Methods, workflows, and analysis from the MaterSynq team — covering MLIP theory, DFT best practices, and the state of cathode material discovery.
How machine-learning interatomic potentials trained on DFT data can reproduce formation energies and phonon spectra at orders-of-magnitude lower compute cost — and what that means for cathode material discovery.
Surface terminations govern ionic conductivity, dendrite nucleation, and solid-electrolyte interphase formation. We examine how accurate surface energy prediction reshapes cathode candidate ranking.
A walkthrough of the MaterSynq structure-prediction pipeline: AFLOW-style prototype enumeration, MLIP-driven relaxation, DFT single-point validation, and stability ranking against the convex hull.
Running four hundred Ni-rich layered oxide variants through simulation to identify thermal stability thresholds and oxygen release onset — shrinking experimental scope from months to a targeted campaign.
Even with exascale HPC access, most battery R&D programs still run at the same discovery cadence as a decade ago. We diagnose the real bottleneck: workflow fragmentation, not compute throughput.
Formation energy on the convex hull is necessary but not sufficient. Phonon dispersion screening eliminates dynamically unstable structures before costly synthesis — here's how we integrate it into the screening funnel.
Carnegie Mellon's MSE department, NETL's in-house simulation capabilities, and the Allegheny region's advanced manufacturing legacy create an underrated foundation for computational materials startups.
