micromachines

Article
Analysis of Trapping Effect on Large-Signal Characteristics of
GaN HEMTs Using X-Parameters and UV Illumination
Kun-Ming Chen 1, * , Chuang-Ju Lin 1,2 , Chia-Wei Chuang 1 , Hsuan-Cheng Pai 3 , Edward-Yi Chang 2
and Guo-Wei Huang 1,3, *

1 Taiwan Semiconductor Research Institute, Hsinchu 300091, Taiwan; yakoai@narlabs.org.tw (C.-J.L.);

cwchuang@narlabs.org.tw (C.-W.C.)
2 International College of Semiconductor Technology, National Yang Ming Chiao Tung University,
Hsinchu 300093, Taiwan; edc@nycu.edu.tw
3 Institute of Electronics, National Yang Ming Chiao Tung University, Hsinchu 300093, Taiwan
* Correspondence: kmchen@narlabs.org.tw (K.-M.C.); gwhuang@narlabs.org.tw (G.-W.H.)

Abstract: GaN high-electron-mobility transistors (HEMTs) have attracted widespread attention

for high-power microwave applications, owing to their superior properties. However, the charge
trapping effect has limitations to its performance. To study the trapping effect on the device large-
signal behavior, AlGaN/GaN HEMTs and metal-insulator-semiconductor HEMTs (MIS-HEMTs) were
characterized through X-parameter measurements under ultraviolet (UV) illumination. For HEMTs
without passivation, the magnitude of the large-signal output wave (XFB 21 ) and small-signal forward
S
gain (X2111 ) at fundamental frequency increased, whereas the large-signal second harmonic output
wave (XFB22 ) decreased when the device was exposed to UV light, resulting from the photoconductive
effect and suppression of buffer-related trapping. For MIS-HEMTs with SiN passivation, much
higher XFB S
21 and X2111 have been obtained compared with HEMTs. It suggests that better RF power
performance can be achieved by removing the surface state. Moreover, the X-parameters of the
MIS-HEMT are less dependent on UV light, since the light-induced performance enhancement is
offset by excess traps in the SiN layer excited by UV light. The radio frequency (RF) power parameters
and signal waveforms were further obtained based on the X-parameter model. The variation of RF
Citation: Chen, K.-M.; Lin, C.-J.; current gain and distortion with light was consistent with the measurement results of X-parameters.
Chuang, C.-W.; Pai, H.-C.; Chang, Therefore, the trap number in the AlGaN surface, GaN buffer, and SiN layer must be minimized for a
E.-Y.; Huang, G.-W. Analysis of good large-signal performance of AlGaN/GaN transistors.
Trapping Effect on Large-Signal
Characteristics of GaN HEMTs Using Keywords: GaN; HEMT; large-signal characterization; trapping effect; UV illumination; X-parameter
X-Parameters and UV Illumination.
Micromachines 2023, 14, 1011.
https://doi.org/10.3390/mi14051011

Academic Editor: Yu-Shyan Lin 1. Introduction

GaN-based HEMTs have been widely investigated for next-generation wireless com-
Received: 14 April 2023
munication and power electronics applications [1–4]. Owing to the high breakdown voltage
Revised: 5 May 2023
Accepted: 6 May 2023
and high electron velocity, they exhibit excellent RF power performance at high frequen-
Published: 8 May 2023
cies. GaN HEMTs fabricated on sapphire substrate were demonstrated to be operated
at 94 GHz with an output power density of 5.8 W/mm and a power-added efficiency of
38.5% [3]. Moreover, for the purposes of keeping costs low and integrating with Si-based
devices, GaN-on-Si technology has also been developed [5,6]. A high maximum oscillation
Copyright: © 2023 by the authors. frequency (fmax ) of 270 GHz has been achieved for GaN-on-Si HEMT, suggesting its great
Licensee MDPI, Basel, Switzerland. application potential for millimeter-wave power circuits. However, the reliability issues
This article is an open access article related to gate leakage and current collapse degrades the output power and efficiency of
distributed under the terms and GaN-based devices [7,8]. The current collapse phenomena are associated with the presence
conditions of the Creative Commons of charge trapping states at the surface, or in the GaN buffer [9,10]. One of the popular
Attribution (CC BY) license (https:// methods for reducing the gate leakage current is introducing a SiN gate dielectric layer
creativecommons.org/licenses/by/ under the metal gate. The SiN layer in this MIS-HEMT can also suppress the surface trap
4.0/).

Micromachines 2023, 14, 1011. https://doi.org/10.3390/mi14051011 https://www.mdpi.com/journal/micromachines

