Deep learning assisted exploration of superionic states and melting temperatures in Li-Sn superconductors
Phys. Rev. B 112, 064516 – Published 28 August, 2025
DOI: https://doi.org/10.1103/g9dv-fl6q
Abstract
The unique spatial arrangement of Li atoms in Li-rich compounds induces properties on Li-rich compounds, including interstitial quasiatoms, superconductivity, and superionic states. However, the dependence of increasing Li concentration on electron-phonon coupling (EPC) and superionic behavior at high pressure remains inadequately understood. Here, we systematically investigate the crystal structure of the ( = 1–8) compound at 0–150 GPa using structure prediction and first-principles calculations. Band calculations reveal that the phases of ( = 1–7) commonly exhibits Dirac points, van Hove singularities, and flat bands. Superconducting critical temperatures () were also investigated, with LiSn exhibiting a of 9.3 K at 0 GPa. We further analyzed the relationship between superconductivity and EPC strength, phonon softening, linewidth, and density of states. Deep learning molecular dynamics simulations of LiSn, , and demonstrate that when the Li concentration exceeds twice that of Sn, Li atoms undergo melting while Sn remains solid, indicating the emergence of a superionic state. Among them, exhibits the highest melting temperature of 1013 K. These findings enrich the crystal structure of Li-Sn compounds and provide a theoretical basis for constructing superconductors with a Kagome lattice and superionic states.