Data Driven Strategies for the Discovery and Optimization of Next Generation Energy Materials

The world is shifting toward clean and sustainable energy, and energy materials that are efficient, stable, and economically viable are in demand at an unprecedented rate. Traditionally, the search for these materials required 10-20 years of painstaking research, years of funding, and immense intellectual labor. Revealing energy materials that were efficient, stable economically upbeat was invariably a significant financial cost. Now with artificial intelligence (AI) already a reality in terms of machine learning and deep learning techniques, the discovery of new materials will be more rapid, easier, better, cheaper and more economically viable. Studies have suggested that AI has made it possible to accelerate materials processing by as much as 70%, while reducing costs of research and development by greater than 50%. AI models can, not only predict properties of materials but also optimize composition, design and simulate performance (even before the research is actually trialed) in areas like lithium-ion batteries, supercapacitors, fuel cells and photovoltaics. Energy materials screening platforms such as The Materials Project and AFLOW have already contributed to the screening of over 100,000 candidate materials. Rapidly developing neural networks, deep learning and graph-based predictive models (e.g, CGCNN) have made predictive analysis of properties easier imaginable and yielding more accurate and interpretable results. In summary this paper represents how AI is currently being used to foster the rapid and accessible discovery of energy materials, and the authors suggest that AI will have a large role to play in the science and engineering science revolution, going forward.