Improving the spatial and spectral resolution of 2D X-ray near-edge absorption structure (XANES) has been a decade-long pursuit. This work develops a post-imaging processing method using a deep neural network to improve the signal-to-noise ratio in XANES images. The neural network adapts to new datasets by incorporating physical features and uses self-supervised learning for self-consistency. The model's robustness is demonstrated by determining the valence states of Ni and Co in LiNixMnyCo1-x-yO2 systems.

Zeyuan Li, Thomas Flynn, Tongchao Liu, Sizhan Liu, Wah-Keat Lee, Ming Tang, Mingyuan Ge