Surrogate Models: Materials & chemistry https://surrogatemodels.com/ News, papers, datasets, and funding and deals in Physics AI for engineers: Materials & chemistry. en E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentialshttps://surrogatemodels.com/papers/nequiphttps://surrogatemodels.com/papers/nequipThu, 12 Nov 2026 09:00:00 GMTNequIP is a graph network that predicts the potential energy of a set of atoms and obtains forces as its gradient. Its internal features are scalars, vectors and higher-order tensors that rotate with the atoms (E(3)-equivariant convolutions), where earlier networks used scalar features only.Materials & chemistry Scaling deep learning for materials discoveryhttps://surrogatemodels.com/papers/gnomehttps://surrogatemodels.com/papers/gnomeThu, 12 Nov 2026 09:00:00 GMTGNoME trains graph networks to predict the energy of a crystal, uses them to filter candidate structures and compositions, checks the survivors with DFT, and feeds the DFT results back as training data over six rounds of active learning.Materials & chemistry A Universal Graph Deep Learning Interatomic Potential for the Periodic Tablehttps://surrogatemodels.com/papers/m3gnethttps://surrogatemodels.com/papers/m3gnetThu, 12 Nov 2026 09:00:00 GMTM3GNet is a graph network with three-body interactions that outputs energy, forces and stresses for a crystal. A single model is trained on the structural relaxations stored in the Materials Project and covers 89 elements.Materials & chemistry