In the simulations, the real GDL structure is mapped onto a 3D grid featuring an appropriate voxel size, which is chosen to adequately represent the structural details of the respective material. As an initial state at the time zero, either a certain amount of liquid water can be randomly distributed within the open pores, or the steady-state water distribution obtained from a previous simulation with the standard GCMC model in the same GDL structure can be used. In the KMC model, the dynamic of the system is characterized by subsequent transitions from one state to another. Escaping from a certain state can occur along a number of pathways. Each of these pathways has its own rate constant, which corresponds to the Metropolis rate. All the states which preceded the actual state of the system are not important. This is the defining property of Markov chains. After each movement, a time increment is computed, which is a statistical estimation of the time needed for this event to take place in reality. Having obtained the time step, the amount of water produced as the result of the current density can be calculated. The water is added from the CL side and is removed from the opposite side, when it reaches the boundary region of the simulation box.
The KMC simulation results show how different current densities and boundary conditions effect the amount of liquid water in the structure, and how water distributes and moves in the GDL depending on the structural and wetting properties. Figure 1 shows an example of the simulation results at four different times after starting from a random water distribution: t = 4.6×10-8 s, t = 0.29 s, t = 0.75 s and t = 1.26 s, respectively. The initial water content corresponds to 20% occupation of the available pore space. In this simulation the current density was set to a constant value of 0.9 A cm-2 and the voxel size had a value of 4.348 μm. The local temperature and relative humidity within the structure were obtained from a previous CFD study. The surface contact angle of the GDL to water is considered to have a value of 94°. The results show the change in size, amount and distribution of water agglomerations with time. As expected for a hydrophobic surface, the water tends to fill the bigger pores within the structure.
References
1. K. Seidenberger, F. Wilhelm, T. Schmitt, W. Lehnert, and J. Scholta, J. Power Sources, 196, 5317–5324 (2011).
2. K. Seidenberger et al., J. Power Sources, 239, 628–641 (2013).
3. A. Pournemat et al., in ECS Transactions,, vol. 75 (2016).
Figures
Figure 1. Water distribution in the PEMFC GDL as the result of a Kinetic Monte Carlo simulation with a voxel size of 4.348 µm and current density of 0.9 A cm-2, water enters the GDL from z = 0.