Micromachines 2023, 14, 1011 2 of 9

states [11]. Although the MIS-HEMT shows lower gate leakage and less surface charge trap-
ping effect, it still suffers from buffer-induced current collapse. Furthermore, the existence
of traps in SiN gate dielectric will result in threshold voltage instability [12].
In order to optimize the device RF power performance and build a large-signal model
that takes the trapping effect into account, the charge trapping behavior of GaN-based
HEMTs has been investigated through S-parameter and load-pull measurements [8,13,14].
Various large-signal models including the trapping-related components were proposed
to simulate the device nonlinear characteristics. The common modeling methods are
based on equivalent circuits with current and charge sources represented by closed-form
equations [15–17]. The extraction of model parameters often relies on pulsed current–
voltage (I–V) and multi-bias S-parameter measurements, and the model development
is time consuming. For fast and efficient model extraction, measurement-based models
utilizing a lookup table [18,19] or neural network techniques [20,21] have been developed to
simulate the large-signal characteristics of GaN devices. Other measurement-based models
using X-parameters were also presented in [22,23] for GaN device modeling, but they
did not describe the charge trapping effect. X-parameters are based on the polyharmonic
distortion (PHD) behavioral model and can be regarded as an extension of the well-known
S-parameters under large-signal conditions. They capture the device behavior at the
fundamental and harmonic frequencies in a single measurement for a given bias point.
Therefore, the nonlinear microwave 2-port networks can be described by X-parameters.
Recently, X-parameters have been utilized in device nonlinear characterization, large-signal
modeling, and circuit design [22–25].
In this study, we investigate the influence of charge trapping on the large-signal
characteristics of AlGaN/GaN HEMTs and MIS-HEMTs by analyzing X-parameters. In
order to change the charge status of traps, a 365 nm UV LED source was used to illuminate
the wafer during X-parameter measurements. Therefore, we can use the UV illumination
method to examine the influence of different trap types on the device characteristics. The
main contents are organized as follows. Section 2 describes the device structures used
in this study and outlines the concept of X-parameters. In Section 3, the X-parameter
measurement results of AlGaN/GaN HEMTs and MIS-HEMTs under UV illumination are
demonstrated. The RF signal waveforms simulated based on the X-parameter model are
also presented. Section 4 provides the conclusion.

2. Materials and Methods

The test devices used in this study were depletion-mode unpassivated Schottky-gate
AlGaN/GaN HEMTs and passivated AlGaN/GaN MIS-HEMTs. The schematic cross
sections of device structures are shown in Figure 1. The epitaxial layer was grown by metal-
organic chemical vapor deposition (MOCVD) on 6-inch silicon substrate. It consisted of an
unintentionally doped 3 µm thick GaN buffer and a 25 nm thick Al0.22 Ga0.78 N barrier layer.
For MIS-HEMTs, an additional 25 nm thick SiN layer was deposited on top of the AlGaN
barrier by plasma-enhanced chemical vapor deposition (PECVD) at 300 ◦ C to passivate the
surface trap states and serve as a gate dielectric. After mesa isolation, a Ti/Al/Ni/Au stack
was formed by electron-beam evaporation and annealed at 800 ◦ C for 60 s as source/drain
ohmic contact. Finally, the Ni/Au gate electrode was evaporated and patterned by a metal
lift-off process. The fabricated devices consisted of two gate fingers with a gate width
of 2 × 25 µm and a gate length of 3 µm. The gate–source spacing (LGS ) and gate–drain
spacing (LGD ) were 2.5 µm and 10 µm, respectively.
X-parameter measurements were performed using the Keysight nonlinear vector network
analyzer (NVNA). The system is based on a Keysight N5242B PNA-X dual-source network
analyzer with a phase reference comb generator for phase calibration and a NVNA application
software for instrument control for highly automated X-parameter characterization and
extraction. The expression for X-parameters used in this study is given by [26]

pm (|A11 |)P + ∑ Xpmqn (|A11 |)P aqn + ∑ XTpmqn (|A11 |)Pm+n aqn
m−n ∗
bpm = XFB m S
(1)
q,n q,n
Micromachines 2023, 14, 1011 3 of 9

where bpm is the scattered wave, aqn is the small-signal incident wave, q and p are the
port numbers, n and m are the harmonic orders, A11 is the large-signal incident wave at
port 1 with fundamental frequency, and P is the unit length phasor with the same phase
as A11 . The XFB S T
pm , Xpmqn , and Xpmqn parameters are characterized as a function of A11 and
used to generate a lookup table model to provide the X-parameter representation of the
device. XFB S
pm represents the response of the device to a large input signal, while Xpmqn and
XTpmqn depict the linear mapping of small-signal incident waves into scattered waves.3 The
Micromachines 2023, 14, x FOR PEER REVIEW of 9
characterization was carried out by applying to each port and each harmonic the reference
resistance of 50 Ω.

(a) (b)
Figure
Figure 1.
1. Schematic
Schematic cross
cross sections
sections of
of (a)
(a) AlGaN/GaN
AlGaN/GaNHEMT
HEMTand
and(b)
(b)AlGaN/GaN
AlGaN/GaNMIS-HEMT.
MIS-HEMT.

