AI framework finds superhard carbon, exotic allotropes
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- Zhibin Gao and colleagues at Xi'an Jiaotong University created a closed‑loop AI framework that couples CrystaLLM‑generated carbon structures with rapid physics‑based stability testing.
- CrystaLLM generated thousands of candidate carbon allotropes, which were screened using a hybridization Shannon entropy descriptor to prioritize mixed sp–sp2–sp3 configurations.
- Superhard carbon phase was identified with calculated hardness exceeding that of diamond, based on a dense sp3‑dominant network.
- C12 phase exhibits metallic conductivity and a negative Poisson’s ratio, combining sp‑sp2‑sp3 hybridization in a 12‑atom unit cell.
- Thermal‑anisotropic material shows direction‑dependent thermal conductivity and ultra‑low shear stiffness, allowing lattice regions to reorient under shear.
- Stability analysis indicates the new allotropes are as stable as known carbon forms such as fullerenes, and could be synthesized via stepwise chemistry or high‑pressure compression.
- Applied Physics Letters published the study, demonstrating that generative AI embedded in a physics‑based loop can explore carbon’s topological landscape with far fewer computational resources than traditional methods.
Why it matters: The AI‑driven framework slashes computational cost, giving researchers a rapid, low‑expense route to novel carbon materials such as a superhard phase, metallic C12 with negative Poisson’s ratio, and shear‑soft, thermally‑anisotropic structures, while traditional exhaustive searches become less competitive for industrial applications.
