The Impact of Machine Learning in Energy Materials Research: The Case of Halide Perovskites

Abstract

RecommendationsD ata is the new gold!Nope, data is the new oil!Surfing the Internet, you may find either proposition.Actually, data is probably neither a gold-nor oil-like asset.Gold exists in finite amounts and is essentially used in its metallic form as it is extracted.Oil is also finite, but differently from gold, it is further processed to obtain different products with enhanced added value.From this perspective, data is both gold and oil, but unlike them, its production increases steadily with no predictable limitation on its amount.In our digital economy, the huge amount of data created every day (some estimates point at >2.5 × 10 18 bytes) helps companies, stakeholders, and governments make decisions about their businesses or future actions.Paralleling the trend of data production, tools to manage big data sets have gained considerable momentum.The field of artificial intelligence (AI) has thus bloomed as a result of big data availability, assisting the user in making optimal decisions or in finding hidden correlations among different data sets, just to cite a few examples.Machine learning (ML) is a sub-field of AI where the ultimate goal is to automatically learn from a given known training set to possibly predict novel or undisclosed information beyond the initial data set.Materials science is no different from the digital economy in this respect (it may actually be even considered to be part of it, to some extent), whereby data production related to experiments or simulations has grown enormously in the past decade.This growth is driven by public and private funding programs worldwide, which see significant opportunities in the discovery of new materials or optimization and repurposing of existing ones.In this framework, materials for energy applications represent a significant fraction of investments, given the recognized importance of materials science in achieving alternative and clean energy production and storage, thus contributing to the mitigation of climate change.Recent advances in AI/ML hold the promise of revolutionizing the way materials for energy are discovered and optimized, and their processing engineered.In this Editorial, I briefly outline the progress, possible shortcomings, and future challenges in ML-assisted materials science, borrowing from selected examples in the recent scientific literature about halide perovskites.This family of wonder semiconductors and the associated surge of interest toward them seem to offer an ideal

Topics & Concepts