Physics-informed generalized embedded-atom method potentials for metals and alloys

• Formulated and enhanced a physics-informed interatomic potential that incorporates pair, embedded, three-body, and non-local descriptors; 

• Wrote C and MATLAB codes to train the interatomic potential across binary (Mo-V), ternary (Mo-Ta-W) (in progress), and quaternary (Cr-Mo-Nb-V) (in progress) alloy systems using data from DFT calculations; 

• Adapted and modified C++ code within LAMMPS to support the newly developed interatomic potential for MD simulations; 

• Conducted comprehensive validation of the developed interatomic potentials by analyzing their performance across all relevant chemical spaces in different alloy systems; 

• Utilized MDMC simulations to explore phase diagrams and studied segregation properties in alloy systems.

Outcomes and impact: 

• Demonstrated that our physics-informed interatomic potential effectively captures the properties of alloys using smaller training datasets compared to other machine learning interatomic potentials; 

• Integrated the developed potential into the LAMMPS software, facilitating molecular dynamics simulations for larger systems and enabling the study of intricate atomic behaviors; 

• Employed simulations for phase diagram and segregation analysis in challenging experimental areas. For example, uncover low-temperature phase stability in the Mo-V system, revealing a B2 phase below 800 K and significant V segregation at grain boundaries;

• Open-sourced the training and LAMMPS codes, freely available to the scientific community for collaborative research and development.

Shi, H. & Samanta, A.† (2024, in preparation). Chemical segregation and strengthening mechanisms in the CrMoNbV high-entropy alloy by the generalized embedded atom method potential.

Shi, H.*, Wang, W.*, & Samanta, A.† (2024, in preparation). Systematic design and Fisher information theory analysis of physics-informed interatomic potentials.

Shi, H., Sharma, B., & Samanta, A.† (2024). Analysis of phase stability and chemical segregation in the Mo-V alloys using a generalized embedded atom method potential. Computational Materials Science, 233, 112732. https://doi.org/10.1016/j.commatsci.2023.112732