Impact of the Microporous Layer on Gas Diffusion Layers with Patterned Wettability I: Material Design and Characterization

Freudenberg H23 GDLs were used as base material for all the experiments. The GDLs were coated in-house with 30%_(w/w) FEP following a procedure reported previously.⁴⁰ Before sintering, the MPL was added. The MPL ink was prepared by mixing 70 mg of fluoropolymer dispersion (FEP 55%_w/w, FEPD 121 Dupont) with 11 ml of isopropanol (Sigma Aldrich) and 7.4 ml of ultrapure water (18.2 MΩ·cm). The obtained mixture was slowly added to 116 mg of carbon (Acetylene Black 50% compressed Alfa Aesar). The resulting ink was sonicated for an hour using an ultrasonic bath (VWR Ultrasonic USC100T) and then spray coated onto the coated GDL using a hand airbrush pistol (Sogole HP-200). The samples of 30 cm² were taped to a flat aluminum table and the ink was sprayed in layers. After each layer compressed air was blown over the sample to help evaporate the solvent. Each layer was sprayed in a direction perpendicular to the previous one and the staring corner was change each time to improve homogeneity. For the highly porous MPL, an ink load of 0.16 ml·cm⁻² was applied. For the crack-free MPL, the total ink load of 0.33 ml·cm⁻² was applied and an intermediate drying step was required. In a first step one quarter of the total ink load was sprayed and the sample and was dried in an oven at 100 °C for an hour. Afterward the GDM was rapidly cooled and the rest of the ink was sprayed. Both the highly porous and crack-free MPL were subsequently sintered at 280 °C following the temperature program reported in our earlier publication.⁴¹

After coating and adding the MPL, the materials were modified using electron induced radiation grafting to achieve the patterned wettability. The activation was done using an electron beam EBLab 200 sealed laboratory emitter system (Comet AG, Switzerland). The characteristics of the final pattern modification depends on the incident radiation beam energy as shown previously for GDLs without MPLs.⁴² To study this dependency for full GDMs in this work, the samples were placed in a metallic trade over a polyethylene film with approximately 1 mm thickness with the MPLs facing down and different samples were irradiated using 6 different electron beam energies: 100 keV, 120 keV, 130 keV, 140 keV, 170 keV and 200 keV. The samples were irradiated under air using a dose of 50 kGy and a trade transport speed of 12 m·min⁻¹.

After irradiation, the GDMs were immersed in reactors containing a solution of 15%_(w/w) of acrylic acid (Sigma Aldrich) and ultrapure water (18.2 MΩ·cm). The solution was bubbled with N₂ during 1 h to prevent the presence of O₂ during the reaction. Afterwards, the sealed reactor was placed in a water bath at 60 °C for 30 min. The samples were removed from the reactors and subsequently cleaned by applying vacuum through a Buchner funnel using three different solvents: first ethanol, then isopropanol and finally ultrapure water (18.2 MΩ·cm). After cleaning, the samples were dried in a vacuum oven overnight at 80 °C and 10 mbars.

For the purpose of analysis, the protons of the acidic groups were exchanged with Na⁺ using the following procedure. The GDMs were placed for 8 h in a 0.05 M NaOH (Sigma Aldrich) solution using a mixture of 40%_(v/v) EtOH (94% Alcosuisse AG) and 60%_(v/v) ultrapure water (18.2 MΩ·cm) as a solvent. The samples were then rinsed with water and dried overnight at 80 °C and 10 mbars. The Na⁺ stained GDM were cut in pieces of 3 × 3 mm² and placed in a holder fixing them with conductive tape. The two sides of the GDM were measured using two different pieces, one placed with the GDL facing up and another piece with the MPL facing up.

The sample termed as "0 keV" was not exposed to radiation and was used as a reference. It was subjected to all other reaction steps, (including cleaning and ionic replacement), to quantify the possible residual element content in the samples.

A similar procedure to the one described above was used to prepare the samples with patterned wettability. In this case, the samples were irradiated using a stainless steel mask that partially blocks the radiation. Three different radiation configurations were used, as summarized in Table I.

