AMMBER: The AI-enabled Microstructure Model BuildER

Abstract: Properties of materials frequently depend on their microstructures, which are features at scales of tens of nanometers to hundreds of micrometers. Thermodynamic free energy, which provides the driving force for evolution, and kinetic properties, which determine how quickly the evolution can occur, together govern how a material evolves at the microscale. This project aims to develop algorithms and software that automate and optimize the selection of thermodynamic and kinetic parameters for simulations of microstructure evolution in multicomponent materials. In its current iteration, our software automatically sets up simulations of microstructure evolution for alloys based on thermodynamic information from empirical databases or atomistic simulations. Our framework relies on phase-field models that generalize easily to consider different multi-component, multi-phase systems. We are developing machine learning approaches for determining unknown thermodynamic and kinetic parameters from known values for similar material systems and from feedback between simulations and experiments.