Multiscale modeling strategies for chemical vapor deposition

Abstract

In order to predict the quality of the fabricated devices as a function of growth conditions in chemical vapor deposition (CVD) reactors, a model should describe multiple time and length scales. These scales include the reactor scale ([approx]0.1-1 m), the feature scale ([approx.]0.1-100 [mu]m), and the atomistic morphology evolution scale ([approx.]10 nm). At present, good reactor and feature scale models are available. However, the linking between them has been done only for low pressure CVD. Also, the atomistic Kinetic Monte Carlo models have been developed only for deposition on unpatterned substrates or over V-grooves. In this work the linking between reactor and feature scale models for both low and high pressure CVD is achieved by matching concentrations and fluxes across the interface. For low-pressure systems, we improve the convergence of the previously developed linking schemes by applying a flux-split algorithm. We analyze the assumptions underlying the linking, and demonstrate that the size of the feature domain is constrained by these assumptions and not simply by the assumption of collisionless gas phase transport. At high-pressure, mass transport between features complicates solution of the entire feature field. To capture the diffusive inter-feature transport, we develop the overlapping computational domains method. The simulation results obtained with the multiscale method are in excellent agreement with experimental data for selective epitaxy of AlGaAs in the presence of HC1. A KMC model is developed for AlGaAs surface morphology evolution during selective epitaxy. The model takes into account zincblende structure of AlGaAs, and reproduces the c(4x4) reconstruction on (100) surfaces.

(cont.) In order to model selective epitaxy, the mask is represented as a hard wall boundary condition, and overgrowth on (111)A facets is included. With this model, we investigate the effects of the unknown parameters and the growth conditions on film morphology evolution. The observed trends are in agreement with the experimental data. Since KMC simulations are limited to small surfaces and short deposition times we propose algorithms for linking the KMC and mesoscale feature shape evolution models. Finally, the feasibility of linking the coupled KMC-mesoscale model and the reactor or reactor-feature scale models is assessed.