The application of physics-informed neural networks to hydrodynamic voltammetry

Abstract

Electrochemical problems are widely studied in flowing systems since the latter offer improved sensitivity notably for electro-analysis and the possibility of steady-state measurements for fundamental studies even with macro-electrodes. We report the exploratory use of Physics-Informed Neural Networks (PINNs) as potentially simpler, and easier way to implement alternatives to finite difference or finite element simulations to predict the effect of flow and electrode geometry on the currents observed in channel electrodes where the flow is constrained to a rectangular duct with the electrode embedded flush with the wall of the cell. Several problems are addressed including the evaluation of the transport limited current at a micro channel electrode, the transport of material between two adjacent electrodes in a channel flow and the response of an electrode where the electrode reaction follows a preceding chemical reaction. The approach is shown to give quantitative agreement in the limits for which existing solutions are known whilst offering predictions for the case of the previously unexplored CE reaction at a micro channel electrode.