Chip Review

MICRO-NANOELECTRONICS DOI: 10.1016/j.chip.2023.100052

On large-signal modeling of GaN HEMTs:

past, development and future
Haorui Luo1,2 , Wenrui Hu1 & Yongxin Guo1,2,∗
1 dard output power were plotted against working frequency of different
Department of Electrical and Computer Engineering, National University of
Singapore, Singapore 117576, Singapore 2 National University of Singapore semiconductor materials and different transistors5 . It can be observed that
(Suzhou) Research Institute, Suzhou 215123, China
the line of wide bandgap third-generation semiconductor materials, repre-
E-mail: [email protected] (Yongxin Guo) sented by GaN, covers both a high power and wide frequency range, indi-
Cite as: Luo, H., Hu, W. & Guo, Y. On large-signal modeling of GaN HEMTs: past, cating the huge potential of GaN in both high-power and high-frequency
development and future. Chip 2, 100052 (2023). applications. Given the excellent high-power and high-frequency charac-
https://doi.org/10.1016/j.chip.2023.100052 teristics of high electron mobility transistor (HEMT), the GaN HEMT has
Received: 21 January 2023 greatly attracted many researchers’ interest since it was invented in 19936 .
Nowadays, GaN HEMT has been widely used in many key areas which
Accepted: 17 April 2023
are closely related to people’s lives, including new mobile communica-
Published online: 16 May 2023 tions, radar systems and object detection, etc.
An accurate and reliable device model is necessary for the efficient
In the past few decades, circuits based on gallium nitride high elec-
and reliable delivery of the advances from GaN HEMT devices to GaN-
tron mobility transistor (GaN HEMT) have demonstrated exceptional
HEMT-based circuits. In addition, the device model also needs to be fea-
potential in a wide range of high-power and high-frequency applica-
tured with good simulation convergence and fast simulation time. Fig. 2
tions, such as the new generation mobile communications, object de-
presents some landmarks of GaN HEMT large-signal models. In this
tection and consumer electronics, etc. As a critical intermediary be-
timeline, the typical models to appear the earliest were empirical mod-
tween GaN HEMT devices and circuit-level applications, GaN HEMT
els, which is mainly ascribed to the fact that empirical models use rela-
large-signal models play a pivotal role in the design, application and
tively simple expressions to characterize HEMT performances. Some rep-
development of GaN HEMT devices and circuits. This review pro-
resentative empirical models are the Statz7 and Crutice8 models, Angelov
vides an in-depth examination of the advancements in GaN HEMT
model9–11 , and EEHEMT model12 . These empirical models mainly ap-
large-signal modeling in recent decades. Detailed and comprehensive
peared in the 1990s. However, the vitality of the empirical models did not
coverage of various aspects of GaN HEMT large-signal model was of-
vanish with the development of the times. Even in recent years, a large
fered, including large-signal measurement setups, classical formula-
number of modified Angelov models13–21 have been appearing to meet
tion methods, model classification and non-ideal effects, etc. In order
the emerging needs of different new application scenarios. The empiri-
to better serve follow-up researches, this review also explored poten-
cal models exhibit the characteristics of being accurate, easy to be used
tial future directions for the development of GaN HEMT large-signal
and highly tunable, thus making them widely applied in many industrial
modeling.
scenarios so that the device model can be developed fast and flexibly.
Keywords: Gallium nitride, High electron mobility transistors, Large- There are many machine-learning (ML)-based GaN HEMT models, in
signal model which the artificial neural network (ANN)-based GaN HEMT models are
widely used. The ANN models first appearing in around early 2000s22,23 ,
is an interesting attempt by introducing machine learning methods to the
GaN HEMT device modeling field. Due to the strong fitting ability and
INTRODUCTION the low deployment cost of the ANN, the ANN-based models are natu-
The electronic information industry with semiconductor devices and cir- rally capable of accurately describing a device’s nonlinearity with a low
cuits serving as the pillars has witnessed rapid development in recent computational cost. One well-known example of the ANN-based mod-
decades, which has greatly promoted the prosperity of the personal con- els is the DynaFET model24 , 25 proposed by Keysight. Later on, in 2010s,
sumer electronics industry, the aerospace field and the smart society. In people began to focus on building GaN HEMT models from the under-
turn, the boom on the demand side has also brought continuous iterative lying physical properties of the device. For example, there are the Ad-
updates and diversified development of semiconductor devices. The ma- vanced SPICE Model (ASM) HEMT model26 , 27 , the MIT Virtual Source
terial properties of some common semiconductor materials are listed in (MVS) HEMT model28 , 29 , the Hiroshima-University Starc Igfet Model
Table 11–4 , the wide bandgap and high saturation velocity make gallium for GaN HEMT30 (HiSIM GaN HEMT model), and the École Polytech-
nitride (GaN) a promising candidate in high power applications. And the nique Fédérale de Lausanne (EPFL) HEMT model31 , Physical models are
satisfactory electron mobility and high frequency dielectric constant en- based on rigorous physical equations. Although some approximations of
sure a good high frequency performance of GaN. The material advantages physical equations need to be done to make physical model robust and
of GaN were embodied clearly in Fig. 1, in which the curves of stan- compact, the physical model still exhibits a much better scalability than

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 1 of 12
Review DOI: 10.1016/j.chip.2023.100052

Fig. 1 | Saturated output power versus frequency for power amplifiers based on different transistor technologies5 . The discrete points are used to represent the
performance of designs in open literatures. The lines are used to represent the trend of performance limit of technologies.

Fig. 2 | Timeline of some typical GaN HEMT large-signal models.

empirical models. The physical models also can easily describe devices poorly obtained. The complexity and poor tunability of physical models
made by different materials. There have been attempts to merge differ- may limit their wide applications.
ent categories of models. For instance, the quasi-physical zone division In this article, the development of GaN HEMT large-signal models
(QPZD) model32 includes both empirical and physical parameters. One in recent decades have been reviewed, including the model classification,
hybrid model33 proposed in 2022 integrates ANN with a classical physical large-signal equivalent circuits and large-signal measurement, etc. The
model. modeling methods for trapping effects and self-heating effects will be cov-
When it comes to the 2020s and beyond, the GaN HEMT large-signal ered as well.
models are also facing many challenges. Some of them come from the
advancing of semiconductor technologies. The continuingly scaling of LARGE-SIGNAL MEASUREMENTS
gate length leads to an increase of operating frequency, while the short-
channel effects become more obvious28 , 34 . The innovation on substrate The validity of large-signal models needs to be examined under large-
materials35 , 36 brings about better device performance and different char- signal conditions. Load-pull measurement is one common item among
acteristics at the same time. It is worth looking forward to whether the many validation metrics. A typical large-signal load-pull measurement
rapid development of artificial intelligence and measurement technology setup in our lab is shown in Fig. 3a. Its core measuring instruments include
can inspire novel device models. For existing large-signal models, the em- the Rohde & Schwarz ZVA50, the Muary AM3200 pulsed IV system,
pirical models are usually lack of physical meaning, which leads to poor the Muary automated source tuner and load tuner, the BONN Elektronik
scalability. The training of ANN-based models requires a high computing power amplifier, attenuator, isolator, 50- terminator, couplers and high-
resource, and it may have a non-convergence issue if the parameters are power bias tees. In addition to the traditional continuous-wave load-pull

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 2 of 12
Review DOI: 10.1016/j.chip.2023.100052

Fig. 3 | Measurement equipment and results. a, Large-signal measurement setup. b, Measured pulsed power-added-efficiency contours. c, Measured pulsed output
power contours. d, Measured pulsed IV curves. e, Measured multi-bias S-parameters (S21).

Fig. 4 | The micrograph of a GaN HEMT sample and its equivalent circuits. Different types of device model can be obtained by using different types of intrinsic
models.

measurement, this system also supports pulse mode measurement with an idea transistor, while the extrinsic part usually refers to the structures
the minimum pulse width of 200 ns. The pulsed power-added-efficiency used to construct the HEMT. Usually, the modeling method of the intrinsic
(PAE) contours and pulsed output power contours measured by this sys- transistor determines the model type, for example, the empirical models,
tem are shown in Fig. 3b and Fig. 3c. The system can also measure the physical models, or the ML-based models.
current-voltage (IV) curves and S-parameters, as shown in Fig. 3d and Since the difference of the equivalent circuit will directly affect
Fig. 3e, respectively. the performance of the model, the extraction of its parameters has
always been one focus in device modeling. A wide variety of ex-
LARGE-SIGNAL EQUIVALENT CIRCUITS traction methods have been reported, including the traditional cold-
condition extraction and its improvements37–41 , the extraction of resis-
The equivalent circuit takes a very important and basic role in device mod- tances14 , 42 , 43 , the algorithm improvements44 , 45 , the structure-based ex-
eling. As shown in Fig. 4, an equivalent can be divided into two parts: traction46 , and the extraction method for novel devices47 , etc. Mature
the intrinsic part and extrinsic one. The intrinsic part usually refers to and successful extraction methods can not only ensure modeling ac-

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 3 of 12
Review DOI: 10.1016/j.chip.2023.100052

curacy, but also naturally create favorable conditions for researches on

scalability46,48–53 .