3. Results and Discussion

X-parameter measurements were performed using the Keysight nonlinear vector net-
3.1. Measured
work analyzerX-Parameters
(NVNA). The of AlGaN/GaN HEMTs
system is based on a Keysight N5242B PNA-X dual-source
FB FB
network Figure analyzer with the
2a displays a phase reference
magnitudes comb
of X 21 , X22 , and XFB
generator 23for phaseinput
versus calibration
power for andan a
NVNA
AlGaN/GaN applicationHEMTsoftware
biased at forgate
instrument GS = −1.5
voltage Vcontrol for Vhighly automated
and drain voltageX-parameter
VDS = 20 V.
characterization
This bias condition and extraction.toThe
corresponds expression
the bias point for forpeak
X-parameters
current-gainused cutoff infrequency
this study(fTis).
X-parameters
given by [26] were measured at an excitation frequency of 1 GHz and the third-order har-
monics were selected. As shown in this figure, the illuminated device exhibited higher XFB 21 ,
lower XFB , and similar X FB
compared with the dark one. Since X FB
, XFB
, and X FB
are related
22
bpm =XFB |A1123
pm second| Pm + ∑ , XSpmqn |A11 | Pm-n aqn ∑q,n X21 T 22
|A | Pm+n 23 *
aqn (1)
to the fundamental, harmonic and third harmonic output pmqn powers,11respectively, these
results bindicate
where pm is thethat the output
scattered wave, power
aqn isandthe linearity
small-signalof HEMT
incidenthavewave,
been qimproved
and p areunder UV
the port
illumination.
numbers, n and Because
m arethethephoton
harmonic energy
orders,of 365
A11nm UVlarge-signal
is the light corresponds incident to the
wavebandgap
at portof 1
GaN, the photons are only absorbed in the GaN buffer layer,
with fundamental frequency, and P is the unit length phasor with the same phase as A11. generating electron–hole pairs
in this
The XFB region. S The photo-generated
T electrons flow into the two-dimensional electron gas
pm , Xpmqn , and Xpmqn parameters are characterized as a function of A11 and used
(2-DEG) channel and increase
to generate a lookup table model to provide the drain current
the IX-parameter
D (i.e., photoconductive
representation effect),
of thewhile the
device.
generated holes accumulate in the GaN buffer [27]. The photo-generated holes neutralize
XFB
pm represents the response of the device to a large input signal, while Xpmqn and Xpmqn
S T
the trapped electrons in the buffer layer. The suppression of buffer-related charge trapping
depict the linear mapping of small-signal incident waves into scattered waves. The char-
under UV illumination has been reported based on the pulse measurement method [28]. In
acterization was carried out by applying to each port and each harmonic the reference
previous literature [29], it was mentioned that the trapped electrons may also release from
resistance of 50 Ω.
the surface state after gaining the photon energy, but this phenomenon was not noticeable
in our device [30]. The photoconductive effect increases the 2-DEG concentration, which
3. Results and Discussion
enhances the transconductance (gm ) and reduces the source/drain access resistances. When
3.1. Measured
electrons X-Parameters
are captured by theof AlGaN/GaN
traps near the HEMTs
gate edge in the GaN buffer, the 2-DEG concen-
trationFigure 2a displays the magnitudes of XFB
in the gate–drain access region is depleted, FB
leading to FB a high drain resistance. The
21 , X22 , and X23 versus input power for an
suppression of
AlGaN/GaN buffer-related
HEMT biased at charge trapping
gate voltage VGS =by−1.5
UVVlight
and willdrain therefore
voltage lowerVDS = 20theV.drain
This
FB
resistance
bias condition [30].corresponds
As a result, theto theilluminated
bias pointdevice
for peak exhibits higher X
current-gain 21 than
cutoff the dark(fone.
frequency T). X-
FB FB
To explain
parameters werethe variations
measured at anof X 22 and X23
excitation with UVof
frequency light,
1 GHz weandcalculated the transcon-
the third-order har-
ductance (g ), first g derivative (g 0 ) and second g derivative (g ”) from the measured
monics were selected. As shown in this figure, the illuminated device exhibited higher
m m m m m
IDFB-V
X GS curves. FB The results areFB shown in Figure 2b. At VGSone. = −1.5 V, the FB changes
, XFB of gFB m,
210, lower X22 , and similar X23 compared with the dark FB FB
Since X
FB21 22 , and X23
g , and
m related m
are g ” with light are consistent with
to the fundamental, second harmonic and third those of X , X
21 harmonic
22 , and X .
output
23 Therefore,
powers, re-the
linearity of these
spectively, these results
devicesindicate
is highly affected
that by transconductance
the output power and linearity and its derivatives.
of HEMT have For
beena
low-distortion
improved under operation, the magnitude
UV illumination. Because differences
the photon between
energyfundamental
of 365 nm UV and harmonic
light corre-
output powers should be as large as possible. For HEMTs
sponds to the bandgap of GaN, the photons are only absorbed in the GaN buffer layer, under UV illumination, these
generating electron–hole pairs in this region. The photo-generated electrons flow into the
two-dimensional electron gas (2-DEG) channel and increase the drain current ID (i.e., pho-
toconductive effect), while the generated holes accumulate in the GaN buffer [27]. The
photo-generated holes neutralize the trapped electrons in the buffer layer. The suppres-
 Micromachines 2023, 14, x FOR PEER REVIEW 4 of 9

increases the 2-DEG concentration, which enhances the transconductance (gm) and re-
duces the source/drain access resistances. When electrons are captured by the traps near
Micromachines 2023, 14, 1011 4 of 9
the gate edge in the GaN buffer, the 2-DEG concentration in the gate–drain access region
is depleted, leading to a high drain resistance. The suppression of buffer-related charge
trapping by UV light will therefore lower the drain resistance [30]. As a result, the illumi-
nated device
differences areexhibits
larger higher XFB
than the thaninthe
21ones thedark one.indicating that the linearity has been
dark,
improved by the photoconductivity effect and suppression of buffer-related trapping.

