Paper Overview
- Field: Machine Learning
- Authors: Juno Nam, Bowen Deng, Xiaochen Du
- arXiv: 2508.08536
- Extends interatomic potentials from static energies to Gibbs free energy models, thermodynamically consistent by construction.
- Thermodynamic responses (entropy, volume, etc.) obtained via automatic differentiation with respect to temperature and pressure.
- TIP[UMA]: built on the universal potential UMA, trained on free energies from quasi-harmonic to MD-level fidelity, calibrated to higher-resolution computations or experiment.
- A single evaluation yields a crystal's equation of state and locates phase transitions among competing branches, including dynamically stable phases.
- Fine-tuning covers alloy solubility limits and miscibility gaps.
Abstract (English)
Free energies govern solid-state phase stability, yet computational materials discovery still relies largely on ground-state energies because free energy calculations require ensemble averages. We introduce the thermodynamic interatomic potential (TIP), which extends an interatomic potential from its static energy to a thermodynamically consistent Gibbs free energy model, with thermodynamic responses following from temperature and pressure by automatic differentiation. We implement TIP[UMA] using the universal potential UMA, train it on free energies from quasi-harmonic to molecular dynamics fidelity, and calibrate it to higher-resolution calculations or experiment. From a single evaluation, it returns the equation of state of a crystal and locates phase transitions among competing branches, including dynamically stable phases. Fine-tuning extends the model to alloy solubility limits and miscibility gaps. TIP makes free energies as accessible as potential energies, opening the door to high-throughput discovery of finite-temperature phase stability.
Key Contributions
Significance
By making free energies as cheap to evaluate as potential energies, TIP enables high-throughput screening of finite-temperature phase stability — a long-missing capability in computational materials discovery.
---
*Auto-collected on 2026-08-18*