Abstract

Drawing on examples from alloy catalysis and lithium-ion batteries, I will demonstrate how machine learning can be used to significantly accelerate the design and discovery of new materials. In the first example, I will present a way to create complete maps of the surface structures and catalytic properties of an alloy catalyst at every point in its phase diagram. The key to this approach is the generation of surface cluster expansions that depend on the lattice parameter of the bulk material and are trained using a Bayesian method. I will discuss how these maps reveal insights into the catalytic properties of intermetallic phases and demonstrate how the maps can be used to rapidly identify the target synthesis conditions for stable and active catalysts. In the second example, I will demonstrate how training an interatomic potential model using on-the-fly machine learning can accelerate the calculation of activation energies for lithium diffusion by seven orders of magnitude, resulting in significantly improved agreement with experiment compared to pure density functional theory. Using the machine-learned interatomic potential models, we have identified new candidate coating materials designed to improve interfacial stability in solid state lithium-ion batteries.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call