Electrolyte design through physics-driven machine learning

PI: Venkat Viswanathan
Co-PI(s): Jay Whitacre
University: Carnegie Mellon University

The aim of the proposed project is to enable a rational design approach for electrolytes via a combination of physics-driven models that are coupled with large datasets and machine learning. The design of an electrolyte for properties such as conductivity or voltage stability depends on a large number of properties. While physics-driven models can identify simple descriptors, these tend to be inadequate for actual material selection. The advent of big data and machine learning allows the opportunity to couple physics-driven descriptor selection for machine learning. This effort will leverage our current capabilities around SEED (System for Electrolyte Exploration and Discovery), which contains an exhaustive dataset on liquid electrolytes. In partnership with Citrine Informatics, we aim to leverage the Citrination platform to carry out advanced electrolyte discovery.