A novel analysis method for electrochemical impedance spectra using deep learning

Abstract

Recent advancements in electrochemical impedance spectroscopy has enabled to acquire large amounts of spectral data within shorter timeframes. However, the analysis methods designed to handle the increasing number of spectra cannot keep pace with instrumental developments. To address this challenge, a new method is proposed in this study, which applies deep learning techniques to analyze a set of electrochemical impedance spectra obtained upon changing potential. Specifically, the study utilizes an auto-encoder, a type of deep neural network, to process the impedance data. The auto-encoder effectively encodes the impedance as two-dimensional latent vectors, the key factor that represents the electrochemical processes that change in response to the electrode potential. The latent vectors contain information related to charge transfer, mass transfer, and electric double-layer charging, and produce quality-enhanced impedance spectra with improved quality when they are decoded. The results and conclusions of this study provide a proof of concept for the automated analysis of electrochemical impedance spectra using machine learning.