📝 SummarXiv

Abstract

Disordered (amorphous) materials, such as glasses, are emerging as promising candidates for applications within energy storage, nonlinear optics, and catalysis. Their lack of long-range order and complex short- and medium-range orderings, which depend on composition as well as thermal and pressure history, offer a vast materials design space. To this end, relying on machine learning methods instead of trial and error is promising, and among these, inverse design has emerged as a tool for discovering novel materials with desired properties. Although inverse design methods based on diffusion models have shown success for crystalline materials and molecules, similar methods targeting amorphous materials remain less developed, mainly because of the limited availability of large-scale datasets and the requirement for larger simulation cells. In this work, we propose and validate an inverse design method for amorphous materials, introducing AMDEN (Amorphous Material DEnoising Network), a diffusion model-based framework that generates structures of amorphous materials. These low-energy configurations are typically obtained through a thermal motion-driven random search-like process that cannot be replicated by standard denoising procedures. We therefore introduce an energy-based AMDEN variant that implements Hamiltonian Monte Carlo refinement for generating these relaxed structures. We further introduce several amorphous material datasets with diverse properties and compositions to evaluate our framework and support future development.