Toolkit for spectral and multivariate statistical analysis of XPEEM images for studying composite energy materials

Description: this dataset was described and used in a case study on the analysis method of XPEEM datasets [1]. The Dataset A (used to illustrate the chemical analysis) is accessible here. The data organisation is as follows: Masks.7z : a folder containing the masks used for the fine segmentation of the LPSC, NCM and carbon materials. XAS_XX.7z : a folder containing the raw, aligned (I_xyE) and processed E-stack data (D_xyE) at the 'XX' absorption edge. The folder contains... dsA_XX_E_I0.csv : the energy and I0 value corresponding to each image in the I_xyE E-stack. dsA_XX_Estack_raw : the raw E-stack data. dsA_XX_Estack_aligned : the aligned (I_xyE) E-stack data. dsA_XX_Estack_D_xyE : the processed (D_xyE) E-stack data. For the Ni L-edge: dsA_Ni_2E_images contains the 2E images at the pre-edge/Ni2+ and pre-edge/Ni3+ peaks, with energies of 849/851.3 and 849/853.1 eV, respectively. In the manuscript, these images are used to build Figure 1.d,e. The Dataset B (used to illustrate the non-linear image alignement) is accessible as a test set in https://github.com/blelotte/XPEEM_toolkit. Abstract: X-ray photoemission electron microscopy (XPEEM) is an advanced technique particularly well suited for elucidating the complex and buried interfacial electrochemical reactions in energy storage or conversion electrodes. Synchrotron-based (post mortem or operando) XPEEM measurements provide very high lateral spatial resolution (down to ~50 nm) to monitor the elemental and chemical surface change upon oxidation and reduction processes. A typical XPEEM acquisition involves recording X-ray absorption images across a range of incident photon energies, generating thousands of spectra. To take full advantage of the measured datasets, an automated process for data correction and analysis for efficient interpretation is highly desirable. In this work, we provide a comprehensive toolkit consisting of a series of processing steps aiming at facilitating the analysis of complex XPEEM datasets, including image drift correction and removal of non-linear distortions to align data acquired on the same region of interest across different absorption edges; local energy scale correction for energy dispersion and background subtraction for accurate spectral analysis; streamlining the analysis using threshold-based segmentation; construction of chemical maps using an unmixing algorithm, where we find that peak-ratios and peak-ratio initialised non-negative matrix factorisation (NNMF) methods achieve optimal results; building a projection of the elemental and chemical maps to discriminate the various chemistries using Gaussian kernel probability density estimation and k-means clustering. We demonstrate the utility of those procedures through several examples obtained on the chemical deconvolution of the LiNi0.6Co0.2Mn0.2O2/Li6PS5Cl interphase layers in all-solid-state batteries [2], highlighting their capability for better data visualisation, to accelerate the analysis and to resolve complex interfacial phenomena. References: [1] Lelotte, B.; Siller, V.; Pelé, V.; Jordy, C.; Gubler, L.; El Kazzi, M.; Vaz, C. A. F.; Toolkit for Spectral and Multivariate Statistical Analysis of XPEEM Images for Studying Composite Energy Materials. Surfaces and Interfaces 2025. DOI: https://doi.org/10.1016/j.surfin.2025.107490. [2] Lelotte, B.; Vaz, C. A. F.; Xu, L.; Borca, C. N.; Huthwelker, T.; Pelé, V.; Jordy, C.; Gubler, L.; El Kazzi, M.; Spatio-Chemical Deconvolution of the LiNi0.6Co0.2Mn0.2O2/Li6PS5Cl Interphase Layer in All-Solid-State Batteries Using Combined X-ray Spectroscopic Methods. ACS Appl. Mater. Interfaces 2025, 17, 9, 14645–14659. DOI: https://doi.org/10.1021/acsami.4c19857.