Abstract:Recent advances in deep learning generative models (GMs) have created high capabilities in accessing and assessing complex high-dimensional data, allowing superior efficiency in navigating vast material configuration space in search of viable structures. Coupling such capabilities with physically significant data to construct trained models for materials discovery is crucial to moving this emerging field forward. Here, we present a universal GM for crystal structure prediction (CSP) via a conditional crystal diffusion variational autoencoder (Cond-CDVAE) approach, which is tailored to allow user-defined material and physical parameters such as composition and pressure. This model is trained on an expansive dataset containing over 670,000 local minimum structures, including a rich spectrum of high-pressure structures, along with ambient-pressure structures in Materials Project database. We demonstrate that the Cond-CDVAE model can generate physically plausible structures with high fidelity under diverse pressure conditions without necessitating local optimization, accurately predicting 59.3% of the 3,547 unseen ambient-pressure experimental structures within 800 structure samplings, with the accuracy rate climbing to 83.2% for structures comprising fewer than 20 atoms per unit cell. These results meet or exceed those achieved via conventional CSP methods based on global optimization. The present findings showcase substantial potential of GMs in the realm of CSP.

What problem does this paper attempt to address?

The paper aims to address the problem of Crystal Structure Prediction (CSP). Specifically, the authors propose a general generative model based on Conditional Crystal Diffusion Variational Autoencoder (Cond-CDVAE) to generate physically plausible crystal structures under user-defined chemical compositions and pressure conditions. The main issues this paper attempts to solve are as follows: 1. **Efficient generation of crystal structures**: Traditional CSP methods face high computational costs and low efficiency when dealing with complex systems. The Cond-CDVAE model proposed in this paper can efficiently generate high-quality crystal structures without performing Density Functional Theory (DFT) local optimization. 2. **Crystal structure prediction under high-pressure conditions**: In materials research, it is often necessary to determine the ground state structure of a specific chemical system under given external pressure, which is crucial for mapping phase diagrams and discovering high-pressure structures with special properties. However, existing generative models are mostly limited to structure generation under zero pressure conditions. By integrating data under high-pressure conditions, this paper enables the model to handle crystal structure prediction under different pressure conditions. 3. **Model generalization capability**: By training on a dataset containing a large number of local minimum structures (including high-pressure structures), the model gains stronger generalization capabilities, allowing it to generate diverse and physically plausible crystal structures. 4. **Experimental validation and performance evaluation**: The authors conducted extensive benchmarking of the Cond-CDVAE model, demonstrating its superior performance in generating high-fidelity crystal structures and significantly higher success rates in predicting experimental structures compared to traditional CSP methods. In summary, this paper is primarily dedicated to developing an efficient and reliable crystal structure prediction model to overcome the limitations of existing methods in complex systems and extend its application range under high-pressure conditions.

Deep learning generative model for crystal structure prediction

Crystal Structure Prediction Using Generative Adversarial Network with Data-Driven Latent Space Fusion Strategy

Organic Crystal Structure Prediction Via Coupled Generative Adversarial Networks and Graph Convolutional Networks

An effective method for generating crystal structures based on the variational autoencoder and the diffusion model

Enhancing crystal structure prediction by combining computational and experimental data via graph networks

Physics guided deep learning for generative design of crystal materials with symmetry constraints

Self-Supervised Generative Models for Crystal Structures

Con-CDVAE: A method for the conditional generation of crystal structures

Latent Conservative Objective Models for Data-Driven Crystal Structure Prediction

Crystal Structure Prediction by Joint Equivariant Diffusion

Data-driven discovery of 2D materials by deep generative models

End-to-End Crystal Structure Prediction from Powder X-Ray Diffraction

Predictive crystallography at scale: mapping, validating, and learning from 1,000 crystal energy landscapes

Generative Design of Crystal Structures by Point Cloud Representations and Diffusion Model

Data-driven discovery of novel 2D materials by deep generative models

CrystalFlow: A Flow-Based Generative Model for Crystalline Materials

Crystal structure prediction by combining graph network and optimization algorithm

Concurrent learning scheme for crystal structure prediction

An Adaptive Genetic Algorithm for Crystal Structure Prediction.