Power density
EXAMPLES OF CLASSICAL LARGE-SIGNAL

Medium
MODELS

High

Low
Low
Low
In this section, several widely used large-signal models will be introduced

Melting point (K)

in detail. We will elaborate the formulation method of one common model
from the empirical models, physical models and the machine-learning
2791 based models, respectively.
2730
1062
1238
1415
Empirical Angelov model The Angelov model is one widely used empir-
ical model developed by Dr. Iltcho Angelov in 199210 . This model was
developed based on the Statz7 and Curtice8 models. Later, it went through
conductivity
(W/cm·K)

two major improvements in 19969 and 199911 . The validity of the An-
Thermal

gelov model has been verified in many publications, including the high-
0.55

frequency applications54 , trapping effects modeling17 , electrothermal ef-

1.5
3.8
0.7

1.5

fects modeling16 , 17 , 21 , 55 and its good applicability to machine-learning

based modeling56 .
dielectric constant
High frequency

The Angelov model adopts independent current source and capaci-

tance source with flexible empirical equations, so that it has a strong tun-
ability. The drain current is formulated as follows:
10.9
11.8

Ids = Ipk (1 + tanh (sinh (ψ ) ) )(1 + λVds )tanh(αVds )

5.3
6.7
9.6

(1)
where, the Ipk is the maximum drain current at maximum transconduc-
tance point, λ is the channel length modulation parameter, α governs the
Static dielectric

transition from linear to saturation region. The ψ is the power series pa-
rameter, and it is defined as:
constant

   2  3
ψ = P1m Vg − Vpk + P2 Vg − Vpk + P3 Vg − Vpk (2)
12.5
12.9
11.8
9.5
9.7

where, P1m is defined as:

 
B1
P1m = P1 1 + (3)
breakdown field

cosh (B2 Vds )

The Angelov model uses nonlinear Cgs and Cgd models, and they are
(MV/m)
Critical

expressed as follows:
0.23

    
3.3
2.2
0.5
0.4

Cgs = Cgsp + αgsCgs0 · tanh k ’ gs Vds + V ’ sgs + 1 ·

Table 1 | Material properties of common microwave semiconductors1–4 .

     
tanh kgs Vgs + Vsgs + P111Vds + 1 (4)
(1e7 cm/s)

    
Saturation

Cgd = Cgdp + αgdCgd0 · tanh k ’ gd Vds + V ’ sgd + 1 − P111 ·

velocity

      
tanh kgd Vgs + Vsgd − P111Vds + 1 + 2P111
2.5

0.6

(5)
2

2
1

Physical MVS HEMT model The MVS HEMT model28 , 29 is an industrial

Electron mobility

standard physical model developed based on the virtual source theory57 .

Different from another very popular physical model, which is the surface-
(cm2 /V·s)

potential based ASM model26 , 27 , the MVS model is charge-based, i.e.,

1500

5400
8500
1400

it describes device property based on the 2DEG property. The physical

950

models are significantly different from the empirical models, e.g. the An-
gelov model. The physical models do not have independent drain current
Bandgap (eV)

sub-model and capacitance (charge) sub-model. Instead, the drain current

equation and capacitance (charge) equation are depended on each other.
In other words, the capacitance model is determined once the drain cur-
3.39
3.26
1.34
1.42
1.12

rent equation is determined, vice versa. There are many recent studies
At 300 K.

based on the MVS-HEMT model, including the subthreshold swing im-

Property#

provement58 , gate capacitance improvement59 , circuit-level validation29 ,

4H-SiC

GaAs

flicker noise improvement60 and the combination with ANN33 , 61 , etc. In

GaN

InP

#
Si

addition to the model based on MVS theory, there are the improved MVS-
2 model62 , 63 , self-consistent field effect transistor (FET) model64 , carbon
nanotube FET model65 , etc.

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 4 of 12
Review DOI: 10.1016/j.chip.2023.100052

The drain current equation of MVS HEMT model can be written as:  
4Q3is + 6Q3id + 8Q2is Qid + 12Qis Q2id
QD = W Lng f (18)
IDS = W ng f Qi (x )v(x ) (6) 15(Qis + Qid )2
Here, W is the width of each finger of the HEMT, ngf is the number Then, the Cgs and Cgd can be obtained by taking the partial derivative
of gate-fingers, Qi (x) means the channel charge density at a position x in of the charge with respect to voltage.
the channel, and lastly the carrier velocity is the v(x). It is worthy to be ∂Q
emphasized that the unit of IDS is not A but A/mm. C= (19)
∂V
Assuming µ is the carrier mobility and ψ(x) is the potential at the
corresponding position, then v(x) = µEx = µdψ(x)/dx, so the IDS is: ML-Based DynaFET GaN HEMT model The ANN-based HEMT model
dψ (x ) mostly adopts a feedforward neural network (FNN) with a simple architec-
IDS = W ng f Qi (x )μ (7)
dx ture (usually one or two hidden layers) to construct the nonlinear current
Considering the influence of velocity saturation, Eq. (7) could be source and capacitance equations. One example of the widely used ANN-
modified as: based models is the DynaFET model24 , 25 proposed by Keysight. Fig. 5
dψ (x ) shows the DynaFET model topology, where three subcircuits are used
IDS = W ng f Qi (x )μ  dx
 dψ (x) β 1/β (8) to model the self-heating and trapping effects. In the DynaFET model,
ANNs are used to model the bias- and temperature-dependent currents
1+ dx
vsat
and charges. The equations are expressed as follows24 , 25 , 66 :
μ
 
In order to get a charge-based formulation for current, the inversion IG (t ) = IG VGS (t ), VDS (t ), Tj (t ), φ1 (t ), φ2 (t )
capacitance Cinv is used to establish the connection between Qi (x) and d  
+ QG VGS (t ), VDS (t ), Tj (t ), φ1 (t ), φ2 (t ) (20)
ψ(x) as follows: dt
d Qi ( x )
Cinv = (9)
dψ (x )  
ID (t ) = ID VGS (t ), VDS (t ), Tj (t ), φ1 (t ), φ2 (t )
By using Eq. (7) and Eq. (9), the partial differential of channel dis-
tance x can be expressed as: d  
+ QD VGS (t ), VDS (t ), Tj (t ), φ1 (t ), φ2 (t ) (21)
dt
W ng f Qi (x )μ
dx = d Qi ( x ) (10) where, VGS and VDS are the instantaneous voltages, φ1 and φ2 are the trap
IDSCinv
state voltages, and Tj is the junction temperature. The ANN-based large-
Integrating Eq. (10) from x = 0 (drain) to x = L (source), and then signal model including electrothermal and trapping dynamics is trained
substituting Qi (0) = Qid and Qi (L) = Qis , Eq. (8) can be written as: and tested by the DC, S-parameters, and large-signal waveform measure-
Q2is − Q2id ments at different temperatures. In 2016, Zhang et al.67 introduced a par-
IDS = W ng f Qi (x )μ   β 1/β (11)
allel training technique for DynaFET modeling with large datasets, which
Qis −Qid
1 + Cinv vL μ adopted cluster systems to calculate the ANN feedforward and derivative
μ
in parallel to reduce model training time.
To make the current expression look similar to the classical virtual Traditionally, the current source sub-model and capacitance sub-
source model, reformat Eq. (11) as: model are independent, which are quite similar to the empirical mod-
els. Differently, the ANNs rather than the empirical equations are used
Qis + Qid
IDS = W ng f v Fvsat (12) to model the nonlinear components for better model accuracy. The drain
2 current and capacitances can be modeled using similar ANN structures.
where, For example, they can exhibit two inputs (Vgs and Vds ), one output (Ids ,
  μ
v = vsat 1 − Ff + 2φt Ff (13) Cgs or Cgd ), and one or two hidden layers. The network configuration can
L
be adjusted based on application scenarios. However, a trade-off between
Qis −Qid network complexity and its training costs should be considered. Usually,
Cinv VDSAT an ANN with more neurons and more hidden layers exhibit higher fitting
Fvsat =   1/β (14)
β ability (higher accuracy), while it is also easier to obtain local optimum
Qis −Qid
1+ Cinv VDSAT training results.
The source-end and drain-end charge density is:
     ADVANCES IN EMPIRICAL MODELS
VGSi − VT − αφt Ff s
Qis = Cinv 2nφt ln 1 + exp (15)
2nφt Although the idea of empirical models has been proposed for a long
     time, empirical models still play important roles in GaN HEMT model-
VGDi − VT − αφt Ff d ing. Many new developments in the field of modeling are based on empir-
Qid = Cinv 2nφt ln 1 + exp (16)
2nφt ical models, such as the construction of dynamic68 and non-dynamic17 , 69
where, the Ffs and Ffd are the Fermi functions to realize a smooth transition trapping model, self-heating effects16 , thermal effects14 , 55 , scalability13 ,
between two states. circuit designs13 , 21 , model simplification20 machine-learning based re-
The charge density is linked to the charge by the following equations: search56 and the innovation of extraction method40 , etc.
 3  In 2007, Jardel et al.68 proposed a widely used trapping sub-circuit to
6Qis + 4Q3id + 12Q2is Qid + 8Qis Q2id characterize the dynamic trapping effects and verified it based on an em-
QS = W Lng f (17)
15(Qis + Qid )2 pirical drain current model. The RC-based trapping sub-circuit shown in
Fig. 6a can be used to represent both drain-lag and gate-lag effects. With

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 5 of 12
Review DOI: 10.1016/j.chip.2023.100052

Fig. 5 | DynaFET model with thermal and trapping subcircuits66 .

