A Comparison of Felt-Type and Paper-Type Gas Diffusion Layers for Polymer Electrolyte Membrane Fuel Cell Applications Using X-Ray Techniques PUBLIC ACCESS

Gas diffusion layers (GDLs) in polymer electrolyte membrane (PEM) fuel cells are porous materials placed between the bipolar plate and the catalyst layer of a membrane electrode assembly (MEA). GDLs must facilitate electrical and thermal transport, reactant transport to the catalyst layer, and gaseous and liquid water transport to and from the catalyst layer. In addition to its role in facilitating these transport mechanisms, the GDL must also provide structural support for the catalyst coated membrane [1–4]. Paper-type GDLs and felt-type GDLs are the most commonly used GDLs in PEM fuel cells for automotive applications [5]. Additionally, state-of-the-art GDLs have a bilayered structure that is composed of a carbon fiber-based macroporous substrate and a microporous layer (MPL) [1,5–7]. Several detailed reviews of GDLs have been published in literature [1,2,6,8].

The primary difference between paper-type and felt-type GDLs is in the microstructure of the materials arising from their manufacturing techniques. In the paper-type GDL, an external binder is used to affix the straight carbon fibers to each other, leading to an apparently two-dimensional (planar) structure. In the felt-type GDL, however, the carbon fibers are bound together through a hydro-entangling process, which results in a three-dimensional (3D) structure with highly entangled carbon fibers [5,9]. This hydro-entangling process results in felt-type GDLs with higher permeabilities despite lower porosities [10]. However, it has been shown by several authors that the surface roughness (of the substrate) does not vary significantly between paper-type and felt-type GDLs [10,11].

A variety of ex situ and in situ techniques exist for the characterization of the GDLs for their transport properties. These include microstructure analysis through X-ray computed tomography (CT) [7,11–22], measurements of breakthrough pressure [23–26], diffusivity [27–29], permeability [2,30,31], and thermal conductivity [32–37]. Researchers have also investigated water accumulation behavior in GDLs using a variety of imaging techniques [38,39], such as neutron radiography [40–45], X-ray radiography [46–55], and magnetic resonance imaging [56,57]. Additionally, high-speed optical imaging has also been used to investigate the accumulation of water in the gas channels [58–62].

In comparing the structural differences between GDL materials, it is of paramount importance to examine the internal microstructure (using X-ray CT) as well as the liquid water transport behavior within the GDLs. Escribano et al. [63] conducted mechanical characterization on cloth-type, felt-type, and paper-type GDLs and concluded that felt-type GDLs showed the highest resilience returning to the same thickness under consecutive compression cycles. They also presented scanning electron microscopy images of the different types of GDLs to show how the felt-type GDL had a more three-dimensional structure compared to the paper-type GDL. These findings have been further supported by the literature [1,5,45].

Fishman et al. [12] compared felt-type GDLs with paper-type and cloth-type GDLs using X-ray CT. They reported that felt-type GDLs showed a more uniform porosity in the core region of the GDL and attributed this characteristic to the hydro-entanglement process used in the fabrication of these materials. Similar observations about uniform porosity in the core region of the GDL were also made by Banerjee et al. [7]. Rofaiel et al. [9] explored the polytetrafluoroethylene (PTFE) distribution in felt-type and paper-type GDLs and concluded that a similar bimodal distribution of PTFE was observed for both felt-type and paper-type GDLs. High PTFE concentrations were observed near the outer surfaces, but low concentrations of PTFE were observed in the GDL core. The aforementioned works [1,5,7,9,11–15] have exclusively investigated the microstructure differences between felt-type GDLs and paper-type GDLs. Direct visualizations of liquid water accumulation would elucidate the relationship between these microstructural differences and liquid water management in the PEM fuel cell.

