CMU & Partners Receive ARPA-E Award for Machine Learning-Accelerated Discovery of Energy Materials
Carnegie Mellon University, Citrine Informatics, Julia Computing and MIT were awarded $0.6 million in funding from the U.S. Department of Energy’s Advanced Research Projects Agency-Energy (ARPA-E). The funding will be used to speed up the process of materials innovation for electrochemical devices, batteries for next-generation electric vehicles and new routes to making chemicals and fuels.
“The common wisdom is that taking a material innovation from lab to market is around 15 years; we want to fundamentally alter this paradigm by bringing this time down by an order of magnitude—enabling many material-related innovations to escape the valley of death,” said Venkat Viswanathan, associate professor of Mechanical Engineering, energy fellow at the Wilton E. Scott Institute for Energy Innovation at Carnegie Mellon University and lead principal investigator of the project.
"We are excited to work with an outstanding team using machine learning to drive a combination of experiments and physics-based computation. This close coupling of machine learning and domain knowledge can fundamentally change the timescale associated with materials development," said Bryce Meredig, founder and chief science officer at Citrine Informatics.
The team will create detailed designs for catalyst systems for electrochemical reactions that convert electrical energy into carbon-neutral chemicals and fuels and electrolyte systems for next-generation batteries. Designing electrochemical systems capable of high turnover and efficiency is a challenge to enable the cost-effective production of carbon-neutral chemicals and fuels. Designing liquid electrolytes for next-generation batteries will provide an alternative transportation fuel to petroleum by improving energy density, thus enabling long-range electric vehicles.
Carnegie Mellon University, Citrine Informatics, Julia Computing and MIT received this competitive award from ARPA-E’s Design Intelligence Fostering Formidable Energy Reduction (and) Enabling Novel Totally Impactful Advanced Technology Enhancements (DIFFERENTIATE) program, which aims to enhance the productivity of energy engineers in helping them to develop streamlined solutions to next-generation energy challenges.
The program identified three general mathematical optimization problems that are common to many design processes. The selected projects then conceptualized machine learning and artificial intelligence-based solutions to one of the three general mathematical optimization problems that are anticipated to help engineers execute and solve these problems in a manner that dramatically accelerates the pace of energy innovation.
The core innovation of the team’s work involves pairing machine learning based filtering of candidate materials with accelerated high-fidelity modeling to efficiently search a large design space for high-performance materials under realistic operating conditions. In particular, the project will develop Julia-based software and hardware-accelerated methods for high-fidelity objective function evaluation, and an efficient global optimization approach using sequential learning and design of experiments to achieve their ambitious goals.
“Julia is the fastest, most productive programming language for machine learning and artificial intelligence, and we are excited to partner with MIT, Carnegie Mellon and Citrine Informatics to conduct research to speed materials innovation,” said Viral Shah, founder and CEO of Julia Computing.