Fig. 6 | a, Schematic of the drain-lag model68 . b, Measurement of a drain-lag related current transient68 . Vgs is fixed at -6 V, Vds is pulsed from 30 to 20 V. c, Measured
(crosses) and modeled average output current in different conditions68 . d, Self-heating thermal subcircuit utilizing multiple time constants17 . e, Long duration pulsed IV
measurement pulsing from Vgsq = -2.64 V to Vgsq = -1.7 V for 5 ms at Vds = 28 V comparing single (X = 1) and three time constant (X = 3) thermal models17 . f,
Photograph of fabricated continuous class-F amplifier on PCB, designed with CREE CGH40010F transistor21 . g, Continuous class-F PA measured and modeled results21 .
h, Large-signal measurement setup and photograph of GaN HPA MMIC13 . i, Measured and modeled input power sweep of GaN MMIC HPA at 25°C and 6 GHz with
100-us pulse width and 10% duty cycle13 . j, Measured and modeled frequency sweep of GaN MMIC HPA at 25°C and Pin = 23 dBm with 100-us pulse width and 10%
duty cycle13 . k, Measured and simulated gain, output power, and PAE versus the available power from the power source at -25°C55 . l, Measured and simulated gain, output
power, and PAE versus the available power from the power source at 150°C55 . Reprinted with permission from refs.13,17,21,55,68 . © 2007, 2009, 2013, 2017, 2021 Institute
of Electrical and Electronics Engineers Inc.

this RC-based sub-circuit, good performances were achieved by the de- tribution of the work by Yuk et al. is the proposal of a trapping modeling
vice model in time-domain measurement and large-signal output current method based on quiescent voltage. This modeling method exhibits good
measurement, as shown in Fig. 6b and Fig. 6c, respectively. accuracy in the pulsed Ids - Vds /Vgs test.
In 2009, Yuk et al.17 proposed an improved Angelov model with the In 2011, Jarndal et al.40 proposed an improved large-signal model-
consideration of self-heating effects and trapping effects. In this paper, the ing method of GaN on SiC devices. The buffer related parasitic conduc-
authors proposed a novel self-heating equivalent circuit with three time tion effect was taken into consideration in their work. The correspond-
constants, so that the time domain self-heating effects can be accurately ing equivalent circuit and extraction method were also proposed. Jarndal
modeled, as shown in Fig. 6d and Fig. 6e. Such a self-heating equivalent et al. also developed the model of investigated GaN-on-SiC device and
circuit is also used by King et al.21 Fig. 6f and g present the associated verified its performance in a class-AB power amplifier (PA) on plastic
circuit along with the measured and modeled results. An additional con- substrate.

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 6 of 12
Review DOI: 10.1016/j.chip.2023.100052

Fig. 7 | a, Die micrograph of a dual-biased GaN PA MMIC96 , designed using the MVS-GaN HEMT model. b, Die micrograph of a CS-CG GaN PA MMIC97 , designed
using the MVS-GaN HEMT model. c, An ANN model with 2 inputs, 2 hidden layers, and 1 output for terminal charges33 . Reprinted with permission from refs.33,96,97 . ©
2015, 2017, 2022 Institute of Electrical and Electronics Engineers Inc.

In 2013, based on the improved Angelov model17 , King et al.21 devel- (CW) condition, the device model presented satisfactory simulation re-
oped a GaN HEMT model and applied this GaN HEMT model into power sults at different temperatures, which proves the accuracy of the developed
amplifier design. This work also proposed a nonlinear thermal resistance, thermal resistance model and the accuracy of MVS-GaN HEMT model in
which is a function of dissipation power, as shown in Fig. 6e. Fig. 6f and large-signal simulations.
g present the designed class-F PA on printed circuit board (PCB) and its In 2019, Radhakrishna et al.29 summarized the modeling method
corresponding modeling and measurement results, respectively. of the MVS-GaN HEMT model and reported its application in GaN-
In 2016, Chen et al.70 , 71 introduced a statistical model for GaN based RF- and HV-circuits. The dual-biased GaN PA MMIC96 shown
HEMT large-signal modeling. This method relied on extensive model- in Fig. 7a and the common-source (CS) common-gate (CG) GaN PA
ing of batches of devices and, which thereforeoffers valuable guidance for MMIC97 shown in Fig. 7b are designed with the adoption of the MVS-
optimizing device processes and enhancing circuit yields. Later, in 2021, HEMT model.
Mao et al.72 applied the statistical model to circuit designs and investi- Recently, in 2022, it was also pointed out that the machine-learning
gated the impact of process fluctuation in design procedure. can be applied to physical models to reduce the difficulty of model de-
In 2017, Xu et al.13 investigated the scalability of GaN HEMT based velopment and enhance modeling accuracy33 . The core idea of this work
on its Angelov model and applied the self-develop model into a GaN high is to use ANN to fit the source-end and drain-end terminal charge densi-
power amplifier (HAP) microwave monolithic integrated circuit (MMIC) ties (qis and qid ), which are originally modeled by complex semi-physical
design. This proposed model exhibited scalable drain current source, scal- equations. The configuration of the ANN model for qis and qid is given as
able thermal resistance and scalable thermal capacitance. Fig. 6h, i, and j Fig. 7c.
show the designed class-AB HPA MMIC and its two corresponding mod-
eling and measurement results, respectively. Advances in surface-potential-based models In addition to the charge-
In 2021. Luo et al.55 improved the temperature dependence of non- based modeling method, the surface-potential-based modeling methods
linear capacitances in Angelov model. The temperature dependence are also widely adopted to construct device models.
of subthreshold swing73 and threshold voltage74 was reflected in the One famous representative of the surface-potential-based models is
temperature-dependent modeling of intrinsic capacitances (Cgs and Cgd ). the ASM-HEMT model26 , 27 . Avirup et al. refined the ASM model’s flicker
This work can be regarded as an extension of many previous studies re- noise82 and thermal noise91 performances. Ahsan et al.88 , 89 studied the
garding temperature dependence75–80 . Fig. 6k and l are the large-signal field-plate capacitance modeling of the ASM model. The modeling of DC,
performances at different temperatures simulated with original intrinsic CV, and RF characteristics86 , trapping effects90 , and gate current87 within
capacitance models and improved intrinsic capacitance models. the framework of the ASM model were fully elaborated in the subsequent
published articles.
ADVANCES IN PHYSICAL MODELS There are also many other GaN HEMT models built with surface-
potential theory. Some of them have been verified with excellent large-
Physical models are relatively new, however, they have witnessed rapid signal performance. Chen et al.85 proposed a model for intrinsic GaN
and remarkable development in the past decade. Their modeling meth- HEMT, in which the Fermi-potential is obtained from charge-density-of-
ods are close to device physics, so that the built models are very reliable. states function and Poisson equations. In 2018, Wu et al.84 studied the
Their excellent scalability can also well meet the needs of modeling large scalability of the surface-potential-based large-signal model.
quantities of devices with different sizes. Therefore, in the past ten years,
modeling methods or improvements of physical models have been contin- Other advances In addition to the well-known ASM model and MVS-
uously proposed, and the large-signal simulation capabilities of physical GaN HEMT model, the emergence of some other physical models is
models have also been continuously improved. According to the different also worthy to be noted. The HiSIM GaN HEMT model30 is developed
formulation methods of the physical model, the advances of the phys- based on the well-established industry-standard HiSIM model98 . It is ob-
ical model can be divided into the advances in the charge-based mod- tained by iteratively solving Poisson equation and can accurately predict
els29 , 33 , 58–61 , 81 , the advances in the surface-potential-based models82–92 , current collapse effects. The QPZD model32 , proposed by University of
and the other advances31,34,93–95 . Electronic Science and Technology of China, combined the zone divi-
sion method with the surface potential theory for model construction.
Advances in the charge-based models One representative of the charge- It includes some empirical parameters in the drain current source model
based physical models is the MVS-GaN HEMT model28,29 . and shows excellent accuracy in various verifications. It has also been
In 2018, Jia et al.81 proposed a scalable thermal resistance model for demonstrated to be capable of predicting the performance of the two-
GaN HEMTs based on the MVS-GaN HEMT model. At continuous-wave dimensional hole gas diamond MOSFET99 . Another model worthy to be

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 7 of 12
Review DOI: 10.1016/j.chip.2023.100052

Fig. 8 | a, Single-tone power sweep measurements (symbols) and simulations at Vdsq = 28 V at Idsq = 30 mA and f0 = 2.7 GHz, with 50-Ohm source and load
impedance102 . b, The measured and simulated pulsed IV104 . c, Measured (solid lines) and simulated (dashed lines) load-pull contours of the output power and PAE at
Vdsq = 30 V at Idsq = 44 mA and f0 = 3.5 GHz104 . d, Outlier detection results of the extracted Cgs and Cds , including the normal points (circle symbols), outliers
(diamond symbols), and decision boundary (dashed lines)105 . e, Measured (symbols) and simulated (lines) CW S-parameters from 0.5 to 40 GHz with the reference plane
at the probe tips. Biases are (from left to right) Vgs = −3 V and Vds = 0 V; Vgs = −2 V and Vds = 20 V; and Vgs = −1 V and Vds = 20 V105 . f, Measured (solid lines)
and simulated (dashed lines) load-pull contours of the output power and PAE at Vdsq = 20 V, Vgsq = -1.68 V, and f0 = 2.8 GHz105 . Reprinted with permission from
refs.102,104,105 . © 2016, 2020, 2021 Institute of Electrical and Electronics Engineers Inc.

