1. Introduction
Polymer electrolyte fuel cells (PEFCs) are electrochemical devices which can convert the chemical energy of hydrogen into electrical energy. PEFCs are attracting considerable interest due to their carbon-free emission and high efficiency [
1]. The main component in a PEFC is a membrane electrode assembly (MEA), which consists of catalyst layers (CLs), a polymer electrolyte membrane (PEM), and gas diffusion layers (GDLs). The typical catalyst are Pt-based nanoparticles supported on a conductive carbon carrier. Iron-base catalysts are also promising due to comparative performance [
2]. During the operation of PEFCs, the following chemical reaction occurs on the surface of Pt-based nanoparticles:
and thus water is produced in the cathode catalyst layer. Under the saturated vapor pressure, water exists as a liquid. In PEFCs, the proton conductivity of PEM depends on the water content, and excessive water disturbs the chemical reaction at the catalyst site. In addition, liquid water existing in the cathode GDL blocks up the oxygen flow [
3]. Therefore, water control in PEFCs is crucially important [
4,
5,
6].
Water generation and its transport are inhomogeneous and random since the materials are porous and complex. Under an operating condition, water transport is governed the complex two-phase flow with an evaporation and condensation process, which depends on the local humidity and temperature in a PEFC [
6,
7]. In order to understand the water behaviors in an operating PEFC and then optimize the design for an improved performance, direct experimental information is indispensable for the simulation and modeling of PEFCs.
A number of experimental studies have been carried out using X-ray computed tomography (CT) [
3,
8,
9,
10,
11,
12], or neutron radiography and CT [
13,
14,
15,
16,
17,
18,
19]. These studies have non-destructively visualized the liquid water distributions in PEFCs under operating conditions. Neutron radiography has been used for the observation of PEFCs, but three-dimensional information, such as a cross-sectional image, inside the PEFCs is difficult to obtain. X-ray CT has been employed for obtaining three-dimensional information, but the targets are limited to model cells since the X-ray CT requires sample rotations. In addition, in conventional X-ray CT it is difficult to probe the liquid water in the vicinity of a catalyst layer because of large X-ray absorption by platinum, except for the use of high energy X-rays around 30 keV or higher. Until now, the application of X-ray CT techniques has been limited to the water behavior in GDLs.
The distribution of water in an operating PEFC has been studied using a combination of small-angle neutron scattering and neutron imaging techniques [
20]. The neutron imaging experiment shows that liquid water tends to accumulate around the channel and rib interface in the GDL and the channel, while the small-angle neutron scattering reveals a smooth distribution of water in the PEM. Due to the lack of one-on-one correlation in the water distribution between the GDL and the PEM, the two-phase flow model with water evaporation and condensation is proposed. The combined techniques, however, are unable to reveal the water content and distribution in the CLs because of the limited spatial resolution. Observing the water behavior throughout the PEM, CLs, GDLs, and channel is crucially important to build a water flow model, associated with water flooding in CLs.
Compton scattering imaging (CSI) has been developed as a non-destructive cross-sectional imaging technique. The advantages of the CSI technique are (1) non-destructive observation because of using high energy X-rays such as 100 keV or higher, (2) direct observation of cross-sectional images without sample rotation, (3) sensitivity to light elements such as water, and (4) capability of simultaneous measurements with fluorescent X-rays of heavy elements. The CSI technique has advanced using intense high-energy synchrotron X-rays. For the last decade, the synchrotron-based CSI has been applied to lithium-ion behaviors in rechargeable batteries [
21,
22,
23]. Recently, its application was extended to fuel cells, and the liquid water distribution in the porous carbon-based materials of GDLs was successfully visualized [
24].
In this work, we have performed a feasibility test of the CSI technique for liquid water observation in an operating PEFC. The incident X-ray energy is as high as 115 keV, which can reduce the X-ray absorption of platinum and thus increase Compton scattered X-rays. Measuring Compton-scattered X-ray distributions at current densities from 0 to 900 mA/cm2, the current density dependency of generated water distributions in the CL, PEM layer, GDL, and channel have been investigated. In addition, simultaneous measurements of fluorescent Pt-Kα X-rays have been carried out. With the Pearson correlation coefficient, the quantitative correlation between generated water and Pt-based catalyst in the cathode CL has been examined. This result indicates that, in a local area, the amount of generated liquid water is proportional to that of Pt-based catalysts at current densities above 500 mA/cm2.
