Pore Structural Characterization of Fuel Cell Layers using Integrated Mercury Porosimetry and Computerized X-ray Tomography

The pore structure of the cathode catalyst layer of proton-exchange membrane (PEM) fuel cells is a major factor influencing cell performance. The nanostructure of the catalyst layer has been probed, using a novel combination of mercury porosimetry with computerised X-ray tomography (CXT), even though the nanopores were below the nominal CXT resolution. The method allows probing of the macroscopic spatial variability in the accessibility of the nanostructure. In particular, mercury entrapment within the catalyst layer showed a pronounced regular spatial patterning corresponding to the already higher X-ray absorbing regions of the fresh catalyst layer. The initial, greater X-ray absorption was due to a higher local concentration of carbon-supported platinum catalyst. This was due to segregation of ionomer away from these areas, caused by the particular screen printing catalyst layer deposition method used, which both enhanced the accessibility of the origin regions and, concomitantly, reduced the accessibility of the destination regions.


For proton-exchange membrane (PEM) fuel cells (FCs) the highest losses in cell voltage
occur for the cathode-side catalyst layer. Hence, the performance of the cathode-side catalyst layer strongly affects the overall performance of the PEMFC. On the cathode side the reaction catalysed is the production of water from hydrogen nuclei migrating across the PEM and oxygen diffusing in from the exterior. The microstructure of the catalyst layer controls the rate of oxygen ingress and water removal, and, thence, the performance of the cathode layer. It is thus important to understand the structure of the cathode catalyst layer as it influences transport rates 1 .
The deposition method can significantly impact the microstructure of the cathode catalyst layer. Common deposition methods include spray-coating, gravure coating, slot die coating, screen printing, and ink jet printing of the ink containing the catalyst directly or indirectly onto the PEM polymer membrane. Previous work has shown that the pore structure of the catalyst layer affects its performance 1 . It is thus important to understand exactly what type of pore structure a particular fabrication process leads to. This work will consider a case study of a specifically engineered catalyst layer structure produced by screen-printing, a commonly employed laboratory-scale deposition technique. The catalyst layer produced during screenprinting is affected by the process parameters such as pressure, mesh size, scan speed and number of applications (screenings) 2 .
Computerised X-ray tomography (CXT) has previously 3,4 been used to look at the agglomerate structure of the primary particles making up the catalyst layer, and at the secondary pore structure. The 50 nm resolution in this earlier work was not fine enough to distinguish 30-40 nm carbon support materials, 3-5 nm platinum nanoparticles, and 5-10 nm ionomer films within the agglomerates. However, CXT can be combined with other techniques to probe certain aspects of the smaller-scale nano-structure below the nominal CXT resolution limit. For example, CXT can be combined with mercury porosimetry, which can probe length-scales down to ~3 nm 5 . Mercury porosimetry is based on the non-wetting nature of mercury to most materials, which means an increased hydrostatic pressure is required to force it into ever smaller pores. Once within small pores mercury can become entrapped even after the pressure is reduced back down to ambient. The high electron density of mercury makes it highly visible in CXT images. Hence, when entrapment occurs, CXT can be used to image the CXT resolution-scale spatial distribution of pores that mercury has entered at a given pressure. The integrated mercury porosimetry and CXT technique allows the study of variations in nanoscopic structure over macroscopic dimensions. This has the advantage, over other techniques that study nanoscale structure like electron tomography and dual-beam microscopy, that the sample volume is much larger and more likely to be statistically representative than the small volumes necessary for these other techniques. CXT, in combination with mercury porosimetry or other metal intrusion methods, has been used to study porous media previously. Mann and co-workers studied low melting point alloy intrusion into the macropore network of an alumina catalyst pellet using X-ray tomography 6 .
Scheiba and co-workers used a Wood's metal intrusion process to study the electrode-gas diffusion layer (GDL) interface, though this method requires slicing and polishing sections and thus is destructive 7 . Hellmuth and co-workers studied mercury entrapped in macroporosity in rocks 8 . However, previous use of CXT combined with mercury porosimetry has generally looked at macroporosity, rather than lower scale properties of materials.
Mercury extrusion can be used to obtain further substantial information on the nature of a porous structure. It has been used to obtain pore connectivity 9 and the macroscopic spatial arrangement of pore sizes 10,11 . This is because the entrapment process, which is a characteristic feature of extrusion in many materials, is particularly sensitive to specific aspects of the pore structure, such as the spatial relations of pores of different sizes and the inter-connectedness of the pore network 12 . In this work the entrapment process will be used to derive novel information about the spatial variation in pore accessibility. This paper describes a case study of the application of the integrated mercury porosimetry and CXT method to a screen-printed catalyst layer to demonstrate the capabilities of the technique in revealing issues with the pore structure. Complementary information was also obtained from electron microscopy techniques including imaging and Energy-dispersive X-ray spectroscopy (EDX).