The pattern used in these experiments consists of 500 μm wide hydrophilic parallel lines separated by 940 μm wide hydrophobic regions. After activation, two different chemistries were employed to graft the materials. In the case of grafting with N-vinylformamide (FEP-g-PNVF) the samples were submerged in a reactor with pure N-vinylformamide (NVF) (Sigma Aldrich) and bubbled with N₂ for an hour. After this, the grafting reaction was conducted by heating the sealed reactor to 70 °C for an hour. For the samples grafted with poly(styrene sulfonic acid) copolymerized with poly(acrylic acid) (FEP-g-(PSSA-co-PAA)), a 0.5 M sodium styrene sulfonate (SSA) and 0.5 M acrylic acid (AA) solution in ultrapure water (18.2 MΩ·cm) was used. The reactor containing the solution and the sample was also bubbled with N₂ for an hour, after which the grafting reaction was conducted during 36 h at 60 °C. The samples grafted with NVF were then cleaned as stated previously. The FEP-g-(PSSA-co-PAA) samples were cleaned with a different cleaning sequence. The homopolymer formed during this reaction is polar and therefore more soluble in water. Thus, the samples were first placed in water for 2 h and then cleaned using the filter based method described previously. The sequence of cleaning solutions was first water, then ethanol and finally water again.

The samples used for EDX analysis were stained in order to be able to visualize the grafted regions. To this purpose, the FEP-g-PNVF samples were first hydrolyzed. The samples were placed in a 2 M NaOH in 80%_(w/w) ultrapure water and 20%_(w/w) EtOH (94% Alcosuisse AG) at 80 °C for 8 h. After the reactions the samples were immediately placed in a 0.1 M HCl in 80%_(w/w) ultrapure water and 20%_(w/w) EtOH (94% Alcosuisse AG) solution for 8 h. The immersion of the samples in acid is necessary to remove the sodium formate residues from the basic hydrolysis⁴³ and HCl was purposely selected to stain the samples with Cl⁻. The samples were then rinsed with ultrapure water and dried in vacuum following the parameters previously described. The FEP-g-(PSSA-co-PAA) samples were stained with Na⁺ using the procedure described in the previous section.

The samples with micro-holes punched into the MPL were irradiated using a specific holder/mask system to ensure that the hydrophilic lines and the punched holes were well aligned. The samples were placed in a holder made from polyethylene with the MPL facing down, covered with the stainless-steel mask and irradiated. Afterwards they were grafted with NVF and cleaned as described before. Using the same holder, the GDMs were positioned with the MPL facing up, and micro-holes were punched through the slits of the mask. To control the depth, diameter, and spacing of the holes, a special tool was created. The shape of the hole is determined by the characteristics of the needle, tip diameter and rod diameter, and by the penetration depth. For this experiment, stainless steel insect pins size 000 (Fine Science Tools) were used. The pins have a 30 μm tip diameter with a rod of 250 μm diameter and 40 mm length. The head of the punching tool was designed and calibrated to precisely control the penetration depth into the material. The punching tool was guided by mounting it on a computer numeric controlled (CNC) mill to precisely control the positioning of the micro-holes through the slits of the mask using a repetitive NC-code. The micro-holes had a final diameter of about 50 μm and 30 μm depth and a separation center-to-center of 150 μm.

A scanning electron microscope (FE-SEM Ultra 55, Carl Zeiss, Germany) was used with a compatible accessory (EDAX TSL, AMETEK) to obtain EDX spectral data and elemental mapping. The analysis was performed using a voltage of 10 kV, an aperture of 120 μm and a gun-to-sample distance of 9.7 mm. For the spectral analysis, three different spots were measured for each sample and averaged.

Since the additional elements present in the grafted monomers (O and N) are difficult to distinguish from the carbon and fluorine of the coating, we stained the grafted regions with Na⁺ or Cl⁻ as described in the previous sections, increasing the detection limit and improving quantification. The sodium or chlorine signal were normalized with the fluorine signal (representative of the fluoropolymer coating). The coating of the GDL is not necessarily homogeneously distributed over the sample and due to the high porosity of the GDM the amount of material measured on a given spot can vary. Therefore the Na/F signal (or Cl/F signal when staining with chlorine) is a better representation of the hydrophilic grafting in the GDM than the corresponding absolute signal.