2. Compton Scattering Imaging (CSI)
The interaction of X-rays with matter includes the photoelectric effect, Thomson scattering, and Compton scattering [
25]. The photoelectric effect varies greatly with the atomic number of substances and dominates the interaction when the incident X-ray energy is slightly above the electron binding energy. As the X-ray energy increases beyond the binding energy, the photoelectric effect decreases rapidly with the X-ray energy and thus Compton scattering becomes more easily detectable. Thomson scattering decreases as the scattering angle increases, while Compton scattering is almost constant over the whole scattering angle. Therefore, Compton scattering becomes dominant at high-energy X-rays around 100 keV and a large scattering angle of 90 degrees.
Figure 1 shows the schematic of the CSI technique using a single-element X-ray detector. The intensity of Compton-scattered X-rays
is given by
where
is the intensity of incident X-rays,
is the incident X-ray transmittance from the entrance surface to the position of the probing volume in an object,
is the scattered X-ray transmittance from the position of the probing volume to the exit surface,
is the average electron density over the probing volume
, and
is the Klein–Nishina differential cross section [
25]. The probing volume
is defined by the two slits system for the incident and scattered X-rays with a scattering angle of 90 degrees. In an experiment,
is monitored so that
depends on
,
, and
. When Compton scattering is dominant and the transmittances are constant, the detected intensity
represents the average electron density
after a calibration. Scanning the incident X-ray beams, an intensity map of Compton-scattered X-rays can be obtained. Since the electron density depends on constituent elements and compositions, the obtained images can display the internal structure of the object and the dynamical behavior of materials associated with chemical reactions.
High-energy X-ray Compton scattering has been used as the basis for an imaging technique under an in situ or operando condition, fully utilizing the high penetration power of high-energy X-rays into substances. The technique of CSI with an X-ray tube [
26] or a γ-ray source [
27] was used to observe composition variations and was applied to medical diagnosis [
28] and archaeological objects [
29]. Because of the limited performance of these sources, however, the attained spatial resolution was sub-millimeter and the measurements suffered from low counting rates. Recently, these drawbacks have been overcome by using synchrotron high-energy X-rays. Taking advantage of the intense, highly parallel X-ray beams, the spatial resolution of 70 µm in a plane and 10 µm in its normal direction have been achieved. In addition, the installation of a two-dimensional X-ray detector, combined with a pinhole (see
Figure 2a), has enabled much faster, direct acquisition of cross-sectional images [
23,
24]. Moreover, using a two-dimensional detector with an energy resolution, which can distinguish between Compton-scattered X-rays and fluorescent X-rays, a multimodal imaging is demonstrated in this work.
Future research directions in the CSI technique are toward the further improvement of data acquisition efficiency and the higher spatial resolution. Both directions conflict with each other since a smaller pinhole aperture or a narrower slit opening can improve the spatial resolution but weakens the Compton-scattered X-ray signals. A solution for both improvements is to install a coded aperture system [
30] into the CSI instrument.
Current applications of the CSI technique include the in operando observation of lithium-ion batteries [
21,
22,
23]. In this work, the CSI technique has been applied to the in operando observation of a PEFC for the first time. Among the hydrogen-related energy technology, possible applications include the in operando observation of a hydrogen-fueled internal combustion engine, the in situ observation of a model furnace for hydrogen-iron manufacturing, and others.
3. Experimental Details
The CSI measurements were carried out on the High-energy Inelastic X-ray Scattering beamline (BL08W), SPring-8. The incident X-ray energy was 115 keV.
