Determination of Local GDL Saturation on the Pore Level by ...

Determination of Local GDL Saturation on the Pore Level by in-situ Synchrotron based X-Ray Tomographic Microscopy

F.N. Büchia, J. Ellera, F. Maroneb, M. Stampanonib,c

a Electrochemistry Laboratory, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland

b Swiss Light Source, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland c Institute for Biomedical Engineering, University and ETH Zürich, Gloriastrasse 35,

CH-8092 Zürich, Switzerland

In-situ synchrotron-based tomographic microscopy (SRXTM) with a spatial resolution in the order of 1μm and sensitivity for carbon and liquid water, has the potential to provide fundamental information for the understanding of the wetting properties of gas diffusion layer (GDL) materials on the pore level. This is important for the understanding of the solid-water interactions in the porous structures since water transport in GDLs is considered a key transport mechanism polymer electrolyte fuel cells (PEFC). However SRXTM of PEFC is a major experimental challenge. To obtain quantitative results, a complete cell needs to be operated under realistic conditions in the constrained space of the small field of view on the beamline sample stage without disturbing the sample rotation.

Introduction In polymer electrolyte fuel cells (PEFC) gas diffusion layers (GDL) play an important role for the mass transport on the microscopic scale and its related losses. The gas transport at high current densities and under condensing conditions as well as the water management of the electrolyte membrane are significantly influenced by the transport properties of the porous GDL. Understanding the mass transport properties of the GDL is therefore essential to understand and improve the water management as well as power density in the high current domain.

Three techniques are known today for imaging water in PEFC: i) neutron imaging has a very high sensitivity for liquid water but is almost not sensitive to carbon and other materials used in PEFC. Due to the moderate spatial resolution (typically 20 μm) and the limited contrast for carbon, neutron imaging is mainly being used for investigations on the cell scale [1, 2].; ii) magnetic resonance imaging (MRI) has mainly been developed for the visualization of soft tissue in the human body. Although it is extremely sensitive to water it requires the absence of magnetically inductive materials. Still it has been used to visualize the water content of membranes and small cells [3, 4]. But spatial resolution and contrast are limited; iii) X-ray imaging is sensitive to water and carbon. The spatial resolution is about an order of magnitude higher than for neutron imaging and MRI. With the high flux available at synchrotron-based installations, an exposure time for a radiogram of considerably below 1 s is achieved and therefore tomographic microscopy is feasible within relatively short measurement times (few minutes). X-ray radiography

ECS Transactions, 33 (1) 1397-1405 (2010)10.1149/1.3484631 © The Electrochemical Society

1397Downloaded 08 Mar 2011 to 129.49.56.80. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

with a pixel size of 3 μm [5, 6] and tomography with 10 μm pixel size [7] have earlier been used to study water in GDL’s. However, radiography has inferior contrast and is averaging along the beam while tomographic microscopy with 10 μm pixel size can not resolve single carbon fibers and the related microstructure in GDL’s. In previous ex-situ work Becker et al. [8] have shown that when using phase contrast tomographic microscopy the GDL structure can be imaged with sub-micrometer pixel size.

Here we report on the development of the experimental set-up to achieve sufficient spatial resolution and carbon to water contrast for visualizing and quantifying the water in GDL's with synchrotron based X-ray tomographic microscopy (XTM) on the pore scale level. First results are presented.

Method and Experimental

For a given set of materials in the cell, the water management in the GDL is determined by the local current density, the temperature, the dew point and gas velocity in the channel as well as the channel and rib dimensions. Therefore, to perform measurements under technically realistic conditions and to obtain results, relevant for applications, operating conditions close to the conditions in a technical cell are required. This means that a cell operated at realistic thermal and fluid dynamic conditions must be run at the beamline endstation.

Computerized tomography uses a series of 2D projections at different angles to

reconstruct the 3D spatial distribution of a physical property [9]. When the field of view is smaller than the sample size (local tomography) the measurement is not quantitative and inhomogeneous gray-scales can arise within a single phase, making the separation of different phases during post-processing more difficult, if not impossible. Therefore it is preferred that the entire sample fits into the field of view (see Table I) which strongly limits the sample size.

