Improving Crystal Structure Prediction in Multicomponent Solids

为了识别潜在的多晶型，CSP方法通过优化描述分子构象和晶胞几何形状的变量，搜索晶体能量图景来识别晶格能量(通常用作自由能的替代)中的最小值。为了可靠地进行这种搜索，它需要使用精确的能量模型来描述晶体，并且需要广泛地搜索晶体结构景观，因为多晶型结构通常是先验未知的。因此，对可靠性的渴望与将问题保持在合理的计算时间内的需要是不一致的。为了进一步发展CSP方法，有两个方面需要探索：(1)改进用于描述晶体的能量模型；(2)提高搜索算法的效率。当然，这两个领域不是独立的，必须同时研究。

To identify potential polymorphs, CSP methods search the crystal energy landscape to identify minima in the lattice energy (often used as a proxy for free energy) by optimising the variables which describe molecular conformation and unit cell geometry. For this search to be reliable, it requires that accurate energy models be used to describe the crystal, and that the crystal structure landscape be extensively searched, since it is often the case that polymorph structures are unknown a priori. The desire for reliability is thus at odds with the need to keep the problem tractable within reasonable computational times. To further advance CSP methods, there are two broad areas which should be explored: (1) Refining the energy models used to describe crystals; (2) Improving the efficiency of the search algorithms. Certainly, these two areas are not independent and must be studied in tandem.