Material
The sample was a prototype fuel cell layer with ink consisting of carbon-supported platinum and ionomer screen-printed onto a PTFE decal transfer sheet. The catalyst ink used to form the catalyst layers consisted of a 60% Pt on carbon catalyst with an intentionally relatively poor Pt dispersion (to enhance X-ray contrast) and aqueous Nafion solution to give a loading of 120% Nafion:C. The ink was prepared using a high shear mixing method to produce an ink with a smooth consistency, which was then coated onto a PTFE substrate by screen printing through a 123 mesh screen with a rubber blade squeegee. The coated PTFE samples were dried at >110°C, under conditions designed to deliberately form a specific catalyst layer structure with a preponderance of micro-cracking, to enable the capability of the integrated techniques to be best demonstrated.

Electron Microscopy
The machine used was a FEI XL30 FEG-ESEM. The SEM was done once for two different forms of the FC layer sample. The full FC layer sample with the PTFE backing layer was used for the first run. For the second run, the FC layer, i.e. carbon, platinum and ionomer binder, was scraped off from the backing layer (PTFE) to see structure at the back, and determine whether it was the same as the top surface. Both samples were placed on a stud and glued, and then placed in the SEM chamber. For the full sample a 10 keV beam was used to minimise the distortion of the PTFE layer, which could be irradiated by higher voltages during scanning. For the scraped sample, since the PTFE was removed, the beam was increased to 15 keV for increased resolution. Both experiments were done in a high vacuum environment.

Energy-dispersive X-ray spectroscopy (EDX)
The FEI XL30 FEG-ESEM was fitted with an Oxford Instruments ISIS unit to analyse the sample by EDX. The software SEMQuant was used to monitor the sample and the ZAF quantitative method was applied to smooth the sample signal. A system energy of 116 eV was used. Randomly chosen spots were imaged at different magnifications and analysed. For each spot an energy-peak diagram was produced where each peak was measured against the energy indices within the software to identify the individual elements in the sample.
Additionally, the software was used to quantify the element weight percentage and atomic percentage of each element in the sample.

Mercury Porosimetry
A Micromeritics Autopore IV 9500 machine was used which ran up to a pressure of 60,000 psia. The rig works in two stages where the first stage involves a low pressure stage and the second is a high pressure stage. The fuel cell layers were cut up into small slices and were weighed prior to inserting them as a 'swiss roll' form inside a penetrometer. The penetrometer was then sealed, lubricated and weighed again to gain a combined weight measurement of the whole penetrometer and the sample. The software was then run and input variables such as sample name, sample weight, temperature, mercury density and pressure point intervals were inserted. On finalizing these inputs the low pressure mercury intrusion stage was initiated.
The software then recorded the volume uptake by the sample with increasing pressure. The low pressure stage involved evacuation to remove any air, moisture absorbed from the environment or any other liquids and unnecessary fluids which otherwise could create contamination issues or air pockets. The pressure of termination for the low pressure stages was ~30 psia. Once this first stage was completed the penetrometer was taken out and weighed again with the mercury inside. The compounded weight was then inserted into the software and the penetrometer was then placed into the high pressure port. The high pressure intrusion process was then initiated and the software continued recording the volume intruded into the sample up to 60,000 psia. The extrusion process was a reversal of the pressure direction to lower pressures. The experiment would terminate when the machine was unable to extrude any further mercury from the sample at the lowest pressure possible. Correction methods were applied by running a blank penetrometer. Thus any intrusion seen from the penetrometer, acting as a baseline, could then be removed from the sample data. The area of interest in the material was just the catalyst layer and so the experiment had to be performed twice on the two different materials in the layer. The first run was done using the backing layer only, and the second run was done on the area where the ink was printed onto the sheet.
This was done because when the mercury intrudes into the catalyst layer, it also intrudes into the surface roughness of the backing and compresses the PTFE layer. Therefore, to remove the effects of the PTFE, the backing layer was analysed separately and the data could then be subtracted from the composite sample. This then provided the data corresponding to the mercury intruded into the catalyst layer only. Two repeats were done on different parts of the catalyst layer and two repeats were done on the backing layer.