mentioned is the EPFL model31 , which has undergone recent studies and In 2017, Liu et al.120 proposed a Wiener-type dynamic neural network,
improvements, including its transcapacitances94 , short-channel effects34 , which exhibited a better convergence performance compared with a time
high-temperature behavior93 , non-quasi-static behavior95 , etc. However, delay neural network.
the large-signal properties of the EPFL model have not been studied In 2018, Na et al.118 proposed the nonuniform grid formulation and
yet. It is anticipated that the EPFL model has great potential for future multidimensional cubic polynomial extrapolation to obtain the extrapo-
research. lated data for ANN training and to improve the model predictions outside
the measurement region.
In 2019, Jarndal103 adopted multilayer perceptrons (MLPs) to repre-
ADVANCES IN THE ML-BASED MODELS sent the bias- and temperature-dependent drain-source current, the bias-
dependent power dissipation, and the bias-dependent incremental current
Many investigations on ML-based modeling have been made, such as due to trapping effects.
small-signal parameter extraction100,101 , nonlinear current and charge In 2020, Du et al.104 developed a large-signal model using an ANN
modeling102–107 , spacing mapping techniques108–111 , ML-based behavioral to model the bias-dependent drain current and empirical equations to de-
modeling112–114 , network architecture optimization of ANN-based mod- scribe the breakdown issue, trapping effects, and self-heating effects. The
els115–117 , extrapolation techniques118 , 119 , new types of neural networks120 , combination of ANNs and empirical equations benefits from fewer mea-
and model extraction from large-signal measurements5,121,122 . surement data and less training time. The modeling results of IV and load-
In 2016, Huang et al.102 proposed an ANN-based large-signal model pull contours are shown in Fig. 8b and Fig. 8c.
including self-heating, surface trapping and buffer trapping effects. The In 2020, Zhao et al.111 presented a space mapping technique for
drain–source current and intrinsic capacitances were modeled by three- GaN HEMT modeling including trapping effects, where separate map-
layer feedforward neural networks. The charge models are obtained by ping modules were used to map different behaviors. The space mapping
integrating the capacitance functions with respect to the terminal volt- technique can make the existing equivalent circuit model match the new
ages. The large-signal model is trained and tested by the PIV measure- transistor behavior.
ments at different temperatures and quiescent biases and multi-bias S- In 2021, Hu et al.105 proposed a consistent gate charge model for
parameter measurements at different temperatures. The modeled and sim- GaN HEMTs. The gate charge was calculated by integrating the ex-
ulated single-tone power sweeps are shown in Fig. 8a, where good agree- tracted intrinsic capacitances, andan MLP was subsequently used to accu-
ment can be observed. rately model the bias and temperature dependence of the gate charge. The

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 8 of 12
Review DOI: 10.1016/j.chip.2023.100052

charge-conservative capacitance models are obtained by taking the partial stantly being expanded, but also many new models are continuously being
derivatives of the gate charge model function with respect to voltages. An proposed. The physical model is endowed with many advantages, such as
outlier detection method for capacitance extractions was proposed as well better scalability, being close to the physical nature of the device, high pre-
in this work, and the results can be found in Fig. 8d. The comparisons cision and so on. However, it cannot be ignored that there are still many
between measured and simulated S-parameters and load-pull contours are difficulties limiting the application of physical models. For example, due
given in Fig. 8e and Fig. 8f. to the complex expressions and high development difficulties of physical
models, some of them have not been applied to the large-signal condi-
MODELING OF NON-IDEAL EFFECTS tions yet. Therefore, in the future, for some physical models, how to apply
them to actual circuit design is a problem which remains to be explored.
Trapping effects and self-heating effects are two well-known significant In addition, some physical models only model the intrinsic transistor. The
non-ideal effects in GaN HEMTs. For the completeness of the large-signal trapping effects and the self-heating effects, which significantly influence
model, it is necessary to review some common modeling methods for GaN HEMT performances, are necessary to be taken into consideration
them. and modeled in the future.
The modeling methods for trapping effects can be classified into the The ML-based large-signal models show a diversified development
non-dynamic15 , 69 and dynamic methods68 , 123 , 124 . These methods all mod- trend with the continuous development of ML algorithms and the im-
ify gate voltage to reflect the influence of trapping effects on thresh- provement of computing power. They participate in modeling in the way
old voltage125 . The non-dynamic methods focus more on the non-ideal of ANN, space mapping based, support-vector-regression, etc. In the fu-
IV performance caused by the trapping effects. For example, the quies- ture, if the application of ML-based models in industry is to be increased,
cent voltage method uses gate and drain quiescent voltages to charac- it is necessary to solve their scalability and extrapolability issues. In the
terize the current collapse in the pulsed IV measurement. This model- ANN-based models, the overfitting issue is another topic that is worth
ing method for trapping effects is easy and direct. Subsequently, Zhao working on in the future.
et al.14 , Xu et al.13 , and Hu et al.105 aopted this method or its im- It is necessary to discuss the circuit design capabilities of different
proved version to model trapping effects. One problem of this method large-signal models. In circuit designs, the accuracy and convergence
is that it is difficult to characterize the trapping effects in the time- of a device model are two key considerations that may somewhat con-
domain. In other words, it is not dynamic. One example of the dy- strain each other and determine the predictive and application abilities of
namic method was proposed by Jardel et al.68 , who used a series RC that model, respectively. In general, the empirical model is of good ac-
subcircuit to model the trapping and de-trapping process, as shown in curacy and convergence, and it has been widely used in circuit designs
Fig. 7a and b. The voltage drops or rises caused by the RC circuit will for a long time, with mainstream foundries continuing to use them. For
directly change the effective gate voltage. This method is claimed to a single device, when new technology or discovery appears, the orig-
predict the average output current more accurately, as given in Fig. 7c. inal empirical model can be quickly adjusted by the model designers,
Later, Huang et al.102 , Wu et al.84 , and Luo et al.20 , 33 used this mod- therefore the accuracy of the new model can quickly meet the needs of
eling method to model trapping effects. However, this method needs a circuit verification. However, the scalability of the empirical models is
lot of efforts to model the pulsed IV performance with non-zero qui- poor. As such, the accuracy of the empirical model is often unsatisfac-
escent voltages. Another dynamic trapping modeling method126 wor- tory when modeling a batch of devices with different sizes. The con-
thy to be mentioned is based on the Shockley-Read-Hall (SRH) model. vergence of the empirical model mainly lies in the problem of charge
This method has been widely used90 , 117 , 127 , 128 including the ASM-HEMT conservation, but the charge conservation problem can be solved by us-
model. ing the charge model rather than the capacitance model. Physical mod-
The self-heating also influences GaN HEMT significantly, while its els have been used by more and more manufacturers recently, which also
modeling methods are relatively consistent: a parallel RC network is com- exhibit satisfactory accuracy and convergence. The accuracy of the phys-
monly adopted to characterize the increase of channel temperature due to ical models for single-size device modeling may not be as high as that
self-heating effects. The gap between different methods may lie in the dif- of the empirical models, but owing to its excellent scalability, overall
ference in the time constants. For example, Yuk et al. and King et al. used high-precision modeling can be well achieved by the physical modelfor
three RC parallel networks (three time constants) to realize an accurate a batch of devices with different sizes. At the same time, due to the
time-domain response, as shown in Fig. 7d and e. direct modeling of channel charges, charge conservation issue of phys-
ical models is naturally assured. One problem for the physical models
CONCLUSION AND PROSPECT is that the initial verification is expensive. When new technologies ap-
pear, it is often necessary to make complex adjustments to the physi-
This article comprehensively reviews the large-signal models for GaN cal models for preliminary circuit verification. In addition, if the accu-
HEMT, including their classical formulation methods and recent ad- racy or convergence of the physical model is not good, it is also difficult
vances. The material properties of GaN, large-signal equivalent circuits, to make adjustments. The accuracy of the ML-based models is excel-
large-signal measurements, trapping effects and self-heating effects were lent, but currently, there is no large-scale application of the ML-based
introduced as well. model in circuit designs. One reason accounting for the phenomenon
Empirical large-signal models exhibit excellent stability, satisfactory above is the convergence problem caused by charge conservation. Ad-
accuracy and low development difficulty. It can be seen, therefore, that ditionally, if the simulation range exceeds the modeling range, the accu-
they have played an important role in helping researchers verify various racy of some ML-based models will drop significantly. Fortunately, so-
innovations related to trapping effects, electrothermal effects and scalabil- lutions to these problems are being proposed continuously, as mentioned
ity, etc. in the past years. In the future, it is foreseeable that the empirical above.
models will still widely participate in various innovations related to mod- To sum up, currently the large-signal GaN HEMT models are in a
els and circuits. blooming era: the improvements of model equations are constantly re-
The physical large-signal models are currently developing at full fining existing models; the introduction of various subcircuits is greatly
speed: not only the application scenarios of the classical models are con- broadening the application scenarios of device models; and the successive

