Enhancing Materials Discovery with Valence Constrained Design in Generative Modeling

2025-07-26

AI Summary

Existing diffusion models for inverse materials design neglect fundamental chemical constraints—such as oxidation-state balance—resulting in chemically invalid generated structures. To address this, we propose CrysVCD, the first framework to explicitly embed valence-balanced chemical rules into the diffusion generative process. It comprises two stages: (1) a Transformer-based language model that generates compositionally feasible, oxidation-state-balanced chemical formulas; and (2) a conditional diffusion model that produces crystal structures, jointly optimized for thermodynamic and phononic stability. The framework supports function-driven conditional generation and offers plug-and-play integration with external property predictors. After fine-tuning, generated structures achieve 85% thermodynamic stability (convex hull distance ≤ 50 meV/atom) and 68% phononic stability (no imaginary modes). CrysVCD successfully discovers novel functional material candidates, including high-thermal-conductivity semiconductors and high-κ dielectrics.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Web data generation and simulation
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Diffusion-based deep generative models have emerged as powerful tools for inverse materials design. Yet, many existing approaches overlook essential chemical constraints such as oxidation state balance, which can lead to chemically invalid structures. Here we introduce CrysVCD (Crystal generator with Valence-Constrained Design), a modular framework that integrates chemical rules directly into the generative process. CrysVCD first employs a transformer-based elemental language model to generate valence-balanced compositions, followed by a diffusion model to generate crystal structures. The valence constraint enables orders-of-magnitude more efficient chemical valence checking, compared to pure data-driven approaches with post-screening. When fine-tuned on stability metrics, CrysVCD achieves 85% thermodynamic stability and 68% phonon stability. Moreover, CrysVCD supports conditional generation of functional materials, enabling discovery of candidates such as high thermal conductivity semiconductors and high-$κ$ dielectric compounds. Designed as a general-purpose plugin, CrysVCD can be integrated into diverse generative pipeline to promote chemical validity, offering a reliable, scientifically grounded path for materials discovery.

Problem

• Ensures chemical validity in generative materials design
• Integrates valence constraints into crystal structure generation
• Improves efficiency and stability in materials discovery

Innovation

• Integrates valence constraints into generative modeling
• Uses transformer-based model for balanced compositions
• Combines diffusion model for crystal structure generation