Design topological materials by reinforcement fine-tuned generative model.

Area of Science:

• Materials Science
• Condensed Matter Physics
• Artificial Intelligence

Background:

• Topological insulators and topological crystalline insulators possess unique electronic properties, such as robust surface states and insulating bulk.
• These materials are critical for next-generation technologies, including quantum computing and spintronics.
• Discovering new topological materials, especially those with wide band gaps, is hindered by traditional screening methods.

Purpose of the Study:

• To develop and apply a novel AI-driven approach for the targeted discovery of topological materials.
• To overcome the limitations of conventional material screening methods.
• To identify new candidate topological insulators and topological crystalline insulators with desirable properties.

Main Methods:

• Utilized reinforcement fine-tuning on a pre-trained generative model.
• Employed targeted material generation ensuring chemical validity and structural stability.
• Screened generated materials for topological properties and band gaps.

Main Results:

• The fine-tuned generative model significantly increased the probability of generating topological insulators and topological crystalline insulators.
• Identified numerous novel candidate topological materials.
• Discovered Ge2Bi2O6 as a strong topological insulator with a substantial band gap of 0.26 eV.

Conclusions:

• Reinforcement-based generative design is a powerful strategy for discovering materials with specific topological properties.
• The developed AI approach accelerates the identification of promising topological materials.
• Ge2Bi2O6 represents a significant advancement in the search for practical topological materials.