(a) (b)
FB FB FB
Figure
Figure 2. 2.(a)(a)Magnitudes
Magnitudesof
of XXFB
21,, XX22,, and
FB and XXFB versus input
23 versus input
power for
power for an
anAlGaN/GaN
AlGaN/GaN HEMT
HEMTat at
21 22 23
VGS = −1.5 V and VDS = 20 V. (b) Magnitudes of gm, gm′, and
0 gm″ versus gate voltage at VDS = 20 V.
VGS = −1.5 V and VDS = 20 V. (b) Magnitudes of gm , gm , and gm ” versus gate voltage at VDS = 20 V.

To explain
Figure 3 shows the variations
the magnitudesof XFB
22 and XFB
of XS1111 XS2111UV
23 , with , XS2121
light, we X
, and T
calculated the transcon-
2121 of an AlGaN/GaN
ductance (g
HEMT. We found that X2111 decreases rapidly with increasing input powermeasured
m), first g m derivative
S (g m′) and second g m derivative (g m ″) from the due to the
ID-VGS curves.
trapping effect. The
Thisresults are shown
observation in Figure
of gain 2b. At Vcannot
compression GS = −1.5 V, the changes of gm, gm′,
be carried out using conven-
FB FB FB
and gS-parameter
tional m″ with light are consistent with those of X21 , X22 , and X23 . Therefore, the linearity
measurements, since they only reveal the small-signal gain S21 . When
ofdevice
the these devices
is exposedis highly
to UVaffected
light,by transconductance
the magnitude of Xand S its derivatives. For a low-dis-
2111 increases, indicating that the
tortion operation, the magnitude differences between
AlGaN/GaN HEMT will have a higher power gain under UV illumination. fundamental and harmonic output
For the design
powers should be as large as possible. For HEMTs under UV
of input and output matching networks of a power amplifier, X1111 , X2121 , and XT2121 ofillumination,
S S these differ-
ences
the power aretransistor
larger thanare theimportant
ones in theparameters
dark, indicating that the linearity
[31]. However, as depictedhas been improved
in Figure 3, the
by the photoconductivity
T effect and suppression of buffer-related trapping. S
magnitudes of X2121 in dark and light conditions are much smaller than X2121 , indicating
that theFigure
output 3 nonlinear
shows thebehavior
magnitudes is of XS1111 , XS2111
insignificant. , XS2121 , and
Therefore, in XT2121case,
our of an
theAlGaN/GaN
circuit design
S
HEMT. We found that X 2111 decreases rapidly with increasing
can ignore the response of the device to the conjugate input signal, and we only input power due consider
to the
trapping effect.S This observation S of gain compression cannot be carried out using conven-
the values of X1111 and X2121 to optimize the source and load impedances, respectively.
tional S-parameter measurements, since they only reveal the small-signal gain S21. When
From Figure 3, we also observed the change in XS1111 Sand XS2121 under UV light, which
the device is exposed to UV light, the magnitude of X2111 increases, indicating that the
means the input and output reflection coefficients at fundamental frequency have changed,
AlGaN/GaN HEMT will have a higher power gain under UV illumination. For the design
owing to the reduced resistance of the channel, source, and drain regions [26]. From S-
of input and output matching networks of a power amplifier, XS1111 , XS2121 , and XT2121
Micromachines 2023, 14, x FOR PEER REVIEW ofof9
parameter measurements, we can obtain the channel, source, and drain resistances,5 which
the power transistor are important parameters [31]. However, as depicted in Figure 3, the
are 3533, 19, and 186 Ω for the HEMT in the dark, and change to 2681, 17, and 134 Ω under
magnitudes of XT2121 in dark and light conditions are much smaller than XS2121 , indicating
UV illumination, respectively. The reduction in channel resistance will reduce XS2121 and
that the output
S nonlinear behavior is insignificant. Therefore, in our case,S the circuit de-
increase X1111 . The reduction in source resistance will further increase X1111 .
sign can ignore the response of the device to the conjugate input signal, and we only con-
sider the values of XS1111 and XS2121 to optimize the source and load impedances, respec-
tively. From Figure 3, we also observed the change in XS1111 and XS2121 under UV light,
which means the input and output reflection coefficients at fundamental frequency have
changed, owing to the reduced resistance of the channel, source, and drain regions [26].
From S-parameter measurements, we can obtain the channel, source, and drain re-
sistances, which are 3533, 19, and 186 Ω for the HEMT in the dark, and change to 2681, 17,
and 134 Ω under UV illumination, respectively. The reduction in channel resistance will
reduce XS2121 and increase XS1111 . The reduction in source resistance will further increase
XS1111 .

Figure 3. Magnitudes of XSS , XS2111 S

S , X2121
T
S , and X2121
T versus input power for an AlGaN/GaN HEMT.
Figure 3. Magnitudes of X1111
1111 , X2111 , X2121 , and X2121 versus input power for an AlGaN/GaN
HEMT.

3.2. Measured X-Parameters of AlGaN/GaN MIS-HEMTs

The surface trap density in HEMTs can be reduced by depositing a SiN passivation
 Figure 3. Magnitudes of XS1111 , XS2111 , XS2121 , and XT2121 versus input power for an AlGaN/GaN
HEMT.

