Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
X-ray Investigations of PEMFC Gas Diffusion Layers (GDLs)
by
Pradyumna R. Challa
A thesis submitted in conformity with the requirements for the degree of Master of Applied Science
Mechanical & Industrial Engineering University of Toronto
© Copyright by Pradyumna R. Challa, 2012
ii
X-ray Investigations of PEMFC Gas Diffusion Layers (GDLs)
Pradyumna R. Challa
Master of Applied Science
Mechanical & Industrial Engineering
University of Toronto
2012
Abstract
In this thesis, synchrotron radiography was utilized to image liquid water distributions in the
porous polymer electrolyte membrane fuel cell (PEMFC) gas diffusion layers (GDLs). GDLs
were compressed in an ex situ flow field apparatus with 1mm x 1mm channels, and injected with
liquid water to study the effect of current density and microstructure on through-plane GDL
liquid water distributions. The effect of the size of the water inlet on GDL liquid water
distribution was also investigated. Micro-computed tomography was employed to characterize
the effect of flow field compression on commercial and non-commercial GDLs. Porosity
distributions of compressed GDLs were compared with those of uncompressed GDLs, and the
effect of microstructure on the porosity was discussed. The experimental techniques documented
in this thesis will inform future research, while the results will help modellers generate realistic
GDL pore structures for multiphase flow simulations and validate their models.
iii
Acknowledgments
I am very grateful to my research advisor, Dr. Aimy Bazylak, for her never ending support and
encouragement throughout these past two years. I am also very grateful for the support of all my
good friends and colleagues at the microscale energy systems transport phenomena (MESTP)
laboratory – thank you for all the wonderful discussions and good memories. I would like to
thank James Hinebaugh, Jongmin Lee, and Ronnie Yip, in particular, for their help with
conducting experiments at the Canadian Light Source Inc., Saskatoon, SK. Additionally, I would
like to thank the research scientists, Dr. George Belev and Dr. Adam Webb of the Canadian
Light Source Inc., for their help with my imaging experiments. I would also like to thank Mr.
Shiang Law, a research associate at the Forestry department at the University of Toronto, for
accommodating my schedule on the CT scanner. Finally, I would never have been able to
complete my MASc without the vision and the emotional support of my parents, Subbarao BVR
Challa and Rukmini Challa, and my brother, Anirudh Challa. Thank you all.
iv
This thesis is dedicated to my parents, Dr. Subbarao BVR Challa and Rukmini Challa, for
inspiring me to pursue engineering.
v
Table of Contents
Abstract ........................................................................................................................................... ii
Acknowledgments .......................................................................................................................... iii
Table of Contents ............................................................................................................................ v
List of Tables ............................................................................................................................... viii
List of Figures ................................................................................................................................ ix
1. Introduction ................................................................................................................................ 1
1.1 Motivation and Objectives ................................................................................................... 1
1.2 Organization of the Thesis ................................................................................................... 2
2. Background and Literature Review............................................................................................ 4
2.1 Pore Structure of a GDL ...................................................................................................... 4
2.2 Quantification of GDL Porosity ........................................................................................... 4
2.3 Effect of Compression on the Porosity of a GDL ................................................................ 6
2.4 Effect of Water Transport on the Porosity of a GDL ........................................................... 7
2.4.1 Review of Visualization Techniques for Water Transport ........................................ 7
2.4.2 Water Transport Investigations using X-ray Imaging ............................................... 8
2.5 Conclusion.......................................................................................................................... 10
3. Ex situ Synchrotron Investigations of Water Distribution in PEMFC GDLs – Flooded Inlet . 12
3.1 Introduction ........................................................................................................................ 12
3.2 Methods .............................................................................................................................. 13
3.3 Image Analysis ................................................................................................................... 14
3.3.1 Normalization .......................................................................................................... 14
3.3.2 Error Analysis .......................................................................................................... 15
3.4 Results and Discussion ....................................................................................................... 16
3.4.1 GDL Microstructure ................................................................................................. 16
vi
3.4.2 Surface Treatment of Felt GDLs .............................................................................. 17
3.4.3 Water Injection Rate ................................................................................................ 18
3.5 Conclusion.......................................................................................................................... 18
3.6 Figures and Tables ............................................................................................................. 20
4. Ex situ Synchrotron Investigations of Water Distribution in PEMFC GDLs – Single Point
Injection ........................................................................................................................................ 24
4.1 Introduction ........................................................................................................................ 24
4.2 Methods .............................................................................................................................. 24
4.2.1 Compression Calibration ......................................................................................... 25
4.2.2 GDL Edge Isolation ................................................................................................. 25
4.3 Results and Discussion ....................................................................................................... 26
4.3.1 Repeatability ............................................................................................................ 26
4.3.2 Water Injection Rate ................................................................................................ 27
4.3.3 GDL Microstructure ................................................................................................. 29
4.4 Conclusion.......................................................................................................................... 29
4.5 Figures and Tables ............................................................................................................. 31
5. Quantifying the Effect of Compression on the Through-plane Porosity Distributions of
PEMFC GDLs with Micro-computed Tomography (μCT) .......................................................... 39
5.1 Introduction ........................................................................................................................ 39
5.2 Methods .............................................................................................................................. 39
5.2.1 Compression Calibration ......................................................................................... 40
5.2.2 GDL Surface Isolation ............................................................................................. 40
5.2.3 Image Binarization ................................................................................................... 41
5.2.4 Image Processing ..................................................................................................... 41
5.3 Results and Discussion ....................................................................................................... 43
5.3.1 Uncompressed Porosity Distributions ...................................................................... 43
5.3.2 Effect of Compression ............................................................................................. 43
vii
5.4 Conclusion.......................................................................................................................... 44
5.5 Figures and Tables ............................................................................................................. 46
6.0 Conclusions & Future Work ................................................................................................... 55
6.1 Water distribution visualization ......................................................................................... 55
6.2 Microstructural Investigations ........................................................................................... 56
References ..................................................................................................................................... 59
viii
List of Tables
Table 3.1: Flooded Inlet Investigations: GDL material properties along with their breakthrough
pressures and average water contents.
Table 4.1: Single-point Injection Investigations: GDL material properties, experimental
conditions and the breakthrough pressures.
Table 5.1: Effect of rib/channel compression on GDL porosity.
Table 5.2: Effect of rib/channel compression on GDL thickness.
ix
List of Figures
Figure 1.1: A Schematic of a PEMFC. The membrane electrode assembly (MEA) consists of a
polymer electrolyte membrane (PEM) coated with catalyst layers (shown in black) and porous
cathodic and the anodic GDLs (black circles represent solid carbon fibers). Water (shown in
blue), which is generated at the cathodic GDL, hinders oxygen diffusion (shown in red).
Figure 1.2: SEM Images of various GDL microstructures: a) paper (Toray TGP-H-090), b) felt
(Freudenberg H2315), and c) cloth (AvCarb 1071).
Figure 3.1: Experimental setup: (a) A schematic of the apparatus with the following
components: compression plate (1), plastic gasket (2), GDL (3), rubber gasket (4), and base
plate (5). (b) This assembly of (1), (2), (3), (4), and (5) is bolted together with four bolts (20in-
lb/bolt). (c) An image of the apparatus and the experimental setup in the hutch at the Canadian
Light Source Inc., Saskatoon, Sk. The orientation of the beam is parallel to the channels.
Figure 3.2: Synchrotron radiographs: (a) An example radiograph obtained from synchrotron X-
ray radiography. This particular radiograph is of Freudenberg H2315 I6 when injected with
liquid water at 8μL/min. The region of interest is highlighted with a dashed line. Rib (R) and
channel (c) placement is also shown. (b) A normalized radiograph of 3.2(a) with the region of
interest highlighted (inverted for clarity).
Figure 3.3: 1D Water thickness distributions at breakthrough of (a) Toray TGP-H-090 (10 wt. %
PTFE), (b) Freudenberg H2315, (c) Freudenberg H2315 I6 (10 wt. % PTFE), (d) Freudenberg
H2315 I3 C1(10 wt. % PTFE with MPL), and (e) AvCarb 1071 (10 wt. % PTFE) at liquid water
injection rates of 1μL/min and 8μL/min.
Figure 4.1: Experimental setup: (a) A schematic of the apparatus with a close-up view of the
GDL set-up with the following components: (1) top plate, (2) GDL, (3) compression plate, and
(4) base plate. (b) This assembly of (1), (2), (3), and (4) is bolted together with four bolts. The
torque on the bolts can be adjusted to compress the GDL to a required pressure. The synchrotron
beam is incident into the plane of the page. (c) A schematic of the experimental setup.
Figure 4.2: Pressure-torque calibration curve for GDL A.
x
Figure 4.3: GDL Edge Identification Process: (a) A typical radiograph obtained from imaging
the setup with synchrotron X-rays. This particular radiograph is of DURA-GDL ST400TC. (b)
The highlighted region of 4.3(a) used to determine the edges of the GDL. (c) Image after
applying the Sobel operator. (d) The edge profile of 4.3(c) with edges of the GDL identified as
points (denoted by asterisk) of maximum intensity.
Figure 4.4: Image normalization and analysis: (a) A typical radiograph obtained from imaging
the setup with synchrotron X-rays. This particular radiograph is of DURA-GDL ST400TC when
injected with liquid water at 2μL/min. (b) A normalized radiograph of 4.4(a) with the region of
interest highlighted. The GDL is highlighted along with the regions under the ribs. (Refer Image
Analysis Section for the normalization process). (c) 1D water thickness profile with error bars
obtained from the 2D water thickness map in 4.4(b).
Figure 4.5: Water thickness contour maps for GDL A (paper) obtained at the moment of
breakthrough: (a) water injection rate of 1μL/min, (b) water injection rate of 2μL/min.
Figure 4.6: Water thickness contour maps for Freudenberg H2315 I3 C1 (felt) obtained at the
moment of breakthrough: (a) water injection rate of 1μL/min, (b) water injection rate of
2μL/min.
Figure 4.7: Water thickness contour maps for DURA-GDL ST400TC (cloth) obtained at the
moment of breakthrough: (a) water injection rate of 1μL/min, and (b) water injection rate of
2μL/min.
Figure 5.1: A schematic of the (a) uncompressed and (b) compressed GDL sample holders for
μCT.
Figure 5.2: Compression calibration results: (a) pressure distribution on a Prescale pressure film
after applying a torque of 10 in-oz in the compression device. (b) average pressure calibration
curves as a function of applied torque for GDL B. The calibration was repeated twice to ensure
repeatability.
Figure 5.3: Performing edge detection on μCT images: (a) A portion of a typical through-plane
μCT slice obtained after reconstruction of GDL B compressed between the flow field and the
compression plate of the compression device. (b) The output of 4(a) after applying Canny’s edge
xi
detection technique. (c) Surface isolation based on the rate of change in the surface profile. The
black compounded line is the surface profile of the GDL, while the red line denotes the rate of
change in the surface profile. Refer 5.2.2 for a discussion on GDL surface isolation.
Figure 5.4: Through-plane uncompressed porosity profiles of investigated GDLs.
Figure 5.5: Through-plane porosity distributions of AvCarb GDS3250.
Figure 5.6: Effect of incremental compression: (a) Through-plane porosity distributions of GDL
B. (b) Figure that illustrates the effect of incremental compression on the thickness of the GDL,
and the average porosity under land and channel.
Figure 5.7: Through-plane porosity distributions of Dura-GDL ST400TC.
