Perovskite-Based Materials for Photovoltaic Applications: A Machine Learning Approach

Abstract

The future of our planet depends greatly on sustainable energy sources and environmental preservation. Our modern society’s primary energy source is fossil fuels, which emit enormous amounts of carbon dioxide and contribute significantly to global warming. Due to global concerns about the environment and the increasing demand for energy, technological advancement in renewable energy is opening up new possibilities for its use. Even today, solar energy continues to be the most abun-dant, inexhaustible, and clean form of renewable energy. In this context, scientists and engineers across the world are working toward the development of highly effi-cient and cost-effective photovoltaic devices. As we move from the first to the third generation of solar cells, although their production cost decreases, their efficiency is also reduced. In the past few years, perovskites emerged as outstanding mate-rials for photovoltaic applications. Halide perovskites have been reported to exhibit a power efficiency of 25.5% due to their excellent defect tolerance, high optical absorption, the minimization of recombination, and long carrier diffusion lengths. Furthermore, halide perovskite materials are more affordable and easier to construct than silicon-based classic solar cells. Thus, it is of paramount significance to design advanced perovskite materials with higher photovoltaic efficiency. Although mixed lead-free and inorganic perovskites have been established as promising photovoltaic materials, their enormous composition space makes it difficult to find compositions with desired bandgap and photovoltaic parameters. The bottleneck impeding this advancement can be addressed by either (1) following a trial-and-error approach to collect enormous experimental data and designing advanced materials with desired properties leading to the development of high-efficiency photovoltaic devices or (2) combining the strengths of experimental materials science and machine learning to understand the underlying compositional and structural descriptors governing the efficiency of these devices. This chapter includes a brief discussion of perovskite materials and the developments made in the lead-free perovskite for photovoltaics. Further, this discussion will turn to the collection and analysis of materials data and extend to the descriptors used to describe the performance and properties of lead-free perovskites. The overarching aim of this chapter is to discuss how ML can be used to design advanced lead-free perovskites with desirable bandgaps and stabilities.