Identifying stable hybrid double perovskites with high-performance photovoltaic and thermal properties via machine learning

Abstract:
Hybrid organic-inorganic perovskites have emerged as promising semiconductor materials especially for photovoltaic applications due to the rapidly increasing power conversion efficiencies. However, currently the efficiencies and longterm stabilities remain the main obstacles for their commercial successes. In this work, based on the machine-learning regression models and a large scale of materials data, we built high-efficiency regression models for formation energies, bandgaps and Debye temperatures of hybrid organic-inorganic double perovskites (HOIDPs), and used them for the highthroughput screening over a huge fabricated chemical space of potential HOIDP candidates. 3629 HOIDP candidates with good chemical stabilities, appropriate bandgaps and high Debye temperatures were filtered out, among which lead-free AgIn/AgSb/AgBi-HOIDP candidates were most frequently chosen. Our work provides rich HOIDP candidates materials for future experimental verifications.