3.2. Measured X-Parameters of AlGaN/GaN MIS-HEMTs

Micromachines 2023, 14, 1011 5 of 9
The surface trap density in HEMTs can be reduced by depositing a SiN passivation
layer upon the AlGaN barrier. This SiN insulating layer also acts as a gate dielectric to
suppress gate leakage.
3.2. Measured The ofmeasured
X-Parameters AlGaN/GaN main X-parameters of the passivated AlGaN/GaN
MIS-HEMTs
MIS-HEMT in the dark
The surface and under
trap density UV illumination
in HEMTs can be reduced arebydepicted
depositinginaFigure 4. The test de-
SiN passivation
vicelayer
is biased at the condition that the peak f is obtained. Compared
upon the AlGaN barrier. This SiN insulating layer also acts as a gate dielectric
T with HEMT, to MIS-
FB FB FB
HEMTsuppress
showsgate leakage.
higher X21The measured
, and lower X main
22 andX-parameters of the passivated
X23 . It demonstrates thatAlGaN/GaN
the device output
MIS-HEMT in the dark and under UV illumination are
power and linearity can be apparently improved by mitigating the surface depicted in Figure 4. The test trap-
charge
device is biased at the condition that FB the peak fT is obtained. Compared with HEMT,
ping. However, the photo-induced X21 enhancement in MIS-HEMT was not observed
MIS-HEMT shows higher XFB FB FB
21 , and lower X22 and X23 . It demonstrates that the device
clearly when it was exposed to UV light. This is because excess charge trapping centers in
output power and linearity can be apparently improved by mitigating the surface charge
the trapping.
amorphous SiN layer
However, are created by
the photo-induced XFBphotons [32], which degrades gm and thus offsets
21 enhancement in MIS-HEMT was not observedFB
the clearly
benefitwhen
of the photoconductive effect.
it was exposed to UV light. This Even so, we still
is because excessobserved a decrease
charge trapping in X
centers in23 un-
der the
illumination
amorphous SiN duelayer
to theare decrease
created byin drain[32],
photons resistance caused gby
which degrades thethus
m and suppression
offsets of
FB
the benefit of the
buffer-related trapping. photoconductive effect. Even so, we still observed a decrease in X 23
under illumination due to the decrease in drain resistance caused by the suppression of
buffer-related trapping.

(a) (b)
FB FB
Figure 4. (a)
Figure Magnitudes
4. (a) Magnitudesofof X FB , X
X21 FB , and FBXFB
21 , X2222, and X23 23
versus
versus inputinput
powerpowerfor an for an AlGaN/GaN
AlGaN/GaN MIS-HEMT
MIS-HEMT at
at VVGSGS= =−3−V S XS S, XS S , XS T T
3 VandandVVDS
DS==20
20 V.
V. (b)
(b) Magnitudes
Magnitudes of of
X , X , X
111111112111 2111 , and X,
2121 2121 2121and X
versus
2121 versus
input input
power for power
for this
this AlGaN/GaN MIS-HEMT.
AlGaN/GaN MIS-HEMT.

As with the XFB parameters, the XS and XT coefficients of the AlGaN/GaN MIS-HEMT
also show less dependence on UV light, as illustrated in Figure 4b. It should be noted
that the trend of XS1111 variation with UV light is opposite to that in HEMT. This result
might be due to increased gate leakage caused by photo-induced traps in the SiN gate
dielectric. The gate leakage can be denoted by a resistance in parallel with the gate-to-source
capacitance. Higher leakage means lower resistance, resulting in a lower input reflection
coefficient. From Figure 4b, we found many improvements in the properties of MIS-HEMTs
compared to HEMTs due to the reduced surface charge trapping. Firstly, MIS-HEMT
exhibits higher XS2111 and lower XT2121 coefficients than HEMT under both dark and light
conditions, indicating it has higher power gain and better output linearity. Secondly, the
XS2111 gain compression at high input power is not significant in MIS-HEMT. Finally, XS1111
and XS2121 do not vary with the power level, suggesting matching networks are easier to
design when the transistor is used in power amplifier circuits.

3.3. Large-Signal Modeling Results

The large-signal characteristics of AlGaN/GaN HEMTs and MIS-HEMTs were simu-
lated based on the X-parameter model using the ADS circuit simulation tool. The circuit
configuration for large-signal simulation is shown in Figure 5a. The dc biases, input power
(Pin ), and frequency are set to match the X-parameter measurement conditions. The source
 3.3. Large-Signal Modeling Results
The large-signal characteristics of AlGaN/GaN HEMTs and MIS-HEMTs were simu-
lated based on the X-parameter model using the ADS circuit simulation tool. The circuit
configuration for large-signal simulation is shown in Figure 5a. The dc biases, input power
Micromachines 2023, 14, 1011 (Pin), and frequency are set to match the X-parameter measurement conditions. The 6source of 9

and load impedances (ZS and ZL) are 50 Ω. Typical simulated power characteristics are
shown in Figure 5b. The output powers measured by a load-pull system with 50 Ω load
and load impedances
impedance (ZS and ZinL ) this
are also displayed are 50 Ω. Typical
figure simulated
to verify power characteristics
the accuracy are
of the X-parameter
shownItin
model. Figure
can 5b. that
be seen The output powers curves
the simulated measured
agreeby well
a load-pull system
with the with 50
measured Ω load
data; there-
impedance
fore, are also displayed
the X-parameter in this
model can figure the
describe to verify
devicethelarge-signal
accuracy of the X-parameter model.
performance.
It can be seen that the simulated curves agree well with the measured data; therefore, the
X-parameter model can describe the device large-signal performance.

