
cm/s), and,
for this reason, the nitride-based materials are very
attractive for fabrication of high power and high per-
formance electronic devices.
AlGaN/GaN HFET Amplifier Performance
The dynamic load line for an AlGaN/GaN HFET
X-band amplifer is shown in Figure 3. The transistor
has a gate length of
L
g
= 0.5 µ
m and gate width of
W =
1 mm. The dynamic load line is superimposed
upon the dc I–V characteristic for the transistor. The
decrease in the dc current with increasing current and
voltage is due to thermal heating. The RF output power
of the amplifier is directly relat-
ed to the drain bias that is
applied, and the ability to apply
high drain bias is key to the abil-
ity of the nitride HFETs to pro-
duce high RF output power. For
example, the RF performance of
the amplifier biased at
V
ds
= 40
v, operated in a Class A circuit,
and tuned for maximum PAE is
shown in Figure 4. At 10 GHz,
the amplifier produces RF out-
put power of
P
o
= 10
W,
PAE
=
50%, and linear gain
G =
15 db. As indicated, this
amplifier should produce good
RF power performance up to 30
GHz. Use of a field plate permits higher drain bias to be
applied, and the RF performance as a function of
V
ds
at
10 GHz is shown in Figure 5. The RF output power
Figure 4. RF performance for an AlGaN/GaN HFET class
Aamplifier (
V
ds
= 40 v
,
L
g
= 0.5 µm
, W
=
1 mm).
60
50
40
30
20
10
0
35302520151050
P
o
(dbm), PAE(%), G(db)
Frequency (GHz)
G
P
o
PAE
Figure 3. Dynamic load line superimposed upon the dc I–V characteristics for an
AlGaN/GaN HFET amplifier at 10 GHz.
High-Field
Nonlinearities
High-Injection
Nonlinearities
I
d
(A)
V
ds
(V)
020406080100
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Figure 5. RF performance for a 10-GHz Class A AlGaN/GaN HFET amplifier as a
function of drain bias.
P
o
(W/mm), PAE(%), G(db)
PAE
60
50
40
30
20
10
0
80 100 1206040
V
ds
(V)
G
P
o

62 March 2005
increases directly with
V
ds
and achieves a power densi-
ty of over 30 W/mm at a drain bias of
V
ds
= 100 v
, with
a PAE
=
50%. Redesign of the device produces the RF
performance at 100 GHz shown in Figure 6. This device
had a gate length of
L
g
= 0.25 µ
m and a gate width of
W = 100 µ
m. The amplifier produces RF output power
of
P
o
= 400
mW (RF power density of 4 W/mm) with
PAE = 17
% and linear gain of
G =
5 db. This indicates
that the AlGaN/GaN HFETs have excellent potential for
millimeter-wave applications.
The load line shown in Figure 3 and the RF perfor-
mance shown in Figures 4–6 are produced by an
advanced microwave device/circuit simulator, and the
simulations do not include several physical effects that
have been found to be important to the operation of
these devices. High-injection
and high-field nonlinearities
have been found to limit
device performance, and they
occur during high-current and
high-voltage operation condi-
tions, as indicated in Figure 3.
The areas within the device
where the high current and
high-voltage physical effects
occur are shown in Figure 7.
There are three fundamen-
tal effects that limit the perfor-
mance of these devices: 1)
space charge effects in the
gate-source region under high
current injection, 2) electron
leakage from the gate to the
surface of the semiconductor,
and 3) high field breakdown
effects at the edge of the gate
electrode and also in the con-
ducting channel. There can
also be trap charging/dis-
charging effects at the epitaxi-
al layer/substrate interface,
but these tend to be less
important than the other three.
High Current Effects
and Space-Charge
Limitations
During the high current por-
tion of the RF cycle, the current
density can achieve very high
magnitude. For example, the
quantum well thickness for the
conducting channel at the
AlGaN/GaN heterointerface is
about 25
˚
A
[9]. Measured cur-
rent for these devices is on the
order of 1–1.2 A for a 1-mm
wide device. Therefore, if all electrons were confined to
the quantum well, the current density would be on the
order of
J ∼ 50 × 10
6
A/cm
2
. Space-charge effects occur
when the injected electron density is on the order of the
background charge. Under these conditions, the electric
field is affected by the injected charge according to
Poisson’s equation
dE
dx
=
q
ε
(
N
d
− n
o
− δn
)
∼
=
−
q
ε
δn,(7)
where the equilibrium charge density
n
o
is essentially
equal to the background doping
N
d
, and
δn
is the
injected excess charge.
