Unveiling Hidden Patterns in Electrode-Electrolyte Interphase Chemistry and Architecture Through Machine Learning-Assisted Mass Spectrometry Analysis.

Electrode-electrolyte interphases (EEIs) govern battery performance, yet the interplay between their chemical composition and spatial organization remains challenging to resolve. Time-of-flight secondary ion mass spectrometry (TOF-SIMS) provides rich 3D chemical depth profiles of these interfaces, yet conventional analyses simplify the complexity by collapsing the data into 1D depth profiles or 2D maps, obscuring critical spatial features. Here, we apply a machine-learning-based approach directly to 3D TOF-SIMS depth profiles, leveraging their full spatial and chemical richness to investigate the interphase structure and chemistry more comprehensively than conventional analyses. Principal component analysis (PCA) identifies fragments that dominate chemical heterogeneity across the EEI, reveals correlated and anticorrelated spatial distributions, and enables direct comparison of average interphase chemistry and chemical variability. Building on this, t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) resolve stratified structures and fine-scale spatial interphase variations. Finally, applying PCA to fragment-specific thickness maps enables comparison of interphase thickness heterogeneity and identifies the chemical species responsible for local thickness variations. Together, the approach creates a generalized framework for extracting chemically and spatially meaningful information from complex TOF-SIMS datasets, linking interphase chemistry to electrochemical behavior and guiding the design of uniform interphases for advanced battery systems.