1
1. Introduction
1.1 Motivation and Objectives
In the age of global warming, clean alternative energy solutions are critically needed for
achieving a sustainable energy future. Polymer electrolyte membrane fuel cells (PEMFCs), with
major applications in automotive propulsion and back-up power systems, provide a zero-local-
emission alternative to internal combustion engines and diesel generators [1-5]. PEMFCs are
electrochemical devices that can convert the stored chemical energy in hydrogen into electrical
energy. The reactions that liberate the chemical energy stored in fuels occur at the catalyst layer
in the presence of fuel and electrons. At the anode, hydrogen (H2) splits into protons and
electrons (Eq.1.1). The electrons travel to the cathode, where they combine with oxygen and
protons to form water (Eq.1.2).
(1.1)
(1.2)
(1.3)
The by-products of the chemical reactions are water and heat (Eq.1.3). A certain amount of water
generated in a PEMFC should be removed in order to facilitate the diffusion of oxygen through
the GDL (Figure 1.1), the prevention of which can lead to performance degradation. However,
excessive water removal can lead to membrane dehydration and increased protonic resistance
[6].
The porous gas diffusion layer (GDL) plays an important role in providing pathways for
reactants to reach the catalyst layer and product water to reach the exhaust channels. The GDL is
required to exhibit high electronic conductivity and provide structural stability for the membrane
electrode assembly (MEA) [7]. Although the GDL is typically treated to be hydrophobic to
enhance liquid water wicking, excess liquid water tends to accumulate and prevents reactants
from reaching the catalyst layer, thereby degrading the cell performance [7]. Specifically, these
mass-transport related voltage losses dominate other losses such as activation and ohmic losses
when the fuel cell is operated at higher current densities i.e. for applications with higher power
2
requirements [7-10]. Through an improved understanding of GDL material properties, such as
porosity and hydrophobicity, new materials can be designed for enhanced water management
strategies. The motivation for this research is to understand the effect of compression and water
generation on GDL porosity. An important first step in understanding the effect of water
generation on the porosity of a GDL is visualizing the water distribution in a GDL, which is one
of the main objectives of this work. The second objective of this work is to characterize the
effect of GDL compression on the porosity of a GDL.
The direct visualization of liquid water in the GDL is needed to understand the influence of
liquid water flow rate and GDL microstructure on the overall liquid water content. Through ex
situ visualizations, water transport within the GDL can be studied independently of membrane
swelling which occurs in operating PEMFCs. Additionally, the outcomes of ex situ experiments
with adjustable inlet conditions can be compared with operating PEMFCs in order to extract
information about the microscale transport behaviour within the porous materials of the fuel cell
that are challenging to observe in situ. It is also important to note that though the conditions in a
fuel cell are transient, a study of flow through the porous electrode at steady state conditions is a
necessary first step to develop a technique to visualize water in the GDL.
1.2 Organization of the Thesis
This thesis is divided into six chapters. Chapter 1 provides a general introduction and motivation
for the research, in addition to listing the objectives of this research. Chapter 2 contains a
literature review of the water distribution visualizations in GDL, in addition to a literature review
on calculating the porosity and the effect of compression on the porosity of a GDL. The third and
fourth chapters provide a discussion of the ex situ experiments that were conducted to visualize
the water distributions in GDLs with varying microstructures using synchrotron X-ray
radiography. In Chapter 3, the liquid water invasion of a GDL from a flooded inlet condition is
reported, and in Chapter 4, the liquid water invasion of a GDL from a single-point injection is
presented. In Chapter 5, the effects of compression on GDL porosity are quantified using
computed micro-tomography. In addition to summarizing the contributions of this work,
suggestions to extend this work are presented in Chapter 6.
3
Figure 1.1: A Schematic of a PEMFC. The membrane electrode assembly (MEA) consists of a
polymer electrolyte membrane (PEM) coated with catalyst layers (shown in black) and porous
cathodic and the anodic GDLs (black circles represent solid carbon fibers). Water (shown in
blue), which is generated at the cathodic GDL, hinders oxygen diffusion (shown in red).
V
4
2. Background and Literature Review
2.1 Pore Structure of a GDL
The pore structure of a GDL is important as it provides pathways for reactants to reach the
catalyst layer and product water to reach the exhaust channels. Both compression and water
generation result in a decrease in the available GDL pore space for oxygen transport [7].
GDLs are typically made of polyacrylonitrile (PAN)-based carbon fibers. Commercially
available GDLs can be classified into three categories based on their microstructure: paper, felt
and cloth (Figure 2.1). In paper, individual carbon fibers are typically cut to lengths of 3–12mm,
bound using a polyvinyl alcohol (PVA) binder, impregnated with resin, and heat-treated at over
2000°C [7]. Felt GDLs are manufactured by replacing the above-mentioned binding step with a
hydro-entanglement step, resulting in fibers oriented in both the in-plane and the through-plane
directions. Cloth GDLs consist of interwoven strands of carbon fibres. The woven GDL structure
results in both macro-pores and micro-pores, whereas the random orientation of fibers in paper
and felt GDLs typically results in strictly micro-sized pores [7]. Pore sizes of both felt and paper
morphologies are typically between 10μm and 30μm [7], and the average pore diameter of the
cloth GDL is approximately 80μm [11]. The microstructure has been shown to influence the
transport properties of the GDL [12]; therefore, it is important to study the effect of compression
and water generation on these GDL microstructures.
2.2 Quantification of GDL Porosity
The porosity of a porous material can be measured using a variety of techniques, some of which
are destructive in nature. Destructive techniques including Capillary Flow Porometry (CFP) and
Mercury Intrusion Porosimetry (MIP) render samples unusable due to contamination.
Porosimetry refers to the technique of invading a porous material by a non-wetting fluid at high
pressures, whereas porometry refers to the technique of preferentially invading the pores with a
wetting liquid and subsequently forcing the liquid out of the sample with a high pressure gas [7].
Pore size distributions can be obtained from the tracked pressure and volume of the invading
fluid as pressure is increased as a function of time. MIP and CFP have been used to obtain the
characteristics of a GDL pore structure, such as pore size distribution and pore volume [7,13-18].
Although the information obtained from MIP and CFP can be used to estimate the porosity of a
5
GDL, this information must be treated with caution, due to the underlying assumption in the
methodology that the porous material can be represented as a bundle of capillary tubes.
A non-destructive technique that can be used to measure the porosity of a material is gas
pyncometry [19,20]. In gas pycnometry, a sample of known solid volume ( = + ; where
is the volume of the void space of the sample, and is the volume of the material) is enclosed
in an evacuated container of known volume ( ). The evacuated container (along with the
sample) is connected to another container that is filled with a gas of known volume ( ) and
pressure ( ). When a valve connecting the two containers is opened, gas from the second
container expands into the evacuated container occupying the entire empty volume of the
container until a uniform pressure distribution ( ) is attained. Using ideal gas law, the volume
of the pores can be calculated as
(
) (2.1)
In prior investigations of the GDL, it was observed that the GDL porosity may be non-uniform
[15,21,22]. Though gas pyncometry allows us to calculate the porosity of a GDL without
destroying the sample, it does not enable the quantification of any spatial variation in the GDL
porosity.
Another non-destructive technique that has recently been successfully utilized to measure
porosity of a GDL is computed micro-tomography [23-32,32-34]. In this technique, a sample of a
GDL is imaged to obtain a three-dimensional (3D) reconstruction of its random microstructure,
which can subsequently be analyzed using image processing algorithms to quantify its porosity.
A significant advantage of employing this technique is that the information obtained from the
reconstructions can be used to study the heterogeneous pore structure. The information can also
be utilized to inform numerical models.
Ostadi et al. [32-34] utilized computed nano-tomography (nCT) to obtain a GDL reconstruction
at a spatial resolution of 680 nm and provided a detailed overview of the tomography analysis
process. They also calculated the bulk porosity of a GDL from the obtained reconstructions, in
addition to calculating the permeability and diffusivity in the in-plane and through-plane
directions. Büchi and coworkers [23,35,36] were the first group to report the use of computed
6
micro-tomography (µCT) to measure through-plane porosity distributions of an uncompressed
paper GDL (Toray TGP-H-060 with 20. wt% PTFE). Through-plane porosity distribution refers
to the porosity distribution of the GDL along its thickness. They correlated the through-plane
porosity distribution with water saturation distribution in a GDL and observed that the water is
retained in the denser regions near the surfaces of the GDL.
Fishman et al. [28] employed µCT to compare the through-plane porosity distributions of paper,
felt, and cloth GDLs. They noticed that felt and paper GDLs possessed surface and core regions
in their through-plane porosity distributions, while the surface and core regions could not be
identified in a cloth GDL. When compared to a paper GDL, the felt GDL was observed to exhibit
a more uniform core region. Fishman and Bazylak [26,27,37], in their subsequent papers, also
investigated the effect of PTFE treatment and MPL application on the through-plane porosity
distributions. The PTFE treatment was observed to be non-uniform, and it predominantly
resulted in a decrease in the porosity at the surfaces of a GDL. The decrease in near-surface
porosity due to the MPL was also observed from their µCT investigations.
These investigations provided valuable insights into the heterogeneous porosity distributions of
uncompressed paper, felt, and cloth GDL materials. Although these heterogeneous porosity
distributions of uncompressed GDLs were subsequently utilized to model GDLs in numerical
models that predicted liquid water distributions in a GDL [38] or thermal conductivity of a PEM
GDL [39], the effect of compression on the through-plane porosity distributions is not yet
experimentally understood.
2.3 Effect of Compression on the Porosity of a GDL
While over-compression can damage the GDL [40,41], it is known that an optimal amount of
compression is needed to maximize the fuel cell performance [42-47]. Ge et al. [43] observed
that there is an optimal GDL compression ratio for maximizing the fuel cell performance in both
paper and cloth GDLs. They also noted that compression affected the performance of paper GDL
to a greater extent than that of the cloth GDL. Roshendal et al. [48] found that a decrease in the
average porosity led to a decreased oxygen consumption thereby decreasing the fuel cell
performance. Additionally, they noted that the change in porosity had a greater effect on the
performance of a fuel cell operating at a higher current density, compared to that at a lower
current density.
7
Becker et al. [25] employed µCT imaging to investigate the effect of varying compression on the
GDL transport properties such as porosity, diffusivity, and permeability. They observed that the
porosity values decreased with compressed GDL thicknesses. However, the effect of
compression on spatially varying material properties has yet to be determined. Wang and Chen
[49] modelled the liquid water distribution in GDLs using spatially varying GDL properties and
uniform land compression ratio. However, direct experimental investigations to understand the
effect of flow field compression on the through-plane porosity are yet to be conducted.
2.4 Effect of Water Transport on the Porosity of a GDL
Water generation at the catalyst layer also has an impact on the effective porosity of a GDL. To
characterize the impact on the effective porosity of a GDL, it is important to visualize the liquid
water distribution in a GDL.
2.4.1 Review of Visualization Techniques for Water Transport
Various visualization techniques [50] including soft X-ray radiography [51,52], micro-computed
tomography [23,31,35,36,53-55] , neutron radiography [56-73], and synchrotron radiography
[54,56,74-83] have been utilized to understand water transport in GDLs. The refined focus of this
review is to discuss the use of X-ray imaging including synchrotron radiography, soft X-ray
radiography, conventional X-ray micro-tomography, and synchrotron X-ray micro-tomography.
