How Can Machine Learning Help Address Climate Change? (2021-02-12)

Abstract

Solving the climate change crisis requires dramatic changesthat can be achieved by policy, technology, or personalbehavior. Significant improvements in the cost of technologicalsolutions for renewable energy are still required to make themeconomically competitive with fossil fuels. This, in turn, requiresmodifications to the device design and materials employed.Parameter space for the discovery of new materials is vast andimpossible to explore with conventional brute-force sampling,neither computationally nor experimentally. Machine learning isa new tool that allows us to learn the empirical rules connectingdisparate experimental observations. Such interpolation allowsus to explore the parameter space more efficiently and evenemploy the inverse design, i.e. predicting new materials withdesired properties. I will highlight how we use machine learning in our group forpredicting and synthesizing new materials for solar cells, Li-ionbatteries, and various catalysts.