硅晶体中的氧杂质会严重影响硅基半导体器件的电学性能，降低晶硅太阳能电池的转换效率。氧杂质在凝固界面处的分凝是影响硅晶体中氧杂质含量及分布的关键因素之一。本项目从原子层面研究硅晶体生长中氧杂质在凝固界面处分凝的模型及影响机理。根据热力学及相平衡理论，构建杂质在凝固界面处分凝的微观热力学模型，提出高效、普遍适用的杂质分凝系数计算方法，以机器学习方法为基础，优化出能正确预测硅熔点温度的硅氧原子相互作用势函数，以准确描述氧杂质在凝固界面处的分凝等热力学性质。基于以上分凝系数计算的数值模型及优化的原子势函数，研究氧杂质分凝随温度、杂质浓度、点缺陷密度等物理量的变化规律，进而阐明这些物理量对杂质分凝的影响机理，为预测硅晶体中氧杂质分布及控制氧杂质含量奠定科学理论基础。

Oxygen impurity in silicon is known to detrimentally impact the electrical performance of silicon based semiconductor devices and reduce the conversion efficiency of crystalline silicon solar cells. The segregation of oxygen atoms at the solidification interface is one of the key factor that influences the concentration, distribution of oxygen in the crystal. In this project, a microscale thermodynamic model of oxygen segregation at the solidification interface is established and an efficient, broadly applicable method based on atomic simulation is proposed for calculation of segregation coefficient. We employ the machine learning algorithm to optimize the empirical parameters of Si-O interatomic potential function which is aimed to predict accurate silicon melt temperature and oxygen thermodynamic properties in the melt and at the interface. With the segregation model and optimized potential, we study the influence mechanisms of key parameters such as temperature, impurity concentration and point defects on segregation, which can provide a theoretical basis for the prediction of oxygen distribution and controlling of oxygen concentration in silicon crystal.