New model speeds crystal structure prediction with fewer steps
uFlowCSP: Crystal Structure Prediction using Mean flow generative models
Artificial IntelligenceMachine Learning
Summary
Predicting crystal structures is important for discovering new materials but can be slow because traditional methods require many steps to find stable arrangements. The authors present uFlowCSP, a model that predicts crystal structures much faster with only a few steps, while keeping or improving accuracy. It uses a special transformer to understand chemistry and crystal symmetry, making it efficient even without extra information at prediction time. This can help researchers explore materials more quickly and efficiently.
What this means in practice
- •For materials engineers: Predict stable crystal structures faster to accelerate materials design with less computational demand.
- •For computational chemists: Generate multiple candidate crystal structures quickly for more efficient screening and analysis in chemical research.
Authors
Sourin Dey, Dipannoy Das Gupta, Lai Wei, Sadman Sadeed Omee, Jianjun Hu
Abstract
Crystal structure prediction (CSP) is fundamental to computational materials discovery. Generative models including CDVAE, DiffCSP, FlowMM, and CrystalFlow learn stable-crystal distributions directly, but diffusion and flow-matching inference requires tens to thousands of sequential network evaluations per candidate. We introduce uFlowCSP, a MeanFlow-based CSP model that learns the average, rather than instantaneous, probability-flow velocity. It generates a complete structure in one to five evaluations, delivering 5x-58x faster inference with equal or better performance. A chemistry- and symmetry-aware Transformer uses canonical atom ordering, global composition, and per-token chemistry embeddings. A coarse crystal-system token is used only during training; it provides additive gains, particularly improving space-group agreement despite being absent at inference, which remains formula-only. On MP-20 with 20 candidates per target, one step matches CrystalFlow (78.38% vs. 78.34%) with 100x fewer evaluations and about 10x lower wall-clock time. Five steps reach 83.64%, exceeding CrystalFlow (78.34% at 2,000 evaluations) and DiffCSP (77.93% at about 20,000), while using 20x fewer evaluations. uFlowCSP generates 10,000 structures in 0.39-1.31 minutes, versus 6.5 for CrystalFlow and 76.1 for DiffCSP. Under CSPBench's energy-ranked top-five structure-and-space-group criterion, five-step uFlowCSP reaches 72%/72%/65% structure, space-group, and consensus match rates. CrystalFlow reaches 78%/73%/68% at 100 steps but falls to 49%/32%/31% at five. Thus, uFlowCSP improves accuracy per network evaluation, not merely peak accuracy.