Application of Bayesian optimization to the synthesis process of BaFe2(As,P)2 polycrystalline bulk superconducting materials

Abstract

This study is the first application of Bayesian optimization to the synthesis process of superconducting materials. As a model case, the phase purity of BaFe2(As,P)2 polycrystalline bulks, which affects their superconducting properties, was improved by optimizing only the heat-treatment temperature using Bayesian optimization. We determined the optimal temperature among 800 candidates in 13 experiments, and a phase purity of 91.3 % was achieved. Moreover, the phosphorus doping level of the best sample approached the optimal doping level owing to a reduction in the impurity phase. Visualization of the Bayesian optimization process showed that a well-balanced global search and local optimization allowed us to obtain a rough correlation between the superconducting properties and experimental conditions and finely optimal experimental conditions over a wide range. These results demonstrate that Bayesian optimization is promising for optimizing the synthesis process of superconducting materials.