(a) (b)
Figure
Figure5.5.(a)(a)Circuit
Circuitconfiguration
configuration for
for device simulationusing
device simulation usingan
anX-parameter
X-parameter model.
model. (b)(b) Measured
Measured
and simulated RF power characteristics of an AlGaN/GaN
and simulated RF power characteristics of an AlGaN/GaN MIS-HEMT.MIS-HEMT.

Figure66shows
Figure showsthethe gate
gate and drain
drain current
currentwaveforms
waveformsininthe thetime
timedomain
domain obtained
obtained
bybyX-parameter
X-parametermodel
model simulation.
simulation. For HEMT
HEMT devices,
devices,we weobserved
observedsignificant
significantcurrent
current
distortionatat
distortion anan input
input power
power above
above −10−dBm
10 dBm in the
in the dark dark
andand it becomes
it becomes moremore serious
serious when
when devices operate at higher power. Moreover, the average drain current
devices operate at higher power. Moreover, the average drain current decreases with in-decreases with
increasing
creasing RF power
RF power level,
level, demonstrating
demonstrating thatthat
the the distortion
distortion is caused
is caused by the
by the trapping
trapping effect
effect [15]. The drop in average current is a significant consequence of the asymmetric
[15]. The drop in average current is a significant consequence of the asymmetric nature of
nature of the trap capture and emission processes. At high frequencies, when the signal
the trap capture and emission processes. At high frequencies, when the signal period is
period is lower than the charge emission time, the trap status cannot respond quickly
lower than the charge emission time, the trap status cannot respond quickly enough to the
enough to the change of the applied signal, leading to the compression of maximum current.
change of the applied
For MIS-HEMTs, signal, leading
no significant todistortion
current the compression of maximum
was observed current.current
and the average For MIS-
was barely affected by the power level, resulting from the mitigation of surface-related
charge trapping.
When the devices are exposed to UV light, several phenomena are observed in the
current waveforms, owing to the increase in photocurrent and suppression of buffer-
related trapping. (1) The average drain current in HEMT and MIS-HEMT increases. The
light-enhanced drain current is consistent with the dc measurement result [30]. (2) The
peak-to-peak amplitude of drain current, which is associated with XFB21 , for the HEMT under
UV illumination is higher than that in the dark environment. (3) The current distortion in
HEMT is improved under illumination. However, the improvement is minor due to the
serious surface-related trapping still existent in the illuminated HEMT.
For MIS-HEMTs, the UV light will induce excess trap centers in the amorphous SiN
layer, which offsets the performance improvement resulting from the photoconductive
effect and the suppression of buffer charge trapping. Therefore, the peak-to-peak amplitude
of drain current is not influenced by light. Furthermore, the photo-induced traps in SiN
slightly increase the gate current of MIS-HEMTs.
 For MIS-HEMTs, the UV light will induce excess trap centers in the amorphous SiN
layer, which offsets the performance improvement resulting from the photoconductive
effect and the suppression of buffer charge trapping. Therefore, the peak-to-peak ampli-
tude of drain current is not influenced by light. Furthermore, the photo-induced traps in
Micromachines 2023, 14, 1011 SiN slightly increase the gate current of MIS-HEMTs. 7 of 9

(a) (b)

(c) (d)
Figure
Figure 6.
6. (a,b)
(a,b)Gate
Gatecurrent
currentand drain
and current
drain waveforms
current of an
waveforms of AlGaN/GaN HEMT.
an AlGaN/GaN (c,d) (c,d)
HEMT. Gate Gate
cur-
rent and drain current waveforms of an AlGaN/GaN MIS-HEMT.
current and drain current waveforms of an AlGaN/GaN MIS-HEMT.

4. Conclusions
In this work, we have investigated the impact of trapping effect on the large-signal
characteristics of AlGaN/GaN HEMTs and MIS-HEMTs based on X-parameter analysis.
From the X-parameter measurements, it can be seen that the output power, power gain, and
linearity of the MIS-HEMT are much better than those of the HEMT because it has fewer
surface traps. Therefore, traps on the AlGaN surface are the major concern for devices used
in RF power applications and must be removed during fabrication process. Although the
surface state can be removed by passivation technology, the traps in the GaN buffer still
degrade the device power performance. In addition, the quality of the passivation layer
must be taken into account to avoid excess traps induced in this film. Furthermore, we
have used the X-parameter model to simulate the large-signal RF current waveform. It is
more convenient than the equivalent circuit model for assessing the device large-signal
performance. From the modeling result, serious current distortion in HEMTs without
passivation was observed. In addition, the RF current gain in HEMTs was much lower than
that in MIS-HEMTs with passivation. After suppressing the buffer-related trapping effect
with UV illumination, the signal distortion and gain were improved.
Micromachines 2023, 14, 1011 8 of 9

Author Contributions: Conceptualization, K.-M.C. and G.-W.H.; methodology, K.-M.C.; validation,

