Generative Inversion of Spectroscopic Data for Amorphous Structure Elucidation

2026-03-24

AI Summary

Traditional approaches to resolving the atomic structure of amorphous materials rely heavily on expert intuition or accurate potential energy surfaces, often struggling to simultaneously satisfy experimental consistency and structural realism. This work proposes GLASS, a novel framework that achieves, for the first time, an end-to-end inversion from multimodal spectroscopic data to atomic structures without requiring a potential energy surface. Leveraging a score-based generative model, GLASS learns structural priors from low-fidelity data and integrates multimodal experimental targets—including the pair distribution function (PDF), X-ray absorption spectroscopy (XAS), and diffraction—through differentiable simulations to directly generate experimentally consistent atomic configurations. The method successfully elucidates long-standing controversies, revealing the paracrystalline nature of amorphous silicon, the liquid–liquid phase transition in sulfur, and the microstructural origins of ball-milled amorphous ice, with generated structures closely reproducing experimental observations and offering new physical insights.

Technology Category

• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Multi-modal Vision

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Determining atomistic structures from characterization data is one of the most common yet intricate problems in materials science. Particularly in amorphous materials, proposing structures that balance realism and agreement with experiments requires expert guidance, good interatomic potentials, or both. Here, we introduce GLASS, a generative framework that inverts multi-modal spectroscopic measurements into realistic atomistic structures without knowledge of the potential energy surface. A score-based model learns a structural prior from low-fidelity data and samples out-of-distribution structures conditioned on differentiable spectral targets. Reconstructions using pair distribution functions (PDFs), X-ray absorption spectroscopy, and diffraction measurements quantify the complementarity between spectral modalities and demonstrate that PDFs is the most informative probe for our framework. We use GLASS to rationalize three contested experimental problems: paracrystallinity in amorphous silicon, a liquid-liquid phase transition in sulfur, and ball-milled amorphous ice. In each case, generated structures reproduce experimental measurements and reveal mechanisms inaccessible to diffraction analysis alone.

Problem

• amorphous structure
• spectroscopic inversion
• atomistic reconstruction
• multi-modal spectroscopy
• structure elucidation

Innovation

• generative inversion
• score-based model
• amorphous structure
• multi-modal spectroscopy
• atomistic reconstruction