Machine Learning Prediction of Superconducting Critical Temperature through the Structural Descriptor

Abstract

Superconductivity allows electric conductance with no energy losses when the ambient temperature drops below a critical value (T c ). Currently, the machine learning (ML)-based prediction of potential superconductors has been limited to chemical formulas without explicit treatment of material structures. Herein, we implement an efficient structural descriptor, the smooth overlap of atomic position (SOAP), into the ML models to predict the T c values with explicit atomic structural information. Using a data set containing 5713 compounds, our ML models with the SOAP descriptor achieved a 92.9% prediction accuracy of coefficient of determination (R 2 ) score via rigorous multialgorithm cross-verification procedures, exceeding the 86.3% accuracy record without atomic structure information. Several new high-temperature superconductors with T c values over 90 K were predicted using the SOAP-assisted ML model. This study provides insights into the structure−property relationship of high-temperature superconductors.

Inverse Design of Next-Generation Superconductors Using Data-Driven Deep Generative Models

“…After superconductivity was discovered in 1911 by Onnes, the efforts to identify novel superconducting materials with high transition temperatures ( T c ) has been an intense area of research in materials science and condensed matter physics. , There have been systematic computational efforts to identify Bardeen–Cooper–Schrieffer (BCS) conventional superconductors , with high T c prior to costly experimental investigation, − where density functional theory-perturbation theory (DFT-PT) calculations have been performed to obtain the electron–phonon coupling (EPC) parameters. In addition, various machine learning approaches have been utilized to accelerate the search for high- T c superconductors. ,− However, these typical funnel-like screening-based approaches are not sufficient for inverse materials design, where, instead of engineering from structure to property, the goal is to engineer from a target property to the crystal structure.…”

Linear Jacobi-Legendre expansion of the charge density for machine learning-accelerated electronic structure calculations

“…Machine learning (ML) has recently emerged as a surrogate for solving DFT KS equations and possibly replacing them [9][10][11]. For instance, trained ML models can be used as predictors for properties such as the energy gaps [12][13][14], superconducting critical temperatures [15][16][17][18][19], thermodynamic stability [20], topological invariant [21,22], just to name a few. These models learn a direct map between the structure/composition and the target property, thus avoiding one or many computationally expensive calculations.…”

Discovering Chemically Novel, High-Temperature Superconductors

“…There have been various applications in this work each with promising discoveries of new superconductive materials and properties with most of them predicting critical temperature. For example, superconducting phase diagrams were predicted using text mining [8], superconducting hydrogen compounds were found using a genetic algorithm and genetic programming [9] critical temperature and pressure were predicted for hydrides [10], critical temperatures of doped Fe-based superconductors were predicted based on structural and topological parameters [11], and critical temperature was predicted on a structure based model using a structural descriptor [12], and superconductor materials and properties have been automatically extracted from literature [13]. An ML-guided discovery will hopefully replace the "serendipitous discovery paradigm" that has existed in this last century of superconductor research [14].…”

“…In this work, we use the SuperCon data set for training, similar to what has been done in other implementations [12], [15]- [19]. Unique, reduced chemical formulae are curated [20] from the NOMAD data set [21] and used for testing.…”