MACHINE LEARNING INTERATOMIC POTENTIALSMACE
FORCE FIELDS
Many-body Atomic Cluster Expansion — equivariant message-passing neural networks for fast, accurate interatomic potentials across the periodic table.
Foundation Models
Run MACE-MP-0, MACE-OFF, or upload your own fine-tuned .model files for custom potentials.
Scientific Visualization
Parity plots, error histograms, and energy convergence charts with publication-quality exports.
Model Benchmarking
Compare your fine-tuned model against MACE foundation models on standard benchmarks.
3D Structure Viewer
Interactive molecular visualization with force vectors, trajectory animation, and dual rendering engines.
Powered by
The MACE framework — created by Ilyes Batatia, David P. Kovacs, Gregor N. C. Simm, and the group of Gabor Csanyi at the University of Cambridge.
Batatia et al., "MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields," NeurIPS 2022.
Built by Zicheng Zhao · Northeastern University