TABLE I. Specifications of the TOMCAT beamline at SLS.

Magnification Field of View [mm] Pixel Size [μm] 4 x 3.78 1.85

10 x 1.5 0.74 20 x 0.75 0.37

In order to be able to perform tomography with resolution on the fiber diameter level,

the entire set-up needs to fit into the field of view. As shown in Table 1 for a pixel size of 1.85 μm, at the TOMCAT beamline [10] of the Swiss Light Source (SLS) the field of view has a dimension of only 3.8 mm. In addition, a tomographic scan requires a 180° rotation of the sample. In our case, the operating cell with its gas and electric (cell load and heating) connections has to be rotated without interfering with the beam and disturbing the cell itself.

The small field of view allows for a cell with one, or at maximum two channels.

Principally a horizontal or vertical arrangement of the cell is possible. Both arrangements offer advantages and disadvantages.

ECS Transactions, 33 (1) 1397-1405 (2010)


Figure 1: Schematic of (a) horizontal and (b) vertical channel arrangement for cell during tomography scan.

In order to allow for good SRXTM reconstruction quality it is preferable to use

samples of cylindrical shape, where the x-ray beam penetrates the same amount of material independently of sample rotation angle. Then beam energy and exposure time fit well for every radiography projection of the XTM scan. A vertically aligned membrane electrode assembly (MEA) has not a cylindrical shape, but the flow field plates still can be cylindrical.

Two new cell designs, as shown schematically in Fig. 1, were realized for in-situ

XTM investigations. Fig. 2 shows the two set-ups mounted on the rotation stage of the TOMCAT beam line. Both cells were designed such that the regions of interest completely fit into the field of view of the 4x objective (3.78 x 3.78 mm2) or reduce material outside the field of view to a minimum.

The cell with a horizontally aligned MEA (Fig. 2a) allows only for a short linear flow

field channel of 2.3 mm (active area 4.9 mm2). Although this is the preferred orientation from the tomography acquisition point of view (similar absorption for all rotation angles) due to the short length channel, the fluid dynamics might not be representative for those in a technical cell.

The cell with the vertically aligned MEA (Fig. 2b) allows the realization of a much

longer channel. A cell with a channel length of 13.4 mm was realized (active area 33.5 mm2). Images for the entire channel length can be stitched together by moving the cell vertically in front of the detector. The vertical orientation is less preferred from the



imaging point of view (different absorption for different rotations) but it allows for realistic fluid dynamics in the channel.

Both cell designs have been realized with a single channel (width 0.8 mm; depth 0.3

mm) between flow field ribs. In both cells the temperature is measured by temperature sensors and can be controlled by small heating pads.

(a) (b)

Figure 2: Cells installed at the beamline endstation (a) horizontal cell and (b) vertical cell.

In both cells Umicore H200 catalyst coated membrane (CCM) and Toray T060 GDL

with 20% PTFE without MPL were used. Cells were operated in differential mode (stoichs > 10) using dry or at room temperature humidified gases at atmospheric pressure (ca. 98.5 kPa).

For the SRXTM scans with the 4x fold objective and absorption mode the beam energy was set to 10 to 13.5 keV. The CCD-camera was read out in unbinned mode, and exposure time was 220 ms. By cropping of the CCD-image the frame time could be reduced from 459 ms to 259 ms. Therefore scan times for scans with 1001 projections decreased from 11 min (full CCD readout) to 4 min (cropped readout).



