Design Topological Materials by Reinforcement Fine-Tuned Generative Model
Researchers used Reinforcement Fine-Tuning (ReFT) on a pre-trained DiffCSP++ generative model to design new topological insulators (TIs) and topological crystalline insulators (TCIs). The fine-tuned model successfully identified 15 novel materials, including Ge2Bi2O6, a TI with a full band gap of 0.26 eV, demonstrating the model's ability to generate high-quality materials with targeted electronic properties.