Synchrotron radiography refers to the process of imaging using high intensity synchrotron X-
rays. These X-rays are produced from a cyclic electron accelerator (synchrotron), in which the
magnetic field and the electric field are synchronized with a travelling electron beam. Of the
methods mentioned above, synchrotron radiography has the highest intensity [56] of 1011
- 1015
photons/s/cm2 with monochromatic beams ideal for imaging fuel cells. Synchrotron radiation is
also capable of providing images with a high spatial resolution of up to 0.7μm [56]. The
temporal resolution of the images obtained using synchrotron radiography can be as low as 1s
[56], enabling researchers to gain a detailed insight of the dynamic liquid water transport
phenomena inside a gas diffusion layer (GDL).
Conventional X-ray radiography refers to the process of imaging using X-rays ranging in
energies from 100 – 225keV [51,52]. Conventional X-rays are released when an electron beam
8
bombards a target (typically tungsten). Conventional X-ray radiography is an economical
imaging technique, which can be produced using a desktop machine. However their biggest
disadvantage when compared to synchrotron radiography is the conical nature of the produced
beam. In synchrotron radiography, a near-parallel beam is used for imaging. A conical beam
causes non-uniform magnification effects, which would be detrimental for investigating liquid
water in the GDL. The maximum spatial and temporal resolutions possible with conventional X-
rays are 1μm and 10s, respectively [56].
Soft X-ray radiography involves X-rays with energies less than 40keV [56]. Soft X-rays or low
energy X-rays are produced when an electron beam is bombarded onto a vanadium target. A
spatial resolution of 0.5μm and a temporal resolution of 1s have been reported with soft X-ray
radiography [51,52].
2.4.2 Water Transport Investigations using X-ray Imaging
Though various numerical models have been developed to predict the distribution of liquid water
within the PEMFC GDL [84-86], there is little availability of experimentally determined liquid
water distributions with microscale resolution due to the imaging challenges associated with the
opacity of fuel cell materials [50].
Using synchrotron X-ray micro-tomography, Flückiger and coworkers [35] visualized ex situ
water injection into a Toray TGP-H-060 GDL (20 wt. % PTFE) and obtained water saturation
distributions at a range of water invasion pressures. They observed that the water accumulated in
the denser regions of the GDLs, and a significant amount of water was retained within the GDL
even after purging. This retention of water was attributed to the non-homogenous porosity
distribution inside of a GDL, which led to the entrapment of water within the highly porous inner
layer [35]. Sinha et al. [55] also conducted an ex-situ experiment to quantify the amount of
residual water in a GDL using conventional X-ray micro-tomography. They noted that after
purging a saturated GDL with nitrogen, a drying-time of 25min was required to remove water
from the GDL. The rate of water removal was also observed to slow down at after 6min of
purging. This slowdown in the rate of water removal was attributed to the lack of evaporation
taking place from the surface of the GDL due to the lack of a temperature gradient between the
GDL and the surrounding environment.
9
Manke and coworkers [78] visualized in-plane water evolution in the GDL of an operating
PEMFC by bombarding X-rays in the through-plane direction of the fuel cell. The authors
observed an eruptive water ejection mechanism, where water was ejected onto the surface of the
cathodic GDL in a cyclic manner from distinct pores. This localization of the eruptive water
ejection mechanism was attributed to the existence of a two-phase equilibrium between liquid
water and water vapour in those pores. Sasabe et al. [52] also visualized in-plane water evolution
using soft X-ray radiography. Though Sasabe et al. [52] did not report the occurrence of eruptive
water ejections; they observed that the spatial distribution of ejected water was uneven [52].
Hartnig et al. [74] observed two major liquid water agglomerations along the cross-section of the
GDL. The first accumulation was located next to the MPL, while the second agglomeration was
observed under the landing of the flow field. Water agglomeration near the MPL on the cathodic
side was attributed to the condensation of water vapour caused by the temperature gradient
across the MPL layer. Water agglomeration under the landing at the cathode was attributed to the
mass limitation of evaporation caused by excessive water accumulation. The occurrence of two
liquid agglomerations on both anodic and cathodic GDLs at higher operating current densities
has also been observed by Sasabe et al. [52].
Sasabe et al. [51,52] also suggested that the water generated at the catalyst layer was drained
from the surface cracks on the surface of the GDL. They also observed that the channel wall
wettability influenced the saturation profile of the GDL. According to the Sasabe and coworkers,
the copper flow field with hydrophilic walls resulted in increased capillary-dominated water
removal and reduced GDL water retention, thereby influencing the saturation profile of the GDL.
Lee et al. [77] investigated the dynamic distribution of water in the GDL using in-plane, in-situ
synchrotron X-ray imaging. They noted that the instant water generation at both the anodic and
the cathodic GDLs was cyclic, with approximately the same time period. Damping in the
frequency of cyclic instant water generation was observed at the cathodic GDL towards the end
of the time period for which the fuel cell was monitored. This damping phenomenon was
attributed to a decrease in the electrochemical activity at the cathode due to flooding.
Despite the temporal and spatial resolution advantages of synchrotron radiography, Schneider et
al. [81] observed that the performance of a fuel cell decreased after minutes of exposure to
synchrotron radiation. Although this effect was only observed when the entire active area of a
10
fuel cell was exposed to synchrotron radiation, the authors noted that the localized effect of
synchrotron radiation on the GDL, the catalyst, and the membrane should be further investigated.
2.5 Conclusion
In this chapter, a critical literature review of microstructural investigations and water distribution
visualizations using X-ray radiography has been presented. Though significant insights into the
process of water evolution were obtained for some GDLs, a scarcity of experimental works exists
addressing the influence of GDL microstructure on porosity and liquid water transport behaviour.
Hence, the effects of compression, multiphase flows, and thermal gradients need to be
characterized using controlled experiments to understand the complex water transport
phenomena inside a fuel cell.
11
Figure 2.1: SEM Images of various GDL microstructures: a) paper (Toray TGP-H-090), b) felt
(Freudenberg H2315), and c) cloth (AvCarb 1071).
12
3. Ex situ Synchrotron Investigations of Water Distribution in PEMFC GDLs – Flooded Inlet
3.1 Introduction
Although a coating of polytetrafluoroethylene (PTFE) is generally applied to the GDL to
minimize water retention, liquid water accumulation is observed in operating PEMFCs
[23,35,87]. This accumulation of liquid water prevents reactants from effectively reaching the
catalyst layer and leads to performance degradation, which in turn affects the reliability of
PEMFCs. Though different microscopic modeling techniques such as pore network modeling
[38] and lattice Boltzmann modeling [88] have been developed to predict water saturation in
GDLs, experimental visualization of water distributions in GDLs is necessary for validating the
modeling results. Given that several assumptions regarding the size of the inlet are made while
modeling water transport through a GDL [89], it is also important to experimentally study the
effect of the size of the inlet on the GDL water distribution.
In this chapter, the investigations that were conducted to visualize water injected into a GDL
with a flooded inlet condition are presented. Synchrotron radiography was employed to visualize
liquid water injection through PEMFC GDLs in an ex situ flow field apparatus, to study the
dependence of liquid water content on GDL microstructure, surface treatment, and simulated
current density (water flow rate). In the microstructure study, liquid water was quantified after
injection via an ex situ apparatus for the following GDLs: paper (Toray TGP-H-090 with 10 wt.
% PTFE), treated felt (Freudenberg H2315 I6 with 10 wt. % PTFE), and cloth (AvCarb 1071
with 10 wt. % PTFE). In the surface treatment study, liquid water distributions of untreated felt
(Freudenberg H2315), treated felt (Freudenberg H2315 I6 with 10 wt. % PTFE), and treated felt
with an MPL (Freudenberg H2315 I3 C1 with 10 wt. % PTFE), were compared to determine the
effect of surface treatment on liquid water content in GDLs. In the flow rate study, the GDLs
were invaded at flow rates representative of current densities, 0.2A/cm2 and 1.6A/cm
2, to
determine the effect of water production rate on the liquid water accumulation in a GDL.
13
All visualizations were performed at the Biomedical Imaging and Therapy (05B1-1) beamline at
the Canadian Light Source Inc. (Saskatoon, Canada). Labview (National Instruments, Texas,
USA) was employed, along with a syringe pump (Harvard 11Plus: Harvard Instruments, MA,
USA) and a pressure transducer (PX309-005G5V Omega Engineering, Connecticut, USA), to
remotely control the liquid water injection into the GDL while measuring the liquid water
pressure during invasion. The liquid water pressure is measured to monitor the water invasion
process in the GDL. The X-ray beam was directed in parallel to the channels of the compression
plate, which facilitated the capture of through-plane liquid water distributions (beam orientation
into the page of Fig. 3.1(a)) with a pixel resolution of 4.4μm/pixel and a temporal resolution of
1s/frame. It has to be noted that pixel resolution is not equivalent to spatial resolution. Pixel
resolution corresponds to the width or height of the imaged region divided by the number of
pixels that span that orientation in the digital image; whereas, spatial resolution corresponds to
the minimum discernible length in the obtained raw radiographs. The obtained spatial resolution
was approximately 20μm.
3.2 Methods
The GDLs were placed in an ex situ polycarbonate sample holder, simulating compression from
a flow field plate and facilitating liquid water injection from the base of the GDL (Fig. 3.1(a)).
The compression plate consisted of parallel, 1mm square channels, spanning the width of the
GDL sample. As shown in Fig. 3.1(a), a compressible silicon gasket (375μm in thickness) and a
non-compressible clear polyethylene gasket (200μm in thickness) were included with the
material assembly in order to prevent leakage. The gasket-GDL assembly was compressed using
four bolts (diameter of 5.7mm, each torqued to 20in-lb). Liquid water was injected through the
base of the apparatus at two flow rates: 1μL/min and 8μL/min. The area of the GDL exposed to
the flooded base plate was 1cm2. These flow rates represented current densities of 0.2A/cm
2 and
1.6A/cm2, respectively, in an operating PEMFC.
Injected liquid water volume data, liquid water pressure measurements, and radiographic
visualizations were simultaneously acquired during invasion, from the initial injection of liquid
water into the apparatus until past breakthrough. A breakthrough event corresponds to the first
formation of a sample-spanning cluster of liquid-water invaded pores [18]. In total, five
commercially available GDLs (Toray TGP-H-090 with 10 wt. % PTFE, Freudenberg H2315,
14
Freudenberg H2315 I6, Freudenberg H2315 I3 C1, and AvCarb 1071 with 10 wt. % PTFE) were
injected with liquid water at two distinct flow rates of 1μL/min and 8μL/min. The results shown
below are obtained from a single parametric investigation dataset per GDL. Between each
invasion experiment, the apparatus was disassembled and dried using compressed air for
subsequent use.
3.3 Image Analysis
3.3.1 Normalization
Water content in the GDL can be quantified from the raw radiographs using Beer-Lambert’s law.
According to the Beer-Lambart equation, the intensity, I, of the beam after traveling a distance,
X, through a material is given by [77]:
(3. 1)
where I0 is the incident intensity of the beam, and μ is the attenuation coefficient of the material.
According to equation (3.1), the intensity of the beam, upon passing through the dry GDL, is
given by:
(3. 2)
where µnon-water is the attenuation coefficient of all materials combined without water present, and
Xnon-water is thickness of those materials.
Similarly, the captured intensity of the beam that passed through the wet GDL at a given instant,
j, after the initialization of invasion is given by:
(3. 3)
where µwater is the attenuation coefficient of water, and Xwater is the thickness of the water.
Combining equations (3.2) and (3.3), with the assumption that the incident intensity I0 is
constant, yields the following expression for water thickness (Xwater) within the GDL for each
pixel of our image:
15
[ ] (3. 4)
where μwater for monochromatic beams with energy of 23 keV is 0.644/cm [90].