Figure 6. RF performance for an AlGaN/GaN HFET class amplifier at 100 GHz
(
V
ds
= 30 v
, Class A).
02468101214 16 18 20 22 24 26 28
0
5
10
15
20
25
30
35
P
o
(dBm), Gain(dB), PAE(%)
Input Power (dBm)
Gain
P
out
PAE
Figure 7. AlGaN/GaN HFET structure showing sources of physical effects limiting RF
performance.
n
+
n
+
Gate Leakage and
Surface Traps
2DEG
Undoped AlGaN
Undoped GaN
Substrate (SiC, Sapphire)
Space-Charge
Effect
Interface Effects
High-Electric Field
Charge Dipole
Domain
+ + +
+ + +
+ + +
- - - -
- - -
- - -
- - -
- - - - -
- - - -

March 2005 63
The level of injected charge that will produce pertur-
bations in the electric field scales with the background
doping, as shown in Figure 8. For background doping
of
N
d
∼ 10
16
cm
−3
the critical current for the onset of
space charge effects is
J
sc
∼ 30 × 10
3
A/cm
2
, and for
N
d
∼ 10
19
cm
−3
, it increases to
J
sc
∼ 30 × 10
6
A/cm
2
.
The measured operating current density in practical
AlGaN/ GaN HFETs is above the critical field for the
onset of space-charge effects during the high current
portion of the RF cycle. Once space-charge effects occur,
the resistance of the semiconductor increases very
rapidly. For example, it has been shown that a doubling
of the current produces an order of magnitude increase
in the semiconductor resistance [10]. The space-charge
effect has the most effect in the gate-source region and
produces a current-modulated source resistance [11].
The result is that the dynamic load line cannot achieve
the full extent of the potential current swing indicated in
the dc I–V characteristics. The dynamic load line is shift-
ed to higher voltage values, as shown in Figure 9. As the
device is driven to higher RF power levels, the source
resistance continues to increase, producing an “RF
walkout” phenomenon illustrated in Figure 10.
High Voltage Effects
The high voltage portion of the RF cycle introduces
nonlinearities due to breakdown phenomena. There are
two major breakdown phenomena that affect HFET
performance: 1) reverse conduction of the gate due to
high electric field at the gate edge, and 2) RF break-
down in the conducting channel due to the high electric
field that exists across the charge dipole domain.
It is well known that a very high magnitude electric
field exists at the edge of the gate due to the two-
dimensional geometry of FETs. The magnitude of the
electric field at this location varies with the voltage
applied to the gate, increasing as the gate voltage
approaches pinch-off, and also with increasing drain
voltage. For large drain voltage, which occurs during the
high voltage portion of the RF cycle, and for large-gate
voltage, the magnitude of the electric field at the edge of
the gate can easily exceed
E∼ 10
6
V/cm. This is suffi-
cient to produce reverse conduction of the gate by an
electron tunneling mechanism [12]. A model for this
phenomenon is shown in Figure 11 [13], [14].
The electrons that tunnel to the semiconductor sur-
face can accumulate on the surface near the gate, cre-
ating a “virtual gate” effect, where the gate length
appears to increase with increasing tunnel leakage
(Figure 12). Under these conditions, the gate length
appears to be a dynamic function of RF drive, and
increases with increasing drive. This generates a drive-
dependent decrease in the current drive capability of
the device, and a reduction in device
f
T
. The electrons
that accumulate on the surface in trap states move in a
sluggish manner due to high effective mass resulting
from the surface conduction band. They move with
low mobility and slowly dissipate. Therefore, as the
electrons tunnel to the surface, the depletion region
under the surface electrons at the gate edge is less able
to respond to rapid modulation of the gate voltage,
Figure 10. RF walkout of the knee of the dynamic load line
due to source resistance modulation under high current
conditions.
I
ds
V
ds
RF "Knee" Voltage
Dynamic Load Line
P
in1
P
in2
P
in3
P
in3
>P
in2
>P
in1
Figure 9. Source resistance modulation under high current
conditions.
i
RF
v
RF
R
L
V
dc
V
ds
∆i
RF
∆v
RF
I
ds
I
dc
Figure 8. Threshold current density versus background
doping for the on-set of space-charge effects.