Figure 2a shows the schematic drawing of the experimental setup. A PEFC was placed on the sample stage with the MEA in a horizontal (X-Y) orientation. The cathode was on the upper side. Cross-sectional images at the PEM layer, CL, and GDL were obtained after positioning the PEFC with a stepping motor along the Z direction. The incident X-ray beams were focused vertically to 10 µm using a Ni compound refractive lens [
31] and horizontally to 1 mm using an asymmetric Johan monochromator [
32]. The Compton-scattered X-rays and fluorescent X-rays from the PEFC were detected by a two-dimensional CdTe X-ray detector (HEXITEC) through a tungsten pinhole with a diameter of 70 µm. The scattering angle is 90 degrees. The pixel size of the detector is 25 µm, and the number of the pixel is 80 × 80 (6400 in total). The energy range of the detected X-rays is 4–200 keV, and the energy resolution is 1 keV at 60 keV. Since each pixel has this energy resolution, Compton-scattered X-rays and fluorescent X-rays can be separated, so that both CSI and XRF imaging can be performed by setting the region of interest (ROI) in the energy spectra. Here, we set the ROI with the energy range of 81.6–104.0 keV for Compton-scattered X-rays.
The size of the MEA is 30 × 30 mm. The observation area of the X-ray detector through the pinhole is 4.2 × 4.2 mm. Since the horizontal width of the X-ray beams is 1 mm, the visible area is effectively limited (see
Figure 2b). The left side of
Figure 2c shows the perspective view of the effective area of all measured images from the X-ray detector, together with the channel and rib structure of the cathode-side separator. The right side of
Figure 2c shows the perspective view of the model PEFC from the incident X-rays, displaying the stacked structure of GDLs, CLs, and PEM along the Z direction.
The MEA was manufactured by FC Development Co, Ltd. (Tokyo, Japan) with an area of 3 × 3 cm
2. The catalysts are TECRu(ONLY)E50 for the anode and TEC10E50E for the cathode (Tanaka Kikinzoku Kogyo, Tokyo, Japan) [
33]. The loading weight of Ru of TECRu(ONLY)E50 is 0.61 mg/cm
2, and that of Pt of TEC10E50E is 0.60 mg/cm
2. The membrane is Nafion
TM NR-212, and the GDLs are Toray Carbon Paper TGP-H-60. The thicknesses of the anode CL, cathode CL, PEM, and GDLs are 20, 20, 50, and 230 µm, respectively. The PEFC was assembled by sandwiching the MEA using gas flow channels and stainless steel separators.
Before the CSI measurements, the gas flow channels were purged using gas N2 to replace the residual air, followed by aging for about 1.5 h. The supplied gases for the anode and cathode channels were H2 and air, respectively. The flow rates of H2 and air were 150 and 500 mL/min, respectively. The voltage was approximately 0.6 V during the aging with a current density of 500 mA/cm2. The PEFC was heated to 80 °C and humidified to 85%RH. The temperature was measured using a thermocouple and the relative humidity was calculated from the temperature of the water tank for the humidification.
The CSI measurements were performed with the following procedure. After aging, the PEFC was operated under open circuit (OC) with a current density of 0 mA/cm2, and then operated under closed circuit (CC) with current density loadings. The loadings were increased to 100, 200, 300, 400, 500, 700, and 900 mA/cm2 in this order. Here the current density is given as the geometric area. After the voltage became stable under each current density condition, the CSI measurement was carried out at the position of PEM, cathode CL, cathode GDL, and channel. During the operation, water is generated in the cathode CL and then transported to the channel through the cathode GDL. The measurement time was 1 h for each image. The in-plane spatial resolution was 75 mm for all the images. The in-plane resolution is determined by the pinhole size. A smaller pinhole can improve the resolution.
4. Results and Discussion
Figure 3 shows the images of Compton-scattered X-rays in the vicinity of the cathode CL at the different current densities. Since the width of the incident X-rays in the horizontal direction is 1 mm, the vertical width of the observed images in
Figure 3 is 1 mm (see
Figure 2b). These images show the intensity distribution of the Compton-scattered X-rays from all compositions, including water, platinum, and carbon. The differences in the images between OC and CC are due to water generation by the chemical reaction (see Equation (1)) since the platinum and carbon contributions remain unchanged. In order to obtain the images of liquid water distributions, the Compton scattering images of OC are subtracted from those of CC. The subtracted images are shown in
Figure 4, where positive intensity values are indicated in blue and negative values in white. The results show that the intensity of subtracted images increase with increasing current density up to 400 mA/cm
2, indicating that the amount of water is proportional to the current density. This fact demonstrates that X-ray Compton scattering is capable of imaging the liquid water distribution in a PEFC. Although the intensities are proportional to the current density up to 400 mA/cm
2, they are found to be nearly constant in the range of 400–900 mA/cm
2. This finding suggests that there is a threshold amount of liquid water that can be contained in the cathode CL. When it exceeds the threshold, liquid water flows into the GDL or channel.
