Research Experiences for Undergraduates in Computational Methods with Applications in Materials Science
MSU Affiliation
James Worth Bagley College of Engineering; Michael W. Hall School of Mechanical Engineering; Center for Advanced Vehicular Systems; Center for Computational Sciences
Major
Physics
Research Mentor
Doyl Dickel
Creation Date
7-27-2026
Abstract
Machine learning interatomic potentials (MLIPs) bridge the gap between high-accuracy quantum mechanics and high-speed classical force fields, enabling quantum-accurate, large-scale molecular dynamics. Cutoff functions, which smoothly truncate atomic interactions beyond a specific radial neighborhood are crucial due to their local scaling and computational efficiency. However, geometric constraints and boundary discontinuities—particularly at the short- range inner cutoff can cause severe force spikes and simulation instabilities. This work investigates Moment Tensor Potentials (MTP) with different cutoff functions using a diverse benchmark suite encompassing three distinct structural phases: bec metals (Li, Mo), fcc metals (Cu, Ni), and diamond-structure semiconductors (Si, Ge) to span the range of crystal structures. The purpose of this work is to improve the potentials developed using the moment tensor methodology by shifting from a set cutoff function to a fitted spline cutoff function both built upon the Chebyshev polynomials for our radial basis functions. Using LAMMPS, we'll compare Density-Functional Theory (DFT) data with our results from molecular dynamics simulations for lattice parameter, elastic constants, bulk modulus, migration energy, and vacancy formation energy.
Presentation Date
Summer 7-30-2026
Keywords
interatomic potential, machine learning
Recommended Citation
Wlochal, Michael and Dickel, Doyl, "Developing a Fitted Cutoff Function for MTP" (2026). Research Experiences for Undergraduates in Computational Methods with Applications in Materials Science. 23.
https://scholarsjunction.msstate.edu/ccs-reu/23
