Li Qi, Jiao Rui, Wu Liming, Zhu Tiannian, Huang Wenbing, Jin Shifeng, Liu Yang, Weng Hongming, Chen Xiaolong
The Beijing National Laboratory for Condensed Matter Physics, Institute of Physics, Chinese Academy of Sciences, Beijing, China.
School of Physical Sciences, University of Chinese Academy of Sciences, Beijing, China.
Nat Commun. 2025 Aug 11;16(1):7428. doi: 10.1038/s41467-025-62708-8.
Accurate crystal structure determination is critical across all scientific disciplines involving crystalline materials. However, solving and refining crystal structures from powder X-ray diffraction (PXRD) data is traditionally a labor-intensive process that demands substantial expertise. Here we introduce PXRDGen, an end-to-end neural network that determines crystal structures by learning joint structural distributions from experimentally stable crystals and their PXRD, producing atomically accurate structures refined through PXRD data. PXRDGen integrates a pretrained XRD encoder, a diffusion/flow-based structure generator, and a Rietveld refinement module, solving structures with unparalleled accuracy in seconds. Evaluation on MP-20 dataset reveals a record high matching rate of 82% (1-sample) and 96% (20-samples) for valid compounds, with Root Mean Square Error (RMSE) approaching the precision limits of Rietveld refinement. PXRDGen effectively tackles key challenges in PXRD, such as the resolution of overlapping peaks, localization of light atoms, and differentiation of neighboring elements.
在所有涉及晶体材料的科学学科中，准确确定晶体结构至关重要。然而，从粉末X射线衍射（PXRD）数据中解析和精修晶体结构传统上是一个劳动密集型过程，需要大量专业知识。在此，我们介绍PXRDGen，这是一种端到端神经网络，它通过从实验稳定晶体及其PXRD中学习联合结构分布来确定晶体结构，生成通过PXRD数据精修的原子级精确结构。PXRDGen集成了一个预训练的XRD编码器、一个基于扩散/流的结构生成器和一个Rietveld精修模块，能在数秒内以无与伦比的精度解析结构。对MP - 20数据集的评估显示，有效化合物的匹配率创纪录地高达82%（单样本）和96%（20样本），均方根误差（RMSE）接近Rietveld精修的精度极限。PXRDGen有效应对了PXRD中的关键挑战，如重叠峰的分辨、轻原子的定位以及相邻元素的区分。
