Abstract:Discovering functional crystalline materials through computational methods remains a formidable challenge in materials science. Here, we introduce VQCrystal, an innovative deep learning framework that leverages discrete latent representations to overcome key limitations in current approaches to crystal generation and inverse design. VQCrystal employs a hierarchical VQ-VAE architecture to encode global and atom-level crystal features, coupled with a machine learning-based inter-atomic potential(IAP) model and a genetic algorithm to realize property-targeted inverse design. Benchmark evaluations on diverse datasets demonstrate VQCrystal's advanced capabilities in representation learning and novel crystal discovery. Notably, VQCrystal achieves state-of-the-art performance with 91.93\% force validity and a Fréchet Distance of 0.152 on MP-20, indicating both strong validity and high diversity in the sampling process. To demonstrate real-world applicability, we apply VQCrystal for both 3D and 2D material design. For 3D materials, the density-functional theory validation confirmed that 63.04\% of bandgaps and 99\% of formation energies of the 56 filtered materials matched the target range. Moreover, 437 generated materials were validated as existing entries in the full database outside the training set. For the discovery of 2D materials, 73.91\% of 23 filtered structures exhibited high stability with formation energies below -1 eV/atom. Our results highlight VQCrystal's potential to accelerate the discovery of novel materials with tailored properties.

Unveiling the physical mechanisms driving delafossite crystal (ABX2) formation through interpretable machine learning

Formation Energy Prediction of Material Crystal Structures using Deep Learning

Towards machine learning for microscopic mechanisms: a formula search for crystal structure stability based on atomic properties

Crystal Structure Assignment for Unknown Compounds from X-ray Diffraction Patterns with Deep Learning.

CrystalX: Ultra-Precision Crystal Structure Resolution and Error Correction Using Deep Learning

Exploring the Crystal Structure and Electronic Properties of γ-Al 2 O 3 : Machine Learning Drives Future Material Innovations

Crystal Structure Assignment for Unknown Compounds from X-ray Diffraction Patterns with Deep Learning

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

VQCrystal: Leveraging Vector Quantization for Discovery of Stable Crystal Structures

AlphaCrystal-II: Distance matrix based crystal structure prediction using deep learning

Accelerating materials discovery using integrated deep machine learning approaches

Data-driven prediction of complex crystal structures of dense lithium

A Machine-Learning-Assisted Crystalline Structure Prediction Framework To Accelerate Materials Discovery

Noncovalent Interactions and Crystal Structure Prediction of Energetic Materials

Crystal Structure Prediction of Binary Alloys via Deep Potential

Machine learning approaches for feature engineering of the crystal structure: Application to the prediction of the formation energy of cubic compounds

A Deep-Learning Potential for Crystalline and Amorphous Li-Si Alloys

Predicting emergence of crystals from amorphous matter with deep learning

Machine learning aided atomic structure identification of interfacial ionic hydrates from AFM images

Competing nucleation pathways in nanocrystal formation

Topological representations of crystalline compounds for the machine-learning prediction of materials properties