Learning ab initio phase-field models

Abstract

Simulating microstructure evolution requires quantum-mechanical accuracy and mesoscopic reach in length and time scales, a combination that no current method achieves. Classical phase-field models provide this reach, but their accuracy is limited by phenomenological free energies and mobilities. Here we develop a framework for learning ab initio phase-field models, where the mesoscopic equation is not postulated but derived from a Mori-Zwanzig projection of molecular dynamics onto species-density fields under explicit assumptions. The nonlocal free energy and mobility left unspecified by this equation are parametrized by neural networks and learned from short molecular dynamics trajectories generated with machine-learning interatomic potentials of ab initio accuracy. We demonstrate the framework on an iron-boron melt and on hydrogen-helium mixtures under planetary conditions. For iron-boron, the model shows that the melt at the FeB₄ composition is spinodally unstable at ambient pressure but stabilized at 10 GPa, offering a thermodynamic rationale for why FeB₄ has been synthesized only under high pressure. For hydrogen-helium, the model predicts the immiscibility boundary and captures droplet nucleation and growth in helium-rain simulations of a column corresponding to 2.2 million atoms, far beyond the scale of atomistic modeling at comparable accuracy. Trained across compositions and conditions, such models could provide a mesoscopic counterpart to ab initio molecular dynamics.

Momentum conservation for ab initio calculations with pseudopotentials via a phase space electronic structure formulation
Guy Ohad, Nadine C. Bradbury, Joseph E. Subotnik
arXiv13 Oct 2026PreprintarXiv licenceFull textCondensed matterarXiv
High-Field Electron Transport in AlGaN alloys: A Full-Band Monte Carlo Study Based on Ab Initio Supercell Calculations
Animesh Datta, Matinehsadat Hosseinigheidari, Uttam Singisetti
arXiv7 Sep 2026PreprintCC BY-NC-ND 4.0Full textApplied physics & instrumentation Condensed matterarXiv
Fracture-Bench: A Controlled Benchmark Suite for Machine Learning Surrogates in Phase-Field Fracture Mechanics
Meenu Krishnan U., Maryam Hakimzadeh, Indrashish Saha +3
arXiv2 Nov 2026PreprintCC BY 4.0Full textComputational physicsarXiv
Bottom-Up Prediction of Amorphous Poly(1,3-dioxolane) using Ab Initio Reactive Machine-Learning Force Fields
Yizhi Song, Bharatha K. Gunawardana, Kriti Alam +3
arXiv7 Oct 2026PreprintarXiv licenceFull textPhysical chemistry Condensed matterarXiv
Material and thermal properties of MgCl2 molten salt by ab initio and machine-learning molecular-dynamics simulations
Roberto Llovera, María Andrea Barral, Verónica Vildosola +2
arXiv24 Sep 2026PreprintCC BY-NC-SA 4.0Full textCondensed matterarXiv

Cite

BibTeX RIS

BibTeX

@misc{chen2026learning,
  title = {{Learning ab initio phase-field models}},
  author = {Mengyi Chen and Peichen Zhong and Zihan Zhang and Qianxiao Li},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.01432},
  eprint = {2610.01432},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.01432},
}

RIS

TY  - UNPB
TI  - Learning ab initio phase-field models
AU  - Mengyi Chen
AU  - Peichen Zhong
AU  - Zihan Zhang
AU  - Qianxiao Li
PY  - 2026
DA  - 2026/10/01
PB  - arXiv
DO  - 10.48550/arxiv.2610.01432
UR  - https://arxiv.org/abs/2610.01432
AB  - Simulating microstructure evolution requires quantum-mechanical accuracy and mesoscopic reach in length and time scales, a combination that no current method achieves. Classical phase-field models provide this reach, but their accuracy is limited by phenomenological free energies and mobilities. Here we develop a framework for learning ab initio phase-field models, where the mesoscopic equation is not postulated but derived from a Mori-Zwanzig projection of molecular dynamics onto species-density fields under explicit assumptions. The nonlocal free energy and mobility left unspecified by this equation are parametrized by neural networks and learned from short molecular dynamics trajectories generated with machine-learning interatomic potentials of ab initio accuracy. We demonstrate the framework on an iron-boron melt and on hydrogen-helium mixtures under planetary conditions. For iron-boron, the model shows that the melt at the FeB_4 composition is spinodally unstable at ambient pressure but stabilized at 10 GPa, offering a thermodynamic rationale for why FeB_4 has been synthesized only under high pressure. For hydrogen-helium, the model predicts the immiscibility boundary and captures droplet nucleation and growth in helium-rain simulations of a column corresponding to 2.2 million atoms, far beyond the scale of atomistic modeling at comparable accuracy. Trained across compositions and conditions, such models could provide a mesoscopic counterpart to ab initio molecular dynamics.
ER  -

Abstract

Related

Cite