Space Group Informed Transformer for Crystalline Materials Generation
arXiv:2403.15734 · doi:10.1016/j.scib.2025.09.035
Abstract
We introduce CrystalFormer, a transformer-based autoregressive model specifically designed for space group-controlled generation of crystalline materials. By explicitly incorporating space group symmetry, CrystalFormer greatly reduces the effective complexity of crystal space, which is essential for data-and compute-efficient generative modeling of crystalline materials. Leveraging the prominent discrete and sequential nature of the Wyckoff positions, CrystalFormer learns to generate crystals by directly predicting the species and coordinates of symmetry-inequivalent atoms in the unit cell. We demonstrate the advantages of CrystalFormer in standard tasks such as symmetric structure initialization and element substitution over widely used conventional approaches. Furthermore, we showcase its plug-and-play application to property-guided materials design, highlighting its flexibility. Our analysis reveals that CrystalFormer ingests sensible solid-state chemistry knowledge and heuristics by compressing the material dataset, thus enabling systematic exploration of crystalline materials space. The simplicity, generality, and adaptability of CrystalFormer position it as a promising architecture to be the foundational model of the entire crystalline materials space, heralding a new era in materials discovery and design.
29 pages, 12 figures
References in corpus (17)
- Automatic chemical design using a data-driven continuous representation of molecules
- Crystal Structure Prediction via Particle Swarm Optimization
- Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals
- Design of Lead-Free Inorganic Halide Perovskites for Solar Cells via Cation-Transmutation
- A Universal Graph Deep Learning Interatomic Potential for the Periodic Table
- Learning Properties of Ordered and Disordered Materials from Multi-fidelity Data
- Atom2Vec: learning atoms for materials discovery
- Inverse design of 3d molecular structures with conditional generative neural networks
- An invertible crystallographic representation for general inverse design of inorganic crystals with targeted properties
- PyXtal: a Python Library for Crystal Structure Generation and Symmetry Analysis
- Targeted free energy estimation via learned mappings
- Atomic positions independent descriptor for machine learning of material properties
- CRYSPNet: Crystal Structure Predictions via Neural Network
- Coevolutionary search for optimal materials in the space of all possible compounds
- Free energy calculation of crystalline solids using normalizing flow
- of two-dimensional electron gas: a neural canonical transformation study
- Predicting the Volumes of Crystals