30 MA/cm
2
30 kA/cm
2
10
19
cm
−3
10
16
cm
−3
Doping
J
sc

64 March 2005
introducing a time lag into the current-voltage behav-
ior and a reduction in the high-frequency performance
of the device. Also, the depletion region under the sur-
face electrons produce an electrostatic barrier in the
path of the channel electrons, and since this barrier is
not as readily modulated, the channel electrons tend to
be trapped in the gate-source region, as shown in
Figure 12. This helps produce a charge accumulation
and charge storage effect in the gate-source region. The
trapped charge enhances the space-charge effect and
increases the source resistance when the current
exceeds the space-charge limited current condition.
Also, at high fields the electrons can tunnel from the
gate into the semiconductor with sufficient energy to
produce impact ionization. When this occurs, gate cur-
rent increases and light emission occurs at the gate
edge. The reverse conduction characteristics of the
gate electrode are a major factor in determining the
saturation behavior of the device when operating
under large-signal conditions. The effects of RF break-
down upon the dynamic load line for the device are
illustrated in Figure 13.
The gate leakage phenomenon can be greatly
reduced by application of field-plate technology. The
field plate suppresses the electric field at the gate edge,
thereby reducing the gate leakage by a significant fac-
tor. The field-plate technology
is used with both GaAs
MESFETs and HEMTs, and
AlGaN/GaN HFETs. Field-
plate technology permits large
drain bias voltages to be
applied, and drain voltages in
the range of
V
ds
= 30
–40 v for
GaAs FETs, and
V
ds
= 120
v
for AlGaN/GaN HFETs have
been applied. This, of course,
permits very large RF voltage
to develop across the device
and high RF output power to
be generated.
IMPATT-Mode Operation
RF breakdown in the conduct-
ing channel can also occur, par-
ticularly during the high voltage, low cur-
rent portion of the RF cycle. When channel
breakdown occurs the drain current increas-
es, providing an increase in both the RF and
dc currents. The increase in dc current effec-
tively self-biases the device so that, in practi-
cal operation, devices are biased for Class
A–B conditions, and when they are driven
into saturation the current increases so that
an essentially Class A operation condition is
obtained. The physics for RF channel break-
down are not well understood, but RF chan-
nel breakdown generally occurs at voltages
below the dc breakdown voltage.
When a field-plate FET is biased at high
drain bias, an interesting phenomena can
Figure 11. Model for tunnel leakage of the gate during
high voltage operation.
Gate Drain
n
ss
d
s
d
1
n(y)
E
g
= d
s
N
D
(y)dy − n
ss
(x)dx
I
t
= f
G
L
g
W
g
J
t
(E
g
) = I
tf
−
I
tr
q
εf
G
L
g
d
1
⌠
⌡
0
d
s
⌠
⌡
0
n
t
=
J
t
(E
g
)
qv
s




Figure 12. AlGaN/GaN HFET showing charge storage and surface electron phenomena.
++++++++++++++++++++++++++++++
δn(x) > n
d
− n
o
C
gs
(Charge Storage)
Trapped Electrons
Charging/Discharging Time, τ
Surface Conduction
2DEG
AlGaN
GaN
I = JA
n > n
d
Source
V
g
(dc + RF)
x = 0
δn(0)
n
+
E
High-Field Region
Figure 13. Dynamic load line for a FET showing effects of RF breakdown.
i
RF
with Breakdown
RF Breakdown
i
RF
I
dc
v
RF
I
ds
R
L
V
dc
V
ds
∆i
RF
τ

March 2005 65
occur. First, at high drain bias voltages a significant deple-
tion of electrons in the conducting channel under the gate
occurs. This also occurs in standard FETs, but is restricted
to near pinch-off bias conditions. For the high-voltage
devices, simulations indicate that the channel becomes
partially depleted, even for high current, open channel
conditions. The depletion region for the device becomes
defined primarily by the drain bias, rather than the gate
length, as occurs for standard FETs. The frequency
response of the device is defined by the transit-time of
electrons as they move from the source to drain, and this
time is determined by transit through the depletion
region. For standard FETs, the depletion region is deter-
mined by the gate length, and a reduction in gate length
produces an increase in device frequency response, as
indicated by the device
f
T
. High-voltage FETs demon-
strate a much reduced dependence of frequency response
on gate length, and the device
f
T
does not demonstrate the direct
dependence upon gate length
observed in lower voltage oper-
ation devices. The reason for this
is the determination of the
depletion region by the drain
bias, rather than gate length.
A second major impact of
the use of field-plate technolo-
gy is a modification to the
channel electric field. That is,
the field plate reduces the elec-
tric field magnitude at the gate
edge, but introduces a high
field region in the conducting
channel, as shown in Figure 14.
