Author : Andrew Philip Lange
Publisher :
ISBN 13 : 9781369615760
Total Pages : pages
Book Rating : 4.6/5 (157 download)
Book Synopsis Aluminum Nitride Buffer Layer Growth for Group III-nitride Epitaxy on (111) Silicon by : Andrew Philip Lange
Download or read book Aluminum Nitride Buffer Layer Growth for Group III-nitride Epitaxy on (111) Silicon written by Andrew Philip Lange and published by . This book was released on 2016 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: This dissertation examines the growth of aluminum nitride (AlN) on (111) silicon by metalorganic chemical vapor deposition. AlN is commonly used as a buffer layer for the growth of gallium nitride on silicon templates. This makes the development of growth protocols for high quality, smooth AlN films on silicon critical to improving the performance and reliability of III-nitride on silicon devices such as light emitting diodes and high power transistors. The optimal nucleation conditions for AlN on silicon have been heavily disputed. Some crystal growers expose the substrate to aluminum prior to AlN deposition, which has been shown to improve crystal quality and decrease surface roughness of both AlN buffer layers and overgrown gallium nitride. However, others adopt an ammonia-first approach, in which the substrate is nitrided prior to AlN deposition. Both can be effective depending on the growth conditions, which has resulted in considerable controversy regarding how aluminum, nitrogen, and silicon interact during these initial "predoses" and how the resulting morphology influences subsequent AlN and gallium nitride growth. The structure and morphology of aluminum predose layers deposited directly on (111) silicon at ~970 °C both with and without subsequent ammonia exposure were studied using electron microscopy, atomic force microscopy, and X-ray photoelectron spectroscopy. Three morphological features were identified -- trenches, islands, and patches. When the predose layer was not exposed to ammonia, a roughening of the substrate was observed, similar to what occurs when gallium reacts with silicon. This gave rise to aluminum rich surface trenches, which suggests that silicon is dissolved by liquid aluminum and the resulting aluminum-silicon liquid solution evaporates. When the predose layer was exposed to ammonia, faceted patches were observed with small islands near their edges. The islands were composed of both zinc-blende and wurtzite AlN polytypes, while the patches consisted of diamond cubic silicon with dilute concentrations of aluminum. A model was proposed to explain these features in which the liquid aluminum-silicon surface layer is converted into AlN and silicon upon nitridation. Low temperature and high temperature AlN growth was examined after varied aluminum predoses using electron microscopy and atomic force microscopy with the aim of explaining anomalous AlN-silicon interface structures observed by others. AlN formed small, three-dimensional islands when grown directly on the substrate at ~970 °C with no predose. When the substrate was first exposed to a predose at ~970 °C, AlN nucleated on both island and patch features causing them to grow laterally and eventually coalesce. The morphologies of films grown with and without predoses were nearly identical after coalescence. This suggests that growth at this temperature is kinetically limited and does not depend on the nucleation surface. At high temperatures (~1060 °C), enhanced lateral growth on patch features formed during the predose was observed. The AlN-silicon interface was found to be predominantly amorphous when no predose was used, consistent with previous reports. The interface was structurally abrupt when aluminum was deposited prior to growth, but contained an additional phase consistent with the zinc-blende islands observed in predose layers. It was proposed that the amorphous SiN[subscript x] interfacial layer formed between nucleation sites when no predose was used as the substrate was exposed to an ammonia ambient prior to lateral growth of the nuclei. When the substrate was first exposed to a predose, aluminum rich silicon patches covered the surface. The presence of aluminum in the patches may limit the reaction between silicon and nitrogen during the early stages of growth. Dislocations in buffer layers grown both with and without aluminum predoses were studied using weak beam dark field transmission electron microscopy. A mosaic microstructure was observed which consisted of clustered dislocations along subgrain boundaries. Many of these subgrains were not bounded by dislocations on all sides, which suggests they did not form by the coalescence of misaligned islands. It was proposed they formed instead by the clustering of dislocations due to attractive and repulsive interactions. Dislocation densities were lower in films grown with a predose, which resulted in the formation of fewer subgrains. It was also found that buffers grown with a predose had a smoother surface. The surface of buffer layers grown without a predose contained small pits along the edges of surface terraces. The separation and geometry of these terraces was consistent with the subgrain structure, indicating surface step bunching may occur around subgrains where dislocation densities are high. Consistent with III-nitride growth on alternative substrates, a-type threading dislocations with line directions normal to the basal plane were found to terminate within highly defective, low temperature nucleation layers. C-type threading dislocations were found to terminate near the AlN-Si interface. It was suggested that the former originate from the climb of basal plane dislocations which form through the dissociation of Shockley partials or the coalescence of I1 type stacking faults. It was suggested that the latter nucleate from surface steps on the substrate. The observed improvement in crystal quality of buffer layers grown with a predose may be due to dislocation annihilation events, rather than the nucleation of fewer threading dislocations. This is corroborated by the presence of voids in the substrate when the buffer was grown with a predose, which indicates that point defects diffuse across the abrupt interface during growth. The presence of amorphous interfaces in films grown without an aluminum predose may inhibit the diffusion of point defects and thereby deter dislocation climb. If this mechanism is active as evidence in this dissertation suggests, an appropriate objective of any nucleation process for AlN buffer layers on silicon may be to improve the structural coherence of the interface.