In the case of the elemental mapping analysis, an image processing procedure as described in our previous publication was used.⁴² The Na/F (respectively Cl/F) signal is computed only for pixels where the F signal is sufficiently high, and interpolated in between. The presented images use a color scale representation where green represents non-grafted regions (Na/F or Cl/F signal of 0.0) and blue represents grafted regions (threshold value Na/F signal of 0.05 or Cl/F signal of 0.04).

The XPS data was acquired using a VG ESCALAB 220iXL spectrometer (Thermo Fisher Scientific) with a monochromatic Al Kα X-ray radiation (1486.6 eV) and a spot with 500 μm diameter (power of 150 W). The spectrometer was calibrated using the Ag 3d_5/2 peak at the binding energy of 368.3 eV. All spectra were recorded in constant analyzer energy mode at a pass energy of 30 eV. The binding energy calibration was done based on the C-C (GDL) peak at 284.5 eV.

The MPL surfaces were imaged using a Leica VZ700 C optical microscope including a polarization filter. The resulting images are presented as obtained from the microscope without any further processing.

The XTM was performed using a Nanotom-m Lab-CT scanner (GE Measurements & Control) at PSI. The images were acquired using an acceleration voltage of 60 kV and a current of 280 μA. For each scan, 2400 projection were recorded during the 360° sample rotation. Each projection consisted of 3 averaged images with an exposure time of 500 ms each. A 30 fold magnification was used, which resulted in a final voxel size of 3.3 μm. To probe the location of the hydrophilic regions, water was injected from the GDL side. The wet sample was equilibrated during 16.5 h and then scanned.

Water injection experiments

Three different methods were used for water injection into GDMs as illustrated in Fig. 2. The description is given below for the three different methods.

Zoom In Zoom Out Reset image size

Water injection by capillary pressure

Electrochemical water injection

Electrochemical injection with hydrophobic barrier

Water generation protocol

Two different acquisition modes were employed for the experiment: through-plane and in-plane imaging. For the through-plane imaging configuration, the neutron beam was perpendicular to the MEA plane, and for the in-plane configuration the neutron beam was parallel to the MEA plane. The neutron radiography experiments were performed at the NEUTRA beamline at the Swiss Spallation Source (SINQ) at PSI.⁴⁴

For the through-plane imaging mode the neutrons were captured using a 30 μm scintillator screen made of Gd₂O₂S. The resulting light output image was collected through a mirror and a commercial optical lens (Zeiss Makro Planar T100) by a CCD camera (Andor Ikon-L, 2048 × 2048 pix). The acquisition time was set to 20 s and images were acquired continuously during the measurements, with an acquisition rate of approximately 3 images per minute.

For the in-plane imaging mode, a gadolinium sheet forming a 2 mm wide vertical slit was used to increase the beam collimation in horizontal direction to a L/D ratio of 1750, in order to limit the blurring due to beam divergence. The anisotropic resolution enhancement method by detector tilting developed at PSI⁴⁵ was used. The neutrons were captured using a 10 μm thick scintillator screen made of Gd₂O₂S, tilted to form an angle of 10° with the beam axis. The resulting light output image was collected through a set of two mirrors and a commercial optical lens (Zeiss Makro Planar T100) by a CCD camera (Andor Ikon-L, 2048 × 2048 pix). The acquisition time was set to 30 s and images were acquired continuously during the measurements, with an acquisition rate of approximately 2 images per minute.

Precise and extensive information about the general methodology employed for image processing can be found elsewhere.⁴⁶ In particular, for these experiments, the images were corrected for camera thermal noise and outliers where removed using a 3-dimensional filter (3 × 3 × 3 pixels) on the measured time sequences. Subsequently, the images were corrected for the spatial inhomogeneities of the neutron beam, for displacements of the cell compared to the reference image of the dry cell and for the background related to scattered neutrons,⁴⁷ and a Gaussian blur filter (sigma = 0.75) was applied to remove the noise in high spatial frequencies. Finally, the images were divided pixel-wise by an image of the dry cell and averaged over 5 min to improve the signal-to-noise ratio.

