998 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 56, NO. 5, MAY 2009 Importance of the Gate-Dependent Polarization Charge on the Operation of GaN HEMTs Ashwin Ashok, Dragica Vasileska, Senior Member, IEEE, Stephen M. Goodnick, Fellow, IEEE, and Olin L. Hartin Abstract—We investigate the influence of the gate-voltage de- pendence of the polarization charge on the electron sheet charge density in the channel and how it reflects on the device transfer and output characteristics in GaN HEMTs. We find that a 10% increase in the polarization charge is needed to match the experi- mental data when the gate-voltage dependence of the polarization charge is included in the theoretical model. This information is important for calibration in commercial device simulators and for better understanding of the quality of the GaN/AlGaN interface. Index Terms—GaN devices, gate-voltage dependence, polariza- tion charge. I. INTRODUCTION A. Overview III–V nitrides have recently attracted intense interest for applications in high-temperature high-power electronic devices that are operating at microwave frequencies [1]–[3]. Gallium nitride (GaN), which has a much larger band gap than gallium arsenide, has drawn recent interest in industry for use in blue laser diodes [4] and microwave power field-effect transistors [5]. Although GaN has three times higher effective electron mass than GaAs, which results in a low-field mobility that is less than that of GaAs, it has some distinct advantages in vari- ous applications. Some of these are larger band gap, larger peak electron velocity, and higher thermal stability—all of which enable it to be a very promising material for high-power, high- temperature, and high-frequency applications. Table I shows a comparison of the important material properties of GaN and other conventional semiconductors. GaN-based metal–semiconductor field-effect transistors, GaN/AlGaN modulation doped field-effect transistors or HEMTs, and GaN/AlGaN superlattice structures have been demonstrated by many groups in the past decade. In these heterostructures, there is a difference in the polarization field between the top layer (AlGaN) and the bottom layer (GaN). This polarization in wurtzite crystals is due to the bulk proper- ties with asymmetric lattice structure and ionicity of the bonds. In addition, strain in one or both layers leads to additional Manuscript received September 2, 2008; revised December 29, 2008. First published March 24, 2009; current version published April 22, 2009. The review of this paper was arranged by Editor S. Bandyopadhyay. A. Ashok is with Intel Corporation, Sta. Clara, CA 95052-8119 USA, and also with the Arizona State University, Tempe, AZ 85287-5706 USA. D. Vasileska and S. M. Goodnick are with the Arizona State University, Tempe, AZ 85287-5706 USA. O. L. Hartin is with Freescale Semiconductors, Tempe, AZ 85284 USA. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TED.2009.2015822 built-in fields due to the polarization effect. As a result, the discontinuity of this polarization field at the interface leads to much higher carrier densities than conventional GaAs/AlGaAs heterostructures. The typical charge density in these structures is as high as 2 × 10 13 cm −2 , which is about ten times higher than in AlGaAs/GaAs HEMTs. B. Summary of Wurtzite GaN Bulk Transport Simulations The first transport simulation using Monte Carlo (MC) meth- ods for GaN materials was reported by Littlejohn et al. [6]. This simulation included a single valley (gamma valley) with both parabolic and nonparabolic bands. Acoustic scattering, polar optical phonon scattering, piezoelectric scattering, and ionized impurity scattering were taken into account in these cal- culations. Velocity saturation and negative differential transcon- ductance in GaN were predicted. Gelmont et al. [7] pointed out that intervalley electron transfer played a dominant role in GaN in high electric field leading to a strongly inverted electron distribution and to large negative differential conduc- tance. They used a nonparabolic two-valley model, including Γ and U valleys. Polar optical phonon, piezoelectric, defor- mation potential, and ionized impurity scattering mechanisms were taken into account. The intervalley coupling coefficient of GaAs was utilized in these calculations. Mansour et al. also used a two-valley model to simulate the high-temperature dependence of the electron velocity. They included acoustic phonon, polar optical phonon, intervalley phonon, and ionized impurity scattering. Bhapkar and Shur [8] came up with an improved multivalley model that included a second Γ valley in addition to the Γ and U valleys. The energy gap between the two valleys was modified to 2 eV from the earlier 1.5 eV used in all the previous simulations. The scattering mechanism taken into account included acoustic phonon, polar optical phonon, ionized impurity, piezoelectric, and intervalley scattering. All the simulations mentioned above used analytical nonpar- abolic band structures. A full-band MC simulation is another approach to get more accurate results at higher electric fields. Full-band MC simulations have previously been reported by the Georgia Tech Group. Kolnik et al. [9] reported the first full-band MC simulation for both wurtzite and zinc-blende GaN. They considered acoustic, polar optical, and intervalley scattering in their calculations. Brennan et al. [10] performed full-band MC simulations and compared the results for different III–V materials. They reported a higher electron velocity for wurtzite GaN than the previous simulation data. Both these simulations could not verify their results as no experimental velocity data were available until then. Yamakawa et al. [11] 0018-9383/$25.00 © 2009 IEEE