Figure 3.2(a) is a raw radiograph of Freudenberg H2315 I3 C1 obtained during water injection at
a flow rate of 8μL/min. Figure 3.2(b) depicts the output of the normalization process from which
a 2D water thickness (Xwater) distribution in the GDL was obtained. To facilitate comparison
among the GDLs, these 2D water thickness distributions were converted to 1D water thickness
distributions which show the variation of average Xwater along the thickness of a GDL. In the
water thickness distributions, the GDL through-plane position of 0μm represents the inlet face of
the GDL, with the other end representing the flow field face of the GDL.
3.3.2 Error Analysis
Beam fluctuations and noise from the thermal fluctuations of the CCD camera can result in false
positive or negative 2D water thicknesses, which when converted into 1D water thickness
distributions, could potentially result in non-physical negative water thickness values.
Uncertainties in the reported water thickness distributions can be accounted for by quantifying
the effect of synchrotron beam fluctuations, methods of which were reported in detail by
Hinebaugh et al. [75]. The procedure outlined in that paper was employed for all images obtained
in this work; the technique is briefly summarized below.
An algorithm that matched the beam position of the raw wet-state radiograph to that of a raw
dry-state radiograph was developed and utilized. A raw dry-state radiograph refers to the
radiographs that were collected before the liquid-water/air interface reached the GDL surface,
and raw wet-state radiograph refers to the radiographs that were obtained after the water touched
the inlet face of the GDL surface. For every raw wet-state radiograph, five raw dry-state
radiographs were selected from a stack of raw dry-state radiographs with matching beam
positions, and an average of these five raw dry-state radiographs was utilized to normalize the
raw wet-state radiograph, as described in the Image Analysis section to obtain the 2D water
thickness map.
The confidence interval associated with the 1D water thickness distributions can be determined
by analyzing the dry-state radiographs separately. Through a similar process, water thickness
16
distributions can be obtained for every dry-state radiograph by matching its beam position to five
other radiographs in the dry-state stack. In the absence of noise, liquid water should not be
detected through this dry-state radiograph analysis. However, due to the noise and errors from
the beam position matching procedure employed [75], a finite water thickness was calculated and
is shown as the confidence intervals presented in the subsequent results.
3.4 Results and Discussion
The uncompressed thickness and the liquid water pressure at breakthrough for the GDLs
employed, and the average through-plane water content values of the GDLs are listed in Table
3.1. Average through-plane water content values were obtained by calculating the average 1D
water thickness values over the entire thickness of the GDL.
3.4.1 GDL Microstructure
Figures 3.3(a), 3.3(c), and 3.3(e) illustrate the water thickness distributions for Toray TGP-H-090
(10 wt. % PTFE), Freudenberg H2315 I6, and AvCarb 1071 (10 wt. % PTFE) at breakthrough
for water injection at a flow rate of 1μL/min. For this flow rate, the breakthrough pressures of
Toray TGP-H-090 (10 wt. % PTFE), Freudenberg H2315 I6, and AvCarb 1071 (10 wt. % PTFE)
were 0.98psi, 0.92psi, and 0.18psi, respectively (Table 3.1). Although Toray TGP-H-090 (10 wt.
% PTFE) and Freudenberg H2315 I6 exhibited comparable breakthrough pressures, the resulting
water thickness distributions varied significantly. At the inlet face of the GDL, Freudenberg
H2315 I6 (water thickness = 2.1mm) exhibited nearly double the water content of Toray TGP-H-
090 (10 wt. % PTFE) (0.9mm). On the other hand, the water content at the inlet face of AvCarb
1071 (10 wt. % PTFE) was approximately 30% of Toray TGP-H-090 (10 wt. % PTFE) and 15%
of Freudenberg H2315 I6.
Throughout the thickness of the GDL, Freudenberg H2315 I6 also exhibited a greater through-
plane average water content compared to both Toray TGP-H-090 (10 wt. % PTFE) and AvCarb
1071 (10 wt. % PTFE), as noted in Table 3.2. Similarly, the average through-plane water content
of Toray TGP-H-090 (10 wt. % PTFE) was 167% higher compared to AvCarb 1071 (10 wt. %
PTFE) at 1μL/min.
17
The average water content was highest in Freudenberg H2315 I6 (10 wt. % PTFE), followed by
Toray TGP-H-090 (10 wt. % PTFE) and AvCarb 1071 (10 wt. % PTFE). This could imply that
there was substantially more interconnected pores in Freudenberg H2315 I6 compared to Toray
TGP-H-090 (10 wt. % PTFE), and AvCarb 1071 (10 wt. % PTFE). It is possible that the binder
(polyvinyl alcohol [1]) used in the paper-making step of Toray TGP-H-090 (10 wt. %
PTFE) may have blocked pore connections. The presence of this binder may lead to fewer
interconnections in Toray TGP-H-090 (10 wt. % PTFE) as compared to those formed
in Freudenberg H2315 I6.
3.4.2 Surface Treatment of Felt GDLs
Figures 3.3(b)-3.3(d) illustrate the water thickness distributions for Freudenberg H2315,
Freudenberg H2315 I6 (10 wt. % PTFE), and Freudenberg H2315 I3 C1 (10 wt. % PTFE and
MPL) at breakthrough. As shown in Table 3.2, Freudenberg H2315 had the lowest average
through-plane water content of the three GDLs at a water injection rate of 1μL/min. Although the
breakthrough pressure at 1μL/min of Freudenberg H2315 (0.41psi) was approximately 25% of
that of the Freudenberg H2315 I3 C1 (1.66psi), the average through-plane water content was
similar: 0.41mm and 0.42mm, respectively.
When the 1D water thickness value at the inlet face of Freudenberg H2315 I3 C1 (10 wt. %
PTFE and MPL) was compared to that of Toray TGP-H-090 (10 wt. % PTFE) and AvCarb 1071
(10 wt. % PTFE) (0.26 mm), it was noted that the 1D water thickness value of non-MPL coated
GDLs, Toray TGP-H-090 (10 wt. % PTFE) and AvCarb 1071 (10 wt. % PTFE), were
respectively 21% and 79% of the MPL coated Freudenberg H2315 I3 C1. The counter-intuitively
high water contents observed at the inlet face of Freudenberg H2315 I3 C1 may be explained by
a delamination of the GDL from the rubber gasket, which allowed water to spread along the
surface of the rubber gasket. Since the MPL had a lower liquid water permeability, the injected
water spread laterally until there was enough pressure to penetrate the GDL, resulting in the
formation of a secondary reservoir, and therefore, a higher water content at the inlet face of
Freudenberg H2315 I3 C1. The unusually high 1D water thickness values at the inlet faces of the
other felt GDLs (Table 3.2) could have also resulted from the delamination of the GDL from the
apparatus. In this case, the cause of the delamination is attributed to the insufficient compression
18
of the untreated Freudenberg H2315 and Freudenberg H2315 I6 (10 wt. % PTFE), which
resulted in the formation of a secondary reservoir.
Among the felt GDLs, Freudenberg H2315 had the lowest average through-plane water content
(Table 3.1). More water accumulated at the inlet face of Freudenberg H2315 I6 (1.89mm)
compared to Freudenberg H2315 (1.00mm) and Freudenberg H2315 I3 C1 (1.22mm). Again, it
was counter-intuitive that the GDL with no treatment (most hydrophilic) had the lowest average
through-plane water content.
3.4.3 Water Injection Rate
The impact of increasing the liquid water injection rate from 1μL/min to 8μL/min at
breakthrough on water thickness profiles was also seen in Figs. 3.3(a)-3.3(e). Upon increasing
the rate of liquid water injection, Freudenberg H2315 showed the maximum decrease in the
average through-plane liquid water content at 51%, followed by AvCarb 1071 (10 wt. % PTFE)
(41%), Toray TGP-H-090 (10 wt. % PTFE) (24%), Freudenberg H2315 I6 (10%), and
Freudenberg H2315 I3 C1 (8%).
The water content at the inlet face (for the exception of Freudenberg H2315 I6 and AvCarb 1071
(10 wt. % PTFE)) was observed to decrease upon increasing the water injection rate from
1µL/min to 8µL/min (Figs. 3.3(a)-3.3(e)). This may be related to the variation in breakthrough
pressures at 1µL/min and 8µL/min. As can be seen from Table 3.1, the breakthrough pressures of
all the GDLs are higher at 8µL/min compared to 1µL/min. The increased pressure in the injected
water may lead to a decreased probability of liquid water spreading laterally within the GDL.
Recall that only a single GDL sample was injected at each flow rate and hence further
investigations should be conducted to conclusively identify the effect of flow rate on the GDL
saturation.
3.5 Conclusion
In this chapter, a systematic technique was outlined to obtain the 1D water thickness distributions
as a function of through-plane positions for various GDLs. Synchrotron radiography was
employed to measure the water content at breakthrough after invading GDL samples with liquid
water to simulate liquid water production in the PEMFC. The liquid water content at
19
breakthrough for five commercially available GDLs was presented, including: Freudenberg
H2315, Freudenberg H2315 I6 (10 wt. % PTFE), Freudenberg H2315 I3 C1 (10 wt. % PTFE
with MPL), Toray TGP-H-090 (10 wt. % PTFE), and AvCarb 1071 (10 wt. % PTFE). This is the
first comparison of liquid water distributions in commercially available GDLs.
Freudenberg GDLs exhibited greater average through-plane water content than Toray TGP-H-
090 (10 wt. % PTFE) and AvCarb 1071 (10 wt. % PTFE) at breakthrough. At a liquid water
injection rate of 1µL/min, the difference in the average through-plane water content between
AvCarb 1071(10 wt. % PTFE) and a) Toray TGP-H-090 (10 wt. % PTFE) was 167%, b)
Freudenberg H2315 was 340%, c) Freudenberg H2315 I6 was 545%, and d) Freudenberg H2315
I3 C1 was 352%. The average through-plane water content at 1µL/min was higher than that at
8µL/min. Upon injecting the GDLs with a higher water injection rate, Freudenberg H2315 (51%)
showed the maximum decrease in the average through-plane liquid water content, followed by
AvCarb 1071 (10 wt. % PTFE) (41%), Toray TGP-H-090 (10 wt. % PTFE) (24%), Freudenberg
H2315 I6 (10%), and Freudenberg H2315 I3 C1 (8%).
Among the felt GDLs, Freudenberg H2315 (no treatment or coatings) had the lowest average
through-plane water content. The difference in the average through-plane water content between
Freudenberg H2315 and Freudenberg H2315 I6 (10 wt. % PTFE treatment) was 46%, whereas
the difference in the average through-plane water content between Freudenberg H2315 and
Freudenberg H2315 I3 C1 (10 wt. % PTFE and MPL) was 2.7%. A larger amount of water also
accumulated at the inlet face of Freudenberg H2315 I6 (1.89mm) when compared to
Freudenberg H2315 (1.00mm) and Freudenberg H2315 I3 C1 (1.22mm).
20
3.6 Figures and Tables
Table 3.1: Flooded Inlet Investigations: GDL material properties along with their breakthrough
pressures and average water contents.
