Logistic regression analysis for the material design of chiral crystals

Abstract

Crystal design is one of the biggest dreams of scientists. Currently, statistical methods and artificial intelligence (AI) technology are rapidly developing and suitable for material design. Large-scale crystallographic databases like ICSD are also gathered. Here we demonstrate design of chiral crystals by logistic regression analysis. 689 chiral compounds and 1000 achiral AxBy type compounds (A and B are elements) were used with a true or false chirality dataset. When logistic regression model analysis was applied to all data, it became clear that crystals containing Group 14 or 16 element elements in the crystal become chiral crystals with high probability. These results are consistent with the bonding modes of the elements. Elements in groups 7 (Mn group), 8 (Fe group), 10 (Ni group) and 12 (Zn group) also tended to grow chiral crystals, consistent with the composition of a new chiral magnetic material (MnCoZn alloy) that has been recently reported (see text). This result implies that logistic regression analysis is useful for the design of chiral crystals.