(Invited) Band Line-up of High-k Oxides on GaN

Abstract

GaN high electron mobility transistors (HEMTs) have been commercially available for over 10 years, however gate leakage limits their performance. The HEMT has the advantages of offering simple associated circuit design and fail-safe operation. Currently the GaN based Metal-Insulator-Semiconductor (MIS)-HEMT device is seen to demonstrate superior performance in power electronics applications over the Schottky gate counterpart, due to its inherently lower gate leakage current, together with the ability to provide larger forward gate voltage swing by engineering the threshold voltage between depletion and enhancement mode operation and also an improved gate-drain breakdown voltage. High band gap gate dielectric materials are preferable as they can provide higher tunnelling barriers for electrons and holes, which result in lower gate leakage current. On the other hand, high dielectric constant (high-k) material is also necessary for improved electrostatic control over the channel and improved on-current, which in-turn results in higher transconductance. The quality of the gate dielectric and the dielectric/GaN interface plays a central role in device performance due to potential problems arising from fixed oxide charge, border and interface traps. The leakage current issue has been mitigated using Al2O3, SiO2 and Si3N4, but comes at a cost of device transconductance degradation and undesirable threshold voltage shifts. A number of high-k dielectrics such as HfO2, ZrO2, Ta2O5, LaLuO3 and TiO2 have been investigated to resolve this issue.In this paper, engineered high-k oxide approach will be presented where Al2O3 is doped with Ti to boost oxide permittivity value while preserving band offsets, as well as Ga2O3 doped with Al to increase band gap and maintain good interface quality with GaN. X-ray photoelectron spectroscopy, inverse photoemission spectroscopy and variable angle spectroscopic ellipsometry were used to estimate the band alignment and interfacial properties. TiO2 is very attractive due to having reported k of 20-86 but has a small band gap of 3.4 eV for amorphous and 3.26 eV for anatase TiO2, being far too narrow to obtain a sufficient band offset with GaN of >1 eV. Al2O3 on the other hand has a sufficient band offset of 1.8-2 eV with AlGaN but suffers from a low dielectric constant of 7.5-9.6 depending on the growth method. The previous studies of Al2O3/TiO2 nanolaminates show favourable properties, in particular the optimum between the rather high-k (~30) and low leakage current has been observed for 30% Ti. No band offset study has been reported for TixAl1-xOy/GaN system. We will show full band alignment analysis of TixAl1-xOy/GaN fabricated by atomic layer deposition (ALD) on GaN for the range of Ti composition of up to 40%. Furthermore, a trivalent Ga2O3 is a promising oxide due to its band gap of 4.4 - 4.9 eV and a moderate permittivity of 10-14. Thermally oxidized Ga2O3 has shown valence band offset of 1.4 eV to GaN. A drawback of thermal oxidation is a growth of non-stoichiometric oxide at GaN interface reported to be as Ga(x+2)N3xO(3-3x). In contrast, ALD has been shown to produce Ga2O3 with no interfacial layer with GaN and low density of interface states of 3.62x1011 cm-2eV-1. Despite the good interface with GaN, the issue with using Ga2O3 is a small conduction band offset (< 1 eV) leading to high leakage current. In this paper, Al doped Ga2O3 have been fabricated using ALD on GaN. The results point to substantial increase of the band gap from ~4.6 eV for Ga2O3 to 5.9 eV for the 1:19 Ga:Al doped sample and a strong suppression of leakage current for Al doped Ga2O3 MIS capacitors. The results are of importance for future GaN HEMTs. Acknowledgement. The work has been conducted under the University Grants Commission - UK-India Education Research Initiative (UGC-UKIERI) joint research programme UKIERI III, project numbers IND/CONT/G/17-18/18 and F.No.184-1/2018(IC), titled “Dielectric engineering on GaN for sustainable energy applications” funded by the British Council. The authors also acknowledge UKRI GCRF GIAA award 2018/19 and “Digital in India” project no. EP/P510981/1, both funded by the EPSRC, UK.