Due to the intermittency of renewable energy sources, alkaline water electrolyzers are typically operated at partial load compared to the nominal design value. It is well-known that gas crossover is dominant at low current densities leading to higher anodic hydrogen content and higher cathodic oxygen content in the separator tanks. High anodic hydrogen content is tantamount to loss of product hydrogen which results in an explosive atmosphere in the gas phase if the volumetric hydrogen content in oxygen exceeds 4%. We have developed a transient model of a multi-cell stack which can describe the operation of the electrolyzer with mixed electrolyte flows (anolyte and catholyte), separated flows, or a combination thereof (dynamic switching). This is a major extension of the steady-state model developed by Haug et al. (International Journal of Hydrogen Energy, 2017, 42, 15,689–15707). In sharp contrast to the steady-state model by Haug et al., the transient model can calculate the gas crossover as the operating conditions (e.g. electrolyte flow cycles) dynamically change in time. Depending on the size of the stack and the separator tanks, the model estimates different rates for impurities to build up. The transient model is validated using independent experimental results by Haug et al. and Brauns et al. (Electrochimica Acta, 2022, 404, 139,715) The results show that the dynamic model can follow experimental results for fluctuating current densities for a period of several days. We found that the dynamic response and transition time to steady state depend significantly on the geometrical volume of the gas separators with respect to the single-cell stack. For a multi-cell stack, we find that the impurities build-up faster when increasing the number of cells in the stack. This model serves as a tool for sizing and process management of the electrolyzer system and the separator tanks especially with respect to explosion safety.

One of the important parameters in water management of proton exchange membranes is the electro-osmotic drag (EOD) coefficient of water. The value of the EOD coefficient is difficult to justify, and available literature data on this for Nafion membranes show scattering from in experiments and simulations. Here, we use a classical all-atom model to compute the EOD coefficient and thermodynamic properties of water from molecular dynamics simulations for temperatures between 330 and 420 K, and for different water contents between λ = 5 and λ = 20. λ is the ratio between the moles of water molecules to the moles of sulfonic acid sites. This classical model does not capture the Grotthuss mechanism; however, it is shown that it can predict the EOD coefficient within the range of experimental values for λ = 5 where the vehicular mechanism dominates proton transfer. For λ > 5, the Grotthuss mechanism becomes dominant. To obtain the EOD coefficient, average velocities of water and ions are computed by imposing different electric fields to the system. Our results show that the velocities of water and hydronium scale linearly with the electric field, resulting in a constant ratio of ca. 0.4 within the error bars. We find that the EOD coefficient of water linearly increases from 2 at λ = 5 to 8 at λ = 20 and the results are not sensitive to temperature. The EOD coefficient at λ = 5 is within the range of experimental values, confirming that the model can capture the vehicular transport of protons well. At λ = 20, due to the absence of proton hopping in the model, the EOD coefficient is overestimated by a factor of 3 compared to experimental values. To analyze the interactions between water and Nafion, the partial molar enthalpies and partial molar volumes of water are computed from molecular dynamics simulations. At different water uptakes, multiple linear regression is used on raw simulation data within a narrow composition range of water inside the Nafion membrane. The partial molar volumes and partial molar excess enthalpies of water asymptotically approach the molar volumes and molar excess enthalpies of pure water for water uptakes above 5. This confirms the model can capture the bulklike behavior of water in the Nafion at high water uptakes.