C.-W.C.; investigation, K.-M.C. and H.-C.P.; resources, E.-Y.C. and G.-W.H.; data curation, C.-J.L. and
H.-C.P.; writing, K.-M.C.; supervision, G.-W.H.; funding acquisition, G.-W.H. All authors have read
and agreed to the published version of the manuscript.
Funding: This work was supported in part by the “Center for the Semiconductor Technology Research”
from The Featured Areas Research Center Program within the framework of the Higher Education
Sprout Project by the Ministry of Education (MOE) in Taiwan, in part by the National Science and
Technology Council, Taiwan, under Grant NSTC-111-2634-F-A49-008, and in part by the National
Chung-Shan Institute of Science & Technology, Taiwan, under Grant No. NCSIST-701-V104 (112).
Data Availability Statement: The data used to support the findings of this study are available within
the article.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Harrouche, K.; Kabouche, R.; Okada, E.; Medjdoub, F. High Power AlN-GaN HEMTs with record power-added-efficiency 70% at
40 GHz. In Proceedings of the IEEE MTT-S International Microwave Symposium, Los Angeles, CA, USA, 4–6 August 2020.
2. Nakajima, S. GaN HEMTs for 5G base station applications. In Proceedings of the IEEE International Electron Devices Meeting,
San Francisco, CA, USA, 1–5 December 2018.
3. Li, W.; Romanczyk, B.; Akso, E.; Guidry, M.; Hatui, N.; Wurm, C.; Liu, W.; Shrestha, P.; Collins, H.; Clymore, C.; et al. Record
94 GHz performance from N-polar GaN-on-sapphire MIS-HEMTs: 5.8 W/mm and 38.5% PAE. In Proceedings of the IEEE
International Electron Devices Meeting, San Francisco, CA, USA, 3–7 December 2022.
4. Wei, J.; Jiang, H.; Jiang, Q.; Chen, K.J. Proposal of a GaN/SiC hybrid field-effect transistor for power switching applications. IEEE
Trans. Electron Devices 2016, 63, 2469–2473. [CrossRef]
5. Cui, P.; Jia, M.; Chen, H.; Lin, G.; Zhang, J.; Gundlach, L.; Xiao, J.Q.; Zeng, Y. InAlN/GaN HEMT on Si with fmax = 270 GHz. IEEE
Trans. Electron Devices 2021, 68, 994–999. [CrossRef]
6. Marti, D.; Lugani, L.; Carlin, J.F.; Malinverni, M.; Grandjean, N.; Bolognesi, C.R. W-band MMIC amplifiers based on AlInN/GaN
HEMTs grown on silicon. IEEE Electron Device Lett. 2016, 37, 1025–1028. [CrossRef]
7. Chu, R.; Shen, L.; Fichtenbaum, N.; Chen, Z.; Keller, S.; DenBaars, S.P.; Mishra, U.K. Correlation between DC–RF dispersion and
gate leakage in deeply recessed GaN/AlGaN/GaN HEMTs. IEEE Electron Device Lett. 2008, 29, 303–305.
8. Bergsten, J.; Thorsell, M.; Adolp, H.D.; Chen, J.-T.; Kordina, O.; Sveinbjörnsson, E.O.; Rorsman, N. Electron trapping in extended
defects in microwave AlGaN/GaN HEMTs with carbon-doped buffers. IEEE Trans. Electron Devices 2018, 65, 2446–2453. [CrossRef]
9. Strenaer, R.; Guhel, Y.; Gaquière, C.; Boudart, B. Trapping effect in AlInN/GaN HEMTs: A study based on photoionization and
pulsed electrical measurements. IEEE Trans. Electron Devices 2022, 69, 6010–6015. [CrossRef]
10. Lin, Y.; Kao, M.-L.; Weng, Y.-C.; Dee, C.-F.; Chen, S.-C.; Kuo, H.-C.; Lin, C.-H.; Chang, E.-Y. Buffer traps effect on GaN-on-Si
high-electron-mobility transistor at different substrate voltages. Micromachines 2022, 13, 2140. [CrossRef]
11. Lu, X.; Yu, K.; Jiang, H.; Zhang, A.; Lau, K.M. Study of Interface traps in AlGaN/GaN MISHEMTs using LPCVD SiNx as gate
dielectric. IEEE Trans. Electron Devices 2017, 64, 824–831. [CrossRef]
12. Meneghesso, G.; Meneghini, M.; Santi, C.D.; Ruzzarin, M.; Zanoni, E. Positive and negative threshold voltage instability in GaN
based transistors. Microelectron. Reliab. 2018, 80, 257–265. [CrossRef]
13. Liu, J.; Mi, M.; Zhu, J.; Liu, S.; Wang, P.; Zhou, Y.; Zhu, Q.; Wu, M.; Lu, H.; Hou, B.; et al. Improved power performance and the
mechanism of AlGaN/GaN HEMTs using Si-rich SiN/Si3 N4 bilayer passivation. IEEE Trans. Electron Devices 2022, 69, 631–636.
[CrossRef]
14. Aoki, H.; Sakairi, H.; Kuroda, N.; Nakamura, Y.; Chikamatsu, K.; Nakahara, K. AlGaN/GaN MIS HEMT modeling of frequency
dispersion and self-heating effects. In Proceedings of the IEEE International Symposium on Radio-Frequency Integration
Technology (RFIT), Melbourne, VIC, Australia, 15–17 August 2018.
15. Jardel, O.; Groote, F.D.; Reveyrand, T.; Jacquet, J.-C.; Charbonniaud, C.; Teyssier, J.-P.; Floriot, D.; Quéré, R. An electrothermal
model for AlGaN/GaN power HEMTs including trapping effects to improve large-signal simulation results on high VSWR. IEEE
Trans. Microw. Theory Tech. 2007, 55, 2660–2669. [CrossRef]
16. Jarndal, A.; Kompa, G. Large-signal model for AlGaN/GaN HEMTs accurately predicts trapping- and self-heating-induced
dispersion and intermodulation distortion. IEEE Trans. Electron Devices 2007, 54, 2830–2836. [CrossRef]
17. Yuk, K.; Branner, G.R.; McQuate, D. An improved empirical large-signal model for high-power GaN HEMTs including self-heating
and charge-trapping effects. In Proceedings of the IEEE MTT-S International Microwave Symposium Digest, Boston, MA, USA,
7–12 June 2009.
18. Martín-Guerrero, T.M.; Santarelli, A.; Gibiino, G.P.; Traverso, P.A.; Camacho-Peñalosa, C.; Filicori, F. Automatic extraction
of measurement-based large-signal FET Models by nonlinear function sampling. IEEE Trans. Microw. Theory Tech. 2020, 68,
1627–1636. [CrossRef]
Micromachines 2023, 14, 1011 9 of 9

