Elements: Software to enable first-principles multi-scale simulations

The project supports the development of a scale-bridging software infrastructure to predict non-equilibrium behavior of materials from first principles. In the course of its development, the software infrastructure is being applied to predict the thermodynamic and kinetic properties of important lithium-metal alloys. Lithium alloys are currently of great interest to serve as anodes in all-solid-state Li batteries, where they can replace the graphite anodes of commercial battery technologies and thereby enable significant increases in energy densities.

Beyond its application to lithium alloys, the software tools are designed to empower scientists and engineers to generate fundamental and mechanistic insights about the dynamic response of materials in both functional and structural applications.

A multi-scale approach is pursued that relies on first-principles statistical mechanics to calculate the essential thermodynamic and kinetic ingredients of generalized phase-field models that describe morphological evolution of a material out of equilibrium. A key component of the approach is the use of cluster expansion based surrogate models to interpolate first-principles electronic structure calculations within Monte Carlo and molecular dynamics simulations.

The infrastructure consists of libraries, executables and jupyter notebooks that expand upon the CASM software package, a first-principles statistical mechanics code suite, and thereby enable the parameterization of machine-learned cluster expansion surrogate models for both crystalline and non-crystalline states of matter. The software infrastructure consists of (i) enumeration tools to generate a rich database of crystallographic and non-crystallographic models to train machine-learned interatomic potentials (MLIPs); (ii) software to sample surrogate models for uncertainty quantification in calculated thermodynamic and kinetic properties; and (iii) software to enable coarse-graining schemes that map predictions of MLIPs onto crystal-based cluster expansion Hamiltonians.

The ability to link macroscopic non-equilibrium behavior to properties at the electronic structure level enables the formulation of powerful design principles with which new materials can be discovered through subsequent high-throughput first-principles calculations.

This award by the NSF Office of Advanced Cyberinfrastructure is jointly supported by the Division of Materials Research.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.