Multiscale studies of nanoconfined charging dynamics in supercapacitors bridged by machine learning

Hualin Zhan, Richard Sandberg, Zhiyuan Xiong, Qinghua Liang, Ke Xie, Lianhai Zu, Dan Li, Jefferson Zhe Liu

Published 2020-10-16Version 1

The energy-delivery performance of supercapacitors is fundamentally determined by the dynamics of ions confined in nanoporous materials, which has attracted intensive interest in nanoscopic research. Many nanoscopic understandings, including changes in in-pore ion concentration and mobility during dynamical processes, are continuously reported. However, quantitative scale-up of these nanoscopic understandings for evaluation of the macroscopic performance of supercapacitors is difficult due to the absence of links between these scales. Here we demonstrate that machine learning can be used to establish such links. Starting from nanoscale, we first reveal a diffusion-enhanced migration of ions in nanopores using primarily modified Poisson-Nernst-Planck model, unlike in bulk electrolyte where diffusion counteracts migration. Using machine learning, we discover a dynamically varying ionic resistance and its equation, resulting from the in-pore ion concentration change contributed by diffusion-enhanced migration. The obtained equation is used to construct a nano-circuitry model (NCM), which describes both the macroscopic performance of supercapacitors and nanometre-resolved ionic behaviour. We demonstrate that NCM can provide additional perspectives to understand cyclic voltammograms. A Faradaic-like current peak can show in non-Faradaic processes, and an asymmetric charging/discharging can occur without ion desolvation. These is because the dynamically varying resistance delivers ions effectively for storage. The demonstrated use of machine learning could extend to other ionic systems including batteries and desalination, paving the route towards rational design.