Revealing interstitial energetics in Ti-23Nb-0.7Ta-2Zr gum metal base alloy via universal machine learning interatomic potentials
Journal of Materials Research and Technology, 2026
Please see the full published article: https://doi.org/10.1016/j.jmrt.2026.02.204
Highlights:
- Universal MLIPs enable high-throughput mapping of interstitial energetics in Ti-23Nb-0.7Ta-2Zr gum metal base alloy.
- C, N, O, and H show broad energy distributions across thousands of local chemical environments.
- MACE-MATPES-PBE-0 and Orb-v3 recover expected bcc site preferences, with C, N, and O favoring octahedral sites and H favoring tetrahedral sites.
- Ti-rich local environments strongly stabilize interstitials, while close proximity to Nb is consistently destabilizing.
- SevenNet-0 shows a limitation for H, predicting octahedral stability contrary to DFT validation.
- DFT benchmarks confirm the energetic ordering of representative configurations and validate tetrahedral stability of H.
- uMLIPs achieve near-DFT insight at orders-of-magnitude lower computational cost, enabling statistically broad defect analysis.
- The study provides atomistic guidance for understanding and tailoring interstitial behavior in gum metal alloys.
- uMLIP-Interactive supports reproducible uMLIP simulations through an accessible GUI and automatic generation of shareable all-in-one Python scripts