Results and Discussion

Tomographic scans were acquired for the cells with both orientations. As described in the experimental section a scan with 1001 exposures requires 4 to 11 minutes. It is expected that only phases which are stable during (most) of the exposures can be reconstructed without artifacts. This means that only liquid water forming a stable phase during the measurement can be clearly identified. Vertical Cell

Orthogonal slices through the vertical cell are shown in Figure 3. This scan was made

near the top of the cell which is close to the gas outlet. The image obtained is of very good quality, the close-up in Figure 4 shows that the noise level. The contrast for the solid (GDL and channel structure) and void is clear. The different components of the MEA (membrane, catalyst layer) can clearly be separated. In the GDL the fine solid fiber structure is clearly resolved and contained phases of liquid water are precisely identified. Liquid water is also observed at the walls of the flow field channels. While the liquid water in the GDL seems to remain mainly stable during XTM scans (clear phase boundaries), the liquid water in the gas flow channel is not stable (blurred phase boundaries). Transient radiography without sample rotation also showed (the expected) unstable water in the gas flow channels.

Figure 3: Reconstructed slices of vertically oriented cell. Scan at 0.5 A/cm2, 40 °C, gases humidified at 24°C, 13.5 keV, λO2=10.5 and λH2=10.5. Top in-plane cut through cathode GDL. Bottom through-plane cut with cathode GDL on upper side.

The images were analyzed for contrast and noise. Figure 4 shows that the contrast

between void and fiber is in the order of 35’000 16-bit gray scale values. The noise in the void phase has an RMS of ca.1’300 gray scale values.



Figure 4: Vertical cell: (left) reconstructed close-up of cathode GDL; (right) gray scale values across two fibers.

Horizontal Cell Orthogonal slices through the horizontal cell are shown in Figure 5. Liquid water can

also be well detected within the GDL. The analysis in Figure 6 shows that while a similar contrast between void and fibers is obtained, the noise is considerably bigger. At the time this unexpected difference between reconstruction of the horizontal and vertical cell arrangements cannot be explained.

Figure 5: Reconstructed slices of horizontally oriented cell. . Scan at 0.5 A/cm2, 24 °C, dry gases, 10 keV, λO2=20 and λH2=42.

0 20 40 60 80 10010000

20000

30000

40000

50000

60000

70000

Gra

y sc

ale

valu

e

Distance [μm]



Figure 6: Horizontal cell: (left) reconstructed close-up of cathode GDL; (right) gray scale values across two fibers.

The contrast between void and fibers seems slightly lower than for the vertical cell (ca. 30’000 gray scale values) while the noise in the void is a factor of ca. 1.5 higher (RMS 1’900 gray scale values). The difference in image quality between the horizontal and vertical design might be given by the fact that we are comparing sagittal (for the vertical setup) and axial (for the horizontal setup) planes. The axial plane corresponds to the reconstruction plane, which is usually affected by more noise. However, further investigation on this respect are still ongoing. Quantitative evaluation

For the human eye the optical quality of the reconstructed slices allows for easy

discrimination of the three phases: fibers, water and void. However for quantitative evaluation the data needs to be segmented using computer based algorithms.

The most straight forward method is gray scale thresholding, including additional

data treatment such filtering or assignment of single voxels. This method requires homogeneous gray scales of the individual phases with clear differences between the phases and minimum overlap. The gray scale differences of the present data turned out not to be sufficiently reliable for water and solid to apply thresholding. Therefore a data treatment often used in neutron radiography was applied: referencing of water saturated GDL structures to the same GDL structures in dry state. Again including digital filtering, this procedure allows for separating the water phase from void and solid with satisfactory accuracy. The resulting 3D-structures of the water can then be analyzed and quantified for i.e. local saturation.

Figure 7 shows data from the vertical cell at 40 °C with reactants humidified at room

temperature (24°C).The liquid water volume fraction is averaged over 1.2 mm channel length at the cathode side. Saturation inside the GDL and water on channel and GDL surface is quantified. In the channel the highest volume fraction is found in the corners far from the GDL. In the GDL area highest saturation is observed near the catalyst layer and the rib surfaces.

0 20 40 60 80 100 12010000

20000

30000

40000

50000

60000

70000

Distance [μm]

Gra

y sc

ale

valu

e



Figure 7: Color map of local liquid water volume fraction averaged over 1.2 mm channel length of the cathode side in the vertical cell.

