Screening And Predicting Of Novel Perovskite Structures Based On Machine Learning

Solar energy is a clean and pollution-free renewable energy source,and photoelectric conversion is one of the main ways to utilize solar energy resources.Inorganic double perovskites（A₂BB’X₆）have become one of the functional materials for photoelectric conversion due to their high stability and low production cost.However,there are many types of elements that constitute inorganic double perovskites,which can be searched in a wide range.Therefore,the rapid search for inorganic double perovskite structures with good optical properties and high thermal stability has become the key.The machine learning algorithm has low time cost and high search efficiency,and is expected to quickly search for inorganic double perovskite materials with better performance.Therefore,this paper uses a combination of density functional theory（DFT）and machine learning algorithm（ML）to search and predict inorganic double perovskites based on optical properties and thermal stability,respectively.The inorganic double perovskite structure with high performance provides theoretical guidance for the research and development of new perovskite solar cell materials.In the process of searching for inorganic double perovskite structures based on optical properties,the band gap,which measures the photoelectric conversion efficiency of perovskite,is used as a label.First,a machine learning model with a prediction accuracy of 95.9%was constructed according to the PBE（Perdew-Burke-Ernzerhof）band gap,and the PBE band gap value prediction of the inorganic double perovskite was calculated.Then,inorganic double perovskites were screened according to octahedral factor（0.4-1.0）,tolerance factor（0.82-1.08）,PBE band gap range（0.65 e V-1.5 e V）and whether they contained noble metal elements and toxic elements,two alternative inorganic double perovskite structures（K₂Na In I₆,Na₂Mg Mn I₆）were obtained.Finally,HSE（Hybrid Functional）band gap values,optical properties,and thermal stability of these two inorganic double perovskite structures are more precisely verified using DFT calculations.The study shows that HSE band gap values of these two inorganic double perovskite materials are suitable（K₂Na In I₆（E_g=1.46 e V）,Na₂Mg Mn I₆（E_g=1.89 e V））,and has good optical properties and high thermal stability.In the process of searching inorganic double perovskite structures based on thermal stability,the formation energy,which measures the thermal stability of perovskite at room temperature,is used as a label.First,a machine learning model with a prediction accuracy of 93.2%was constructed according to the formation energy,and the formation energy prediction of the inorganic double perovskite was calculated.Then,the inorganic double perovskites were screened according to octahedral factor（0.4-1.0）,tolerance factor（0.82-1.08）,formation energy range（<=0.1 e V/atom）and whether they contained noble metal elements and toxic elements,and obtained 37 stable inorganic double perovskite structures.After that,the band gap values of the screened 37 inorganic double perovskite structures were calculated,screened and verified using PBE functional and HSE functional,and two inorganic double perovskite structures（Cs₂Li In I₆,Rb₂Li In I₆）were obtained.Finally,the optical properties and thermal stability of the two inorganic double perovskite structures are more precisely verified using DFT calculations.The study shows that HSE band gap values of these two inorganic double perovskite materials are suitable（Cs₂Li In I₆（E_g=1.16e V）,Rb₂Li In I₆（E_g=1.14 e V））,and has photoelectric conversion efficiency and high thermal stability.To facilitate researchers to quickly calculate the formation energy of inorganic double perovskites,this research uses Py Qt5 to compile the formation energy prediction software of inorganic double perovskites,using the formation energy prediction module of this software,the formation energy of inorganic double perovskite materials can be calculated in a few seconds.