Materials informatics based on evolutionary algorithms: Application to search for superconducting hydrogen compounds

TL;DR

This paper introduces a materials informatics approach using evolutionary algorithms to efficiently discover superconducting hydrogen compounds, combining genetic algorithms, programming, and first-principles calculations.

Contribution

The study develops a novel iterative method integrating genetic algorithms and programming for predicting superconducting materials, demonstrated on hydrogen compounds.

Findings

Predicted superconducting critical temperature of 122 K at 300 GPa for KScH12.

Predicted 98 K at 180 GPa for GaAsH6.

Validated the approach with first-principles calculations.

Abstract

We present materials informatics approach to search for superconducting hydrogen compounds, which is based on a genetic algorithm and a genetic programming. This method consists of four stages: (i) search for stable crystal structures of materials by a genetic algorithm, (ii) collection of physical and chemical property data by first-principles calculations, (iii) development of superconductivity predictor based on the database by a genetic programming, and (iv) discovery of potential candidates by regression analysis. By repeatedly performing the process as (i) $\rightarrow$ (ii) $\rightarrow$ (iii) $\rightarrow$ (iv) $\rightarrow$ (i) $\rightarrow$ $\dots$, the superconductivity of the discovered candidates is validated by first-principles calculations, and the database and predictor are further improved, which leads to an efficient search for superconducting materials. We applied this method to hypothetical ternary hydrogen compounds and predicted KScH$_{12}$ with a modulated hydrogen cage showing the superconducting critical temperature of 122 K at 300 GPa and GaAsH$_{6}$ showing 98 K at 180 GPa.