Zn K-edge X-ray Absorption Spectroscopy Dataset and Graph Neural Network Models for Aqueous ZnCl2 Solutions

This dataset supports the machine learning prediction of Zn K-edge X-ray absorption spectra (XAS) from atomic structures of aqueous ZnCl₂ solutions. Atomic structures were sampled from molecular dynamics (MD) simulations using a machine learning interatomic potential (MLIP), and target XAS spectra were computed with VASP using the core-hole approach, spanning a range of ZnCl2 concentrations from 0.1 m to 30 m and solvation environments. The repository includes:- Atomic structures (PyMatGen format) and target XAS spectra - Pre-trained graph neural network (GNN) weights using an M3GNet backbone - Python source code and Jupyter notebooks for model training, inference, and interpretability analysis (Integrated Gradients, UMAP clustering) Associated publication: Chuntian Cao et al., Deciphering the Solvation Structure of Aqueous ZnCl₂ Solutions from X-ray Absorption Spectra using Interpretable Graph Neural Network, The Journal of Physical Chemistry B, 2026 (in press).