The water thickness was calculated using the Lambert-Beer law:

$$\begin{eqnarray*}&&\displaystyle \frac{I}{{I}_{{ref}}}={e}^{-{\rm{\Sigma }}x}\end{eqnarray*}$$

where I is the intensity of the beam after the cell containing water and I_ref is the intensity of the beam after the dry cell. Σ is the water attenuation coefficient that has a value of 0.35 mm⁻¹ (for the NEUTRA beam line and for the used detector configuration) and x is the water thickness. To obtain the water profiles of the different tested cells, only the water in the rib regions was taken into account.

As mentioned before, the ability to choose whether the MPL is modified along with the GDL or remains fully hydrophobic is important. Our previous analysis determined that the minimum electron beam energy required to graft through the thickness of a GDL is 160 keV, and numerical simulations predicted that with an electron beam energy of 140 keV, the radiation should be stopped near the interface between GDL and MPL.⁴² The previous experiments were conducted on different GDL substrates (Toray TGP-H-060), but the Freudenberg H23 used here has a similar thickness and porosity. Considering that, besides the additional thickness it induces, the MPL has a higher density than the GDL, the penetration depth of an electron beam of a given energy will be smaller for bi-layer samples. Therefore, to modify the material through its thickness we expect that energies higher than 160 keV are required.

Figure 3 shows the measured amount of grafting from each side of the sample—as a reminder, the sample was exposed to an electron beam coming from the GDL side—using EDX and XPS as detection methods. As explained in the experimental section we replaced the protons of the acid group of the grafted compound by Na⁺ in order to detect it more reliably, and we normalize it to the amount of coating, measured by the F signal. The results obtained without irradiation (displayed as the point at zero energy) confirm that the source of Na signal is resulting from the ionic exchange of protons in the grafting.

Zoom In Zoom Out Reset image size

The analysis by EDX (Fig. 3a) shows a gradual increase of the Na/F signal for the GDL when increasing the electron energy. In theory, this energy should not impact the amount of grafting at the sample side which receives the radiation, as the radiation dose the sample receives is set as a constant. The energy of the electron beam will determine the penetration depth into the sample. Our interpretation of the increase of grafted amount on the GDL side at higher energies is that, since our samples are highly porous, we are probing the depth of the material while analyzing with EDX. The increase of radiation energy modifies deeper areas of the GDL and therefore increases the sample volume contributing to the measured signal.

Since the radiation was applied from the GDL side, we only expect to observe grafting from the MPL side once the electron energy is high enough to penetrate throughout the whole sample. This happens between 140 keV and 170 keV, where we detect a strong increase in the Na/F signal. Similar trends can be observed for the XPS analysis (Fig. 3b): The Na/F signal from the GDL side increases with electron energy—although the values are clearly more scattered than from the EDX analysis—and from the MPL side there is nearly no signal until 140 keV, and a strong increase between 140 keV and 170 keV. Even though the EDX and XPS data sets follow the same trends, the Na/F signal for XPS is one order of magnitude lower than for the EDX when measuring from the MPL side. This could be related to accumulation of fluoropolymer near the MPL surface or different distribution of the grafting degree on the MPL.

In summary, we identified that the most suitably electron energy for modifying the GDL only is 140 keV, as it provides the highest level of grafting in the GDL without modifying the MPL and electron energies higher than 140 keV allow to graft also the MPL.

Once demonstrated that the MPL can be grafted, the next step was to create a patterned GDM. Since higher activation energies lead to a better defined pattern,⁴² the GDMs for which both MPL and GDL were grafted using an electron energy of 200 keV, which was the highest possible at the used electron beam. Figures 5a and 5b show EDX maps of such a modified GDM. It can be seen that the wettability pattern is clearly present in the MPL but less defined than in the GDL. One possible reason is that, due to the fine porosity and structure of the MPL, the grafting solution does not imbibe the material homogeneously, hindering the grafting in certain locations. Another possibility is that the carbon particles compete with the fluoropolymer for the radical formation (or act as radical scavenger) lowering the efficiency of the reaction. This is not an issue in the GDL because, in this case, the surface area of the carbon is much lower than in the MPL, and the fluoropolymer surface to be grafted is not in close vicinity to free carbon surfaces. In the MPL, the carbon particles have a mean diameter of 50 nm—resulting in a much larger surface area than for the GDL fibers having a diameter of 10 μm, and are closely mixed with the fluoropolymer which acts as a binder.

