Machine Learning Combined With First-Principles Calculations To Predict Two-Dimensional Perovskite Photovoltaic Materials

Recently,the search for renewable and clean energy has become a hot topic in current scientific research,because energy resource shortage and environmental pollution caused by fossil energy production came up.The traditional trial-and-error method and the method based on density functional theory have high error rate,high resource consumption and long time consuming.In this work,machine learning prediction and first-principles calculations are combined to successfully predict efficient and stable potential two-dimensional perovskite materials（Ba₂VON₃ and Sr₂VON₃）.Then the band structure,stability,optical absorption spectrum and theoretical maximum photoelectric conversion efficiency of potential materials are studied.At the same time the change of photovoltaic performance after carbon ion implantation for Sr₂VON₃ was studied.The main research contents of this work are as follows:Firstly,material screening was carried out based on machine learning.In order to improve the accuracy of prediction results,three machine learning models（GBR,EXTR,RF）were selected to train perovskite materials,and eight candidate photovoltaic materials were successfully obtained.Secondly,the stability and photovoltaic properties of materials were analyzed based on density functional theory.Through the calculation of molecular dynamics,phonon dispersion spectrum and decomposition enthalpy,it is found that two potential perovskite materials Ba₂VON₃ and Sr₂VON₃ have good dynamic stability and thermodynamic stability.At the same time,four potential materials（Al₂Hf O₄,Ba₂VON₃,Sr₂VON₃,and Sr₃Si₂Se₇）were verified to have excellent photovoltaic performance in the visible light range by calculating optical absorption spectrum and theoretical maximum photoelectric conversion efficiency.Finally,based on the time-dependent density functional theory,the potential two-dimensional perovskite material（Sr₂VON₃）with higher theoretical maximum photoelectric conversion efficiency was studied by ion implantation.The defects of C,Ag,Cu,H,He and N ion implanted into Sr₂VON₃ and the influence of photovoltaic properties were simulated,as well as the electron stopping power and the corresponding instantaneous charge response of C ion implanted material Sr₂VON₃.It is found that the defects are interstitial defects,and the ion implantation can effectively improve the photo-absorbing properties of materials.And the excited electrons located on valence band will affect the electron-phonon coupling,defect evolution and formation,and the photovoltaic properties of materials.Herein,the two kinds of efficient and stable two-dimensional perovskite materials（Ba₂VON₃,Sr₂VON₃）were obtained by using machine learning algorithm combined with the first principles calculation.It provides a theoretical foundation for the rapid discovery of high-efficiency,high-precision and environmentally stable two-dimensional perovskite solar cells and the further improvement of photovoltaic performance through ion implantation technology.