TcESTIME: predicting high-temperature hydrogen-based superconductors

Abstract

Superconductivity can be considered among the most exciting discoveries in material science of the 20th century. However, the hard conditions for the synthesis and the difficult characterization, make the statement of new high critical temperature (T_c) complex from the experimental viewpoint and have recently led to several hot controversies in the literature. In this panorama, theory has become a trustworthy diagnosis. Nevertheless, this comes at an extremely high computational cost. A faster alternative would be to find cheap footprints of superconductivity from the electronic structure. Some of the authors have recently shown that a correlation exists between T_c, the networking value [Nature Communications, 12, 5381 (2021)], and the molecularity index [arXiv:2403.07584v1 (2024)]. The networking value reflects the metallicity of the parent compound as a measure of its electron delocalization channels, by means of the Electron Localization Function topology (its bifurcation trees). Instead, the molecularity index quantifies the presence of H₂ molecules within the system. All in all, these two quantities characterize bonding features that are related to high T_c: high metallicity and low molecularity boost high T_c states. However, the quantification or these bonding characteristics was initially made by a visual approach, which is not scalable for high throughput screening. We have developed a new code, TcESTIME, which allows to determine the networking value for a given hydrogen-based compound. In this contribution, we present such code and the underlying periodic algorithms we have developed. As a reference, the estimation of T_c for LaH₁₀ thanks to this new code amounts to 10 CPU minutes in a computer cluster equipped with Intel Xeon 2.4 GHz processor. Given the new potential for screening, we have applied it to a larger set including ternary hydrogen based superconductors, and have proposed new fits to estimate T_c, leading to errors of ca. 33 K. We believe that this contribution settles the bases for an automatic high-throughput screening of hydrogen-based superconductors.