Research On Perovskite Material Design Based On Data-drive

In the field of materials science,the development and design of new materials are often achieved through traditional experimental "trial and error" methods and modeling calculations,but this requires a lot of manpower and resources,as well as a long material research and development cycle.In recent years,with the continuous development of science and technology,artificial intelligence algorithms are constantly updated and have been widely used in the field of materials.Since the material genome project was proposed,more and more researchers have paid attention to it and explored it,forming a data-driven research paradigm.In the field of material science,this method is mainly through high-throughput first-principles calculation,high-throughput material chemistry experiments and obtaining a large amount of data from existing material databases,and using machine learning algorithms to model and analyze the data,so as to achieve the effect of reverse design of materials.This can not only accelerate the understanding of the properties of materials by scientific researchers,but also greatly decrease the cost of material design and shorten the material development cycle.At present,perovskite materials have been paid close attention by researchers because of their excellent properties and broad application prospects,such as in the field of solar cells and catalysis.Some researchers have also noticed their application in the field of energy storage.However,perovskite materials also have poor stability problem.Therefore,in this paper,perovskite materials are designed using the data-driven paradigm and artificial intelligence algorithm.It mainly includes the following aspects:In the first part,the adsorption energies of 640 different two-dimensional perovskites and different ion combinations were calculated by high-throughput first-principles calculation,filled with inorganic descriptors,and the perovskite material data set was established.A variety of machine learning algorithms were used for modeling,and three groups of perovskite material combinations with excellent ion adsorption energy,stable structure and appropriate band gap were selected from 11976 potential two-dimensional perovskite and ion combinations.It provides a basis for the application of perovskite materials in the field of ion storage.In the second part,this paper tested the passivation effect of 78 groups of different double dyes on perovskite through high-throughput perovskite photoelectrochemistrical experiment,filled the organic descriptors in the database,established the data set,and used machine learning algorithm modeling to design two different combinations of dyes consensitizers to passivate perovskite,which improved the stability of perovskite by 4% and the photoelectronic property by 16.7%,and carried out the symbolic regression for feature engineering.This formulates descriptors that are more relevant to the stability of perovskite,and provides suggestions for the subsequent design of molecular passivators for perovskite stability.