The saturation in the GDL may be quantified for the horizontal and vertical directions.

Figure 8 shows the respective profiles for the data in Figure 7. For the given operating conditions, an analysis of the horizontal distribution shows a higher average saturation (s = 0.15 – 0.35) is observed under the rib area vs. the channel area (s = 0.1 – 0.25). For the vertical distribution maxima of s ≈ 0.4 are observed at top and bottom near catalyst layer and rib/channel surface.

Figure 8: Averaged saturation profiles in the GDL in the horizontal and vertical direction for the data in Fig. 7.

0

0.2

0.4

0.6

0.8

1

Liqu

id F

ract

ion

Liqu

id F

ract

ion

500 500 μm

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2.00

0.1

0.2

0.3

0.4

Mea

n sa

tura

tion

Horizontal position [mm]

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.180.1

0.2

0.3

0.4

0.5

Mea

n sa

tura

tion

Vertical position [mm]

Rib Rib



Conclusions

Two different PEFC cell concepts were developed and tested for in-situ synchrotron based X-ray tomographic microscopy and measurements were carried out at the TOMCAT beam line of the SLS. Both concepts produced successful imaging data when recording tomographic scans with a duration of about 4 minutes during operation of the cells.

Comparing the reconstructions of the horizontal and the vertical cells, images from

the vertical cell reconstructions show better signal to noise ratios, which so far is not clearly understood.

The image data was processed and the local water fraction was evaluated. It is shown

that the liquid water saturation in the GDL can quantitatively be determined when applying appropriate data treatment.

This method development will allow in future parameter studies to better understand

the complex interaction of the liquid water with the porous structure and support the development of materials with improved water management properties.

Acknowledgments

Financial support from the Swiss Federal Office of Energy und grant no. 1540107, carbon material donated by Steinemann AG, Chur, precise machining work by M. Hottiger, software and electronic support by T. Gloor and support at the beamline by M. Zaglio, S. Kreitmeier and J. Bernard (all PSI) is gratefully acknowledged.

References

1. D. Kramer, J. Zhang, R. Shimoi, E. Lehmann, A. Wokaun, K. Shinohara, G. G. Scherer, Electrochim. Acta, 50, 2603 (2005).

2. K. Yoshizawa, K. Ikezoe, Y. Tasaki, D. Kramer, E. H. Lehmann, G. G. Scherer, J. Electrochem. Soc., 155, B223 (2008).

3. J. Bedet, G. Maranzana, S. Leclerc, O. Lottin, C. Moyne, D. Stemmelen, P. Mutzenhardt, D. Canet, Int. J. Hydrogen Energ, 33, 3146 (2008).

4. K. Teranishi, S. Tsushima, S. Hirai, J.Electrochem. Soc., 153, A664 (2006). 5. I. Manke, C. Hartnig, M. Grunerbel, W. Lehnert, N. Kardjilov, A. Haibel, A.

Hilger, J. Banhart, H. Riesemeier, Appl. Phys. Lett., 90, 174105 (2007). 6. C. Hartnig, I. Manke, R. Kuhn, N. Kardjilov, J. Banhart,W. Lehnert, Appl. Phys.

Lett., 92, 134106 (2008). 7. P. K. Sinha, P. Halleck, C.-Y. Wang, Electrochem. Solid St., 9, A344 (2006). 8. J. Becker, R. Flückiger, M. Reum, F.N. Büchi, F. Marone, M. Stampanoni, J.

Electrochem. Soc., 156, B1175 (2009). 9. G. Hounsfield, Brit. J. Radiol., 46, 1016 (1973). 10. M. Stampanoni, A. Groso, A. Isenegger, G. Mikuljan, Q. Chen, A. Bertrand, S.

Henein, R. Betemps, U. Frommherz, P. Meister, M. Lange, R. Abela, Proc. SPIE, 6318, 1605 (2006).



Determination of Local GDL Saturation on the Pore Level by ...

Documents