Abstract
The impact of NH(3) pre-flow duration on the strain development of AlN and its alloy buffer layers, as well as GaN layers, was reproduced through atomistic simulation. The reported method provides access to information that is not obvious from the raw data alone. The growth morphology of AlN utilizing pre-nitridated silicon substrates was analyzed with respect to the mechanisms of defect formation in each deposited monolayer. This report presents two distinct case studies concerning the epitaxy of the III-nitrides. In the first case study, the crystalline quality of AlN deposited on silicon substrates with NH(3) pre-flow durations of 0 s and 30 s was compared. The growth rates of the samples were aligned with those from previous simulation studies published by our group. It was noted that the defect density extracted from the sample with a 30-second NH(3) pre-flow was lower than that of the sample with a 0-second pre-flow. The results obtained from this preliminary case study led to a repetition of the multistep hetero-epitaxy experiment, previously reported by Kadir et al., with modifications to the NH(3) pre-flow duration on the silicon substrate in the subsequent case study. The epitaxial growth of GaN on AlN and three-step graded Al(x)Ga(1-x)N (where x = 0.8, 0.5, and 0.2) strain relief layers was simulated at the atomistic scale using the MOCVD process over silicon (111) substrates, with variations in NH(3) pre-flow times. The strain induced by lattice mismatch between the silicon substrate (both without and with NH(3) pre-flow) and the various buffer layers was examined in terms of dislocation density extracted layer-by-layer. The effect of NH(3) pre-flow time on the generation of threading dislocation density (TDD) in each monolayer of the AlN and AlGaN buffer layers was analyzed. It was determined that the duration of NH(3) pre-flow significantly influences the morphology and quality of each deposited monolayer justifies the experimental observations. A higher likelihood of amorphous SiN(x) formation was observed with no and shorter NH(3) pre-flow times. The lowest TDDs across all strain relief layers were measured at approximately ~ 10(10) cm(- 2). The dislocations generated in the initial buffer layer (AlN) were identified as contributing to the TDD in the subsequent layers. Furthermore, it was noted that the sample lacking an intentional nitridation step displayed a higher TDD and vacancy density compared to those with optimal nitridation.