X-ray radiography is a powerful technique for visualizing liquid water within an in operando fuel cell with high spatial and temporal resolutions [54,64], which allows for accurate quantification of liquid water accumulation within the GDL. The technique has been utilized by a variety of groups to isolate water in the cracks of the MPL [53,65–67], quantify liquid water content in the MPL [4,68–70], investigate water accumulation at the interfaces [4], identify water cluster distributions [71–74], and compare water accumulation under land and channel regions [46,75,76].

In this work, both felt-type and paper-type GDLs from the same manufacturer are compared to identify the differences between the GDL microstructure and corresponding influence on liquid water accumulation through the use of X-ray tomography and X-ray radiography, respectively. Porosity profiles of uncompressed and compressed GDLs are presented and discussed for the two types of GDLs. The liquid water accumulation in the GDLs was measured in operando using synchrotron X-ray radiography, and comparisons between these materials are provided in the context of their microstructural differences.

This section describes the samples investigated in this work and their corresponding properties. Two characterization techniques are employed to understand the microstructure and water accumulation separately.

Four GDL samples from Sigracet^® Fuel Cell Components (SGL Technologies, Meitingen, Germany) were used for the comparison study between felt-type and paper-type GDLs. Details of the GDLs are presented in Table 1, some of which were obtained from the manufacturer's specification sheet. The SGL 10 BA GDL has a felt-type carbon fiber substrate with a nominal thickness of 400 μm and does not include an MPL coating. This GDL is considered to be the base material for felt-type GDLs included in this study. The SGL 25 BA GDL has a paper-type carbon fiber substrate with a nominal thickness of 190 μm (also without an MPL coating) and is assumed to be the base material for the paper-type GDLs included in this study. Both GDLs exhibit a hydrophobic loading of 5 wt % PTFE, rendering them sufficiently hydrophobic for PEM fuel cell applications. The SGL 10 BC GDL is a bilayered GDL that is composed of SGL 10 BA coated with an MPL. Similarly, the SGL 25 BC GDL is a bilayered GDL composed of SGL 25 BA coated with an MPL. These four GDL samples were operated in the fuel cell independently so that their microstructures were compared directly to their influence on liquid water accumulation behavior.

The GDL samples investigated in this work were imaged using a Skyscan 1172 desktop X-ray CT device (Bruker Corp., Billerica, MA). All the scans were performed at a voltage of 36 kV and a current of 222 μA. The equipment and setup were also used in prior studies relating to the microstructural analysis of GDLs [7,12–15]. Figure 1 shows the schematic of the X-ray CT imaging setup, including the device used for sample mounting. Sequential projections were obtained at intervals of 0.3 deg rotations and reconstructed using NRECON^® software (Bruker Corp.).

The output of the reconstruction algorithm is a three-dimensional grayscale structure, which is voxelized with the same spatial resolution as the acquired projections. Two sets of scans were performed for the microstructural analysis: uncompressed and compressed, where the same compression (thickness) was applied during in operando radiography (described in Sec. 2.3). A resolution of 2.88 μm/voxel was achieved for uncompressed GDL tomography, and a resolution of 3.3 μm/voxel was achieved for compressed GDL tomography.

The three-dimensional grayscale representation of the GDL structure is composed of the carbon fiber substrate, MPL (if applicable), and void space. To distinguish the phases and their contributions to the overall porosity of the GDL, the image was processed using an algorithm developed in-house that accounted for the expected volumes of individual components, which was estimated based on the respective areal mass and densities of each phase. Morphological image processing tools were applied in sequence and informed by the areal volumes to uniquely identify the individual phases. The reader is directed to an earlier work where the details of this algorithm were presented [7].

