Deep learning potential molecular dynamics simulates the melting, elastic, and thermodynamic properties of the orthocarbonate
Phys. Rev. B 112, 224317 – Published 23 December, 2025
DOI: https://doi.org/10.1103/sj6h-c8qj
Abstract
The physical properties of the newly discovered Sr carbonate are of great importance for understanding deep carbon storage in Earth's interior. We investigated the melting curve, phase diagram, elastic properties, and thermodynamic behavior of the lower-mantle candidate mineral under high-temperature and high-pressure conditions using a deep learning potential model. The excellent agreement between our ab initio molecular dynamics results and high-precision structural optimizations validates the reliability of the trained potential. The accuracy of the deep learning potential was further verified by calculating the lattice constants and high-temperature and high-pressure equations of state of the Pnma structure using first-principles calculations, molecular dynamics simulations, and the quasiharmonic approximation. Zero-pressure melting temperatures of 1630, 1500, and 1508 K were obtained from the single-phase, void, and two-phase methods, respectively, and the melting transition was further confirmed through radial distribution functions and mean-square displacements. Elastic properties derived via the stress-strain approach reveal pronounced temperature-induced softening of the elastic constants and elastic moduli. Seismological parameters calculated from the elastic moduli indicate that Pnma remains a high-density and low-velocity carbonate throughout mantle conditions, suggesting potential relevance to deep carbon storage. Thermodynamic quantities, including heat capacity and entropy, are additionally evaluated within the quasiharmonic approximation. These results establish key constraints on the melting behavior, elasticity, and thermodynamics of under extreme conditions, providing insights into the high-pressure stability of alkaline-earth orthocarbonates within Earth's interior.