High Critical Transition Temperature Of Lead-based Perovskite Ferroelectric Crystals: A Machine Learning Study

Ferroelectrics undergoes a reversible structural phase transitions from the ferroelectric to the paraelectric phase when its temperature exceeds a critical temperature namely Curie temperature Tc.As ferro-paraelectric phase transition are always accompanied by heat-flow,dielectric and pyroelectric anomaly,the value of Tc is extremely important for ferroelectric applications.Lead-based perovskite ferroelectric crystals represented by PMN-PT have a high voltage coefficient.However,the value of their Curie temperature has become a major obstacle to their application.In this thesis,the Curie temperature of lead-based perovskite ferroelectric solid solution has been studied by machine learning methods.The critical temperature of 205 lead-based perovskites are collected from the published articles.The scaling natures are constructed from the physical and chemical properties for the corresponding B doping lead-based materials in order to describe them.Subsequently,we used Kernel Ridge regression(KRR),Support Vector regression(SVR)and Extremely randomized trees regression(ETR)to learn the Curie temperature of the lead-based perovskite ferroelectric solid solution.We use the 5 run of tenfold cross-validation method to evaluate the machine learning models.The Hyperparameters was also chosen carefully with the cross-validation to avoid overfitting.From our cross-validation,we found that the mean average error(MAE)between the predicted and experimental values of the machine learning methods were 14.4 K,14.7 K,and 16.1 K,respectively.We stack these three machine learning models together by averaging the output of each regression model and build an ensembled model.The MAE of the ensembled model was 13.9 K.Finally,we have predicted the Curie temperature of more than 200,000 lead-based perovskites and provide two ferroelectric materials that may have high Curie temperatures.