The individual GDL microstructures were compared using the porosity profiles of each GDL. The local porosity was calculated at each through-plane location of the GDL, and the contribution of each phase (fiber and MPL) to the solid fraction was also determined. Equations (1)–(3) show the calculation of fiber solid fraction and MPL solid fraction Display Formula

The fiber solid fraction consists of the summation of voxels identified as carbon fiber materials. The MPL voxel solid fraction represents the summation of voxels identified as MPL materials. MPL voxels appear as solid in X-ray CT since the MPL nanostructure cannot be resolved with the scanning resolution used here. To account for the porous nature of the MPL, an MPL porosity of 50% ($εMPL$ = 0.5), as obtained from literature [7,78,79], was assumed. Figure 2 shows a representative porosity profile of an SGL 10 BC GDL with each of the solid fractions shown as a function of the through-plane position. The pore space (void) is represented by the white area in Fig. 2. Figure 3 shows representative cross sections in both the in-plane and through-plane directions for each of the four GDLs described in Table 1. Black represents the fibers in the substrate, orange represents the MPL phase, while white is indicative of the void space.

A fuel cell customized for synchrotron X-ray radiography was used in this study. The fuel cells were assembled with Nafion^® HP membrane (DuPont Company, Wilmington, DE) coated with 0.3 mg Pt/cm² of Pt/C catalyst, as obtained from Ion Power Inc. (New Castle, DE). The same catalyst loading and GDL were used on both the anode and the cathode sides. The four GDLs described in Table 1 were individually tested in separate fuel cells. Constant current density operation was applied to the four fuel cells, and the current density was increased by intervals of 0.5 A/cm² up to 3.0 A/cm². The MEA had an active area of 68 mm² (8.0 mm × 8.5 mm) and was placed in between two graphite flow field plates with parallel channels and lands (square cross sections) that were both 0.5 mm wide. The channels are the square flow fields through which the reactant gases were introduced. The lands are solid regions in the flow field, which provide structural support to the MEA as well as conductive pathways for the electrical current to flow.

For all the cases, the fuel cell was operated at 60 °C with a gas inlet relative humidity of 100% and a backpressure of 200 kPa (absolute), controlled by a Scribner 850 e fuel cell test stand (Scribner Associates Inc., Southern Pines, NC). These conditions were used to intentionally exacerbate the accumulation of liquid water in the GDL to observe the structural influence on liquid water transport in the absence of membrane dehydration effects [80]. The cell was supplied with a constant gas flow rate of 1.0 l/min, and a cathode stoichiometric ratio of 80 was reached at a current density of 1.0 A/cm². The high stoichiometric ratio was prescribed to purge liquid water from the channels at all current densities and ensure that our observations were not influenced by channel liquid water blockage. At each current density, the cell was held for 15 min to reach a steady-state liquid water behavior [54,81].

Water accumulation in the GDL was observed in operando using synchrotron X-ray radiography. The radiography experiments were performed at the Biomedical Imaging and Therapy–Bending Magnet (05B1-1) beamline at the Canadian Light Source Inc., (CLSi), in Saskatoon, SK, Canada [82]. The radiographic imaging was conducted at an energy level of 24 keV, and the images were captured using an AA40 scintillator (Hamamatsu Photonics, Japan) and a Hamamatsu ORCA Flash 4.0 coupled charge device camera (Hamamatsu Photonics, Shizuoka, Japan). A pixel resolution of 6.5 μm and a frame rate of 0.33 frames per second were achieved. Figure 4 shows a schematic of the imaging setup for the fuel cell, with an example radiograph presented as a reference.

Reference images were obtained when the cell was operated at a current density of 0 A/cm² (i.e., open circuit voltage) such that liquid water was not present in the system. Test images were obtained continuously during the cell operation, which were compared with the reference images according to the Beer–Lambert law (Eq. (4)) to quantify the liquid water content present in the system [83,84]. Display Formula

In Eq. (4), t_w is the thickness of water, μ_w is the linear attenuation coefficient of water, and I_dry and I_wet are the intensities of fuel cells operated under dry and wet conditions, respectively. The processing is completed using algorithms developed in-house [83,84].

The data presented in this work were obtained at steady-state. Liquid water content was measured at the end of the 15 min period of steady-state operation, and the data presented here are an average of 20 frames (1 min of operation). The values of liquid water thickness obtained from the radiographs were normalized by the thickness of the GDL traversed by the beam.

