For the solution growth of silicon carbide, solvent inclusions are significant technological issues, and methods to suppress the formation of solvent inclusions are investigated in this study. Experimental observations show that solvent inclusions are formed behind the cellular structures. A phase field model is used to reproduce the formation process of cellular structures and solvent inclusions. Simulation results indicate that slight perturbations of the step front can convert into cellular structures in the case of insufficient supply of carbon, and the overdeveloped cellular structures consequently result in solvent inclusions. Accordingly, several schemes can be suggested by the simulation model to suppress the formation of cellular structures by enhancing the carbon supply. By increasing the carbon diffusion coefficient, cellular structures can be suppressed. Moreover, the step height and the solution flow direction also play an important role in suppressing the cellular structures. This study provides a comprehensive understanding of the formation process of cellular structures and solvent inclusions. A growth process with a high diffusion coefficient and opposite solution flow to the step flow direction was proposed to suppress the formation of cellular structures. The proposed numerical model could be applied in other solution crystal growth methods.

In the solution growth of the SiC crystal, macrosteps with sufficient height on an off-axis substrate are required to reduce defects and achieve a high-quality grown layer. However, overdeveloped macrosteps can induce new defects and adversely affect the crystal quality. To better understand and control the behavior of macrosteps corresponding to the control parameters of the growth system, a simulation method that consists of a global twodimensional computational fluid dynamic (CFD) model, a local three-dimensional CFD model near the growth front, and a kinetics model that describes the movement of macrosteps on the crystal surface is proposed. The simulation method is first applied to investigate the effect of the crystal rotation speed on macrostep morphology. Although the results indicate that a higher crystal rotation speed results in less step bunching, constantly rotating the crystal in one direction is demonstrated to be incapable of yielding a uniform macrostep distribution on the whole surface. Accordingly, a sophisticated control pattern is designed by periodically switching the flow direction underneath the crystal surface, where the proposed simulation method is critical to determine detailed control-parameter values. When the control pattern suggested by the simulation is used, a grown crystal with a uniform macrostep morphology and ideal step height on the whole surface is obtained in the practical experiment.