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 9 of 12
Review DOI: 10.1016/j.chip.2023.100052

application of different modeling methods are continuously increasing the 21. King, J. B. & Brazil, T. J. Nonlinear electrothermal GaN HEMT model applied
model diversity and unveiling new possibilities. More importantly, the par- to high-efficiency power amplifier design. IEEE Trans. Microw. Theory Tech. 61,
444–454 (2013). https://doi.org/10.1109/TMTT.2012.2229712.
ticipation, passion and innovation of researchers from all over the world 22. Zhang, Q.-J., Gupta, K. C. & Devabhaktuni, V. K. Artificial neural networks for
have been injecting a steady stream of impetus into the development of RF and microwave design - from theory to practice. IEEE Trans. Microw. Theory
device modeling. As such, we can rightfully anticipate a more prosperous Tech. 51, 1339–1350 (2003). https://doi.org/10.1109/TMTT.2003.809179.
23. Schreurs, D. et al. ANN model for AlGaN/GaN HEMTs constructed from
era of device modeling in the future. near-optimal-load large-signal measurements. In IEEE MTT-S International Mi-
crowave Symposium Digest, 447–450 (IEEE, 2003). https://doi.org/10.1109/
MWSYM.2003.1210972.
REFERENCES 24. Xu, J. et al. Global dynamic FET model for GaN transistors: DynaFET model
validation and comparison to locally tuned models. In 83rd ARFTG Microwave
1. Boles, T. GaN-on-silicon–present capabilities and future directions. AIP Conf. Measurement Conference, 1–6 (IEEE, 2014). https://doi.org/10.1109/ARFTG.
Proc. 1934, 020001 (2018). https://doi.org/10.1063/1.5024484. 2014.6899526.
2. Sun, Y. et al. Review of the recent progress on GaN-based vertical power Schot- 25. Xu, J., Jones, R., Harris, S. A., Nielsen, T. & Root, D. E. Dynamic FET model -
tky barrier diodes (SBDs). Electronics 8, 575 (2019). https://doi.org/10.3390/ DynaFET - for GaN transistors from NVNA active source injection measure-
electronics8050575. ments. In 2014 IEEE MTT-S International Microwave Symposium (IMS2014),
3. Mukhopadhyay, J., Banerjee, S., Mukherje, M. & Banerjee, J. P. A com- 1–3 (IEEE, 2014). https://doi.org/10.1109/MWSYM.2014.6848293.
parative study on indium phosphide and α-gallium nitride based IM- 26. Khandelwal, S. et al. Robust surface-potential-based compact model for GaN
PATT oscillators for terahertz communication. J. Telecommun. 3,14-21 HEMT IC design. IEEE Trans. Electron Devices 60, 3216–3222 (2013). https:
(2010). https://www.researchgate.net/profile/Soumen-Banerjee/publication/ //doi.org/10.1109/TED.2013.2265320.
260402749_A_comparative_study_on_Indium_Phosphide_and_a-Gallium_ 27. Khandelwal, S., Chauhan, Y. S. & Fjeldly, T. A. Analytical modeling of surface-
Nitride_based_Impatt_Oscillators_for_Terahertz_Communication/. potential and intrinsic charges in AlGaN/GaN HEMT devices. IEEE Trans. Elec-
4. Marino, F. A., Faralli, N., Ferry, D. K., Goodnick, S. M. & Saraniti, M. Figures of tron Devices 59, 2856–2860 (2012). https://doi.org/10.1109/TED.2012.2209654.
merit in high-frequency and high-power GaN HEMTs. J. Phys.: Conf. Ser. 193, 28. Ujwal, R. Modeling gallium-nitride based high electron mobility transistors: linking
012040 (2009). https://doi.org/10.1088/1742-6596/193/1/012040. device physics to high voltage and high frequency circuit design. Massachusetts
5. Hua, W. et al. Power amplifiers performance survey 2000-present. Georgia Tech Institute of Technology (2016). http://dspace.mit.edu/handle/1721.1/7582
Electronics and Micro-System Lab. Revised November 19, 2020. https://gems. 29. Radhakrishna, U., Choi, P. & Antoniadis, D. A. Facilitation of GaN-based RF- and
ece.gatech.edu/PA_survey.html. HV-circuit designs using MVS-GaN HEMT compact model. IEEE Trans. Electron
6. Khan, M. A., Bhattarai, A., Kuznia, J. N. & Olson, D. T. High electron mobility Devices 66, 95–105 (2019). https://doi.org/10.1109/TED.2018.2848721.
transistor based on a GaN-Alx Ga1−x N heterojunction. Appl. Phys. Lett. 63, 1214– 30. Okada, Y. et al. HiSIM_GaN: compact model for GaN-HEMT with accurate dy-
1215 (1993). https://doi.org/10.1063/1.109775. namic current-collapse reproduction. IEEE Trans. Electron Devices 66, 106–115
7. Statz, H., Newman, P., Smith, I. W., Pucel, R. A. & Haus, H. A. GaAs FET (2018). https://doi.org/10.1109/TED.2018.2868284.
device and circuit simulation in SPICE. IEEE Trans. Electron Devices 34, 160– 31. Jazaeri, F. & Sallese, J.-M. Charge-based EPFL HEMT model. IEEE Trans. Elec-
169 (1987). https://doi.org/10.1109/T-ED.1987.22902. tron Devices 66, 1218–1229 (2019). https://doi.org/10.1109/TED.2019.2893302.
8. Curtice, W. R. & Ettenberg, M. A nonlinear GaAs FET model for use in the 32. Wen, Z. et al. A quasi-physical compact large-signal model for AlGaN/GaN
design of output circuits for power amplifiers. IEEE Trans. Microw. Theory Tech. HEMTs. IEEE Trans. Microw. Theory Tech. 65, 5113–5122 (2017). https://doi.
33, 1383–1394 (1985). https://doi.org/10.1109/TMTT.1985.1133229. org/10.1109/TMTT.2017.2765326.
9. Angelov, I., Bengtsson, L. & Garcia, M. Extensions of the Chalmers nonlinear 33. Luo, H., Yan, X., Zhang, J. & Guo, Y. A neural network-based hybrid physi-
HEMT and MESFET model. IEEE Trans. Microw. Theory Tech. 44, 1664–1674 cal model for GaN HEMTs. IEEE Trans. Microw. Theory Tech. 70, 4816–4826
(1996). https://doi.org/10.1109/22.538957. (2022). https://doi.org/10.1109/TMTT.2022.3206442.
10. Angelov, I., Zirath, H. & Rosman, N. A new empirical nonlinear model for HEMT 34. Allaei, M., Shalchian, M. & Jazaeri, F. Modeling of short-channel effects in GaN
and MESFET devices. IEEE Trans. Microw. Theory Tech. 40, 2258–2266 (1992). HEMTs. IEEE Trans. Electron Devices 67, 3088–3094 (2020). https://doi.org/10.
https://doi.org/10.1109/22.179888. 1109/TED.2020.3005122.
11. Angelov, I., Rorsman, N., Stenarson, J., Garcia, M. & Zirath, H. An empiri- 35. Jarndal, A., Du, X. & Xu, Y. Modelling of GaN high electron mobility transistor on
cal table-based FET model. IEEE Trans. Microw. Theory Tech. 47, 2350–2357 diamond substrate. IET Microw. Antennas Propag. 15, 661–673 (2021). https:
(1999). https://doi.org/10.1109/22.808981. //doi.org/10.1049/mia2.12093.
12. EE HEMT1 Model (EEsof Scalable Nonlinear HEMT Model). https://edadocs. 36. Chernykh, M. Y. et al. GaN-based heterostructures with CVD diamond heat
software.keysight.com/pages/viewpage.action?pageId=6063102. sinks: a new fabrication approach towards efficient electronic devices. Appl.
13. Xu, Y. et al. A scalable large-signal multiharmonic model of AlGaN/GaN HEMTs Mater. Today 26, 101338 (2022). https://doi.org/10.1016/j.apmt.2021.101338.
and its application in C-band high power amplifier MMIC. IEEE Trans. Mi- 37. Crupi, G. et al. Accurate multibias equivalent-circuit extraction for GaN HEMTs.
crow. Theory Tech. 65, 2836–2846 (2017). https://doi.org/10.1109/TMTT.2017. IEEE Trans. Microw. Theory Tech. 54, 3616–3622 (2006). https://doi.org/10.
2669984. 1109/TMTT.2006.882403.
14. Zhao, X. et al. Temperature-dependent access resistances in large-signal mod- 38. Jeon, M. Y., Kim, B. G., Jeon, Y. J. & Jeong, Y. H. A technique for ex-
eling of millimeter-wave AlGaN/GaN HEMTs. IEEE Trans. Microw. Theory Tech. tracting small-signal equivalent-circuit elements of HEMTs. IEICE Trans. Elec-
65, 2271–2278 (2017). https://doi.org/10.1109/TMTT.2017.2658561. tron. E82-C, 1968–1976 (1999). https://search.ieice.org/bin/summary.php?id=
15. Yuk, K., Branner, G. R. & McQuate, D. An improved empirical large-signal model e82-c_11_1968.
for high-power GaN HEMTs including self-heating and charge-trapping effects. In 39. Jarndal, A. & Kompa, G. A new small-signal modeling approach applied to GaN
2009 IEEE MTT-S International Microwave Symposium Digest, 753–756 (IEEE, devices. IEEE Trans. Microw. Theory Tech. 53, 3440–3448 (2005). https://doi.
2009). https://doi.org/10.1109/MWSYM.2009.5165806. org/10.1109/TMTT.2005.857332.
16. Wang, C. et al. An electrothermal model for empirical large-signal modeling 40. Jarndal, A., Markos, A. Z. & Kompa, G. Improved modeling of GaN HEMTs on Si
of AlGaN/GaN HEMTs including self-heating and ambient temperature effects. substrate for design of RF power amplifiers. IEEE Trans. Microw. Theory Tech.
IEEE Trans. Microw. Theory Tech. 62, 2878–2887 (2014). https://doi.org/10. 59, 644–651 (2011). https://doi.org/10.1109/TMTT.2010.2095034.
1109/TMTT.2014.2364821. 41. Dambrine, G., Cappy, A., Heliodore, F. & Playez, E. A new method for determin-
17. Yuk, K. S., Branner, G. R. & McQuate, D. J. A wideband multiharmonic empiri- ing the FET small-signal equivalent circuit. IEEE Trans. Microw. Theory Tech. 36,
cal large-signal model for high-power GaN HEMTs with self-heating and charge- 1151–1159 (1988). https://doi.org/10.1109/22.3650.
trapping effects. IEEE Trans. Microw. Theory Tech. 57, 3322–3332 (2009). 42. Li, M., Feng, F., Li, Z., Lin, Q. & Zhang, Q.-J. A new hybrid method for extracting
https://doi.org/10.1109/TMTT.2009.2033299. bias and temperature-dependent source/drain access resistances in passivated
18. Emekar, S. et al. Modified angelov model for an exploratory GaN-HEMT tech- GaN HEMTs. IEEE Microw. Wirel. Compon. Lett. 32, 308–311 (2022). https://doi.
nology with short, few-fingered gates. In 2017 International Conference on Sim- org/10.1109/LMWC.2021.3129338.
ulation of Semiconductor Processes and Devices (SISPAD), 117–120 (IEEE, 43. Colangeli, S. et al. Nondestructive, self-contained extraction method of parasitic
2017). https://doi.org/10.23919/SISPAD.2017.8085278. resistances in HEMT devices. IEEE Trans. Microw. Theory Tech. 68, 2571–2578
19. Zhao, Z. et al. Highly accurate GaN HEMT model based on the Angelov model (2020). https://doi.org/10.1109/TMTT.2020.2988665.
with error compensation. Microw. Opt. Technol. Lett. 62, 3505–3513 (2020). 44. Brady, R. G., Oxley, C. H. & Brazil, T. J. An improved small-signal parameter-
https://doi.org/10.1002/mop.32497. extraction algorithm for GaN HEMT devices. IEEE Trans. Microw. Theory Tech.
20. Luo, H., Zhang, H., Hu, W. & Guo, Y. A simplification method for capacitance 56, 1535–1544 (2008). https://doi.org/10.1109/TMTT.2008.925212.
models in AlGaN/GaN high electron mobility transistors under large drain voltage 45. Hussein, A. S. & Jarndal, A. H. Reliable hybrid small-signal modeling of GaN
using channel analysis. Int. J. RF Microw. Comput.-Aided Eng. 31, 1–10 (2021). HEMTs based on particle-swarm-optimization. IEEE Trans. Comput.-Aided Des.
https://doi.org/10.1002/mmce.22489.

