Rapid advances in information technology have made it possible to generate, analyze, and distribute large data sets of material properties. One of the great challenges in materials research is to effectively make use of these data sets to accelerate the design and development of new materials. To this end, I will demonstrate how data can be leveraged in two ways: to accelerate the design of nanoalloy catalysts, and to reduce the the computational cost of electronic structure calculations on crystalline materials. For nanoalloy catalysts, we have developed a machine learning approach for constructing model Hamiltonians that are capable of rapidly predicting the structure and properties of multi-nanometer nanoparticles with accuracy close to that of density functional theory. I will discuss the unique insights and design guidelines this approach provides for highly active Pt-Ni-based oxygen reduction catalysts. I will show how sub-surface atomic order can play a significant role in determining catalytic activity and discuss likely reasons for the enhance stability of Pt-Ni nanocatalysts after they have been doped with small amounts of Mo. For electronic structure calculations, I will demonstrate how informatics can be used to rapidly generate highly efficient k-point grids, addressing a longstanding problem in computational materials research. Our group has developed a publicly accessible k-point grid server backed by a database of tens of thousands of k-point grids, and we estimate that for a broad range of computational methods the use of this server should nearly double the speed of electronic structure calculations on materials.

Accelerating materials research through the effective use of data

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.