One fact that leads to this theory is that we have not been able to graft the MPL using NVF. (Figs. 5c and 5d). Our studies of grafting reaction kinetics and fourier transform infrared spectroscopy analysis in FEP films show that the grafting reactions with NVF are superficial, while grafting reactions with AA modify the bulk of the film, even over-growing the base polymer and becoming the main polymer.⁴⁰ Reactions with SSA-AA induce a gradient where the grafting is mostly superficial but also result in a mild bulk modification. Our hypothesis is that the chemicals which induce more bulk reactions are able to overcome the reaction competition with the carbon and that this is the reason why we were able to graft with AA and SSA-AA but not with NVF (See S.I.1 is available on at stacks.iop.org/JES/167/064516/mmedia).

As explained in the introduction, it has been shown that the cracks on the MPL can act as preferential water pathways since the capillary pressure through larger pores is smaller and the water permeability through the large pores is larger. The incorporation of a MPL to the GDL is crucial to obtain a PEFC with high performance. Our first attempt to incorporate a self-standing MPL to a grafted GDL showed the expected performance improvement. However, the water separation between the hydrophobic and hydrophilic regions was not as pronounced as when there was no MPL present. This difference in water distribution was attributed to the structural properties of the MPL. Since the MPL has smaller pores and lower permeability the necessary capillary pressure to break through this layer is more important than the capillary pressure differences obtained by changing the surface chemistry of the GDL. The water is transported through the MPL through the paths with lowest resistance to the GDL. If these paths lead to hydrophobic regions, it results in an undesired water accumulation there. Therefore, the incorporation of perforation in the MPL is expected to improve the efficiency of the grafting modifications in the GDL by establishing preferential pathways from the CL to the hydrophilic regions of the GDL.

In Fig. 4, optical microscope images of the surface of the different MPLs are shown (corresponding to the MPLs schematically illustrated in Fig. 1). Due to the opacity of the carbon particles, a polarization filter was required to image the MPL. This makes carbon fiber appear as white in the images (Fig. 4a) since they are more reflective than the MPL. By controlling the spraying methodology we can obtain two different types of MPL, a highly porous one (Fig. 4b) or a crack-free MPL (Fig. 4c). The crack-free MPL was further modified by punching holes or grafting. Studies analyzing the impact of a highly porous MPL³⁴ reported that it can improve cell performance, highlighting the relevance of this type of material. Nevertheless, since one of the goals of our experiment was to incorporate features to the MPL and analyze the impact of such modifications on the water distribution in a GDL with patterned wettability, we casted crack-free MPLs to minimize the effect of the inhomogeneous structure.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The images of the MPL surface of the crack-free MPL show no difference between the unmodified MPL (Fig. 4c) and the chemically modified (grafted) MPL (Fig. 4d), as expected. In Fig. 4e, the mechanically modified MPL with micro-holes of an approximate diameter of 50 μm, a separation of 200 μm and a penetration depth of 30 μm are shown.

To verify that the micro-holes in the MPL were properly positioned over the hydrophilic lines of the GDL, an X-ray computed tomography scan was performed. The sample was partially imbibed with water in order to make the hydrophilic portions of the GDL visible. Figure 6 shows two X-ray tomography slices of a GDM with punched micro-holes. These results confirm that the lines of micro-holes in the MPL (Fig. 6a) are aligned to the hydrophilic areas of the GDL (Fig. 6b).

Zoom In Zoom Out Reset image size

The key characteristic to determine the efficiency of the modifications incorporated to the GDM is how they affect the water distribution. The modifications—chemical and mechanical—are targeted to reduce, for defined pathways, the capillary pressure required for the water to be transported through the material, leading water to flow preferentially through these regions. This behavior can be tested by injecting water by capillary pressure and visualizing the water distribution in the material. We have already published the results and methodology of this testing method for several of our GDL samples with patterned wettability.²⁵ This technique is very helpful to identify the samples and methods that are efficient at creating the water pathways but does not necessarily reflect the water distribution under real fuel cell operation.

