Cgformer Transformerenhanced Crystal Graphnetwork With Global Attention For Material Propertyprediction 102380 December 3 2025 Kehao Tao Jiacong Li Wei He An Chen Yanqiang Han Feiming Huang Fuqiang Huang Jinjin Li by Kehao Tao & Jiacong Li & Wei He & An Chen & Yanqiang Han & Feiming Huang & Fuqiang Huang & Jinjin Li instant download after payment.

PROGRESS AND POTENTIAL Developing next-generation batteries for electric vehicles and grid storage requires new materials that are not only efficient but also safe. Traditional trial-and-error discovery is slow and expensive, especially for complex ‘‘high-entropy’’ materials, which are made by mixing multiple elements and offer vast potential for enhanced performance. The sheer number of possible elemental combinations makes it nearly impossible to test them all in the lab. To solve this, we created an AI model called CGformer. Unlike previous models that only look at immediate neighboring atoms, CGformer uses a ‘‘global attention’’ mechanism to understand how all atoms in a complex crystal interact over long distances. This provides a more complete picture, leading to much more accurate predictions of how well ions, like sodium, can move through the material—a key factor for battery performance. We used CGformer to rapidly screen nearly 150,000 potential high-entropy materials for sodium-ion batteries, a promising low-cost alternative to lithium-ion. Our AI pinpointed six top candidates, which we then successfully created and tested in the lab. These new materials showed significantly better performance than the original, undoped material, validating our AI-driven approach. The immediate impact is a faster, more intelligent way to design advanced materials. This AI framework is not limited to sodium-ion batteries; it can be adapted to discover other high-performance materials for a wide range of applications, from lithium-ion batteries to thermoelectrics. By accelerating the design cycle, CGformer paves the way for faster breakthroughs in energy storage and other technologies vital for a sustainable future.