Critical issues for homoepitaxial GaN growth by molecular beam epitaxy on hydride vapor-phase epitaxy-grown GaN substrates

Highlights

• Critical issues affecting the surfaces of freestanding GaN substrates are reviewed.

• Effect of these issues on homoepitaxial growth by MBE are evaluated.

• Procedures are outlined for obtaining high crystal quality homoepitaxial GaN.

• High crystal quality homoepitaxial GaN grown by MBE is demonstrated.

Abstract

While the heteroepitaxial growth of gallium nitride-based materials and devices on substrates such as SiC, sapphire, and Si has been well-documented, the lack of a cost-effective source of bulk GaN crystals has hindered similar progress on homoepitaxy. Nevertheless, freestanding GaN wafers are becoming more widely available, and there is great interest in growing GaN films and devices on bulk GaN substrates, in order to take advantage of the greatly reduced density of threading dislocations, particularly for vertical devices. However, homoepitaxial GaN growth is far from a trivial task due to the reactivity and different chemical sensitivities of N-polar (0001) and Ga-polar (0001) GaN surfaces, which can affect the microstructure and concentrations of impurities in homoepitaxial GaN layers. In order to achieve high quality, high purity homoepitaxial GaN, it is necessary to investigate the effect of the ex situ wet chemical clean, the use of in situ cleaning procedures, the sensitivity of the GaN surface to thermal decomposition, and the effect of growth temperature. We review the current understanding of these issues with a focus on homoepitaxial growth of GaN by molecular beam epitaxy (MBE) on c-plane surfaces of freestanding GaN substrates grown by hydride vapor phase epitaxy (HVPE), as HVPE-grown substrates are most widely available. We demonstrate methods for obtaining homoepitaxial GaN layers by plasma-assisted MBE in which no additional threading dislocations are generated from the regrowth interface and impurity concentrations are greatly reduced.

Introduction

Gallium nitride-based materials and devices have progressed from the research laboratory to the commercial sector over the past decade, facilitated by improvements in heteroepitaxial growth of group III-N device layers on non-native substrates such as SiC, sapphire, and Si. Heteroepitaxial growth has been necessitated by the limited availability of bulk single-crystal III-N materials that can be used as substrates. However, there have been substantial simultaneous improvements in the growth and availability of bulk nitrides, and a number of groups have already demonstrated a wide variety of devices grown on freestanding GaN substrates, such as light emitting diodes (LEDs) [1], [2], laser diodes [3], [4], [5], [6], [7], [8], [9], solar cells [10], [11], [12], vertical power p–n diodes [13], [14], Schottky diodes [15], and high electron mobility transistors [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27].

While high pressure [28], [29] and ammonothermal [30], [31] growth of bulk GaN have resulted in crystals with very low densities of threading dislocations (TDs), i.e.,<10² cm⁻² and 5×10⁴ cm⁻², respectively, hydride vapor-phase epitaxy (HVPE) has been used most frequently for producing freestanding GaN substrates due to its wider availability, faster growth rates, and larger and more uniform wafer size [32]. The threading dislocation density (TDD) of HVPE-grown GaN substrates is typically 10⁶–10⁷ cm⁻², several orders of magnitude greater than for high pressure- or ammonothermally-grown GaN. Nevertheless, the TDD of heteroepitaxial GaN is typically 10⁸–10¹⁰ cm⁻²; hence, homoepitaxy on HVPE-grown GaN offers significant reductions in TDD compared to heteroepitaxy. Thus, a device measuring 10×10 µm² might reasonably be expected to contain from one to ten TDs on average if fabricated on homoepitaxial GaN layers, but 10² to 10⁴ if fabricated on heteroepitaxial layers.