Figure 7 shows the water distribution within the same type of GDL measured with neutron radiography using two different water injection methods. The GDL was a Freudenberg H23 carbon paper coated with 30% FEP and grafted with NVF with a pattern of 500 μm wide hydrophilic lines separated by 940 μm wide hydrophobic regions (no MPL). The water distribution varies significantly (Figs. 7a and 7b), as the capillary pressure injection method uses a hydrophobic layer that impedes water from leaving the set-up on the side opposite to the injection (see Fig. 2). Due to this, higher water saturation can be achieved and the material even reaches full saturation when using sufficiently high capillary pressures.²⁵ With the electrochemical water injection method, water is produced in the CL and it is transported through the GDL over the pathway representing the lowest capillary barrier. Once this pathway is established, water exits the GDL to the flow channels and no further pressure can be build up. In consequence, much less water is accumulated and the pattern created by the water distribution is not as defined. Another difference between the two methodologies is that, in the capillary injection method, water can only be accumulated in the GDL under test. For the electrochemical water injection method, water can be present not only on the cathode side (where the test sample is), but also in the anode GDM. To clarify this, Fig. 7c shows an in-plane neutron radiograph of the sample tested with the electrochemical water injection method. It can be seen that the water accumulates mainly on the cathode side and under the rib region. Thus, water accumulated in the anode side masking the water on the cathode side cannot be the explanation for the absence of a clear pattern in the water distribution image acquired with the through-plane configuration (Fig. 7b). It must be noted that, with the in-plane imaging configuration, the patterned wettability cannot be analyzed since the pattern is perpendicular to the flow channels.

Zoom In Zoom Out Reset image size

Finally, an important difference between capillary pressure injection and electrochemical injection is the injection points. In both cases, the possible injection points are distributed over an extended area covering the hydrophobic and hydrophilic regions. However, with the capillary pressure injection method, all possible injection points are connected by the water supply channel, which allows the pressure to equilibrate and thus the water goes through the most favorable path in terms of capillary pressure barrier. With electrochemical injection, water needs to partially saturate the GDL in hydrophobic regions to create lateral transport pathways to the hydrophilic regions. As a consequence, a higher water saturation is obtained in the hydrophobic domains than with the capillary pressure injection method.

Figure 8a shows the neutron radiography of an electrochemical water injection experiment for our "baseline" sample, a GDL with patterned wettability and an unmodified hydrophobic crack-free MPL. In comparison to the sample without MPL (Fig. 7b), the separation between hydrophilic and hydrophobic domains is not visible anymore. To analyze in more detail the origin of these differences in water distribution, another electrochemical water injection set-up was built with an extra hydrophobic layer on the cathode side (see experimental, Fig. 2). This layer hinders the water transport to the flow channels and allows to accumulate similar amounts of water as with the capillary injection method. We also compare both electrochemical water injection methods with the water distribution under operation. There is not much difference between the water distribution with the electrochemical injection method without and with the hydrophobic barrier (Figs. 8a and 8b) when the water content is low. The main difference is the possibility to achieve higher water saturations (Fig. 8c) when a hydrophobic barrier is added. In that case, the impact of the wettability pattern is noticeable. Nevertheless, it is not as defined as when the sample is measured using the capillary pressure injection method (Fig. 7a). The results obtained with electrochemical water injection with a hydrophobic barrier are the most similar to the water distribution measured for a fuel cell in operation (Fig. 8d).

Zoom In Zoom Out Reset image size

To understand the differences between these experiments, several factors have to be taken into account. The first of these factors is the impact of the MPL. Unfortunately, our capillary pressure set-up did not allows us to reach capillary pressure high enough to break through the MPL, so this impact will be discussed on the basis of the electrochemical injection results. When a MPL is added, its fine porosity creates a hydrophobic barrier and the necessary capillary pressure for the water to permeate through is significantly higher than when there is no MPL. The produced water permeates through this barrier using the path with less capillary pressure, generating randomly distributed water injection points from the MPL into the GDL.¹⁶ Since there is no preferential pathway through the CL, there is the same probability for injection points to be in a hydrophobic or in a hydrophilic area. If the injection point lands on the hydrophobic region water needs to be transported laterally towards the hydrophilic region. As mentioned previously, this lateral transport is also required when no MPL is used, but it is apparent that the forcing of specific injection points by the MPL further hinders the lateral transport and the formation of the water distribution pattern.