19. Niessen, D.; Gibiino, G.P.; Cignani, R.; Santarelli, A.; Schreurs, D.M.M.-P.; Filicori, F. Charge-controlled GaN FET modeling by
displacement current integration from frequency-domain NVNA measurements. IEEE Trans. Microw. Theory Tech. 2016, 64,
4382–4393. [CrossRef]
20. Xu, J.; Jones, R.; Harris, S.A.; Nielsen, T.; Root, D.E. Dynamic FET model–DynaFET–for GaN transistors from NVNA active source
injection measurements. In Proceedings of the IEEE MTT-S International Microwave Symposium, Tampa, FL, USA, 1–6 June 2014.
21. Barmuta, P.; Płoński, P.; Czuba, K.; Avolio, G.; Schreurs, D. Nonlinear AlGaN/GaN HEMT model using multiple artificial neural
networks. In Proceedings of the International Conference on Microwaves, Radar & Wireless Communications, Warsaw, Poland,
21–23 May 2012.
22. Horn, J.; Root, D.E.; Simpson, G. GaN device modeling with X-parameters. In Proceedings of the IEEE Compound Semiconductor
Integrated Circuit Symposium, Monterey, CA, USA, 3–6 October 2010.
23. Ayari, L.; Xiong, A.; Maziere, C.; Ouardirhi, Z.; Gasseling, T. A robust and reliable behavioral model of high power GaN HEMTS
for RF Doherty amplifier application. In Proceedings of the ARFTG Microwave Measurement Conference, Philadelphia, PA, USA,
10–15 June 2018.
24. Verspecht, J.; Root, D.E. Polyharmonic distortion modeling. IEEE Microw. Mag. 2006, 7, 44–57. [CrossRef]
25. Pelaez-Perez, A.M.; Woodington, S.; Fernandez-Barciela, M.; Tasker, P.J.; Alonso, J.I. Large-signal oscillator design procedure
utilizing analytical X-parameters closed-form expressions. IEEE Trans. Microw. Theory Tech. 2012, 60, 3126–3136. [CrossRef]
26. Baylis, C.; Marks, R.J.; Martin, J.; Miller, H.; Moldovan, M. Going nonlinear. IEEE Microw. Mag. 2011, 12, 55–64. [CrossRef]
27. Zaidi, Z.H.; Houston, P.A. Highly sensitivity UV detection mechanism in AlGaN/GaN HEMTs. IEEE Trans. Electron Devices 2013,
60, 2776–2781. [CrossRef]
28. Nagarajan, V.; Chen, K.-M.; Chen, B.-Y.; Huang, G.-W.; Chuang, C.-W.; Lin, C.-J.; Anandan, D.; Wu, C.-H.; Han, P.-C.;
Singh, S.K.; et al. Study of charge trapping effects on AlGaN/GaN HEMTs under UV illumination with pulsed I-V measurement.
IEEE Trans. Device Mater. Reliab. 2020, 20, 436–441. [CrossRef]
29. Vetury, R.; Zhang, N.Q.; Keller, S.; Mishra, U.K. The impact of surface states on the DC and RF characteristics of AlGaN/GaN
HFETs. IEEE Trans. Electron Devices 2001, 48, 560–566. [CrossRef]
30. Chen, K.-M.; Lin, C.-J.; Nagarajan, V.; Chang, E.Y.; Lin, C.-W.; Huang, G.-W. Analysis of high-frequency behavior of AlGaN/GaN
HEMTs and MIS-HEMTs under UV illumination. ECS J. Solid State Sci. Technol. 2021, 10, 055004. [CrossRef]
31. Root, D.E.; Verspecht, J.; Xu, J. Closed-form solutions to large-signal PA problems: Wirtinger calculus applied to X-parameter. In
Proceedings of the 12th European Microwave Integrated Circuits Conference (EuMIC), Nuremberg, Germany, 8–10 October 2017.
32. Warren, W.L.; Kanicki, J.; Robertson, J.; Poindexter, E.H.; McWhorter, P.J. Electron paramagnetic resonance investigation of charge
trapping centers in amorphous silicon nitride films. J. Appl. Phys. 1993, 74, 4034–4046. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.