Abstract:The ability to rapidly develop materials with desired properties has a transformative impact on a broad range of emerging technologies. In this work, we introduce a new framework based on the diffusion model, a recent generative machine learning method to predict 3D structures of disordered materials from a target property. For demonstration, we apply the model to identify the atomic structures of amorphous carbons ($a$-C) as a representative material system from the target X-ray absorption near edge structure (XANES) spectra--a common experimental technique to probe atomic structures of materials. We show that conditional generation guided by XANES spectra reproduces key features of the target structures. Furthermore, we show that our model can steer the generative process to tailor atomic arrangements for a specific XANES spectrum. Finally, our generative model exhibits a remarkable scale-agnostic property, thereby enabling generation of realistic, large-scale structures through learning from a small-scale dataset (i.e., with small unit cells). Our work represents a significant stride in bridging the gap between materials characterization and atomic structure determination; in addition, it can be leveraged for materials discovery in exploring various material properties as targeted.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is in the field of materials science, how to inversely predict the three - dimensional atomic structure of materials from spectroscopic data of materials. Specifically, the authors introduce a new framework based on the diffusion model, which is a recently developed generative machine - learning method for predicting the three - dimensional structure of disordered materials according to target properties. They take amorphous carbon (a - C) as an example and identify its atomic structure through X - ray absorption near - edge structure (XANES) spectroscopy. XANES is a commonly used experimental technique for probing the atomic structure of materials. In addition, the authors also show that their model can adjust the atomic arrangement according to specific XANES spectra through conditional generation guidance, thereby achieving customization of material properties. The key contributions of the paper are: 1. **Conditional Generation**: A conditional generation framework is proposed, which can generate the three - dimensional structure of materials based on given spectral data, which is helpful for materials design and discovery. 2. **Scale - Independence**: The model has the property of scale - independence and can learn from small - scale data sets and generate large - scale real structures, which is very important for applications that need to predict material properties in a large range. 3. **Inverse Design**: It provides a new method for the inverse design of materials, that is, starting from the desired material properties, deriving the material structures that satisfy these properties, which is an important breakthrough in traditional materials design methods. In conclusion, this paper aims to solve the long - term challenge of reverse - engineering atomic structures from complex spectroscopic data by combining advanced machine - learning techniques and spectroscopic data in materials science, providing new possibilities for the rapid development and customization of materials.

Spectroscopy-Guided Discovery of Three-Dimensional Structures of Disordered Materials with Diffusion Models

Integrating Machine Learning Potential and X-ray Absorption Spectroscopy for Predicting the Chemical Speciation of Disordered Carbon Nitrides

Towards End-to-End Structure Solutions from Information-Compromised Diffraction Data via Generative Deep Learning

Ab Initio Structure Solutions from Nanocrystalline Powder Diffraction Data

A probabilistic deep learning approach to automate the interpretation of multi-phase diffraction spectra

Generative Design of Crystal Structures by Point Cloud Representations and Diffusion Model

Experiment-driven atomistic materials modeling: A case study combining X-ray photoelectron spectroscopy and machine learning potentials to infer the structure of oxygen-rich amorphous carbon

Scalable Diffusion for Materials Generation

Structural Constraint Integration in Generative Model for Discovery of Quantum Material Candidates

Harnessing interpretable and unsupervised machine learning to address big data from modern X-ray diffraction

Atomic structure generation from reconstructing structural fingerprints

Machine learning-accelerated small-angle X-ray scattering analysis of disordered two- and three-phase materials

Auto-resolving atomic structure at van der Waal interfaces using a generative model

Multiscale structure-property discovery via active learning in scanning tunneling microscopy

3D Equivariant Diffusion for Target-Aware Molecule Generation and Affinity Prediction

Revealing the chemical bonding in adatoms arrays via machine learning of 3D scanning tunneling spectroscopy data

3-D Inorganic Crystal Structure Generation and Property Prediction via Representation Learning

Fast extraction of three-dimensional nanofiber orientation from WAXD patterns using machine learning

Uncovering Material Deformations via Machine Learning Combined with Four-Dimensional Scanning Transmission Electron Microscopy

Generative Inverse Design of Crystal Structures via Diffusion Models with Transformers

Decoding Structure-Spectrum Relationships with Physically Organized Latent Spaces