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 10 of 12
Review DOI: 10.1016/j.chip.2023.100052
Integr. Circuits Syst. 37, 1816–1824 (2018). https://doi.org/10.1109/TCAD.2017. 68. Jardel, O. et al. An electrothermal model for AlGaN/GaN power HEMTs includ-
2782779. ing trapping effects to improve large-signal simulation results on high VSWR.
46. Nguyen, T. T.-L. & Kim, S.-D. A gate-width scalable method of parasitic parame- IEEE Trans. Microw. Theory Tech. 55, 2660–2669 (2007). https://doi.org/10.
ter determination for distributed hemt small-signal equivalent circuit. IEEE Trans. 1109/TMTT.2007.907141.
Microw. Theory Tech. 61, 3632–3638 (2013). https://doi.org/10.1109/TMTT.2013. 69. Jarndal, A. & Kompa, G. Large-signal model for AlGaN/GaN HEMTs accurately
2279360. predicts trapping- and self-heating-induced dispersion and intermodulation dis-
47. Raffo, A. et al. A new modeling technique for microwave multicell transistors tortion. IEEE Trans. Electron Devices 54, 2830–2836 (2007). https://doi.org/10.
based on EM simulations. IEEE Trans. Microw. Theory Tech. 68, 3100–3110 1109/TED.2007.907143.
(2020). https://doi.org/10.1109/TMTT.2019.2961078. 70. Chen, Z. et al. A large-signal statistical model and yield estimation of GaN HEMTs
48. Crupi, G. et al. Scalability of multifinger HEMT performance. IEEE Microw. based on response surface methodology. IEEE Microw. Wirel. Compon. Lett. 26,
Wirel. Compon. Lett. 30, 869–872 (2020). https://doi.org/10.1109/LMWC.2020. 690–692 (2016). https://doi.org/10.1109/LMWC.2016.2597196.
3012181. 71. Chen, Z. et al. A GaN HEMTs nonlinear large-signal statistical model and its
49. Resca, D. et al. Scalable nonlinear FET model based on a distributed para- application in S-band power amplifier design. IEEE Microw. Wirel. Compon. Lett.
sitic network description. IEEE Trans. Microw. Theory Tech. 56, 755–766 (2008). 26, 128–130 (2016). https://doi.org/10.1109/LMWC.2016.2517444.
https://doi.org/10.1109/TMTT.2008.918153. 72. Mao, S. et al. A yield-improvement method for millimeter-wave GaN MMIC power
50. Heinz, F., Schwantuschke, D., Leuther, A. & Ambacher, O. Highly scalable dis- amplifier design based on load—pull analysis. IEEE Trans. Microw. Theory Tech.
tributed high electron mobility transistor model. In 2019 IEEE Asia-Pacific Mi- 69, 3883–3895 (2021). https://doi.org/10.1109/TMTT.2021.3088499.
crowave Conference (APMC), 1286–1288 (IEEE, 2019). https://doi.org/10.1109/ 73. Vandamme, E. P., Jansen, P. & Deferm, L. Modeling the subthreshold swing
APMC46564.2019.9038318. in MOSFET’s. IEEE Electron Device Lett 18, 369–371 (1997). https://doi.org/10.
51. Serino, A., Ciccognani, W., Colangeli, S. & Limiti, E. A measurement-based ap- 1109/55.605442.
proach to model scaling properties of FETs. IEEE Microw. Wirel. Compon. Lett. 74. Tsividis, Y. Operation and Modeling of the Mos Transistor. (McGraw-Hill Inc.,
26, 912–914 (2016). https://doi.org/10.1109/LMWC.2016.2615027. 1987). https://dl.acm.org/doi/abs/10.5555/27167
52. Zheng, J. X. et al. A scalable active compensatory sub-circuit for accurate GaN 75. Lee, J.-W. & Webb, K. J. A temperature-dependent nonlinear analytic model for
HEMT large signal models. IEEE Microw. Wirel. Compon. Lett. 26, 431–433 AlGaN-GaN HEMTs on SiC. IEEE Trans. Microw. Theory Tech. 52, 2–9 (2004).
(2016). https://doi.org/10.1109/LMWC.2016.2555940. https://doi.org/10.1109/TMTT.2003.821227.
53. Jarndal, A. & Kompa, G. An accurate small-signal model for AlGaN-GaN HEMT 76. Chen, Y. et al. Temperature-dependent small signal performance of GaN-on-
suitable for scalable large-signal model construction. IEEE Microw. Wirel. Com- diamond HEMTs. Int. J. Numer. Model.: Electron. Netw. Devices Fields 33, e2620
pon. Lett. 16, 333–335 (2006). https://doi.org/10.1109/LMWC.2006.875626. (2019). https://doi.org/10.1002/jnm.2620.
54. Zhang, W. et al. An Angelov large-signal FET model up to 67 GHz. IEEE 77. Alim, M. A. & Rezazadeh, A. A. Thermal influence on multibias small-and
Trans. Electron Devices 66, 2577–2582 (2019). https://doi.org/10.1109/TED. large-signal parameters of GaAs pHEMT fabricated in multilayer 3-D MMIC.
2019.2913395. IEEE Trans. Electron Devices 64, 1511–1518 (2017). https://doi.org/10.1109/
55. Luo, H., Zhong, Z., Hu, W. & Guo, Y. Analysis and modeling of the temperature- TED.2017.2658685.
dependent nonlinearity of intrinsic capacitances in AlGaN-GaN HEMTs. IEEE Mi- 78. Crupi, G. et al. Temperature influence on GaN HEMT equivalent circuit. IEEE Mi-
crow. Wirel. Compon. Lett. 31, 373–376 (2021). https://doi.org/10.1109/LMWC. crow. Wirel. Compon. Lett. 26, 813–815 (2016). https://doi.org/10.1109/LMWC.
2021.3057444. 2016.2601487.
56. Huang, A.-D., Zhong, Z., Guo, Y.-X. & Wu, W. A general dimension reduction 79. Darwish, A. M., Huebschman, B. D., Viveiros, E. & Hung, H. A. Dependence
method for the dispersion modeling of semiconductor devices. IEEE Access 6, of GaN HEMT millimeter-wave performance on temperature. IEEE Trans. Mi-
39422–39434 (2018). https://doi.org/10.1109/ACCESS.2018.2855044. crow. Theory Tech. 57, 3205–3211 (2009). https://doi.org/10.1109/TMTT.2009.
57. Khakifirooz, A., Nayfeh, O. M. & Antoniadis, D. A simple semiempirical short- 2034050.
channel MOSFET current–voltage model continuous across all regions of opera- 80. Alim, M. A., Rezazadeh, A. A. & Gaquiere, C. Temperature effect on DC and
