Machine learning guided discovery of superconducting calcium borocarbides
Chao Zhang, Hui Tang, Pan Chen, Hong Jiang, Huaijun Sun, Kai‐Ming Ho, Cai‐Zhuang Wang
Abstract
Pursuit of superconductivity in light-element systems at ambient pressure is of great experimental and theoretical interest. In this work, we combine a machine learning (ML) method with first-principles calculations to efficiently search for the energetically favorable ternary Ca-B-C compounds. Three layered borocarbides (stable $\mathrm{CaB}{\mathrm{C}}_{5}$ and metastable ${\mathrm{Ca}}_{2}\mathrm{B}{\mathrm{C}}_{11}$ and $\mathrm{Ca}{\mathrm{B}}_{3}{\mathrm{C}}_{3}$) are predicted to be phonon-mediated superconductors at ambient pressure. The stable $\mathrm{CaB}{\mathrm{C}}_{5}$ and the low-energy metastable ${\mathrm{Ca}}_{2}\mathrm{B}{\mathrm{C}}_{11}$ (with formation energy only 9.5 meV/atom above the convex hull) have a superconducting ${T}_{c}$ of 5.2 and 8.9 K, respectively. While the hexagonal $\mathrm{Ca}{\mathrm{B}}_{3}{\mathrm{C}}_{3}$ possesses a ${T}_{c}$ of 26.1 K, it is metastable with formation energy of 153 meV/atom above the convex hull. The ML-guided approach opens up a way for greatly accelerating the discovery of new high-${T}_{c}$ superconductors.