High-Tc superconductor candidates proposed by machine learning

We cast the relation between the chemical composition of a solid-state material and its superconducting critical temperature (T c ) as a statistical learning problem with reduced complexity. Training of query-aware similarity-based ridge regression models on experimental SuperCon data achieves average T c prediction errors of ±5 K for unseen out-of-sample materials. Two models were trained with one excluding high pressure data in training (“ambient” model) and a second also including high pressure data (“implicit” model). Subsequent utilization of the approach to scan ∼153k materials in the Materials Project enables the ranking of candidates by T c while accounting for thermodynamic stability and small band gap. The ambient model is used to predict stable top three high T c candidate materials that include those with large band gaps of LiCuF 4 (316 K), Ag 2 H 12 S(NO) 4 (316 K), and Na 2 H 6 PtO 6 (315 K). Filtering these candidates for those with small band gaps correspondingly yields LiCuF 4 (316 K), Cu 2 P 2 O 7 (311 K), and Cu 3 P 2 H 2 O 9 (307 K).