Prediction of Dynamic Ni Morphology Changes in Patterned Ni-YSZ Anode with Physics-Informed Neural Networks

Abstract Nickel (Ni) film on patterned Ni- yttria-stabilized zirconia (YSZ) anode shows dynamic spreading and splitting during solid oxide fuel cell (SOFC) operation, where wettability of Ni on YSZ is greatly enhanced (Z. Jiao, N. Shikazono, J. Power Sources 396 119–123, 2018). In the present study, a physics-informed neural network (PINN) constrained by Cahn-Hilliard equation of phase field model is proposed to estimate the unknown parameters for predicting dynamic Ni movements of the patterned Ni-YSZ anode. The unknown parameters such as interface thickness and mobility are inversely inferred by PINN using top-view images obtained from the operando experiments. Obtained excess surface diffusivity values were three to four orders of magnitude larger than the values reported for surface diffusion in the literature. It is therefore considered that Ni spreading and splitting of patterned anode cannot be simply explained by surface diffusion, and other mechanisms should be introduced.