Skip to main content
A domain-specific atomistic model for Earth’s core materials

A domain-specific atomistic model for Earth’s core materials

This is a Preprint and has not been peer reviewed. This is version 2 of this Preprint.

Add a Comment

You must log in to post a comment.


Comments

There are no comments or no comments have been made public for this article.

Downloads

Download Preprint

Authors

Tianqi Wan, Liangrui Wei, Zepeng Wu, Renata Wentzcovitch, Yang Sun 

Abstract

Understanding Earth’s core requires accurate descriptions of multicomponent iron alloys under extreme conditions. Here, we develop the Earth’s Core Atomistic Model (ECAM) for Fe–Ni–S–Si–O–C–H alloys through sampling with a foundational atomistic model, targeted ab initio simulations, and iterative exploration. A compact dataset selected from nearly half a million configurations is consistently labeled using high-accuracy first-principles calculations and used to train a lightweight model. The model explicitly includes thermal electronic contributions and runs nearly two orders of magnitude faster than the original foundational model. ECAM corrects artifacts in the foundational model’s description of core materials and reproduces the density, seismic velocity, Gibbs free energy, and phase diagrams of Fe and binary alloys in agreement with previous ab initio calculations. ECAM is expected to enable more realistic modeling of chemical partitioning and crystallization in Earth’s core. Our workflow also provides a practical route from foundational-model sampling to efficient domain-specific simulation.

DOI

https://doi.org/10.31223/X5K23H

Subjects

Earth Sciences, Mineral Physics

Keywords

Earth's core, Machine learning, Atomistic model

Dates

Published: 2026-11-04 14:06

Last Updated: 2026-11-04 14:08

Older Versions

License

CC BY Attribution 4.0 International

Metrics

Views: 16

Downloads: 2