Scale Model Experiment of Liquid Water Transport in Gas Diffusion Layer of PEFC and Optimization of the Structure

Abstract

To improve the performance of a polymer electrolyte fuel cell (PEFC), it is essential to control liquid water behavior in the cell. Gas diffusion layer (GDL) plays an important role for the water transport, and optimization of the structure is one of the major research topics. As the structure of GDL is quite complex and small with the scale order of 200 mm thickness, measurement of liquid water behavior is difficult, and only a few results are reported with X ray measurement. In this study a scale model experiment is proposed to observe the behavior in a model GDL made with a 3D printer in the scale of 300 times of the actual one. The parameters relating to the liquid water motion in a complex fiber structure are Capillary number, Weber number and Bond number. As the pore scale in the GDL is quite small and the speed of the water is very slow, it was assumed that capillary force is the dominant factor for the motion and other effects such as viscosity, inertial and gravity can be ignored. This gives that the behavior in the GDL is organized by Capillary number, Ca = uμ/σ, where, u is velocity, μ is viscosity and σ is interface tension. It represents the ratio of viscous to capillary forces. By using silicone oil as water and water as air, the scale model can be magnified to 300 times of actual size. Gravitational effect can be also removed, as the densities of the two liquids are similar. The scale model was made by 3D printer from the structural data obtained by X ray CT images of actual GDL.Figure 1 shows the GDL made by 3D printer. The length is 25mm ´ 25 mm ´ 45 mm. The bottom of the model faces to a flat plate with a hole assuming a crack of micro porous layer (MPL). Silicone oil representing water is dyed in red. Viscosity ratio of silicone oil to liquid water is 55, which is same as the ratio of air to liquid water. Inter face tension is σ = 3.1´10-2 N/m, equivalent order of water and air. The contact angle of the material and the silicone oil was about 130°, which is the same order with the actual GDL condition. The GDL structure is set in the box made of acrylic and the GDL is immersed in liquid water representing air. Figure 2 shows comparison of liquid water behavior in the scale model and they are compared to the results simulated by Lattice Boltzmann Method (LBM). Length of the simulation area is 80 μm ´ 80 μm ´ 160μm. The interfacial (surface) tension is σ = 7.29´10-2 N/m. Viscosity ratio of liquid phase to gas phase is 55 and the contact angle is set 130° in the simulation. In each case Capillary number, Ca, is adjusted to ~10-2 by inlet velocity. The developments of liquid water in the fibers are quite similar between the scale model and the simulation. Particularly the selection of pores in different scales in the water motion is very similar. Variety of observations in different Ca conditions showed good similarity between the model and the simulation. These results indicate the possibility of the scale modeling. Investigation was made for the water behavior for the wide range of different Capillary numbers. Figure 3 shows pictures with different capillary numbers. In Ca ~10-1 liquid phase immerses small pores in bottom region, whereas the small pores are not filled by the water in Ca less than 10-2. This is due to the fact that viscous and inertial forces become large relative to the capillary force when Capillary number is large, and the capillary force becomes dominant when Capillary number is small. In less than Ca ~10-2, the phenomena were almost same and independent to the Capillary numbers. In this study the characteristics of liquid water behavior was investigated intensively for wide range of Capillary numbers.Using the scale model experiment, optimization of the fiber structures was attempted by comparing the results with LBM simulation. Water motion in the vicinity of libs and channels facing to the GDL was also investigated. Some of the results will be presented to show clues for the optimization of the GDL structure. Figure 1