(372h) Data-Driven Modeling of Electrochemical Systems

A convergent stream of big data, modeling and analytics in electrochemical systems enabled by high fidelity and throughput experimentation and computation creates a unique opportunity for the discovery of novel chemical physics and engineering guidance. In this talk, I will present work on learning from microscopic images, diffraction and atomistic simulations. Examples include 1) quantifying previously inaccessible or inaccurate knowledge of the electrochemical reaction kinetics, free energy, surface kinetic map and chemo-mechanical coupling from high dimensional images and videos of reacting or equilibrated lithium iron phosphate (LFP) particles (3D, x,y and time, or even 4D x,y,k_x, and k_y), 2) extracting autocatalytic reaction kinetics of lithium in nickel manganese cobalt oxide (NMC) out of in-situ X-ray diffraction and electroanalytical methods, and 3) coarsening of molecular simulations, all of which achieved quantitative matching between data and model. These applications show new ways in which experimental and simulation data can be extremely useful in revealing mechanistic insights about fundamental electrochemical processes (such as the suppression and induction of phase separation) and generating knowledge that can be translated into making new predictions through continuum modeling. I will introduce the modeling, inversion and statistical techniques that lie at the core of this endeavor.