Elsevier

Microelectronics Reliability

Volume 71, April 2017, Pages 35-40
Microelectronics Reliability

Operational frequency degradation induced trapping in scaled GaN HEMTs

https://doi.org/10.1016/j.microrel.2017.02.008Get rights and content

Highlights

  • Electric field induced traps in drain access region

  • Improvements in cut-off frequency from lateral scaling

  • Reduction in current density due to vertical scaling

  • Reduced effects of induced traps on fT for a laterally scaled down device

Abstract

Cut-off frequency increase from 12.1 GHz to 26.4 GHz, 52.1 GHz and 91.4 GHz is observed when the 1 μm gate length GaN HEMT is laterally scaled down to LG = 0.5 μm, LG = 0.25 μm and LG = 0.125 μm, respectively. The study is based on accurately calibrated transfer characteristics (ID-VGS) of the 1 μm gate length device using Silvaco TCAD. If the scaling is also performed horizontally, proportionally to the lateral (full scaling), the maximum drain current is reduced by 38.2% when the gate-to-channel separation scales from 33 nm to 8.25 nm. Degradation of the RF performance of a GaN HEMT due to the electric field induced acceptor traps experienced under a high electrical stress is found to be about 8% for 1 μm gate length device. The degradation of scaled HEMTs reduces to 3.5% and 7.3% for the 0.25 μm and 0.125 gate length devices, respectively. The traps at energy level of ET = EV + 0.9 eV (carbon) with concentrations of NIT = 5 × 1016cm 3, NIT = 5 × 1017cm 3 and NIT = 5 × 1018cm 3 are located in the drain access region where highest electrical field is expected. The effect of traps on the cut-off frequency is reduced for devices with shorter gate lengths down to 0.125 μm.

Introduction

A huge potential of gallium nitride (GaN) high electron mobility transistors (HEMTs) in future radio frequency (RF) and power applications has been fuelling their research and development in the recent years. Material properties of wide-band-gap nitrides such as spontaneous (SP) and piezoelectric (PZ) polarizations allow creation of a two-dimensional electron gas (2DEG) with sheet densities above 1012 cm 2, a high electron mobility (up to 2000 cm2V 1s 1 in 2DEG), a large energy band gap (3.4 eV), a good thermal conductivity (160 WK 1m 1) ensuring a good heat dissipation, and a very high breakdown field (3.3 MV cm 1) make it an ideal candidate for all devices requiring fast carrier transport with high breakdown voltage [1], [2]. However, the maximum RF power output is not always reproducible due to the existence of defects/traps in the GaN device structure inhibiting the wide commercial use of these devices. The trap related phenomena, affecting the reliability of device, may result in a reduction of the drain current, virtual gate formation, transconductance frequency dispersion [3], light sensitivity [4] and restricted microwave output power [5]. These effects are largely due to the formation of a non-equilibrium charge distribution in the device. Various proposals have been made regarding the nature, location, and effects of traps in GaN devices [6] such as a correlation between RF output power degradation and bulk traps [7], and a high electrical stress induced traps [6] opposing the traditional explanation by a self-heating [8].

Since experimental techniques are insufficient to provide a detailed understanding of carrier transport process in GaN based devices, physically-based simulations are required to uncover complex transport phenomena [9]. Since the GaN HEMTs for power and RF applications have typically microscale (power/RF) or sub-microscale dimensions (RF), drift-diffusion and hydrodynamics transport models [1] are suitable for fast turn-out modelling while ensemble Monte Carlo simulations are used for in-depth modelling into details of hot-carrier non-equilibrium transport [10]. A compromise is a combination of the Fermi kinetic transport model with Maxwell equations to efficiently capture the essential physics of hot-carriers to model the electromagnetic wave effects in the HEMTs [11].

This paper studies the effects of electric field induced bulk traps on the DC and RF characteristics of GaN HEMTs. The investigations are carried out using Atlas simulation toolbox by Silvaco [12] using the drift-diffusion transport model for all simulations. The study is based on careful calibration of device I-V characteristics using a well-developed methodology [1]. The RF parameters are calculated from the small signal conductance and capacitance values during small signal analysis. The admittance parameters (Y) are obtained directly from the small signal parameters and then converted to scattering parameters (S) using matrix transformation [12].

