1-23. Material design based on motifs and machine learning

Youyong Li
Institute of Functional Nano & Soft Materials, Soochow University

Abstract:Through the development of multi-scale simulation methods and theme design, we design a variety of energy materials, such as OPV, perovskite. A variety of electrocatalysis and photocatalysis materials with excellent performance have been designed through the new theme of "binary element mixed phase", and a variety of new HER electrocatalysis materials, such as silicon and rhodium composite materials, have been designed through silicon and precious metal composite. Based on the "binary element motif", by mixing two elements with different electronegativity on two-dimensional nano materials, we can obtain the charge evenly distributed on the surface, thus enhancing its electrocatalytic activity. The same strategy will introduce the inherent dipole moment and band bending into the two-dimensional nano materials, so as to enhance their photocatalytic activity. 

      The multi-scale method developed by us can be combined with the machine learning method to generate a large number of data to train and optimize the model under different scales and different precision, so as to modify and optimize the parameters of the model, get a reliable model, and realize the efficient optimization design of materials. We propose a molecular packing factor to improve the efficiency of the algorithm prediction.

       基于主题设计和机器学习的材料设计

李有勇 

苏州大学功能纳米与软物质研究院

摘要:我们通过发展多尺度模拟方法和主题设计，设计了多种能源材料，如OPV、钙钛矿材料等。 通过“二元元素混合相”等新主题，设计了多种性能优异的电催化和光催化材料，通过硅和贵 金属复合，设计了多种新型HER电催化材料，如硅和铑复合材料。通过单原子元素和基底配体的 多种组合，我们筛选出了多种性能优异的材料，如用于HER和OER的材料。基于“二元元素混合 相”，通过在二维纳米材料上混合两种具有不同电负性的元素，我们能够获得均匀分布在表面的 电荷，从而增强其电催化活性。同样的策略会在二维纳米材料中引入固有的偶极矩和能带弯曲， 从而增强其光催化活性，我们并可以调整和改善二维纳米材料的带隙和其他特性。

      我们发展的多尺度方法可以和机器学习的方法相结合，在不同尺度、不同精度下产生大量 数据来训练和优化模型，从而对模型的参数进行修正和优化，得到可靠的模型，实现对材料的 高效优化设计，我们提出了分子堆积因子来提高算法预测的效率。