A Numerical Study for the Growth of InGaSb Crystals with a Flatter Interface by Vertical Gradient Freezing under Normal Gravity and Utilizing Bayesian Optimization

Growth of InGaSb crystals with a flatter growth interface by Vertical Gradient Freezing (VFG) under normal gravity was numerically investigated. The growth of high-quality crystals is difficult due to the adverse effect of natural convection under the effect of Earth's gravity. The natural convection developing in the melt leads to an undesirable non-flat growth interface and thus affects the quality of grown crystals. In order to obtain a flatter growth interface and better compositional uniformity in the melt, the applications of crucible rotation and external magnetic field are considered. We carried out a numerical study to optimize growth parameters for these purposes. We also considered the utilization of Bayesian optimization to have a fast search for most favorable control parameters. Such an optimization has led to a significant reduction in the growth interface deflection and a flatter growth interface was maintained during the entire growth process. Strength of natural convection in the melt was minimized and melt solute uniformity was improved. The growth rate was also increased without compromising crystal quality. Simulation results showed that the use of a numerical simulation together with Bayesian optimization is an effective way to provide a better control for the growth parameters involved.