Publication date: 10th April 2024
Phase-field models of the Cahn-Hilliard type (the "Poisson-Cahn" framework) have proven useful for models of electrochemical solid-state interfaces due to their fidelity to microscopic experimental datasets and strong connections to first principles calculations. However, kinetic versions of the theory require the solution of a system of fourth-order initial-boundary value differential equations, which is cumbersome even for skilled computational researchers. This talk will present two main results applying Poisson-Cahn to experimental datasets. In one application, a Bayesian kinetic Poisson-Cahn analysis of Scanning Kelvin Probe microscopy on specially-constructed thin film ceria cells performing electrochemical reduction of carbon dioxide yielded estimates of relevant reaction equilibria, rate constants and gradient energy coefficients. The analysis featured an implementation of the Poisson-Cahn in Idaho National Laboratory's MOOSE framework, which implements solvers and provides a convenient user interface. Relevant and customizable codes, including codes implementing the samplers for Bayesian analysis, are available as free software. In another application, equilibrium Poisson-Cahn models applied to TEM-based Electron Energy Loss Spectroscopy measurements at grain boundaries in gadolinia-doped ceria yielded grain boundary conductivity estimates. In this study a "transformed target" statistical approach enabled easy Gaussian process regression to estimate inhomogeneous excess free energies (including gradient effects). The talk will demonstrate how a simple modification of the publicly available FoKL-GP routines enabled the computationally facile estimation of physical quantities directly from microscope data without the aid of large scale computing resources.
Alejandro Mejia-Giraldo, Will Bowman, Hasti Vahidi, Steven Nonnenmann, Jiaxin Zhu, Jiyang Wang, Angelo Cassiadoro