GDL Breakthrough Pressure
(psi)
Average Water Content
(mm)
Microstructure Substrate Treatment
GDL
Thickness
(µm) [19]
At
1µL/min
At
8µL/min At 1µL/min
At
8µL/min
Felt
Freudenberg
H2315
No PTFE,
No MPL 287 0.41 0.54 0.41 0.27
Freudenberg
H2315 I6
10 wt. %
PTFE 287 0.92 0.94 0.60 0.57
Freudenberg
H2315 I3 C1
10 wt. %
PTFE and
MPL
320 1.66 1.94 0.42 0.37
Paper Toray TGP-
H-090
10 wt. %
PTFE 300 0.98 1.06 0.25 0.17
Cloth AvCarb 1071 10 wt. %
PTFE 400 0.18 0.20 0.09 0.06
21
Figure 3.1: Experimental setup: (a) A schematic of the apparatus with the following
components: compression plate (1), plastic gasket (2), GDL (3), rubber gasket (4), and base
plate (5). (b) This assembly of (1), (2), (3), (4), and (5) is bolted together with four bolts (20in-
lb/bolt). (c) An image of the apparatus and the experimental setup in the hutch at the Canadian
Light Source Inc., Saskatoon, Sk. The orientation of the beam is parallel to the channels.
(a)
(b)
(c)
22
Figure 3.2: Synchrotron radiographs: (a) An example radiograph obtained from synchrotron X-
ray radiography. This particular radiograph is of Freudenberg H2315 I6 when injected with
liquid water at 8μL/min. The region of interest is highlighted with a dashed line. Rib (R) and
channel (c) placement is also shown. (b) A normalized radiograph of 3.2(a) with the region of
interest highlighted (inverted for clarity).
1mm
1mm
(a)
(b)
23
Figure 3.3: 1D Water thickness distributions at breakthrough of (a) Toray TGP-H-090 (10 wt. %
PTFE), (b) Freudenberg H2315, (c) Freudenberg H2315 I6 (10 wt. % PTFE), (d) Freudenberg
H2315 I3 C1(10 wt. % PTFE with MPL), and (e) AvCarb 1071 (10 wt. % PTFE) at liquid water
injection rates of 1μL/min and 8μL/min.
0.0
0.5
1.0
1.5
2.0
2.5
0 100 200 300 400
Wate
r th
ickn
ess (
mm
)
GDL Through Plane-Position (μm)
(a)
1 ml/min
8 ml/min
0.0
0.5
1.0
1.5
2.0
2.5
0 100 200 300 400
Wate
r th
ickn
ess (
mm
)
GDL Through Plane-Position (μm)
(b)
1 ml/min
8 ml/min
0.0
0.5
1.0
1.5
2.0
2.5
0 100 200 300 400
Wate
r th
ickn
ess (
mm
)
GDL Through Plane-Position (μm)
(c)
1 ml/min
8 ml/min
0.0
0.5
1.0
1.5
2.0
2.5
0 100 200 300 400
Wate
r th
ickn
ess (
mm
)
GDL Through Plane-Position (μm)
(d)
1 ml/min
8 ml/min
0.0
0.5
1.0
1.5
2.0
2.5
0 100 200 300 400
Wate
r th
ickn
ess (
mm
)
GDL Through Plane-Position (μm)
(e)
1 ml/min
8 ml/min
1μL/min
8μL/min
1μL/min
8μL/min
1μL/min
8μL/min
1μL/min
8μL/min
1μL/min
8μL/min
24
4. Ex situ Synchrotron Investigations of Water Distribution in PEMFC GDLs – Single Point Injection
4.1 Introduction
Presented in this chapter are synchrotron radiographic visualization employed to visualize single
point liquid water injection through PEMFC GDLs in an ex situ flow field apparatus, for
studying the dependence of liquid water content on GDL microstructure and simulated current
density (water flow rate). In the microstructure study, liquid water was quantified after injection
via an ex situ apparatus for the following three MPL-coated 10 wt. % PTFE-treated GDLs: paper
(GDL A), felt (Freudenberg H2315 I3 C1), and cloth (DURA-GDL ST400TC). In the flow rate
study, the GDLs were invaded at flow rates of 1μL/min and 2μL/min, representative of current
densities of 0.8A/cm2 and 1.6A/cm
2, respectively. This work is motivated by recent numerical
modelling work performed in our group, where it was shown that the size of the inlet has a
significant impact on the liquid water distribution in a GDL [89].
4.2 Methods
Figure 4.1(a) shows an exploded view of the ex situ apparatus used to inject water into a GDL
from a single point, and Fig. 4.1(b) shows a collapsed view with the field of view highlighted.
The apparatus consisted of a: top plate, flow field compression plate, and base plate. A 5mm x
5mm GDL sample was placed in a 1.5mm-deep inset in the center of the top plate. The
compression plate features a configuration of alternating 1mm-wide ribs and 1mm-wide
channels, with an injection hole (0.8mm in diameter) oriented above the center landing. The top
plate and the compression plate were bolted together, and torque was applied to the bolts until a
compression of 250psi (~1.8MPa) was achieved. The bolted top/compression plates were slot
fitted into the base plate, which was secured to the stage inside the experimental hutch. Figure
4.1(c) illustrates the experimental setup. The thicknesses of the top plate (15mm) and
compression plate (15mm) were selected to minimize bowing during compression. The top plate
was machined out of optically clear polycarbonate, and the flow field compression plate and the
base plate were produced with 3D polyjet printing (Nova Product, Toronto, Canada).
All visualizations were performed at the Biomedical Imaging and Therapy (05B1-1) beamline at
the Canadian Light Source Inc. (Saskatoon, Canada). The absorption imaging technique was
25
used to measure the liquid water content in a GDL, using monochromatic X-rays at 18keV.
Labview (National Instruments, Texas, USA) was employed, along with a syringe pump
(Harvard 11Plus: Harvard Instruments, MA, USA) and a pressure transducer (PX309-005G5V
Omega Engineering, Connecticut, USA), to remotely control the liquid water injection into the
GDL while measuring the liquid water pressure during invasion. The liquid water pressure is
measured to monitor the water invasion process in the GDL, and the X-ray beam was directed
parallel to the channels of the compression plate, which facilitated the capture of through-plane
liquid water distributions with a pixel resolution of 4.4μm/pixel and a temporal resolution of
1s/frame. Radiographs were acquired from the initial injection of liquid water until the
breakthrough event had occurred.
4.2.1 Compression Calibration
To relate the applied torque to the average pressure under the ribs of a GDL, a pressure sensitive
film (Fujifilm) was inserted in between the inset surface of the top plate and the GDL. The
colour of the film changed with the applied torque, and it was analyzed using a software package
(Pressure Mapping System, FPD-8010E) developed by Fujifilm. Figure 4.2 shows the obtained
correlation between the applied torque and the average pressure under the ribs for all the GDLs
utilized.
4.2.2 GDL Edge Isolation
Edge identification in a 2D image is the first step in isolating the GDL edge from that of the
apparatus. Matlab has a number of in-built and well documented techniques such as Sobel [91],
Prewitt [92], Canny [93] etc., that can be employed to identify the solid-air interfaces in 2D
images. These techniques can be used to convert a gray scale image into a black and white
image, where white represents a solid-air interface and black represents void space. These edge
detection techniques find the edges by identifying gradients in pixel intensity values. A number
of factors, such as the surface features on the apparatus, X-ray attenuation coefficient of the
GDL-MPL, and the apparatus material, influence the selection of the edge identification
technique. Sobel and Prewitt techniques are commonly used to identify strong edges while
Canny’s technique has been proven useful when the edges are less prominent or weak [93].
26
As the edges are prominent in the raw radiograph (Fig. 4.3(a)), the Sobel operator was used to
extract the GDL edges from a raw radiograph [91]. Figure 4.3(b) shows a highlighted region of
Fig. 4.3(a) that was used to identify the edges of the GDL. Figure 4.3(c) is the resulting image
after applying the Sobel operator to identify the edges. After running the in-built edge
identification technique on the radiographs in Matlab, an edge intensity profile (Fig. 4.3(d)) i.e.,
the variation of the average pixel value along the thickness of the GDL was determined for the
edge identified image (Fig. 4.3(c)). Subsequently, the locations of local-maxima (denoted by
asterisk in Fig. 4.3(d)) in the edge intensity profile were identified as the edges of the GDL.
4.3 Results and Discussion
The uncompressed and compressed GDL thicknesses, along with liquid water breakthrough
pressures are listed in Table 4.1. This section is organized into three sub-sections: In the first
sub-section, the effect of invasion experiments at a single flow rate on distinct 5mm x 5mm
samples of a GDL is investigated. In the 2nd
sub-section, the effect of flow rate on liquid water
distribution is investigated, and in the 3rd
sub-section, the effect of GDL microstructure on the
liquid water distributions is discussed.
Using Equation 3.4 and the analysis described in section 3.3 of Chapter 3, 2D water thickness
maps were obtained for each region (Fig. 4.4(b)), which were subsequently converted into 1D
water thickness distributions (Fig. 4.4(c)). The 0μm-position in the 1D water thickness
distributions represents the inlet face of the GDL. As discussed in Chapter 3, it is important to
note that the negative water thickness values result from thermal noise in the detector, beam
fluctuations, and phase contrast effects. The error bars on the 1D water thickness distributions
(e.g. Fig. 4.4(c)) only account for the thermal noise and the beam fluctuations, and the error due
to phase contrast effects has not been quantified. The phase contrast effects are dominant at the
inlet and the outlet faces of the GDL due to the difference between the X-ray absorption
properties of the GDL and the polycarbonate top plate (Fig. 4.4(b)).
4.3.1 Repeatability
The average water content distribution for each GDL regions is shown in Fig. 4.5. For the GDLs
that have been invaded at the same water injection rate, reasonable repeatability was obtained
from the analysis of multiple samples; however, there were some exceptions, such as the
27
experiment Trial 2 at 2μL/min involving a cloth GDL (Table 4.1). Exceptions such as these can
be attributed to the non-homogenous MPL coating, which may have had a particular influence
since the GDL samples were cut from various locations of a larger supply sheet. Samples
showing high water content may have exhibited more instances of cracks in the surface.
4.3.2 Water Injection Rate
The effects of water injection rate and inlet size on liquid water distributions were investigated;
however, a statistically relevant distinction between injection rates was not observed in the
measured average water content, as shown in Table 4.1. Although limited beam time poses a
challenge, further trials would be needed to fully determine the impact of these particular flow
rates on water content. However, the widely ranging results presented in Table 4.1 suggest that a
closer examination of the characteristic flow regime expected in the GDL should be performed at
these flow rates, in the context of our empirical observations. This examination is presented
below in the form of a discussion of the non-dimensional numbers to identify the influential
forces acting in the system. The capillary number, defined as the ratio of the invading viscous
forces to the surface tension forces acting across the invading-defending fluid interface, is given
by:
(4.1)
where and V are the dynamic coefficient of viscosity and the mean velocity of the invading
fluid, respectively, and is the surface tension of the invading-defending fluid interface. The
Weber number is the product of the capillary number and the Reynolds number, which can be
written as:
(4.2)
where is the density of the invading fluid, and L is the characteristic length of the porous
media. The mean velocity, V, must be defined carefully, and since V is not known a priori, upper
and lower estimations were performed based on our experimental observations. The upper limit
of the mean velocity was determined from the thickness of a GDL (approximately t ~ 200μm
(Table 1)), its tortuosity (τ ~ 1.5 [37]) and the invasion time. The invasion time (T ~ 2s) was
defined by the shortest time over which liquid water penetrated the GDL to reach breakthrough.
28
The upper limit of the mean velocity can be obtained from:
(4.3)
The lower limit of the mean velocity was calculated from the injection flow rate (Q = 1μL/min),
and the cross-sectional area of the GDL (A = 25mm2). The lower limit of the mean velocity can
be obtained from:
(4.4)
From Equations 4.3 and 4.4, the upper and the lower limits of the mean velocity were between
the orders of 10-4
m/s and 10-6
m/s, respectively, resulting in Ca = 10-8
-10-6
and We = 10-9
-10-7
.
