Batch Discovery of Complex Metal Superhydrides via an Effective Machine Learning Method Structured by Chemical...

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Chemistry

Background:

• Metal superhydrides are materials with high hydrogen content and complex hydrogen cage structures.
• They are extensively studied for their potential in high-temperature superconductivity.
• The chemical space of metal superhydrides, especially with noninteger hydrogen ratios, remains largely unexplored.

Purpose of the Study:

• To develop an efficient workflow for discovering new stable metal superhydrides.
• To explore the relationship between hydrogen clathrate structures and superconducting transition temperatures (Tc).
• To expand the search for novel superconducting materials.

Main Methods:

• Integration of the "chemical template effect" with machine learning algorithms.
• Development of a specialized structure discovery workflow.
• Analysis of structural properties and correlation with superconducting transition temperatures.

Main Results:

• Identification of 13 new structural prototypes and 31 stable metal superhydrides, a 23% increase in discoveries.
• A 65% increase in the prediction of high superconducting transition temperatures (Tc).
• Discovery of 19 new superhydrides with Tc > 100 K, many featuring large unit cells and 3D hydrogen clathrates.

Conclusions:

• The developed machine learning workflow significantly enhances the efficiency of superhydride discovery.
• 3D hydrogen clathrate structures are strongly correlated with high Tc, indicating potential for even higher temperature superconductors.
• The method shows promise for large-scale searches of both binary and ternary superhydrides.