Data-Driven Framework for Correlative SECCM–SEM Analysis During Electrodeposition

This work presents a high-throughput, multimodal approach to investigating electrochemical nucleation and growth (EN&G) processes. First, correlative microscopy is performed at identical locations on transmission electron grids, combining scanning electrochemical cell microscopy (SECCM) measurements with field emission scanning electron microscopy (FESEM) characterization. Second, an unsupervised learning approach is employed on the entire acquired electrochemical signal to limit human bias from the analysis and the selection of specific descriptors, with the aim of shedding light on unknown features. To begin, this method is validated using synthetic chronoamperograms to evaluate the effectiveness of various clustering methods in accurately grouping electrochemical curves based on a predefined number of active sites, regardless of added noise or random induction times that could be observed experimentally. A similar analysis is then carried out on the experimental data acquired at the same potential. The clustering performance is discussed using the corresponding FESEM images as proxy labels. Finally, an unsupervised analysis of a combined data set of depositions performed at three different potentials is conducted to investigate the interplay between potential and surface activity that gives rise to the diversity of electrodeposition behaviors observed at the microscale, demonstrating that overpotential alone does not uniquely define EN&G pathways.