Predicting Voltammetry Using Physics-Informed Neural Networks.

Area of Science:

• Electrochemistry
• Computational Science
• Machine Learning

Background:

• Cyclic voltammetry is a crucial electrochemical technique.
• Traditional simulation methods often require complex discretization.
• Developing efficient simulation tools is essential for electrochemical analysis.

Purpose of the Study:

• To present a novel discretization-free approach for simulating cyclic voltammetry.
• To demonstrate the application of Physics-Informed Neural Networks (PINNs) in electrochemical simulations.
• To evaluate the performance of PINNs against established methods.

Main Methods:

• Utilizing Physics-Informed Neural Networks (PINNs) by constraining a feed-forward neural network with the diffusion equation and electrochemically consistent boundary conditions.
• Predicting one-dimensional voltammetry at a disc electrode with semi-infinite or thin layer boundary conditions.
• Solving the two-dimensional diffusion equation for voltammetry at a microband electrode and near the edges of a square electrode.

Main Results:

• PINN predictions for one-dimensional voltammetry quantitatively agree with finite difference methods and analytical expressions.
• Simulations of two-dimensional diffusion at microband electrodes show close agreement with literature data.
• PINNs effectively quantify nonuniform current distribution at electrode edges, demonstrating capability in higher dimensional problems.

Conclusions:

• PINNs provide a discretization-free, potentially faster, and easier alternative for voltammetric simulations.
• The ease of developing PINNs is particularly noted for higher-dimensional electrochemical problems.
• This approach shows promise for advancing electrochemical simulation techniques.