Computerised X-ray Tomography (CXT)
Once the high pressure analysis was completed for the mercury porosimetry the sample was then removed from the penetrometer and cut into ten square pieces. Two square pieces were cut from a fresh sample and added to the stack. The layers were ~2 cm × 2 cm. The stack, thus holding 12 layers, was then clamped by a sample holder to ensure that the alignment of the individual pieces would not be disturbed during the imaging process. The machine used for the X-ray imaging was a high resolution Xradia versa XRM-500 machine which can be operated at 160 kV electrical power and current of 62 μA. The machine comprises an X-ray detector of a high end 1014×1014 pixel resolution with a 64 bit dynamic range. The pixel resolution was 1.93 m. The sample and sample holder were placed onto a rotating stage between the beam source and the detector with a distance of 40 mm. The stage rotated while X-rays passed through the sample and 5-6 images were then taken and recorded at each angular position. The computer software was then used to acquire data from the recorded images to reconstruct the entire sample. Options can be set such that the sample can be reconstructed in a 2D image sequence or 3D visual construct. The software of the machine was used to arrange the data in a stack of 2D images displaying the projections at every angular position. A total of 1013 μ-CT slices were produced. An open-access imaging processing software package, known as ImageJ, was then used to analyse the image stack to determine surface profiles and create 3D video simulations of the stack. Adjustments were made to how the stack was presented. The machine recorded the images in an angular sequence. This angular sequence was loaded into ImageJ and a second reconstruction and stack was generated which showed how the sample was orientated directly through the stack, rather than around the stack. The final image sequence was thus shown to go through the stack layer-by-layer rather than at a significant angle. Each layer (backing and ink) had a 8 thickness equivalent to typically a stack of 75 CXT image slices. Each catalyst layer typically was encompassed by 25 CXT image slices, on average. Figure 1 shows a general overview of the uppermost surface of the catalyst layer. It can be seen that the surface has many crescent-shaped cracks. Figure 1 also shows close-up views of a crack region and the interior of a crack. The close-up views revealed some white spots which were studied in more detail.   (e) (f) Figure 4 shows the data for a typical region of the ink layer examined using EDX. Table 1 shows the elemental and atomic abundances by percentage for carbon, fluorine, and platinum in the regions shown in Figure 4. The set of images in Figure 4 consists of a series of imageacquisitions with increasing magnification, focussing down upon a particular platinum cluster. From the spectra in Figure 4 and the abundances in Table 1 it can be seen that as the image is focussed ever closer onto the platinum cluster the fluorine signal declines relative to the platinum. This suggests less fluorine in the regions occupied by platinum clusters.     The intrusion and extrusion pressures have then ( Figure 5(b)) been converted to pore radius using the standard Kloubek 13 correlation, which removes the effects of contact angle hysteresis. From Figure 5 it can be shown that the modal intrusion radius is ~10-20 nm. This would correspond to the typical neck size at the entrance to the ink layer mesopore network accessible from just the front of the complete composite FC layer. This is the side that would be bound to the fuel cell membrane during a typical lamination process. It can also be seen that, once the contact angle hysteresis is removed, and allowing for random error arising from subtraction of the backing data, intrusion into the ink layer itself is completely reversible for the smallest pore sizes, i.e. smaller than ~6-7 nm. The lack of mercury entrapment in the smallest pores has been confirmed by mercury thermoporometry of the entrapped mercury (data not shown). Hence, mercury entrapment is confined to pores of roughly modal size and larger. It can be seen that the overall entrapment level (residual mercury saturation) is high.