In this section, the results of both the microstructure analysis and the water accumulation studies are presented for the four GDLs. A comparison is made between the felt-type and paper-type GDLs included in this study, and the impact of their microstructural differences on the water accumulation in the GDL is explored.

Figure 5(a) shows the solid fraction profiles of uncompressed GDLs for each of the four GDLs investigated, properties of which have been presented in Table 1. For single-layered GDLs (SGL 25 BA and SGL 10 BA), there is only one profile which shows the overall porosity of the sample. This overall porosity is the result of the fiber solid fraction, which is the only component contributing to the solid fraction. However, for the bilayered GDLs (SGL 25 BC and SGL 10 BC), three profiles have been presented in the figures. The definition of these porosity profiles is given in Equations (1)–(3) as discussed in Sec. 2.2. The white region above the solid fraction profiles represents the void space in the porous medium.

For all of the GDLs, the porosity is observed to be high at the edges of the GDLs and decreases to reach low porosities in the core of the GDL. For SGL 10 BA (felt-type), the porosity profile shows high porosity at the edges and gradually decreases to a value of approximately 0.83 toward the core of the GDL. The core region (from 100 to 300 μm) of the GDL exhibited an average porosity of ε_core = 0.83 ± 0.003. Similarly, SGL 25 BA (paper-type GDL) also exhibited high porosities at the edges, and the porosity decreased toward the core of the GDL. The key difference between the felt-type and the paper-type GDL was observed in the cores of the GDLs, where the paper-type GDL (SGL 25 BA) exhibited a nonuniform porosity profile, with the minima located at 60 μm from one of the outer surfaces and had a value of ε_core = 0.80 ± 0.003 (core region was 50 μm thick).

SGL 10 BC and SGL 25 BC are composed of SGL 10 BA and SGL 25 BA, respectively, as the fibrous substrates with MPL coatings. The MPLs were coated on one side of the substrate (by the manufacturer) so that the MPL acts as the interface with the catalyst layers of the MEA. The bulk MPL is defined as the region of the MPL in which there are very few carbon fibers. This region is smaller in SGL 10 BC compared to SGL 25 BC. This comparative difference in the bulk MPL thickness is further supported by the nominal thickness measurement provided by the manufacturer (Table 1). The SGL 25 BC GDL is 45 μm thicker than SGL 25 BA, which can be attributed to the MPL thickness. The SGL 10 BC GDL is only 20 μm thicker than SGL 10 BA, which is also attributed to the MPL thickness in SGL 10 BC.

In the bilayered GDLs, the thickness of the MPL was categorized into two sections: one section resided outside the boundaries of the fibrous substrate and the other section of the MPL intruded the fibrous substrate (i.e., a mixed zone of MPL and carbon fiber substrate). The total MPL thickness was calculated from the porosity profiles and defined as the thickness of the GDL whereby void voxels within the GDL consisted of less than 50% of the total GDL voxels [75]. Table 2 shows the thickness of the MPL as calculated using this methodology for both the GDLs in uncompressed and compressed states. The extra overlap between the MPL and the substrate was calculated as the difference between the uncompressed and compressed thicknesses of the MPL (also shown in Table 2). For SGL 10 BC, this increased overlap (MPL intrusion into substrate) was calculated as 47 μm, while that for SGL 25 BC is 19 μm. This shows that the paper-type GDL is less prone to MPL intrusion into the fibrous substrate.

Figure 5(b) shows the porosity profile of the GDLs studied under compression. Each GDL was compressed by 20–25% of its uncompressed thickness, a practice that has been shown to provide the best contact between the layers without increasing mass transport resistances [85–87].

The porosity profile of SGL 10 BA shows a relatively uniform porosity throughout the thickness, because the high porosity regions near the edge were eliminated due to the application of compression. The edges still demonstrated higher porosities than the core regions. The porosity profile of SGL 25 BA exhibited similar changes as was observed in SGL 10 BA upon compression. The same low porosity region was observed with a decreased porosity of ε = 0.74, with the minimum located 30 μm from the GDL surface.

