Predicting Perovskite Performance with Multiple Machine-Learning Algorithms

Abstract

Perovskites have attracted increasing attention because of their excellent physical and chemical properties in various fields, exhibiting a universal formula of ABO3 with matching compatible sizes of A-site and B-site cations. In this work, four different prediction models of machine learning algorithms, including support vector regression based on radial basis kernel function (SVM-RBF), ridge regression (RR), random forest (RF), and back propagation neural network (BPNN), are established to predict the formation energy, thermodynamic stability, crystal volume, and oxygen vacancy formation energy of perovskite materials. Combined with the fitting diagrams of the predicted values and DFT calculated values, the results show that SVM-RBF has a smaller bias in predicting the crystal volume. RR has a smaller bias in predicting the thermodynamic stability. RF has a smaller bias in predicting the formation energy, crystal volume, and thermodynamic stability. BPNN has a smaller bias in predicting the formation energy, thermodynamic stability, crystal volume, and oxygen vacancy formation energy. Obviously, different machine learning algorithms exhibit different sensitivity to data sample distribution, indicating that we should select different algorithms to predict different performance parameters of perovskite materials.

Machine learning in energy chemistry: introduction, challenges and perspectives

“…For example, Zhang et al establish the lattice constants model based on cubic perovskites 127,128 . Besides, Li et al predicted formation energy, thermodynamic stability, crystal volume and oxygen vacancy formation energy using a variety of machine learning models 129 . Saidi et al constructed a convolutional neural (CNN) model for deriving relevant physical properties (e.g.…”

“…127,128 Besides, Li et al predicted formation energy, thermodynamic stability, crystal volume and oxygen vacancy formation energy using a variety of machine learning models. 129 Saidi et al constructed a convolutional neural (CNN) model for deriving relevant physical properties (e.g., lattice constants, octahedral tilt angles, etc.) from the given perovskite material.…”

Process Insights into Perovskite Thin‐Film Photovoltaics from Machine Learning with In Situ Luminescence Data

“…[24] Time series models have been developed for accelerated material stability evaluation and performance forecasting in humid environments. [25][26][27] Furthermore, data-driven approaches can be used to generate predictive models for optoelectronic characteristics such as the bandgap [28][29][30][31] based on theoretical physical material features like ionization energy, atomic/molecular sizes, or lattice constant. These material parameters can be extracted from open-access material properties databases.…”

A domain knowledge enhanced machine learning method to predict the properties of halide double perovskite A2B+B3+X6

“…[11][12][13][14][15] However, different machine learning algorithms exhibit different sensitivities to material sample distribution. 16 Therefore, it is necessary to select different algorithms to predict different properties of perovskite materials. For example, XGBoost and ridge regression (RR) can model the linear structure-property relationships in perovskite materials for predicting thermodynamic stability effectively, while random forest (RF), GBR and SVR are more suitable for capturing the nonlinear relationship between band gap and input features.…”