RESULTS
Given the intrusion at the highest pressures does not lead to entrapment, which would provide direct evidence of actual mercury penetration, it might be suggested that that intrusion is only apparent, and is really due to compression of the sample. However, any apparent intrusion cannot be simple, reversible, elastic compression of the pore walls because there is still some hysteresis at the highest pressures. In addition, dimensional change has been suggested previously as a cause of hysteresis in porosimetry when mercury pressure is exerted on the exterior wall of a still to-be-intruded pore 14   (a) (b) Figure 6 shows electron micrographs of general overviews of the top surface of the fuel cell layer following mercury porosimetry. It can be seen that mercury is retained in the crescentshaped macroscopic cracks that cover the surface even though they have direct access to the exterior. This is because atmospheric pressure is sufficient to keep mercury in pores of sizes ~14 microns or larger. In some cases, perhaps due to inadvertent flexing of the layer, mercury has left the macropores.  Figure 7 shows cropped parts of successive ((a) followed by (b)) CXT image slices for roughly the same spatial region of a particular fuel cell layer following porosimetry. In the CXT images the darker regions correspond to low density and the brighter, whiter regions to high density. The images show a periodic 'polka dot' like pattern of dark, low density regions separated by brighter, high density regions. The pattern of white spots is similar to the array of spheres in a single layer of hexagonal close packing. In the images in Figure 7, obtained following mercury entrapment, the mercury trapped in the macroscopic crescent-shaped cracks can be seen superimposed on the regular array of black and white spots. The images also show a crack that has partially lost its mercury from a middle segment, which now has very low intensity in the image. As the electron micrographs in Figure 1   shows that the intensity of the black spot regions has stayed at ~60-80, while that of the white spots has increased to ~100-120. It is noted that the intensity of each peak and trough in the intensity profile is very similar and shows no systematic decrease from peak-to-peak as might be expected if the image slice was tilted and was missing part of the ink layer towards one 21 side. The very high intensity spikes in the intensity profile correspond to the mercury entrapped in the macroscopic crescent-shaped cracks. The results suggest that the 'polka-dot' pattern in the CXT images has become more defined because the intensity of the high density spots has increased further following mercury entrapment. Since mercury is a dense element and is likely to increase absorption, and thus intensity, as it has for crescent-shaped cracks, then the increase in intensity in the white spot regions must be caused by higher mercury entrapment in those regions. This suction may result in a reduced density in the ink at these sites. Increasing snap-off speed (e.g. due to an increased snap-off gap) is known to reduce the ink deposit 15 . The regular array of fibres in the screen and their intersections is thus probably responsible for the regular pattern in density distribution observed in the CXT images. The spacing of the holes in the mesh matches the spacing of the dots in the images obtained above.
The results reported above show that a very particular spatial pattern of entrapped mercury was observed in the CXT images, with mercury becoming preferentially entrapped in the high X-ray absorbing regions for the fresh sample. Due to the very high fractional entrapment of intruded mercury, as seen in Figure 5, the rise in X-ray absorbance density, due to the entrapment, is effectively a tracer for where the most mercury intrusion has occurred. For the fresh sample, the highest electron density element present initially is the platinum catalyst, since all other components are generally made up of elements in the earliest rows of the periodic table. The carbon support and platinum are always associated, so where there is enhanced electron density (Pt) in the unfilled sample there will also be an enhanced amount of carbon. As the ink only consists of catalyst and ionomer, the only reason for enhanced platinum, and therefore catalyst, must be due to a partial depletion of the ionomer from this area. The ionomer in the ink would be expected to be mobile when the ink is printed and still wet. Hence, it seems likely that the white spot areas of enhanced electron density in the CXT images of fresh FC layer are due to a combination of relatively high catalyst concentration coupled with low ionomer concentration, whereas the dark areas are due to low catalyst concentration, but high ionomer concentration. When mercury is intruded into the catalyst layer, the areas that become more intensely X-ray absorbing are the areas not only with reduced amounts of ionomer, but also those with porosity that is still accessible. It is possible that the mercury is not entering the low X-ray absorbing areas because the enriched ionomer there is blocking access to many of the pores within that part of the layer. The low Pt density (dark) regions are enriched with ionomer and therefore may be expected to hold more water, if it can penetrate them.

CONCLUSIONS
The particular fabrication method for the cathode-side catalyst layer studied in this work has resulted in a regular patterning of the spatial arrangement of low and high density regions. A combination of mercury porosimetry and CXT has shown that the high density regions created when using a screen-printing process have relatively more accessible porosity. The variation in initial X-ray density in the fresh layers suggests this is due to segregation of the ionomer during the screen-printing process, which reduces the pore accessibility of the ionomer-enriched regions.