The high-temperature electrolysis of water using solid oxide electrolysis cells (SOEC) offers higher electrical efficiencies than other electrolysis processes, but is less technologically mature [1]. Three-dimensional modeling of cells enables understanding of the relationships between microstructure, operating parameters and performance, including potential fracture formation, e.g., of the electrolyte, which can lead to cell failure. As an alternative to using 3D reconstructions of physical cells, artificial computer-generated microstructures allow more straightforward adjustment of morphological parameters, potentially leading to even higher performance of modern SOECs. Therefore the work of Westhoff et al. [2] is extended for the creation of multilayered microstructures as found in SOECs. The generation consists of several steps: First, a dense packing of spheres is created. After constructing a neighborhood graph defining the contact points of the final particles, spherical harmonics are used to generate realistically shaped particles. For possible subsequent simulations of e.g. nickel coarsening [3], a voxelization is performed and a diffuse interface is set up via a multiphase-field method [4]. The analysis of various morphological measurements, such as volume fractions, tortuosities, or particle size distributions, allows comparison with 3D reconstructions of physical cells. A multi-objective Bayesian optimization yields input parameters for the generation workflow to match the selected measurements of the reconstructions. Such a workflow is applied to single layers, such as the anode, but is also extended to create artificially generated microstructures of the multiple layers of SOECs. The latter is crucial to model the complex behavior of such cells, i.e. cracking due to thermal and chemical shrinkage due to oxygen partial pressure. In addition, Bayesian optimization provides not only an optimal set of parameters, but also insight into the correlation between these parameters and morphological measurements. This allows the microstructure to be tailored without changing the complex manufacturing of SOECs.

References

[1] O. Schmidt, A. Gambhir, I. Staffell, A. Hawkes, J. Nelson and S. Few, 2017 Future cost and performance of water electrolysis: An expert elicitation study. International Journal of Hydrogen Energy, 42(52), 30470-30492.

[2] D. Westhoff, I. Manke and V. Schmidt, 2018 Generation of virtual lithium-ion battery electrode microstructures based on spatial stochastic modeling. Computational Materials Science, 151, 53-64.

[3] P.W. Hoffrogge, D. Schneider, F. Wankmüller, M. Meffert, D. Gerthsen, A. Weber, B. Nestler and M. Wieler, 2023 Performance estimation by multiphase-field simulations and transmission-line modeling of nickel coarsening in FIB-SEM reconstructed Ni-YSZ SOFC anodes I: Influence of wetting angle. Journal of Power Sources, 571, 233031.

[4] B. Nestler, H. Garcke and B. Stinner, 2005 Multicomponent alloy solidification: phase-field modeling and simulations. Physical Review E, Physical Review E.