Establishing baselines for generative discovery of inorganic crystals

2025-01-04

AI Summary

Assessing the practical utility of generative AI for inorganic crystal discovery remains challenging due to inconsistent evaluation protocols and lack of standardized baselines. Method: We conduct a systematic benchmark comparing traditional approaches—including random enumeration and ion substitution—against three generative paradigms: variational autoencoders (VAEs), large language models (LLMs), and diffusion models. We establish the first reproducible, standardized evaluation framework and propose a universal post-generation screening pipeline integrating a pretrained universal interatomic potential (UIP) with a multitask property predictor for joint stability and target-property (e.g., bandgap, bulk modulus) assessment. Contribution/Results: While traditional methods achieve comparable stable-structure generation rates to generative models, they lag significantly in structural novelty and targeted property optimization. Post-generation screening universally boosts success rates by 2–3× across all methods, demonstrating the paradigm’s broad applicability and computational efficiency.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Generative artificial intelligence offers a promising avenue for materials discovery, yet its advantages over traditional methods remain unclear. In this work, we introduce and benchmark two baseline approaches - random enumeration of charge-balanced prototypes and data-driven ion exchange of known compounds - against three generative models: a variational autoencoder, a large language model, and a diffusion model. Our results show that established methods such as ion exchange perform comparably well in generating stable materials, although many of these materials tend to closely resemble known compounds. In contrast, generative models excel at proposing novel structural frameworks and, when sufficient training data is available, can more effectively target properties such as electronic band gap and bulk modulus while maintaining a high stability rate. To enhance the performance of both the baseline and generative approaches, we implement a post-generation screening step in which all proposed structures are passed through stability and property filters from pre-trained machine learning models including universal interatomic potentials. This low-cost filtering step leads to substantial improvement in the success rates of all methods, remains computationally efficient, and ultimately provides a practical pathway toward more effective generative strategies for materials discovery.

Problem

• Generative Artificial Intelligence
• New Inorganic Crystals
• Advantage Over Traditional Methods

Innovation

• Generative AI
• Material Discovery
• Machine Learning Screening