What Is a Machine-Learned Interatomic Potential?
An MLIP gives you near-DFT accuracy at close to classical-MD speed. Here is how these potentials work, where the speed comes from, and how they fail.
Author: Dinesh Varma Mudunuri
PhD researcher in computational materials science at NIT Warangal, with a Master of Materials Technology from UNSW. I work on atomistic simulation and machine-learned interatomic potentials — which ones deserve trust, on which systems, and how you would know. Before the doctorate: four years of sheet-metal forming experiments, and two years building machine learning models in production.
BCC vs FCC vs HCP: Why Crystal Structure Decides Ductility
BCC steel has 48 slip systems and shatters in the cold. FCC copper has 12 and never does. Here is the packing-and-slip physics that decides which.