About this Abstract |
Meeting |
2023 TMS Annual Meeting & Exhibition
|
Symposium
|
Computational Thermodynamics and Kinetics
|
Presentation Title |
H-29: A Machine Learned Spin-Lattice Potential for Bulk Iron |
Author(s) |
Benjamin Seddon, James Elliott, Christoph Ortner |
On-Site Speaker (Planned) |
Benjamin Seddon |
Abstract Scope |
It is well known that magnetic effects stabilise the room temperature body-centred cubic (BCC) phase of iron, and so it is important to take these into account in computer simulations. Here, we describe our contribution to the recent set of machine learning potentials that incorporate magnetic effects by explicitly expressing the energy in terms of the atomic magnetic and spatial degrees of freedom. We use the Lagrangian formulation of Constrained Density Functional Theory (cDFT), as implemented in Abinit, to generate an open-source dataset of cDFT for bulk bcc iron. This dataset provides atomic position and spin configurations with the associated energy and energy derivatives (force, spin torque, pressure, etc.) for bulk bcc iron, including defects and spin fluctuations. We then use this dataset to parameterise a magnetic atomic cluster expansion potential using ACE.jl, and demonstrate its use by studying defect diffusion in BCC iron. |
Proceedings Inclusion? |
Planned: |
Keywords |
Iron and Steel, Magnetic Materials, Machine Learning |