tion and employing only physical parameters. IEEE Trans. Electron Devices 56, equivalent circuit parameters of 0.15-µm gate length GaN/SiC HEMT for mi-
1674–1680 (2009). https://doi.org/10.1109/TED.2009.2024022. crowave applications. IEEE Trans. Microw. Theory Tech. 64, 3483–3491 (2016).
58. Luo, H. & Huang, A. An improved sub-threshold swing model for MIT virtual https://doi.org/10.1109/TMTT.2016.2604815.
source GaN model. In 2018 IEEE MTT-S International Wireless Symposium 81. Jia, Y., Xu, Y. & Guo, Y. A universal scalable thermal resistance model for com-
(IWS), 1–4 (IEEE, 2018). https://doi.org/10.1109/IEEE-IWS.2018.8400855. pact large-signal model of AlGaN/GaN HEMTs. IEEE Trans. Microw. Theory
59. Jia, Y., Xu, Y., Wen, Z., Wu, Y. & Guo, Y. Analytical gate capacitance models Tech. 66, 4419–4429 (2018). https://doi.org/10.1109/TMTT.2018.2854185.
for large-signal compact model of AlGaN/GaN HEMTs. IEEE Trans. Electron De- 82. Dasgupta, A., Khandelwal, S. & Chauhan, Y. S. Compact modeling of flicker
vices 66, 357–363 (2019). https://doi.org/10.1109/TED.2018.2881255. noise in HEMTs. IEEE J. Electron Devices Soc. 2, 174–178 (2014). https://doi.
60. Luo, H., Yan, X., Hu, W. & Guo, Y. The charge-based flicker noise model for org/10.1109/JEDS.2014.2347991.
HEMTs. IEEE Microw. Wirel. Compon. Lett. 32, 125–128 (2022). https://doi.org/ 83. Zhang, A. et al. Analytical modeling of capacitances for GaN HEMTs, including
10.1109/LMWC.2021.3118708. parasitic components. IEEE Trans. Electron Devices 61, 755–761 (2014). https:
61. Luo, H. & Guo, Y. A neural network-based modeling method for terminal charges //doi.org/10.1109/TED.2014.2298255.
in GaN HEMTs. In 2022 IEEE MTT-S International Wireless Symposium (IWS), 84. Wu, Q. et al. A scalable multiharmonic surface-potential model of AlGaN/GaN
1–3 (IEEE, 2022). https://doi.org/10.1109/IWS55252.2022.9978118. HEMTs. IEEE Trans. Microw. Theory Tech. 66, 1192–1200 (2018). https://doi.
62. Rakheja, S., Lundstrom, M. S. & Antoniadis, D. A. An improved virtual-source- org/10.1109/TMTT.2017.2775639.
based transport model for quasi-ballistic transistors—Part I: Capturing effects 85. Cheng, X. & Wang, Y. A surface-potential-based compact model for AlGaN/GaN
of carrier degeneracy, drain-bias dependence of gate capacitance, and nonlin- MODFETs. IEEE Trans. Electron Devices 58, 448–454 (2011). https://doi.org/10.
ear channel-access resistance. IEEE Trans. Electron Devices 62, 2786–2793 1109/TED.2010.2089690.
(2015). https://doi.org/10.1109/TED.2015.2457781. 86. Khandelwal, S. et al. ASM GaN: industry standard model for GaN RF and power
63. Rakheja, S., Lundstrom, M. S. & Antoniadis, D. A. An improved virtual-source- devices—Part 1: DC, CV, and RF model. IEEE Trans. Electron Devices 66, 80–
based transport model for quasi-ballistic transistors—Part II: experimental veri- 86 (2018). https://doi.org/10.1109/TED.2018.2867874.
fication. IEEE Trans. Electron Devices 62, 2794–2801 (2015). https://doi.org/10. 87. Albahrani, S. A. et al. Consistent surface-potential-based modeling of drain and
1109/TED.2015.2457872. gate currents in AlGaN/GaN HEMTs. IEEE Trans. Electron Devices 67, 455–462
64. Wei, L., Mysore, O. & Antoniadis, D. Virtual-source-based self-consistent current (2020). https://doi.org/10.1109/TED.2019.2961773.
and charge FET models: from ballistic to drift-diffusion velocity-saturation oper- 88. Ahsan, S. A., Ghosh, S., Khandelwal, S. & Chauhan, Y. S. Analysis and mod-
ation. IEEE Trans. Electron Devices 59, 1263–1271 (2012). https://doi.org/10. eling of cross-coupling and substrate capacitances in GaN HEMTs for power-
1109/TED.2012.2186968. electronic applications. IEEE Trans. Electron Devices 64, 816–823 (2017). https:
65. Luo, J. et al. Compact model for carbon nanotube field-effect transistors in- //doi.org/10.1109/TED.2017.2654264.
cluding nonidealities and calibrated with experimental data down to 9-nm gate 89. Ahsan, S. A. et al. Capacitance modeling in dual field-plate power GaN HEMT for
length. IEEE Trans. Electron Devices 60, 1834–1843 (2013). https://doi.org/10. accurate switching behavior. IEEE Trans. Electron Devices 63, 565–572 (2016).
1109/TED.2013.2258023. https://doi.org/10.1109/TED.2015.2504726.
66. Pedro, J. C., Root, D. E., Xu, J. & Nunes, L. C. Nonlinear Circuit Simulation and 90. Albahrani, S. A., Mahajan, D., Hodges, J., Chauhan, Y. S. & Khandelwal, S.
modeling: Fundamentals for Microwave Design. (Cambridge University Press, ASM GaN: industry standard model for GaN RF and power devices—Part-II:
2018). https://doi.org/10.1017/9781316492963. modeling of charge trapping. IEEE Trans. Electron Devices 66, 87-94 (2019).
67. Zhang, S., Xu, J., Zhang, Q.-J. & Root, D. E. Parallel matrix neural network doi:https://doi.org/10.1109/TED.2018.2868261.
training on cluster systems for dynamic FET modeling from large datasets. In 91. Dasgupta, A., Khandelwal, S. & Chauhan, Y. S. Surface potential based model-
2016 IEEE MTT-S International Microwave Symposium (IMS), 1–3 (IEEE, 2016). ing of thermal noise for HEMT circuit simulation. IEEE Microw. Wirel. Compon.
https://doi.org/10.1109/MWSYM.2016.7540387. Lett. 25, 376–378 (2015). https://doi.org/10.1109/LMWC.2015.2422472.