Figure 5a shows the Compton scattering images in the cathode GDL. The fibrous structure of carbon fibers is observed in
Figure 5a.
Figure 5b–g show the subtracted Compton-scattering images at the different current densities of 100–700 mA/cm
2. The case of 900 mA/cm
2 is not displayed because of unstable voltage. In 100–400 mA/cm
2, the amount of liquid water is negligibly small in the GDL. On the other hand, liquid water is observed in a limited area at 500 and 700 mA/cm
2. The observed liquid water is considered to be the one which flows from the cathode CL to the channel. The water flow path is located at the void space of the GDL.
Figure 6a shows the image of Compton scattered X-rays in the PEM. Since the membrane consists of bulk polymers, significant structures are not observed with the present spatial resolution.
Figure 6b–h show the subtracted images at the current densities of 100–900 mA/cm
2. Additional liquid water generated by the chemical reaction (see (1)) is negligibly small at all current densities. An explanation for this result is that the water uptake in the membrane is already saturated by the humidified supply gases at the OC condition.
Figure 7a,b show the images of Compton-scattered X-rays in the layer of the cathode channel and rib at 0 and 500 mA/cm
2, respectively.
Figure 7c shows the subtracted images at the current density of 500 mA/cm
2. In the vertical center of
Figure 7a,b, the structure of the gas flow channel with a width of 1 mm wide is observed, together with ribs above and below it. Accumulation of liquid water is observed on the wall surface of the ribs. This observation is consistent with the results of the combined, small-angle neutron scattering, and neutron imaging [
20].
Finally, the relationship between the amount of generated liquid water and that of Pt catalysts is investigated using the Pearson correlation coefficient [
34].
Figure 8 shows both the Compton scattering intensity and the Pearson correlation coefficient in the vicinity of the cathode layer as a function of the current density. Here, the Compton scattering intensity is obtained by integrating the counts over the Compton scattering peak. The intensity increases with current density in 0–500 mA/cm
2, indicating that the amount of generated water increases with the increase in the current densities. In contrast, in 500–900 mA/cm
2, the intensity is slightly decreased. This finding suggests that the amount of water generated in the cathode catalyst layer exceeds a certain threshold, resulting in the formation of paths where the water flows from the cathode CL to the channel through the GDL. This is consistent with
Figure 5f,g and
Figure 7c.
Here, the Pearson correlation coefficient (
) is defined as
where
is the intensity of the X-ray fluorescence of Pt-Kα and
is the differentiated intensity of Compton-scattered X-rays in the
i-th pixel, respectively.
and
are their mean values. If this value is close to 1, it means that there is very high correlation between Pt and water, and they are distributed in the same location. If it is close to 0, it means that there is no correlation between them. As shown in
Figure 8, the Pearson correlation coefficient is about 0.1 at 100 mA/cm
2 and about 0.8 at 300 mA/cm
2 and higher. The correlation maps between Pt and water at 100 and 500 mA/cm
2 are presented in
Figure 9b,c, together with the image of Pt-Ka fluorescent X-ray intensities in
Figure 9a. In
Figure 8, the value of
increases when increasing the current density from 0 mA/cm
2 to 300 mA/cm2 and is almost constant or slightly decreased in 400–900 mA/cm
2. The overall trend is almost the same to that of the generated water. These facts show that water is dispersed in the CL without any correlation with the amount of Pt catalysts at low current densities, and water tends to be distributed in the Pt-rich areas at high current densities. This finding is consistent with the fact that the chemical reaction takes place at the Pt catalysts.
This work has demonstrated that the study of correlations between liquid water and a specific element is feasible with the simultaneous measurements of Compton-scattered X-rays and fluorescent X-rays. This method can be applied to the study of the relationship between Ce scavengers and liquid water in an operating PEFC.