Microstructure Mapping of Complex Perovskite Materials from Machine Learning-Enabled Energy Model

Hsin-An Chen¹, Ping-Han Tang¹, Chun-Wei Pao^1*

¹Research Center for Applied Sciences, Academia Sinica, Taipei, Taiwan

* Presenter:Chun-Wei Pao, email:cwpao@gate.sinica.edu.tw

Revealing the correlations between chemical compositions, microstructures, and material properties is a critical but challenging task for designing advanced chemically complex perovskite materials aiming for enhancing both efficiency and stability. In this work, the combination of machine learning and atomistic simulations enables us to evaluate close to one million atomistic structures of MA_yFA _1-y Pb(Br_xI _1-x) ₃ mixed ion perovskite over the whole composition space in a computationally feasible manner, thereby allowing us to map the microstructures and material properties with chemical compositions. Subsequent correlation analysis reveals the process-structure-property (PSP) relationships of MA_yFA _1-y Pb(Br_xI _1-x) ₃ mixed ion perovskite, suggesting that the compositions give low lattice distortion could retain the microstructure of single-phase solid solution and yield high photovoltaic conversion efficiencies. The computational framework employed in the present study can be readily extended to complex perovskite materials with even more constituents to discover their optimal compositions.

Keywords: mixed ion perovskite, atomistic simulation , machine learning, microstructure