High-quality InP on GaAs

Abstract

In addition to traditional telecommunication applications, devices based on InP have received increased attention for high-performance electronics. InP growth on GaAs is motivated by the fact that InP wafers are smaller, more expensive, and utilize older fabrication equipment than GaAs. High-quality InP on GaAs may also serve as a step towards bringing high-quality InP onto the Si platform. Integrating high-quality InP onto bulk GaAs has proven to be challenging, however. While a number of commercial Molecular Beam Epitaxy (MBE) growth foundries offer InP on GaAs for M-HEMT (Metamorphic High-Electron-Mobility Transistor) applications, the successful demonstration of InP-based, minority-carrier devices on bulk GaAs remains elusive. In this work InP on GaAs suitable for minority carrier devices is demonstrated exhibiting a threading dislocation density of 1.2x1 06/cm2 determined by plan-view transmission electron microscopy. To further quantify the quality of this InP on GaAs, a photoluminescence (PL) structure was grown to compare the quality to bulk InP. Comparable room and low (20K) temperature PL was attained. (The intensity from the PL structure grown on the InP on GaAs was -70% of that on bulk InP at both temperatures.)

(cont.) To achieve this, graded buffers in the InGaAs, InGaP, InAlAs and InGaAlAs materials systems were explored. In each of these systems, under certain growth conditions, microscopic compositional inhomogeneities along the growth direction blocked dislocations leading to dislocation densities sometimes > 109/cm2. Using scanning-transmission electron microscopy, composition variations were observed. These composition variations are caused by surface-driven phase separation leading to Ga-rich regions. As the phase separation blocked dislocation glide and led to high threading dislocation densities, conditions for avoiding phase separation were explored and identified. Composition variations could be prevented in InxGal-,As graded buffers grown at 725 °C to yield low dislocation densities of 9x105/cm2 for x < 0.34, accommodating -70% of the lattice mismatch between GaAs and InP. However, further grading to 53% In is required to attain the lattice constant of InP. Compositional grading in the InyGal_yP (0.8 < y < 1.0) materials system was found to accommodate the remaining lattice mismatch with no rise in threading dislocation density by avoiding phase separation.

(cont.) Consequently, to achieve high-quality InP on GaAs a graded buffer in the InGaAs material system was followed by a graded buffer in the InGaP materials system to reach InP. The research to achieve high-quality InP on GaAs diverged into two paths. The first successful path, using graded buffers in different materials systems, was discussed above. The second path involved the deposition of InP at various temperatures on the high-quality Ino.34Gao.66As platform that was developed to determine if InP deposited on the InGaAs platform with 1.2% misfit relaxed controllably without much dislocation nucleation. To the contrary, rampant dislocation nucleation occurred in this highly-strained InP at all temperatures studied. Interestingly, however, the InP was observed to relax via a secondary-slip system, a/2<110>{1 10}. This secondary-slip system has a Burgers vector typical in semiconductors of a/2<l 10>. Unlike the primary-slip system, where dislocations glide on { 111 }-type planes, the secondary-slip system dislocations glide on { 110}-type planes. Relaxation via the secondary-slip system was found to be a function of stress and temperature.

(cont.) A critical stress, ec, appears to be required for dislocations to glide via the secondary-slip system otherwise all relaxation occurs by the primary-slip system. For e > ec and at all temperatures studied, both the primary- and secondary-slip systems are active with apparent cross-slip from one system to the other. At low temperatures, nearly all of the relaxation was accomplished through the secondary-slip system, however. The amount of relaxation via the primary- and secondary-slip systems at three different temperatures was quantified; the resulting Arrhenius plot suggests a difference in the activation energy for glide between the two systems is 1.5 eV.