Elements: Data Driven Autonomous Thermodynamic and Kinetic Model Builder for Microstructural Simulations

Materials with improved properties can dramatically impact sustainability, human welfare, and national prosperity. As an example, a stronger material can reduce the weight of vehicles and can therefore reduce energy consumption and pollution. Properties of materials frequently depend on their microstructures (features in materials at scales of one micrometer to hundreds of micrometers).

Thermodynamic free energy (providing the driving force for evolution) and kinetic parameters (providing how quickly the evolution can occur) together govern how a material evolves at the microscale. This project develops algorithms and software that automate the extraction of the thermodynamics and kinetic information using artificial intelligence to enable simulation of microstructure evolution for complex mixtures of metals.

The AI-enabled Microstructure Model BuildER (AMMBER) harvests and harnesses data ranging from first-principles calculations, experimental micrographs and associated natural language text, and thermodynamic databases, as well as custom user input. It then produces input to microstructure evolution models that facilitate the fundamental understanding needed to gain control of the microstructure and resulting material properties.

The demonstration of its capability is planned for commercially important alloys (nickel-aluminum-based and aluminum-copper-based alloys), as well as the corresponding high-entropy alloys (alloys with five or more components with near equimolar fractions). AMMBER contributes to the software infrastructure for simulation-based material discovery and development within the context of the Material Genome Initiative.

Training activities, including training workshops for the community to learn about the software and the theory behind it and integration into the undergraduate and graduate thermodynamics and kinetics courses, provide opportunities for education and professional development. Nickel-aluminum-based and aluminum-copper-based alloys are key materials in the aerospace and automobile industries, and thus the results are expected to have a direct impact on manufacturing.

The goal of this project is to develop an artificial intelligence framework for the autonomous determination of input parameters for phase-field models based on a variety of data sources to establish constraints on the model parameters. The AI-enabled Microstructure Model BuildER (AMMBER) leverages automated data-stream pipelines to collect, curate, and tabulate disparate data sources spanning first-principles calculations, experimental micrographs, and associated natural language text, thermodynamic databases, and custom user input.

Then, advanced optimization algorithms iteratively optimize phase-field parameters such that the resulting models reproduce known microstructural characteristics (e.g., the phase fraction and characteristic length scale as a function of time). These models can then be used to simulate the microstructural evolution of materials over a range of conditions that are relevant to engineering and manufacturing.

The demonstration of AMMBER involves commercially important Ni-Al-based and Al-Cu-based alloys, some of which contain more than five components, leading to a complicated high-dimensional parameter space in which thermodynamic and kinetic model parameters must be optimized. The application to high-entropy alloys, which contain near equimolar amounts of five or more components, provides a ground for new scientific discoveries.

By automating the time-consuming initial model parameterization, AMMBER reduces the human bottleneck of materials modeling and paves the way to increased throughput of phase-field simulations. AMMBER complements existing Materials Genome Initiative (MGI) efforts, and it leverages and integrates into existing computational materials research communities built around tools such as open-source phase-field software (PRISMS-PF, MOOSE), an integrated computational materials engineering framework (PRISMS), CALPHAD tools (ESPEI, Thermo-Calc), and a dissemination platform (nanoHUB).

The training workshops and integration of the computational tools and research findings into classrooms facilitate community interaction and engagement. Ni-Al-based and Al-Cu-based alloys are key materials in the aerospace and automobile industries, and thus the results are expected to have a direct impact on manufacturing.

This proposal receives funds through the Office of Advanced Cyberinfrastructure in the Computer and Information Science and Engineering Directorate and the Division of Materials Research in the Mathematical and Physical Sciences Directorate.

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.