For the HFET shown in Figure 14, the device is biased
with a drain voltage of
V
ds
= 100
v. For no field plate,
the electric field magnitude at the gate edge is very high
and well into gate breakdown. The device could not be
operated at this voltage. The
L
FP
= 0.25 µ
m field plate
significantly reduces the electric field at the gate edge,
but introduces a high field region in the conducting
channel, but located a distance from the gate. The
L
FP
= 0.5 µ
m field plate reduces the gate edge electric
field so that it is well below the gate leakage magnitude.
The high field region introduced in the channel adds to
the high field domain that already exists in the channel
due to FET operation. The net high electric field region
in the conducting channel can achieve very high mag-
nitude, and a magnitude sufficient to produce
avalanche ionization, particularly during the high volt-
age portion of the RF cycle.
Figure 14. Electric field under the gate in a FET for different field-plate length for V
ds
=
100 v. (a) No field plate. (b)
L
fp
= 0.25 µm
. (c)
L
fp
= 0.5 µm
14
12
10
8
6
4
2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Length (µm)
× 10
6
(V)
14
12
10
8
6
4
2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Length (µm)
× 10
6
(V)
14
12
10
8
E
6
4
2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Length (µm)
× 10
6
(V)
E Field (V/cm)
Breakdown Electric Field
Gate with Field Plate
Gate
(a) (b) (c)
Figure 15. FET showing avalanche breakdown region and depletion region creating con-
ditions for IMPATT-mode operation.
+++++++
++ −−
++ ++++++++++++
Holes Recombine
with 2DEG Electrons
Increasing Source Resistance
Field Plate
Depletion Region
Electron MotionHole Motion
Generated Holes
Generated Electrons
Impact Ionization Region
i
RF
∼ e
−jωτ
2DEG

66 March 2005
The combination of an avalanche region within the
conducting channel and the existence of a depletion
region provide the conditions for creating an impact
ionization avalanche transit time (IMPATT) mode of
operation. The situation is illustrated in Figure 15. The
avalanche breakdown of the high field region in the
channel generates an electron-hole plasma. The holes
will slowly drift towards the source, where they
recombine with channel electrons. This helps to
reduce the electron density and increase source resis-
tance. The electrons that are generated are injected
into the depletion region where they travel at saturat-
ed velocity towards the drain. The device current is
increased by the generated electrons, thereby enhanc-
ing the self-bias effect. When the combination of the
avalanche breakdown delay time and the transit delay
of the drifting electrons is sufficient to produce a
phase shift between the RF current and voltage that is
greater than
90
◦
, a negative resistance is generated.
The IMPATT mode of operation introduces gain into
the FET operation, and increased RF power results.
Since the IMPATT mode is triggered by the RF signal
applied to the device, the device is essentially a phase-
locked cascade of an FET and IMPATT amplifier. The
added gain compensates for losses in the device and
permits the device to produce high RF power. The
IMPATT mode only exists above a frequency defined
by the depletion region length, which varies with the
magnitude of the RF voltage. The IMPATT mode can
be observed in the output of the device where the
magnitude of S
22
is observed to approach and exceed
unity when the mode is triggered. Also, oscillations
can be observed in the output circuit.
Summary
AlGaN/GaN HFETs show potential for use in
improved RF performance microwave amplifier
applications. Development progress has been rapid,
and prototype devices have demonstrated RF output
power density as high as 30 W/mm. Microwave
amplifier output power is rapidly approaching 100 W
for single-chip operation, and these devices may soon
find application for cellular base station transmitter
applications. Devices are being developed for use in
X-band radars, and RF performance is rapidly
improving. The HFET devices experience several
physical effects that can limit performance. These
effects consist of nonlinearities introduced during the
high-current and high-voltage portions of the RF
cycle. High-current phenomena involve the operation
of the conducting channel above the critical current
density for initiation of space-charge effects. The
source resistance is modulated in magnitude by the
channel current, and high source resistance results.
High voltage effects include reverse leakage of the
gate electrode and subsequent charge trapping effects
on the semiconductor surface, and RF breakdown in
the conducting channel. These effects can produce
premature saturation effects. Also, under certain con-
ditions, high voltage operation of the device can initi-
ate an IMPATT mode of operation. When this occurs
the channel current increases and RF gain is increased.
This phenomenon enhances the RF output power of
the device. The physical limiting effects can be con-
trolled with proper design, and the outlook for use of
these devices in practical applications is excellent.