Additionally it must be consider that when water is produced in the CL it can be transported through the GDM either in liquid or in vapor form. We have already discussed how the transport of liquid water takes place; for the gas transport it depends on the vapor water diffusion through the GDM and is not influenced by the surface chemistry. Nevertheless, due to the temperature differences in the cell the water is subjected to condensation/evaporation events. Where the condensation takes place depends mostly on temperature differences and local cold spots in the GDL. Although this is an important factor in the water distribution in a fuel cell, it is not the main reason of the differences in water distribution seen with the capillary injection method and operando cell test. The water distribution for the sample tested by electrochemical water injection with hydrophobic membrane is very similar to the water distribution for the same sample tested in operando. The capillary pressure test was performed at room temperature, the electrochemical water injection test was performed at 35 °C meanwhile the operando test was performed at 50 °C.

The fact that there is no hydraulic connection between the injection points is the key aspect that makes the water content in the hydrophobic region higher when the electrochemical water injection is used. This hinders the water separation due to the patterned wettability.

We used the electrochemical water injection method with a hydrophobic barrier to compare the three different GDMs with different MPL characteristics (Fig. 9). The in-plane radiographies (Figs. 9d–9f) show that the produced water mostly accumulates in the cathode GDM. It must, however, be noticed that the water content in the anode channels is increased compared to the experiment with no hydrophobic barrier (Fig. 7c). This is to be expected, since due to the higher water pressure in the cathode the water diffuses to the anode side. When comparing the water content for different MPL modifications this fact must be considered since the water in the anode, although lower than in the cathode, can no longer be neglected. Nevertheless, as the anode GDM has no patterned wettability, water accumulating on this side can be regarded as a constant offset, which does not influence the visualization of water distribution in the cathode GDM wettability pattern.

Zoom In Zoom Out Reset image size

There are no significant differences between the water content of the baseline sample (Fig. 9a) and the sample with holes punched into the MPL (Fig. 9b): both samples have an average water thickness of approximately 80 μm, and the impact of the patterned wettability is visible in both samples.

This is not the case for the sample with a grafted MPL (Fig. 9c). The average water content of this sample is slightly higher (approximately 90 μm), but no pattern of water accumulation can be observed. A possible reason to explain why the pattern is not defined for the sample with grafted MPL is that we used a different grafting chemistry. The chemistry of the sample with grafted MPL is FEP-g-(PSSA-co-PAA) which yields a contact angle of 45° (see S.I.1), in comparison to the contact angle of 30° for the two samples grafted with FEP-g-NVF. One of the main elements to ensure an efficient water separation is to have a high enough difference between the contact angle of the hydrophobic and hydrophilic regions. The lower this difference, the less effective is the pattern.

In summary, although the success of the local modifications of the MPL (hydrophilic grafting and punched holes) was confirmed by ex situ characterization, they do not have a clear impact on the water distribution when injecting the water electrochemically.

The reason for using preferential water pathways is to limit the water content in the hydrophobic regions, and our results show that the effect is mostly favoring water accumulation in the hydrophilic regions without reducing the water content in the hydrophobic regions (compare Fig. 8b to S.I.2). This is due to the water being injected to the GDM through localized and hydraulically disconnected small spots (in the order of a few micrometers) when the water is produced electrochemically. These spots are randomly distributed through the MPL. To achieve that the GDM modifications reduce the water content in the hydrophobic regions the water injected here has to be transported laterally through the MPL, CL or membrane to a hydrophilic region before the break through to the GDL. Our hypothesis is that this lateral transport requires a higher pressure than the one required for water break through the GDM in hydrophobic regions. In the future, the MPL modifications (either chemical or mechanical) can focus on promoting water lateral transport through the MPL in order to canalize the water injection in the hydrophilic regions of the GDL.