As expected after compression, the SGL 10 BC GDL demonstrated a significantly lower porosity than when uncompressed. It is interesting to note that the minimum combined porosity of SGL 10 BC decreased to ε = 0.5. Although the SGL 25 BC GDL also demonstrated a decrease in porosity with compression, its lowest porosity region has a combined porosity of ε = 0.6. The combined porosity of uncompressed SGL 10 BC was higher (ε = 0.74) than that of SGL 25 BC (ε = 0.62) as shown in Fig. 5. The fibrous substrate extended throughout the MPL of SGL 10 BC. For SGL 25 BC, the bulk MPL region extended to 30 μm, within which carbon fibers were entirely absent. From observation, the MPL coated on the felt-type GDL is susceptible to penetration by the felt-type carbon fiber region under compression. Carbon fibers embedded in the MPL may come into contact with the catalyst layer and puncture the membrane and lead to durability and performance losses.

Figure 6 shows the normalized liquid water thickness measured in SGL 10 BA and SGL 25 BA during operation at the current densities of 0.5, 1.0, and 1.5 A/cm². The through-plane water accumulation in both the anode and cathode GDLs are shown for an increasing range of current densities (0.5, 1.0, and 1.5 A/cm²) to identify the onset of liquid water accumulation. The water in the GDL regions under the channel and land regions were characterized separately, since significant differences in water accumulation between the two regions were observed, which is in agreement with the literature [75,76,88]. The shaded gray region indicates the position of the polymer electrolyte membrane, which is 26 μm thick. The majority of the conclusions will be drawn from the cathode GDL as the cathode typically exhibits water management challenges [89,90], which results in higher oxygen transport resistances [75,88,91] and larger performance losses compared to the anode.

Figures 6(a) and 6(b) show the normalized water thickness in the GDL region under the channel for the felt-type GDL (SGL 10 BA) and the paper-type GDL (SGL 25 BA), respectively. In Fig. 6(a), the higher water content in the felt-type GDL under the channel is observed, with significant water accumulation near the catalyst layer. This high water content is attributed to the high porosity in SGL 10 BA and large pores that are prone to flooding. This flooding was the primary mode of cell failure for the felt-type GDL. This flooding phenomenon was mitigated in the paper-type GDL as shown in Fig. 6(b), which shows an insignificant accumulation of water in the GDL region under the channel.

Figures 6(c) and 6(d) show the normalized water thickness in the GDL region under the land for the felt-type GDL (SGL 10 BA) and the paper-type GDL (SGL 25 BA), respectively. The average water quantity in the felt-type GDL under the channel is similar to that under the land region while the paper-type GDL shows a significant difference of water quantity between the channel and land regions. The relatively uniform distribution of water between the land and channel regions is attributed to the heterogeneous 3D structure of the felt-type GDL. The water accumulation in the land region of the paper-type GDL is significantly higher than that in the channel region of the GDL. This nonuniform liquid water distribution is attributed to a temperature gradient between the land and channel regions resulting in condensation in the cooler land region [75,76]. The lack of convective drag at the land-GDL interface also inhibits effective water removal. It is noteworthy that the overall water content in the felt-type GDL is still higher than that in the paper-type GDL under the land region. The low porosity region in SGL 25 BA (at 100 μm) as shown in Fig. 5(b) is not a preferential region for liquid water accumulation. The SGL 10 BA GDL does not exhibit a similar low porosity region to discourage liquid water accumulation.

Figure 7 shows the normalized liquid water thicknesses for the bilayered GDLs corresponding to the felt-type and paper-type GDLs, SGL 10 BC and SGL 25 BC, respectively. Figure 7(a) shows the liquid water content in the channel region of the felt-type bilayered GDL (SGL 10 BC), and Fig. 7(b) exhibits the water content in the same region of the paper-type bilayered GDL (SGL 25 BC). The water content in the bilayered GDLs (Fig. 7) is generally lower than that in the single-layered GDLs. This lower liquid water content has been attributed to the enhancement of vapor phase transport [68,92] and back diffusion through the membrane to the anode [93,94].