These values suggest that the inertial forces were negligible compared to the viscous and surface
tension forces. The viscosity ratio at room temperature in the experiments was M = 54 (log M =
1.72).
To qualitatively understand the underlying drainage mechanism occurring in these GDLs, the
capillary numbers and the viscosity ratio (viscosity ratio at room temperature in the experiments
was M = 54 or log M = 1.72) were superimposed onto the phase diagram for flow through porous
media, developed by Lenormand et al. [94]. Three zones with distinct drainage mechanisms,
namely capillary fingering regime, stable displacement regime, and viscous fingering regime,
were identified by Lenormand et al. [94]. In the capillary fingering regime, the capillary forces at
the pores dominated over the viscous forces in the invading fluid, while in the stable
displacement and the viscous fingering regime, the viscous forces dominated over the capillary
forces. With the experimental conditions employed in this thesis, the flows are positioned within
the capillary fingering regime, where the saturation is expected to be independent of flow rate. It
has to be noted that even though in an operating fuel cell (Δop ~ 90°C [6]) the viscosity ratio (M
= 17) is lower than the viscosity ratio (M = 54) at room temperature (Δexp ~ 20°C), the expected
regime according to the phase diagram developed by Lenormand et al. [94] does not change.
While in the literature, it is generally expected that liquid water transport in the GDL exists
within the capillary fingering regime of a porous media phase diagram [84], in the experiments
performed here, a closer inspection is required. Rebai and Prat [95] presented a phase diagram to
illustrate the limit of the invasion percolation regime for drainage in the GDL. If we consider a
29
network thickness of N=10 (a 200m-thick GDL with an average pore size of 20m) combined
with the estimated Ca values (10-8
-10-6
), according to the phase diagram presented by Rebai and
Prat [95], the drainage processes in this paper coincide with the limit of the invasion percolation
regime, where viscous effects may be non-negligible.
Lenormand et al. [94] noted in their work that the phase diagram is material specific, and that the
limits that determine the flow regime change from material to material. Also, the phase diagrams
developed by Lenormand et al. [94] and Rebai and Prat [95] were developed for 2D structured
pore networks, and hence may not precisely represent the flow regime in GDLs that have random
3D pore structures. Hence, further investigations are required to determine if there is an effect of
flow rate on average GDL saturation.
4.3.3 GDL Microstructure
Figures 4.5-4.7 illustrate the water thickness contour maps from a single invasion investigation
of GDL A (paper), Freudenberg H2315 I3 C1 (felt), and DURA-GDL® ST400TC (cloth) at
breakthrough for water injection rates of 1μL/min and 2μL/min, respectively. At a flow rate of
2μL/min, the breakthrough pressures of GDL A, Freudenberg H2315 I3 C1, and DURA-GDL
ST400TC were 3.34psi, 2.14psi, and 1.10psi, respectively (Table 4.1). The cloth GDL (DURA-
GDL ST400TC) had the lowest breakthrough pressure, which was expected due to its large pore
sizes. The lateral spread of water can be defined as the spread of water in the in-plane direction,
which is a good indicator of the interconnectivity in the GDL pore network. Compared
to DURA-GDL ST400TC (Fig. 4.7), the lateral spread was greater in both the Freudenberg
H2315 I3 C1 (Fig. 4.6) and the GDL A (Fig. 4.5). Also, from Figs. 4.5(a) and 4.6(a), it can be
seen that in both felt and paper GDLs, lateral spread was greater at the low flow rate when
compared to the higher flow rate.
4.4 Conclusion
In this chapter, the influences of current density and gas diffusion layer (GDL) microstructure on
liquid water distribution in a polymer electrolyte membrane fuel cell (PEMFC) GDL were
investigated using an ex situ study, where liquid water was introduced into a GDL via a single-
point injection and simultaneously imaged using synchrotron X-ray radiography. A water
injection device was developed to successfully inject water into a GDL from a single point, along
30
with a technique to apply the desired compression on the GDL. The liquid water contents at
breakthrough for three structurally distinct 10 wt. % PTFE-treated, MPL-coated GDLs were
presented, including: GDL A (paper), Freudenberg H2315 I3 C1 (felt), and DURA-GDL
ST400TC (cloth).
Although the average through-plane water content at a given flow rate was generally constant for
all GDLs investigated here, the lateral spread in the in-plane direction was the highest in the felt
and paper GDLs implying greater pore connectivity in their pore network, compared to the cloth.
The average water content in a GDL also remained generally constant when various samples
obtained from the same GDL manufacturing lot were injected with water. The variations
observed in liquid water distributions are expected, given the random nature of the GDL as well
as the heterogeneous MPL-coating.
Current densities of 0.8A/cm2 and 1.6A/cm
2 were simulated using flow rates of 1μL/min and
2μL/min, respectively. A definitive trend could not be identified between the average water
content in a GDL at an injection rate of 1µL/min and the average water content in the GDL
invaded at a flow rate of 2µL/min. Finally, it was also observed that the water content in the
GDL, when injected from a single point, was lower when compared to the observed average
water content, when injected from a flooded inlet condition, as reported in chapter 3. This is in
agreement with what has been seen in the modelling experiments [89].
31
4.5 Figures and Tables
Table 4.1: Single-point injection investigations: GDL material properties, experimental
conditions, and the breakthrough pressures.
GDL Experiment
#
Flow
Rate
(µL/min)
Uncompressed
thickness
(µm)
Compressed
thickness
(µm)
Breakthrough
Pressure
(psi)
Average
Water
Content
(mm)
GDL A
Trial 1 1 300 203 2.21 0.008
Trail 2 1 300 203 2.96 0.004
Trial 3 1 300 203 1.79 0.010
Trial 1 2 300 186 1.73 0.009
Trail 2 2 300 146 3.34 0.117
Trial 3 2 300 129 1.29 0.050
Freudenberg
H2315 I3 C1
Trial 1 1 280 195 1.61 -0.008
Trail 2 1 280 212 4.62 -0.009
Trial 3 1 280 186 2.91 0.061
Trial 1 2 280 186 2.14 0.018
Trail 2 2 280 199 2.13 0.059
Trial 3 2 280 168 3.12 0.080
DURA-GDL
ST400TC
Trial 1 1 400 202 1.30 0.117
Trail 2 1 - - - -
Trial 3 1 - - - -
Trial 1 2 400 185 1.10 0.041
Trail 2 2 400 176 0.74 0.138
Trial 3 2 400 194 1.16 0.043
32
Figure 4.1: Experimental setup: (a) A schematic of the apparatus with a close-up view of the
GDL set-up with the following components: (1) top plate, (2) GDL, (3) compression plate, and
(4) base plate. (b) This assembly of (1), (2), (3), and (4) is bolted together with four bolts. The
torque on the bolts can be adjusted to compress the GDL to a required pressure. The synchrotron
beam is incident into the plane of the page. (c) A schematic of the experimental setup.
(a) (b)
(c)
1
2
3
4
33
Figure 4.2: Pressure-torque calibration curve for GDL A.
y = 44.594x R² = 0.9358
0
50
100
150
200
250
300
0 1 2 3 4 5 6 7
Ave
rag
e P
ressure
(p
si)
Applied torque (in-oz)
34
Figure 4.3: GDL Edge Identification Process: (a) A typical radiograph obtained from imaging
the setup with synchrotron X-rays. This particular radiograph is of DURA-GDL ST400TC. (b)
The highlighted region of 4.3(a) used to determine the edges of the GDL. (c) Image after
applying the Sobel operator. (d) The edge profile of 4.3(c) with edges of the GDL identified as
points (denoted by asterisk) of maximum intensity.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 50 100 150 200 250
Avera
ge P
ixel In
tensity
Through-plane Position, X (pixels)
GDL
Thro
ugh-p
lane P
ositio
n,
X
(a)
(b) (c)
(d)
35
Figure 4.4: Image normalization and analysis: (a) A typical radiograph obtained from imaging
the setup with synchrotron X-rays. This particular radiograph is of DURA-GDL ST400TC when
injected with liquid water at 2μL/min. (b) A normalized radiograph of 4.4(a) with the region of
interest highlighted. The GDL is highlighted along with the regions under the ribs. (Refer Image
Analysis Section for the normalization process). (c) 1D water thickness profile with error bars
obtained from the 2D water thickness map in 4.4(b).
-0.05
0.00
0.05
0.10
0.15
0 50 100 150 200
1D
Wate
r T
hic
kness (
mm
)
GDL Through-plane Position (m)
(a)
(b)
(c)
36
Figure 4.5: Water thickness contour maps for GDL A (paper) obtained at the moment of
breakthrough: (a) water injection rate of 1μL/min, (b) water injection rate of 2μL/min.
(a)
(b)
Water Thickness (mm)
Water Thickness (mm)
GDL In-plane Position (μm)
GDL In-plane Position (μm)
GD
L T
hro
ugh
-pla
ne P
ositio
n (
μm
) G
DL T
hro
ugh
-pla
ne P
ositio
n (
μm
)
37
Figure 4.6: Water thickness contour maps for Freudenberg H2315 I3 C1 (felt) obtained at the
moment of breakthrough: (a) water injection rate of 1μL/min, (b) water injection rate of
2μL/min.
(a)
(b)
Water Thickness (mm)
Water Thickness (mm)
GD
L T
hro
ugh
-pla
ne
Positio
n (
μm
) G
DL T
hro
ugh
-pla
ne P
ositio
n (
μm
)
GDL In-plane Position (μm)
GDL In-plane Position (μm)
38
Figure 4.7: Water thickness contour maps for DURA-GDL ST400TC (cloth) obtained at the
instance of breakthrough: (a) water injection rate of 1μL/min, and (b) water injection rate of
2μL/min.
(a)
(b)
Water Thickness (mm)
Water Thickness (mm)
GDL In-plane Position (μm)
GDL In-plane Position (μm)
GD
L T
hro
ugh
-pla
ne P
ositio
n (
μm
) G
DL T
hro
ugh
-pla
ne P
ositio
n (
μm
)
39
5. Quantifying the Effect of Compression on the Through-plane Porosity Distributions of PEMFC GDLs with Micro-computed Tomography (μCT)
5.1 Introduction
The porosity of a GDL is an important design parameter for GDL manufacturers, as it has a
direct impact on the material permeability and effective diffusivity of reactants to the catalyst
layer [7]. In this chapter, the effect of compression on the pore structures of three MPL-coated
GDLs, paper (AvCarb GDS3250), felt (GDL B), and cloth (Dura-GDL STC 400ST) GDLs, is
investigated using microscale computed tomography (μCT). Through-plane porosity
distributions from this study may be used to inform future GDL models
5.2 Methods
GDL samples of size 5mm x 4mm were first scanned using μCT in the uncompressed state using
a sample holder, and subsequently scanned in a compressed state utilizing a compression device
that consisted of a base plate and two compression plates. The base plate had two channels that
were 4mm long, 1mm wide, and 0.5mm deep, which represented the gas channel in an operating
fuel cell. Figures 5.1(a) and 5.1(b) respectively show the uncompressed GDL apparatus and the
compression setup that consisted of a compression device, gaskets and GDLs. Silicon gaskets
(250μm in thickness) were used in the compression device to ensure an even distribution of
pressure throughout the apparatus. Due to high user fees, the devices were designed to maximize
the number of material samples imaged during each scan. The uncompressed GDL device was
rapid prototyped (Nova Product Development Services Inc., Toronto, Canada), while the
compression device was fabricated with polycarbonate using conventional machining. The
assembled devices were subsequently scanned in a μCT machine (Skyscan 1172, Belgium) to
acquire the raw radiographs necessary for 3D GDL reconstructions.
