Abstract The Top-Seeded Solution Growth (TSSG) method is a promising technique for the production of high-quality SiC single crystal. To achieve a high- and uniform-growth rate in the TSSG process of SiC, the fluid flows developing in the growth solution (melt), due to the applied and induced electromagnetic fields, buoyancy, seed rotation, and free surface tension gradient, need to be controlled. Previous numerical analysis has shown that such complex flows in the TSSG melt can be controlled by the applications of a static magnetic field and seed rotation. However, the requirement of significant computational resources prevented us from carrying out the needed optimization for the process parameters involved. In order to resolve the computational demand issue, in this study, we utilized the Bayesian optimization algorithm for an efficient optimization of the associated control parameters of the TSSG process of SiC. It was shown that the Bayesian algorithm determines the optimal state at about roughly 1/4 of the computational cost of a conventional optimization, and accurately predicts the growth-rate evaluation function around the optimal state. The optimal state obtained by the present optimization process predicts a high- and uniform-growth rate in the TSSG system of SiC considered in this work.

中文翻译：

---

外加磁场和种子轮换下碳化硅顶部种子溶液生长过程中高均匀晶体生长速率的贝叶斯优化

摘要 顶部晶种溶液生长 (TSSG) 法是一种生产高质量 SiC 单晶的有前途的技术。为了在 SiC 的 TSSG 过程中实现高且均匀的生长速率，由于施加和感应的电磁场、浮力、种子旋转和自由表面张力梯度，流体在生长溶液（熔体）中流动，需要被控制。先前的数值分析表明，TSSG 熔体中的这种复杂流动可以通过应用静磁场和种子旋转来控制。然而，对大量计算资源的需求使我们无法对所涉及的工艺参数进行所需的优化。为了解决计算需求问题，在本研究中，我们利用贝叶斯优化算法对碳化硅 TSSG 工艺的相关控制参数进行有效优化。结果表明，贝叶斯算法确定最优状态的计算成本约为常规优化计算成本的 1/4，并准确预测了最优状态附近的增长率评估函数。通过本优化过程获得的最佳状态预测了本工作中考虑的碳化硅 TSSG 系统的高且均匀的增长率。