Figures 7(c) and 7(d) show the normalized liquid water in the GDL under the land region for both bilayered GDLs, corresponding to the felt-type (SGL 10 BC) and paper-type (SGL 25 BC) GDLs, respectively. In both the GDLs, less water was present near the catalyst layer/PEM interface, which was attributed to the presence of the MPL. The water content increased near the land region, which was also observed in the single-layered GDLs. In the bilayered GDLs, the difference in water content between the felt-type and the paper-type GDLs was not significant as the impact of the MPL dominated liquid water accumulation in the GDL (over the structural differences between the carbon fiber substrates).

As shown in Figs. 7(c) and 7(d), more liquid water was observed near the catalyst layer in the case of SGL 25 BC compared to the same region in SGL 10 BC, which is in contrast to the single-layered GDLs. The lower liquid water content at the catalyst layer in SGL 10 BC was attributed to the larger thermal gradient (compared to SGL 25 BC) that arises from the thicker GDL, which promotes vapor phase transport [68] through the MPL.

The observations from Fig. 6 reveal that the main structural differences in the felt-type and paper-type GDLs have a strong impact on the liquid water accumulation in the GDL when used as single-layered GDLs. However, with the application of the MPL, the influence of the MPL dominates over the impact of the structural differences of the fibrous substrate, in relation to the water accumulation in the GDL.

In this work, felt-type and paper-type Sigracet^® GDLs were compared in terms of their microstructure and related capacities to hold liquid water during operation. X-ray computed tomography was used to characterize the GDL microstructure in the form of through-plane porosity profiles. In operando synchrotron X-ray radiography was also performed with a custom fuel cell to observe liquid water accumulation patterns within the GDLs. The main findings are summarized below:

The work presented here uses the microstructure analysis and the water accumulation in the GDL to evaluate the applicability of felt-type and paper-type GDLs in PEM fuel cells. For bilayered GDLs, the impact of the MPL was more significant than the influence of the structural differences of the substrate materials. This work can be used to develop more robust and effective GDLs with enhanced liquid water transport capabilities for PEM fuel cell applications.

Financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC), the NSERC Discovery Accelerator Program, the NSERC Canada Research Chairs Program, the Ontario Ministry of Research and Innovation Early Researcher Award, the Canada Foundation for Innovation, and the R&D Collaboration Program (R-152351) of Hyundai Motor Company are gratefully acknowledged. Graduate scholarships to Jongmin Lee from the Mercedes-Benz Canada Graduate Fellowship in Fuel Cell Research, the HATCH Graduate Scholarship, and the David Sanborn Scott & Ron D. Venter Fellowship are gratefully acknowledged. Graduate scholarships to Hang Liu from the University of Toronto Connaught International Scholarship for Doctoral Students are gratefully acknowledged.

The authors would also like to thank Dr. Shiang Law and Prof. Mark Kortschot from the Department of Chemical Engineering & Applied Chemistry as well as Prof. Mohini Sain from the Faculty of Forestry for providing the generous access to their X-CT equipment.

Research described in this paper was performed at the BMIT facility at the Canadian Light Source, which is supported by the Canada Foundation for Innovation, Natural Sciences and Engineering Research Council of Canada, the University of Saskatchewan, the Government of Saskatchewan, Western Economic Diversification Canada, the National Research Council Canada, and the Canadian Institutes of Health Research. Authors acknowledge the receipt of support from the CLS Post-Doctoral and Graduate Student Travel Support Program. The authors would like to acknowledge Dr. Denise Miller, Dr. Adam Webb, Dr. Ning Zhu, and the BMIT group of the Canadian Light Source Inc. for their generous assistance.