Experimentally reported values for the cut-off frequency for GaN HEMTs have a fT × LG product of 13 GHz μm [13], although this can be improved by increasing the gate recess depth [14] and has been shown to depend on the substrate material [15]. We will show that results from our simulations are in a good agreement with the experimentally reported ones [16], [17]. We give an in-sight, firstly, on the variation of drain current and cut-off frequency with scaling down the device. Thereafter, traps are placed in the device according to Ref. [6] in order to predict their effect on the DC and RF performance of the device. The induced traps are predicted to cause a positive shift in the threshold voltage and a general reduction in the cut-off frequency.

Section snippets

Device structure and simulation methodology

Fig. 1 illustrates the schematic cross-section of an investigated asymmetrical 1 μm gate length GaN HEMT. It consists of a 2 nm GaN cap, 20 nm AlGaN barrier, 1 nm AlN spacer, and 3 μm GaN buffer, all grown on a SiC substrate with source-to-gate and gate-to-drain separations of LSG = 2 μm and LGD = 3 μm, respectively. Fig. 2 shows the conduction band profile overlapped with electron current density showing the role of the thin aluminium nitride (AlN) layer in additional confinement of the 2DEG.

The

Lateral and vertical scaling of the device

The maximum drain current for the investigated 1 μm gate length GaN HEMT is IDmax = 516.7 mA/mm (see Fig. 3) while the extracted cut-off frequency is f = 12.1 GHz which closely agrees with the experimental results of 11.5 GHz (1 μm gate length Al0.22Ga0.78N/GaN HEMT on sapphire) reported in Ref. [17]. We have then scaled down the gate length from LG = 1 μm to LG = 0.5 μm, LG = 0.25 μm, LG = 0.125 μm and, correspondingly, the source-to-gate and gate-to-drain distances in order to predict possible improvements in the

Trapping effects on DC and RF performance

Although degradation mechanisms in GaN HEMTs have yet to be fully understood, a number of explanations have been advanced as possible mechanisms leading to degradation in these devices. Some of these include: (i) virtual gate formation as a result of tunnelling from the gate [25]; (ii) Poole-Frenkel gate leakage surface conduction [26]; (iii) hot electron trapping and interface state creation [27]; (iv) hot electron trapping at the surface and AlGaN barrier [28], [29]; (v) trap generation in

Trapping effects on a laterally scaled device

In order to gain some insight into how device behaviour in a laterally scaled device is affected by the traps, we have scaled down the traps generation region in proportion to the various device dimensions. Assuming a relatively small value of the trap density up to 1017 cm 3 for smaller device dimensions may not, in reality, represent accurately the effects of traps as smaller devices would experience higher electric fields under similar biases. We observe in Fig. 9 that the current is

Conclusions

The detrimental effect of traps generated by a high electric field in the stressed GaN HEMTs has been studied in scaled devices. The study is based on accurate calibration of the TCAD model to experimental transfer and output characteristics of 1 μm gate length GaN HEMT with fT = 12.1 GHz obtained using commercial simulation software Atlas by Silvaco.

The scaled device exhibits improvement in the RF performance by increase in the cut-off frequency (fT = 12.1 GHz) by 118% (fT = 26.4 GHz), 330% (fT = 52.1 GHz)

Acknowledgements

This research is funded by the Sêr Cymru National Research Network in Advanced Engineering and Materials [Grant code: NRN081]. We thank Edward Wasige and Abdullah Al-Khalidi of the University of Glasgow for providing us with the experimental data.

References (33)

  • M. Faqir et al.

    Investigation of high-electric-field degradation effects in AlGaN/GaN HEMTs

    IEEE Trans. Electron Devices.

    (2008)
  • B. Benbakhti et al.

    Effects of self-heating on performance degradation in algan/gan-based devices

    IEEE Trans. Electron Devices.

    (2009)
  • G. Curatola et al.

    Modelling of GaN HEMTs: from device-level simulation to virtual prototyping

  • T. Sadi et al.

    Investigation of self-heating effects in Submicrometer GaN/AlGaN HEMTs using an electrothermal Monte Carlo method

    IEEE Trans. Electron Devices.

    (2006)
  • M. Grupen

    GaN high electron mobility transistor simulations with full wave and hot electron effects

    IEEE Trans. Electron Devices.

    (2016)
  • Silvaco, Atlas User's Manual

    (2015)
  • Cited by (10)

    View all citing articles on Scopus
    View full text