Enhancing Superconductor Critical Temperature Prediction: A Novel Machine Learning Approach Integrating Dopant...

Area of Science:

• Materials Science
• Condensed Matter Physics
• Solid State Chemistry

Background:

• Doping significantly impacts superconductor critical temperature (Tc), but predicting these effects is challenging.
• Existing models struggle to accurately capture the complex relationship between dopants and Tc.

Purpose of the Study:

• To develop a novel doping descriptor for predicting Tc in superconductors.
• To create an accurate predictive model for Tc by integrating doping, elemental, and physical features.
• To identify new doped superconductor candidates with high Tc.

Main Methods:

• Introduced a novel doping descriptor to quantify dopant influence.
• Employed a Mixture of Experts (MoE) model integrating the descriptor with material features.
• Screened existing and hypothetical compounds using the model and a generative approach.

Main Results:

• Achieved a high prediction accuracy with R² = 0.962 for Tc, outperforming previous models.
• Successfully identified optimal doping levels in the Bi2-xPbxSr2Ca2-yCu yO8 system.
• Discovered 40 promising candidates for high-Tc superconductivity.

Conclusions:

• The developed model accurately predicts Tc by explicitly accounting for doping effects.
• This approach serves as a powerful tool for guiding the discovery of new superconductors.
• The findings accelerate research in high-temperature superconductivity and material design.