Fishman et al. [82] observed that the porosity of a GDL is batch dependent. Given the batch
dependent nature of the porosity, care was taken to obtain samples from the same sheet received
from the supplier. Even while sampling the GDL from a single sheet, care was taken to not
sample the GDL from the same region. This was done to obtain a more representative porosity
profile of a GDL.
40
5.2.1 Compression Calibration
Similar to the procedures described in Chapter 4, pressure sensitive films were placed between
the compression plate and the gaskets to obtain pressure distributions, which were analyzed
using the Fuji Digital Analysis System for Prescale (Tekscan Inc., Boston, MA) to obtain the
average compression on the GDL (Fig. 5.2(a)). Fig. 5.2(b) shows the relationship between
applied pressure and applied bolt torque. From Fig. 5.2(b), it can be seen that both the GDL
samples that were compressed together produced similar pressure distributions and that the
obtained results were repeatable.
5.2.2 GDL Surface Isolation
GDL surface isolation from the compressed device was a crucial step in quantifying the effect of
compression on GDL porosity. A number of factors such as the surface features on the apparatus,
and the X-ray attenuation coefficients of the MPL and the apparatus material, influenced the
selection of the surface identification technique. Ultem, Teflon, and polycarbonate were intially
considered as candidate materials, and ultimately, polycarbonate was chosen for its relatively
highest absorption contrast with the MPL.
To detect GDL surfaces, an algorithm has been developed in Matlab utilizing the built-in
Canny’s edge identification technique. Figure 5.3(a) is a portion of a single through-plane slice,
and Fig. 5.3(b) is the resulting image after the solid-air interfaces have been identified using
Canny’s edge identification technique. After identifying the edges in a through-plane slice, the
edge intensity profile was obtained for that slice by averaging the intensity along the through-
plane direction of the GDL. Fig 5.3(c) shows the edge intensity profiles of all the through-plane
slices in the digital GDL reconstruction (~750 through-plane slices). Subsequently, a surface
intensity profile (denoted by a black compounded line in Fig. 5.3(c)) is generated by averaging
all the individual edge intensity profiles shown in Fig. 5.3(c). The surface information profile
was further converted into a rate of change of intensity profile that denoted the instantaneous
change in the surface along the through-plane direction. This rate of change of intensity profile
(denoted by a red line in Fig. 5.3(c)) was utilized to identify the surfaces of the GDL that are in
contact with the polycarbonate device. The first maximum instantaneous change in the surface
(circled in Fig. 5.3(c)), which denoted a shift from a darker region into a brighter region (Fig.
5.3(b)), was identified as the MPL/compression plate interface. Similarly, the last minima
41
(circled in Fig. 5.3(c)) that denoted a shift from a brighter region into a darker region (Fig.
5.3(b)), was identified as the GDL/landing or GDL/channel interface.
5.2.3 Image Binarization
Various thresholding techniques have been proposed in the past to binarize a GDL: Fishman et
al. [96] used Otsu’s method, whereas Jhong et al. [97] used a visually determined threshold and
AMIRA’s filament tracing method to binarize a GDL image. In this work, Otsu’s method was
used to binarize the GDL. Otsu's method involves the use of an iterative algorithm to select a
threshold that minimizes the statistical intra-class variance of the histogram. A threshold divides
the histogram into foreground and background segments, while a statistical intra-class variance
of the histogram is defined as a weighted average of the statistical variances (square of the
standard deviation) of the foreground and background segments [98].
Three-dimensional GDL reconstruction can be binarized in two distinct ways: a) obtaining an
individual threshold for every voxel-thick in-plane slice using the Otsu’s method, or b) obtaining
a single threshold for the entire 3D GDL reconstruction using the Otsu’s method. Since the
greyscale histograms of the GDL substrate and the MPL substrate are not similar, in this work it
was necessary to binarize the in-plane slices individually to obtain a realistic porosity value of
both the MPL and the GDL.
5.2.4 Image Processing
A typical through-plane cross-section of a reconstructed GDL is shown in Fig. 5.3(a). Fishman et
al. [28] observed that the minimum representative cross-sectional area for determining the GDL
through-plane porosity distributions was 1mm2. This minimum representative area ensured that
porosity distributions of GDLs could be determined within a repeatability of 4%. Therefore, the
porosity profiles obtained from compressed GDLs with a cross-sectional area of 2mm2 were
considered reasonable for the calculations presented here.
While the uncompressed GDL reconstructions were analyzed using the technique described by
Fishman et al. [28] to obtain through-plane porosity distributions, the compressed GDL
reconstructions were first analyzed to identify the interfacial locations of the MPL/compression
plate and the GDL/land. An interface identification algorithm was developed using Matlab to
42
identify the GDL surfaces that were in contact with the polycarbonate compression device.
Subsequently, the isolated GDL was binarized based on a threshold greyscale value obtained
from Otsu's technique. Subsequently, the through-plane porosity distribution was obtained by
calculating the porosity values of voxel-thick in-plane slices as a function of their through-plane
position. The process of obtaining the through-plane porosity distributions from compressed
GDLs is summarized below:
1) Step 1: Reconstruction
a) The raw projections are reconstructed using software developed by Skyscan.
2) Step 2: GDL Isolation
a) Edge Identification: The edge identification algorithm was employed on all
through-plane slices to identify the solid/air interfaces.
b) Surface Profile: The edge profiles of all the through-plane slices were averaged to
obtain a surface profile.
c) The rate of change of the intensity of the surface profile was obtained and the
surfaces of the GDL were identified based on the first local maxima and the local
minima.
3) Step 3: Image Binarization
a) The in-plane slices of the isolated GDL were binarized individually using a
threshold obtained from Otsu’s algorithm
4) Step 4: Porosity Calculation
a) The porosities of in-plane slices were calculated based binarized images.
43
5.3 Results and Discussion
5.3.1 Uncompressed Porosity Distributions
Figure 5.4 contains the through-plane porosity distributions of uncompressed paper, felt, and
cloth GDLs, which shows that the through-plane porosity of paper (AvCarb GDS3250) was
consistently higher than felt (Freudenberg H2315 I3 C1) and cloth (Dura-GDL ST400TC). In
these results, the 0μm-position corresponds to the surface of the GDL that would be in contact
with the catalyst layer in an operating PEFMC. Although all the GDLs were coated with MPLs, a
distinct MPL region could only be identified for the felt GDL. For paper and cloth GDLs,
however, the MPL appeared impregnated into the GDL itself. Consequently, the MPL region
shown in Fig. 5.4 was defined only for the felt GDL. Average porosities of all GDLs, which can
be obtained by averaging the porosity along the thickness of a GDL, were summarized in Table
5.1. Paper had the highest average porosity (0.81), followed by cloth (0.79), and felt (0.71).
5.3.2 Effect of Compression
Figures 5.5-5.7 illustrate the effect of compression on the regions under the landings and the
channels of compressed paper, felt, and cloth, respectively. The reduction in the thickness of the
GDL (Table 5.2) at 250psi is the highest for paper GDL (50%), followed by cloth GDL (49%)
and felt GDL (17%). It is interesting to note that although paper GDL has the maximum
thickness change due to compression at 250 psi, the porosity of the compressed GDL in the bulk
regions remain fairly unchanged (Fig. 5.5). This is also the case for the felt GDL (Fig. 5.6(a)),
indicating that the surface regions of the GDL were more susceptible to compression than the
bulk region of the GDL, i.e. compression has a non-uniform effect on the porosity distribution of
paper and felt GDLs. For a cloth GDL, however, a uniform reduction in its through-plane
porosity distribution was observed at 250 psi (Fig. 5.7). Since cloth GDLs were comprised of
strands of woven carbon fibers, the compression was uniformly translated throughout the
thickness of the material, as opposed to the paper and felt GDLs that were comprised of
randomly oriented carbon fibers. Table 5.1 shows that the change in average porosity under the
landing at a compression of 250psi was maximum in cloth (37%), followed by felt (14%) and
paper (9%). Similarly, the maximum change in average porosity under the channel at a
compression of 250 psi was observed in cloth (36%), followed by felt (11%), and paper (1%).
44
The effect of compression (250psi) on the average porosity of paper, felt, and cloth GDLs was
also summarized in Table 5.1. In general, the average porosity was higher under the channel
compared to that under the landing. Moreover, in felt and cloth GDLs, the compressed porosity
distributions under the landing and channel (Figs. 5.6 and 5.7), track each other, i.e. the average
porosity under the land and the channel is within 3% of each other. However, for the paper GDL,
the average porosity under the channel is 8% higher than the average porosity under the landing.
This difference, which can also be seen in the through-plane porosity distributions (Fig. 5.5), can
be attributed to the delamination of a portion of the GDL surface under the channel due to the
applied landing-channel compression scheme. The hydro-entanglement step in felt GDL
manufacturing process and the weaving step in the cloth GDL manufacturing process make the
fibers in felt and cloth GDLs less susceptible for breakage.
Felt was also compressed to 175psi, 250psi, and 325psi to determine the effect of incremental
compression on the porosity. From Fig. 5.6(a) it can be seen that incremental compression did
not significantly affect the bulk or surface regions of the felt. As the compression increased from
175psi to 250psi, the GDL thickness decreased by 5%; thickness further decreased by 3% when
the same GDL was compressed from 250psi to 325psi (Table 5.2). Nevertheless, the average
porosities under the landing and channel (at 250psi and 325 psi) remained within 1% of the
average porosity at 175psi (Fig. 5.6(b)). This suggests that felt may be compressed for the benefit
of enhanced electrical conductivity without compromising the oxygen diffusivity, which can
potentially lead to increased performance of the PEMFC.
5.4 Conclusion
In this work, a systematic technique using micro-computed tomography was developed to
characterize the effect of compression on paper, felt, and cloth GDLs under the influence of a
channel and rib configuration that would be present in an operating fuel cell. Uncompressed
porosity profiles were obtained prior to carrying out the compression investigation with the aid
of a polycarbonate apparatus. A linear torque-compression correlation was employed to
determine the appropriate compression for each GDL. Polycarbonate was identified as a viable
material in order to maximize the absorption contrast between the MPL and the compression
device for ex-situ visualizations. A GDL surface isolating algorithm was developed in Matlab to
separate the GDL from the polymer surface of the device.
45
From the uncompressed scans, it was observed that the porosity of paper GDL was consistently
higher throughout its thickness when compared to the porosities of the felt and cloth GDLs. The
average porosities of uncompressed paper, felt, and cloth GDLs was 0.81, 0.79, and 0.71,
respectively. The uncompressed scans also showed that the effects of compression in a GDL are
not uniform throughout the thickness of the felt and paper GDLs. In paper and felt, the effect of
compression was greater on the surface regions than on the bulk regions of these GDLs. For the
cloth GDL, however, the effects of compression were uniform throughout its thickness. The
reduction in the thickness of the GDL at 250psi was highest for paper (50%), followed by cloth
(49%) and felt (17%). The maximum change in average porosity under the landing at a
compression of 250psi was observed in cloth (37%), followed by felt (14%) and paper (9%).
Similarly, the maximum change in average porosity under the channel at a compression of 250
psi was also observed in cloth (36%), followed by felt (11%), and paper (1%). In addition, in felt
and cloth GDLs, the compressed porosity distributions under the landing and channel tracked
each other, i.e. the average porosity under the landing is within 3% of the average porosity under
the channel. However, for the paper GDL, the average porosity under the channel is 8% higher
than the average porosity under the landing.
