About this Abstract |
Meeting |
2023 TMS Annual Meeting & Exhibition
|
Symposium
|
Computational Thermodynamics and Kinetics
|
Presentation Title |
Diffusivity in a Multicomponent Alloy Using Machine Learning and Variational Approaches |
Author(s) |
Dallas R. Trinkle, Soham Chattopadhyay |
On-Site Speaker (Planned) |
Dallas R. Trinkle |
Abstract Scope |
The diffusivity of multicomponent or high-entropy alloys helps determine the ultimate range of stability of their solid solution phase. Computing the diffusivity of multicomponent alloys presents a challenge to sample the state space accurately and adequately while determining the transport coefficients. The large number of components suggests a density-functional theory-based approach, but the large state space makes kinetic Monte Carlo unattractive. Here, we use a variational method for the transport coefficients to consider two approaches that require orders-of-magnitude fewer calculations: a mean-field Green function method, and a novel hybrid machine-learning method. The methodologies are tested using classical potential simulations of high-entropy alloys, to quantify the computational savings as well as expected accuracy. Moreover, as both approaches are variational, they serve as upper limits to the true diffusivity; this can allow for more rapid screening in alloy design. |
Proceedings Inclusion? |
Planned: |