Abstract:Crystalline materials are a fundamental component in next-generation technologies, yet modeling their distribution presents unique computational challenges. Of the plausible arrangements of atoms in a periodic lattice only a vanishingly small percentage are thermodynamically stable, which is a key indicator of the materials that can be experimentally realized. Two fundamental tasks in this area are to (a) predict the stable crystal structure of a known composition of elements and (b) propose novel compositions along with their stable structures. We present FlowMM, a pair of generative models that achieve state-of-the-art performance on both tasks while being more efficient and more flexible than competing methods. We generalize Riemannian Flow Matching to suit the symmetries inherent to crystals: translation, rotation, permutation, and periodic boundary conditions. Our framework enables the freedom to choose the flow base distributions, drastically simplifying the problem of learning crystal structures compared with diffusion models. In addition to standard benchmarks, we validate FlowMM's generated structures with quantum chemistry calculations, demonstrating that it is about 3x more efficient, in terms of integration steps, at finding stable materials compared to previous open methods.

What problem does this paper attempt to address?

The paper proposes a new method called FlowMM for generating crystalline materials. Crystalline materials play a key role in next-generation technologies, but predicting their stable structures and discovering new materials poses significant computational challenges. Among all possible atomic arrangements, only a few are thermodynamically stable, limiting the materials that can be achieved experimentally. FlowMM is a generative model that performs well in predicting stable crystal structures with known elemental compositions (Crystal Structure Prediction, CSP) and proposing new materials with their stable structures (De Novo Generation, DNG). It is more efficient and flexible compared to existing methods. FlowMM utilizes Riemannian Flow Matching and generalizes for the inherent symmetries of crystals, such as translation, rotation, permutation, and periodic boundary conditions. The challenges mentioned in the paper include: the atomic types are discrete, while the lattice parameters and atomic positions are continuous; dealing with interdependent structures among different variables; and evaluating the stability of crystals. FlowMM simplifies these problems by selecting a base distribution suitable for crystal geometry, enabling more efficient learning of crystal structures. Compared to diffusion models, FlowMM reduces the required steps to find stable materials by about 3 times. Furthermore, FlowMM verifies the stability of generated structures using quantum chemical calculations and compares them with existing open methods. The model also employs a binary representation to reduce the dimensionality of atomic types, thereby improving the accuracy of predicting the number of unique elements in a unit cell. In summary, FlowMM aims to accelerate the discovery of new materials by addressing computational challenges in the design space of crystalline materials.

FlowMM: Generating Materials with Riemannian Flow Matching

FlowLLM: Flow Matching for Material Generation with Large Language Models as Base Distributions

Flow Matching for Accelerated Simulation of Atomic Transport in Materials

Rigid Body Flows for Sampling Molecular Crystal Structures

Efficient Generation of Molecular Clusters with Dual-Scale Equivariant Flow Matching

Efficient Symmetry-Aware Materials Generation via Hierarchical Generative Flow Networks

Flow Matching on General Geometries

Mixed Continuous and Categorical Flow Matching for 3D De Novo Molecule Generation

MOFFlow: Flow Matching for Structure Prediction of Metal-Organic Frameworks

Hessian-Informed Flow Matching

Response Matching for generating materials and molecules

SE(3)-Stochastic Flow Matching for Protein Backbone Generation

Flow Matching for Optimal Reaction Coordinates of Biomolecular System

AFLOW: An automatic framework for high-throughput materials discovery

ET-Flow: Equivariant Flow-Matching for Molecular Conformer Generation

Machine learning, phase stability, and disorder with the Automatic Flow Framework for Materials Discovery

Equivariant flow matching

Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Design

Scalable Diffusion for Materials Generation

Equivariant Flow Matching with Hybrid Probability Transport for 3D Molecule Generation

Efficient 3D Molecular Generation with Flow Matching and Scale Optimal Transport