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 11 of 12
Review DOI: 10.1016/j.chip.2023.100052
92. Wu, Y. et al. An improved surface-potential-based model for MOSFETs consid- 114. Long, X., Jun, Z., Zhang, B., Chen, D. & Liang, W. A unified electrothermal be-
ering the carrier Gaussian distribution. IEEE Trans. Microw. Theory Tech. 68, havior modeling method for both SiC MOSFET and GaN HEMT. IEEE Trans. Ind.
4082–4090 (2020). https://doi.org/10.1109/TMTT.2020.2997819. Electron. 68, 9366–9375 (2020). https://doi.org/10.1109/TIE.2020.3026283.
93. Sahebghalam, N., Shalchian, M., Chalechale, A. & Jazaeri, F. High-temperature 115. Gorissen, D., Zhang, L., Zhang, Q.-J. & Dhaene, T. Evolutionary neuro-space
HEMT model. IEEE Trans. Electron Devices 69, 4821–4827 (2022). https://doi. mapping technique for modeling of nonlinear microwave devices. IEEE Trans.
org/10.1109/TED.2022.3184662. Microw. Theory Tech. 59, 213–229 (2011). https://doi.org/10.1109/TMTT.2010.
94. Jazaeri, F., Shalchian, M. & Sallese, J.-M. Transcapacitances in EPFL HEMT 2090169.
model. IEEE Trans. Electron Devices 67, 758–762 (2019). https://doi.org/10. 116. Ho, Y.-W. et al. Neuroevolution-based efficient field effect transistor compact de-
1109/TED.2019.2958180. vice models. IEEE Access 9, 159048–159058 (2021). https://doi.org/10.1109/
95. Touchaei, B. J. & Shalchian, M. Non-quasi-static intrinsic GaN-HEMT model. ACCESS.2021.3130254.
IEEE Trans. Electron Devices 69, 6594–6601 (2022). https://doi.org/10.1109/ 117. Hu, W. & Guo, Y.-X. An evolutionary multilayer perceptron-based large-signal
TED.2022.3211485. model of GaN HEMTs including self-heating and trapping effects. IEEE Trans.
96. Choi, P. et al. A 5.9-GHz fully integrated GaN frontend design with physics-based Microw. Theory Tech. 70, 1146–1156 (2022). https://doi.org/10.1109/TMTT.2021.
RF compact model. IEEE Trans. Microw. Theory Tech. 63, 1163–1173 (2015). 3132892.
https://doi.org/10.1109/TMTT.2015.2405913. 118. Na, W. et al. Advanced extrapolation technique for neural-based microwave
97. Choi, P., Radhakrishna, U., Boon, C. C., Peh, L.-S. & Antoniadis, D. Linearity modeling and design. IEEE Trans. Microw. Theory Tech. 66, 4397–4418 (2018).
enhancement of a fully integrated 6-GHz GaN power amplifier. IEEE Microw. https://doi.org/10.1109/TMTT.2018.2854163.
Wirel. Compon. Lett. 27, 927–929 (2017). https://doi.org/10.1109/LMWC.2017. 119. Zhang, L. & Zhang, Q.-J. Simple and effective extrapolation technique for neural-
2746673. based microwave modeling. IEEE Microw. Wirel. Compon. Lett. 20, 301–303
98. Miura-Mattausch, M., Mattausch, H. J. & Ezaki, T. The Physics and Modeling (2010). https://doi.org/10.1109/LMWC.2010.2047450.
of MOSFETS: Surface-Potential Model HiSIM. (World Scientific Press, 2008). 120. Liu, W., Na, W., Zhu, L., Ma, J. & Zhang, Q.-J. A Wiener-type dynamic neural
https://doi.org/10.1142/6159. network approach to the modeling of nonlinear microwave devices. IEEE Trans.
99. Fu, Y. et al. A large-signal model for two-dimensional hole gas diamond MOSFET Microw. Theory Tech. 65, 2043–2062 (2017). https://doi.org/10.1109/TMTT.2017.
based on the QPZD. IEEE Access 7, 76868–76877 (2019). https://doi.org/10. 2657501.
1109/ACCESS.2019.2918187. 121. Schreurs, D. M. M.-P., Verspecht, J., Vandenberghe, S. & Vandamme,
100. Khusro, A., Husain, S., Hashmi, M. S., Ansari, A. Q. & Arzykulov, S. A generic E. Straightforward and accurate nonlinear device model parameter-estimation
and efficient globalized kernel mapping-based small-signal behavioral modeling method based on vectorial large-signal measurements. IEEE Trans. Microw.
for GaN HEMT. IEEE Access 8, 195046–195061 (2020). https://doi.org/10.1109/ Theory Tech. 50, 2315–2319 (2002). https://doi.org/10.1109/TMTT.2002.803427.
ACCESS.2020.3033788. 122. Xu, J., Hom, J., Iwamoto, M. & Root, D. E. Large-signal FET model with
101. Marinković, Z., Crupi, G., Caddemi, A., Marković, V. & Schreurs, D. M. M.-P. A multiple time scale dynamics from nonlinear vector network analyzer data. In
review on the artificial neural network applications for small-signal modeling of 2010 IEEE MTT-S International Microwave Symposium (IEEE, 2010). https:
microwave FETs. Int. J. Numer. Model. Electron. Netw. Devices Fields 33, e2668 //doi.org/10.1109/MWSYM.2010.5516843.
(2020). https://doi.org/10.1002/jnm.2668. 123. Jardel, O. et al. A drain-lag model for AlGaN/GaN power HEMTs. In 2007
102. Huang, A.-D., Zhong, Z., Wu, W. & Guo, Y.-X. An artificial neural network- IEEE/MTT-S International Microwave Symposium (IEEE, 2007). https://doi.org/
based electrothermal model for GaN HEMTs with dynamic trapping effects con- 10.1109/MWSYM.2007.379972.
sideration. IEEE Trans. Microw. Theory Tech. 64, 2519–2528 (2016). https: 124. Jardel, O. et al. Modeling of trap induced dispersion of large signal dynamic
//doi.org/10.1109/TMTT.2016.2586055. characteristics of GaN HEMTs. In 2013 IEEE MTT-S International Microwave
103. Jarndal, A. On neural networks based electrothermal modeling of GaN de- Symposium Digest (MTT) (IEEE, 2013). https://doi.org/10.1109/MWSYM.2013.
vices. IEEE Access 7, 94205–94214 (2019). https://doi.org/10.1109/ACCESS. 6697576.
2019.2928392. 125. Jia, Y. et al. Characterization of buffer-related current collapse by buffer potential
104. Du, X. et al. ANN-based large-signal model of AlGaN/GaN HEMTs with accu- simulation in AlGaN/GaN HEMTs. IEEE Trans. Electron Devices 65, 3169–3175
rate buffer-related trapping effects characterization. IEEE Trans. Microw. Theory (2018). https://doi.org/10.1109/TED.2018.2849501.
Tech. 68, 3090–3099 (2020). https://doi.org/10.1109/TMTT.2020.2990171. 126. Rathmell, J. G. & Parker, A. E. Circuit implementation of a theoretical model
105. Hu, W., Luo, H., Yan, X. & Guo, Y.-X. An accurate neural network-based con- of trap centres in GaAs and GaN devices. Proc. SPIE 6798, 67980R (2007).
sistent gate charge model for GaN HEMTs by refining intrinsic capacitances. https://doi.org/10.1117/12.758711.
IEEE Trans. Microw. Theory Tech. 69, 3208–3218 (2021). https://doi.org/10. 127. Gomes, J. L., Nunes, L. C., Goncalves, C. F. & Pedro, J. C. An accurate charac-
1109/TMTT.2021.3076064. terization of capture time constants in GaN HEMTs. IEEE Trans. Microw. Theory
106. Wang, J. et al. Artificial neural network-based compact modeling methodology Tech. 67, 2465–2474 (2019). https://doi.org/10.1109/TMTT.2019.2921338.
for advanced transistors. IEEE Trans. Electron Devices 68, 1318–1325 (2021). 128. Prasad, A., Thorsell, M., Zirath, H. & Fager, C. Accurate modeling of GaN HEMT
https://doi.org/10.1109/TED.2020.3048918. RF behavior using an effective trapping potential. IEEE Trans. Microw. Theory
107. Jarndal, A., Husain, S., Hashmi, M. & Ghannouchi, F. M. Large-signal modeling Tech. 66, 845–857 (2018). https://doi.org/10.1109/TMTT.2017.2748950.
of GaN HEMTs using hybrid GA-ANN, PSO-SVR, and GPR-based approaches.
IEEE J. Electron Devices Soc. 9, 195–208 (2021). https://doi.org/10.1109/JEDS.
2020.3035628.
108. Zhang, L., Xu, J., Yagoub, M. C. E., Ding, R. & Zhang, Q.-J. Efficient analyt- MISCELLANEA
ical formulation and sensitivity analysis of neuro-space mapping for nonlinear
microwave device modeling. IEEE Trans. Microw. Theory Tech. 53, 2752–2767
(2005). https://doi.org/10.1109/TMTT.2005.854190.
Funding This work was supported in part by the National Research Founda-
109. Kabir, H. et al. Smart modeling of microwave devices. IEEE Microw. Mag. 11,
105–118 (2010). https://doi.org/10.1109/MMM.2010.936079. tion (NRF) of Singapore under Grant NRF-CRP17-2017-08.
110. Zhang, Q.-J., Gad, E., Nouri, B., Na, W. & Nakhla, M. Simulation and automated
modeling of microwave circuits: state-of-the-art and emerging trends. IEEE J.
Microw. 1, 494-507 (2021). https://doi.org/10.1109/JMW.2020.3033780. Declaration of Competing Interest The authors declare no competing inter-
111. Zhao, Z., Zhang, L., Feng, F., Zhang, W. & Zhang, Q.-J. Space mapping tech- ests.
nique using decomposed mappings for GaN HEMT modeling. IEEE Trans. Mi-
crow. Theory Tech. 68, 3318–3341 (2020). https://doi.org/10.1109/TMTT.2020.
3004622. © 2023 The Author(s). Published by Elsevier B.V. on behalf of Shanghai
112. Cai, J. et al. Bayesian inference-based behavioral modeling technique for GaN
HEMTs. IEEE Trans. Microw. Theory Tech. 67, 2291–2301 (2019). https://doi.
Jiao Tong University. This is an open access article under the CC BY license
org/10.1109/TMTT.2019.2906304. (http://creativecommons.org/licenses/by/4.0/)
113. Cai, J. et al. Machine learning-based broadband GaN HEMT behavioral model
applied to class-J power amplifier design. Int. J. Microw. Wirel. Technol. 13, 415–
423 (2021). https://doi.org/10.1017/S1759078720001385.

CHIP | VOL 2 | AUTUMN 2023 Luo, H., Hu, W. & Guo, Y. Chip 2, 100052 (2023) 12 of 12