Acknowledgment
This work was partially supported by the U.S. Office of
Naval Research Grant N00014-03-1-0803 and by the U.S.
Army Research Office Grant DAAD19-03-1-0148.
References
[1] Y.F. Wu, B.P. Keller, P. Fini, S. Keller, T.J. Jenkins, L.T. Kehias,
S.P. Denbaars, and U.K. Mishra, “High Al-content AlGaN/GaN
MODFET’s for ultrahigh performance,” IEEE Electron Dev. Let.,
vol. 19, pp. 50–53, Feb. 1998.
[2] A.T. Ping, Q. Chen, J.W. Yang, M.A. Khan, and I. Adesida, “DC
and microwave performance of high-current AlGaN/GaN het-
erostructure field effect transistors grown on p-type SiC sub-
strates,” IEEE Electron Dev. Let., vol. 19, pp. 54–56, Feb. 1998.
[3] G.J. Sullivan, M.Y. Chen, J.A. Higgins, J.W. Yang, Q. Chen,
R.L. Pierson, and B.T. McDermott, “High power 10 GHz operation
of AlGaN HFET’s on insulating SiC,” IEEE Electron Dev. Let.,
vol. 19, pp. 198–200, June 1998.
[4] Q. Chen, J.W. Yang, R. Gaska, M.A. Khan, M.S. Shur, G.J.
Sullivan, A.L. Sailor, J.A. Higgings, A.T. Ping, and I. Adesida,
“High-power 0.25-
µ
m gate doped-channel GaN/AlGaN het-
erostructure field effect transistor,” IEEE Electron Dev. Let., vol.
19, pp. 44–46, Feb. 1998.
[5] L.F. Eastman, V. Tilak, J. Smart, B.M. Green, E.M. Chumbes, R.
Dimitrov, H. Kim, O.S. Ambacher, N. Weimann, T. Prunty, M.
Murphy, W.J. Schaff, and J.R. Shealy, “Undoped AlGaN/GaN
HEMT’s for microwave power amplification,’’ IEEE Trans. Electron
Dev., vol. 48, pp. 479–485, March 2001.
[6] Y.F. Wu, D. Kapolnek, J.P. Ibbetson, P. Parikh, B. Keller, and
U.K. Mishra, “Very-high power density AlGaN/GaN HEMT’s,’’
IEEE Trans. Electron Dev., vol. 48, pp. 586–590, Mar. 2001.
[7] V. Tilak, B. Green, V. Kaper, H. Kim, T. Prunty, J. Smart, J. Shealy,
and L. Eastman, “Influence of barrier thickness on the high power
performance of AlGaN/GaN HEMT’s,’’ IEEE Electron Dev. Lett.,
vol. 22, pp. 504–506, Nov. 2001.
[8] Y.F. Wu, A. Saxler, M. Moore, R.P. Smith, S. Sheppard,
P.M. Chavarkar, T. Wisleder, U.K. Mishra, and P. Parikh, “30-
W/mm GaN HEMTs by field plate optimization,’’ IEEE Electron
Dev. Lett., vol. 25, pp. 117–119, Nov. 2004.
[9] B. Jogai, “Influence of surface states on the two-dimensional elec-
tron gas in AlGaN/GaN heterojunction field-effect transistors,” J.
Appl. Phys., vol. 93, no. 3, p. 1631–1635, Feb. 2003.
[10] R.J. Trew, “AlGaN/GaN HFET amplifier performance and limi-
tations,’’ in 2002 IEEE Int. Microwave Symp. Dig., Seattle, WA, vol.
3, pp. 1811–1814.
[11] R.J. Trew, “SiC and GaN transistors: Is there one winner for
microwave power applications?,’’ Proc. IEEE, Special Issue on Wide
Bandgap Semiconductors, vol. 90, pp. 1032–1047, June 2002.
[12] R.J. Trew and U.K. Mishra, “Gate Breakdown in MESFETs and
HEMTs,” IEEE Electron Dev. Lett., vol. 12, pp. 524–526, Oct. 1991.
[13] T.A. Winslow, R.J. Trew, and A.S. Morris, “An improved gate
breakdown model for studying high efficiency MESFET operation,”
Proc. IEEE/Cornell University Conf. on Advanced Concepts in High Speed
Semiconductor Devices and Circuits, Ithaca, NY, 1993, pp. 401–410.
[14] T.A. Winslow and R.J. Trew, “Principles of large-signal MESFET
operation,” IEEE Trans. Microwave Theory Tech., vol. 42, pp.
935–942, June 1994.