Kristin Persson1,2
UC Berkeley1,Lawrence Berkeley National Laboratory2
Kristin Persson1,2
UC Berkeley1,Lawrence Berkeley National Laboratory2
Entering the era of the fourth paradigm of materials science, we recognize that highly curated, tested, and provenanced data is the fuel for machine learning. As part of the Materials Genome Initiative, the Materials Project (www.materialsproject.org) was founded in 2010 at Lawrence Berkeley National Laboratory to accelerate materials design using supercomputing and software infrastructure together with state-of-the-art quantum mechanical theory. Today, the Project has over 400,000 registered users and supplies millions of data records every day to an increasingly data-hungry community. As one example, early attention to the importance of data, the Project saved and curated millions of relaxation trajectories, which since 2017 has provided the basis of training highly accurate interatomic machine learning potentials. In this talk we will give an update on recent work in the realm of data-driven materials design for energy applications, showcase successes and comment on remaining bottlenecks in accelerating materials innovation.