In the felt GDL, upon applying a compression of 175psi, there was a 13% decrease in the
thickness. Subsequent increases in applied compression yielded further decreases in GDL
thickness of 3% for 250psi, and 5% for 325psi. However, the average porosity under land and
channel remained within 1% of the average porosity at 175 psi. This suggests that felt may be
compressed with the benefit of enhanced electrical conductivity without compromising on the
oxygen diffusivity, which can potentially lead to increased performance of the PEMFC.
46
5.5 Figures and Tables
Table 5.1: Effect of rib/channel compression on GDL porosity.
GDL
Microstructure
GDL
Name
Average
Uncompressed
Porosity
(U)
Average Compressed
Porosity (250psi) % Change
Land
(L)
Channel
(C)
(C-
L)/L (L-U)/U (C-U)/U
Paper AvCarb
GDS 3250 0.81 0.74 0.80 8 -9 -1
Felt GDL B 0.71 0.62 0.63 3 -14 -11
Cloth Dura-GDL
ST400TC 0.79 0.50 0.51 3 -37 -36
47
Table 5.2: Effect of rib/channel compression on GDL thickness.
GDL
Microstructure
GDL
Name
Uncompressed
Thickness
(μm)
(UT)
Compressed Thickness
(μm) (CT)
% Change at
250psi
175psi 250psi 325psi (UT-CT)/UT
Paper AvCarb
GDS 3250 305 N/A 153 N/A 50
Felt GDL B 276 239 228 222 17
Cloth Dura-GDL
ST400TC 415 N/A 210 N/A 49
48
Figure 5.1: A schematic of the (a) uncompressed and (b) compressed GDL sample holders for
μCT.
Uncompressed Sample Holder
GDL
GDL
Base Plate
Gasket
Compression Plate
b) a)
49
Figure 5.2: Compression calibration results: (a) pressure distribution on a Prescale pressure film
after applying a torque of 10 in-oz in the compression device. (b) average pressure calibration
curves as a function of applied torque for GDL B. The calibration was repeated twice to ensure
repeatability.
y = 34.988x R² = 0.8509
y = 32.296x R² = 0.8285
0
50
100
150
200
250
300
350
400
0 5 10 15
Avera
ge P
ressure
(psi)
Torque Applied (in-oz)
Round 1Round 2Linear (Round 1)Linear (Round 2)
b) a)
Avera
ge
Pre
ssure
(psi)
50
Figure 5.3: Performing edge detection on μCT images: (a) A portion of a typical through-plane
μCT slice obtained after reconstruction of GDL B compressed between the flow field and the
compression plate of the compression device. (b) The output of 4(a) after applying Canny’s edge
detection technique. (c) Surface isolation based on the rate of change in the surface profile. The
black compounded line is the surface profile of the GDL, while the red line denotes the rate of
change in the surface profile. Refer 5.2.2 for a discussion on GDL surface isolation.
-0.02
-0.02
-0.01
-0.01
0.00
0.01
0.01
0.02
0.02
0.00
0.05
0.10
0.15
0.20
0.25
0 50 100 150 200 250 300 350 400 450 500
Rate
of
ch
an
ge in
th
e S
urf
acfe
Pro
file
Ed
ge I
nte
nsit
y P
rofi
le
Through-plane Position, X (m)
a) b)
c)
Thro
ugh-p
lane P
ositio
n,
X
GDL Surface
GDL Surface
Compression Plate
Base Plate
GDL
51
Figure 5.4: Through-plane uncompressed porosity profiles of investigated GDLs.
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0 50 100 150 200 250 300 350 400
Po
rosit
y
GDL Through-plane Position (m)
AvCarb GDS3250 (Paper)
GDL B (Felt)
Dura-GDL ST400TC (Cloth)FELTMPL
52
Figure 5.5: Through-plane porosity distributions of AvCarb GDS3250.
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0 50 100 150 200 250 300 350 400
Poro
sity
GDL Through-plane Position (m)
Average Uncompressed
Average under Land
Average under Channel
Uncompressed Compressed under Landing Compressed under Channel
53
Figure 5.6: Effect of incremental compression: (a) Through-plane porosity distributions of GDL
B. (b) Figure that illustrates the effect of incremental compression on the thickness of the GDL,
and the average porosity under land and channel.
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0 50 100 150 200 250 300 350 400
Poro
sity
GDL Through-plane Position (mm)
Average UncompressedAverage under Land (175psi)Average under Channel (175psi)Average under Land (250psi)Average under Channel (250psi)Average under Land (325psi)Average under Channel (325psi)
0
60
120
180
240
300
0.50
0.60
0.70
0.80
0.90
1.00
175 250 325
Co
mp
ressed
GD
L T
hic
kn
ess (m
m)
Avera
ge P
oro
sity
Appiled Compression (psi)
Land Porosity
Channel Porosity
Thickness
b)
a)
Compressed Porosity under a Landing
Compressed Porosity under a Channel
Compressed Thickness
Uncompressed Compressed under Landing (175psi) Compressed under Channel (175psi) Compressed under Landing (250psi) Compressed under Channel (250psi) Compressed under Landing (325psi) Compressed under Channel (325psi)
54
Figure 5.7: Through-plane porosity distributions of Dura-GDL ST400TC.
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0 50 100 150 200 250 300 350 400
Poro
sity
GDL Through-plane Position (m)
Average Uncompressed
Average under Land (250psi)
Average under Channel (250psi)
Uncompressed Compressed under Landing Compressed under Channel
55
6.0 Conclusions & Future Work
6.1 Water distribution visualization
A critical literature review of water distribution visualizations of GDLs using X-ray radiography
revealed a gap in the experimental work addressing the influence of a GDL microstructure and
current density on liquid water transport in the GDL. Additionally, the assumptions regarding the
size of the inlet while modeling water transport through a GDL needed to be verified as well.
Accordingly in this thesis, a systematic technique was outlined to obtain 2D water thickness
distributions and 1D through-plane water thickness distributions for commercial and non-
commercial GDLs with varying microstructures using synchrotron radiography.
In Chapter 3 the design of an ex situ water injection setup to simulate the compression and water
generation occurring in an operating PEMFC was reported. Water was injected into GDLs after
flooding the inlet face of the sample holder to investigate the effect of GDL microstructure and
current density on commercially available GDLs. The liquid water content at breakthrough for
five commercially available GDLs (Freudenberg H2315, Freudenberg H2315 I6, Freudenberg
H2315 I3 C1, Toray TGP-H-090, and AvCarb 1071), with varying microstructures and surface
treatments, was presented. Freudenberg GDLs exhibited greater average through-plane water
content than Toray TGP-H-090 and AvCarb 1071 GDLs at breakthrough. For all the GDLs, the
average through-plane water content at 1µL/min was higher than the average through-plane
water content at 8µL/min. Among the Fruedenberg GDLs, Freudenberg H2315 had the lowest
average through-plane water content. A larger amount of water also accumulated at the inlet face
of Freudenberg H2315 I6 (1.89mm) when compared to Freudenberg H2315 (1.00mm) and
Freudenberg H2315 I3 C1 (1.22mm). This was the first ever experimental comparison of liquid
water distributions in commercially available GDLs.
Subsequently, in Chapter 4, the development of a new water injection device to successfully
inject water into a GDL from a single point was reported. The objective again was to investigate
the influences of current density and GDL microstructure on liquid water distribution in the GDL
using synchrotron radiography. A technique was also developed to measure the applied
compression on the GDL. The liquid water contents at breakthrough for three structurally distinct
PTFE-treated, MPL-coated GDLs were presented, including: GDL A (paper), Freudenberg
56
H2315 I3 C1 (felt), and DURA-GDL ST400TC (cloth). Although the average through-plane
water content at a given flow rate was generally constant for all GDLs investigated here, the
lateral spread in the in-plane direction was the highest in the felt and paper GDLs implying
greater pore connectivity in their pore networks, compared to the cloth. For a given flow rate, the
variations observed in average water contents are expected, given the random nature of the GDL
as well as the heterogeneous MPL-coating. Current densities of 0.8A/cm2 and 1.6A/cm
2 were
simulated using flow rates of 1μL/min and 2μL/min, respectively. A definitive trend could not be
identified between the average water content in a GDL at an injection rate of 1µL/min and the
average water content in the GDL invaded at a flow rate of 2µL/min. The overall average water
content reported in Chapter 4 was significantly lower than the overall average water content
reported in Chapter 3. This variation in the water content was due to the change in the size of the
inlet, which is in agreement with modelling work by Hinebaugh et al. [89].
Although the experiments reported in this thesis were conducted at room temperature, they
provide valuable data that can inform the choice of a GDL microstructure. However, the addition
of thermal gradients across the GDL will help simulate realistic fuel cell operation. The results
presented in Chapters 3 and 4 were obtained after imaging GDLs using the absorption contrast
imaging technique. In the future, phase contrast imaging can be investigated for its applicability
to image water distributions in GDLs. Also in the future, water should be injected from multiple
injection points to better represent the conditions in a fuel cell. Finally, the phase diagrams found
in the literature used to characterize the flow through the GDLs were based on 2D structured
pore network models. Phase diagrams based on GDL materials should be used to determine
operating regimes.
6.2 Microstructural Investigations
A critical literature review of microstructural investigations on GDLs revealed a scarcity of
experimental investigations addressing the effect of compression on the through-plane porosity
distributions, which are necessary for accurately modelling multiphase flow through a PEMFC
GDL. In Chapter 5, the development of a non-destructive technique to characterize the effect of
compression on paper, felt, and cloth GDLs under the influence of a channel and rib
configuration was reported. Uncompressed porosity profiles were initially obtained by imaging
an uncompressed GDL. The same GDL sample was then compressed and imaged. Polycarbonate
57
was identified as viable material when looking to maximize the absorption contrast between
itself and MPL for ex situ visualizations. A GDL surface isolating algorithm was developed in
Matlab to distinguish the GDL from the polymer.
From the uncompressed GDLs, it was observed that the porosity of the paper GDL was
consistently higher throughout its thickness when compared to the felt and the cloth GDLs. The
average porosity of uncompressed paper, felt, and cloth GDLs was 0.81, 0.79, and 0.71,
respectively. From the compressed GDLs, it was observed that the effects of compression were
not uniform throughout the thickness of the felt and paper GDLs. In paper and felt, it was also
observed that the effect of compression was greater on the surface regions than on the bulk
regions of these GDLs. For the cloth GDL, however, the effects of compression were uniform
throughout the thickness of the GDL. Also, in felt and cloth GDLs, the compressed porosity
distributions under the landing and channel followed the same trend i.e., the average porosity
under the landing is within 3% of the average porosity under the channel. However, for the paper
GDL, the average porosity under the channel is 8% higher than the average porosity under the
landing.
Upon applying a compression of 175psi on the felt a 13% decrease in the thickness was
observed, followed by a 5% decrease for an additional 75psi, and a further 3% for subsequent
compression by another 75psi. The average porosity under land and channel, however, was
within 1% of the average porosity at 175 psi. This suggests that felt GDL may be compressed
with the added benefit of enhanced electrical conductivity without compromising the oxygen
diffusivity, which can potentially lead to increased performance of the PEMFC. The results
presented in Chapter 5 will provide the required insights for modelling the effect of compression
on spatially varying porosity of a GDL, and can be used for stochastic generation of GDLs to
model liquid water distributions in PEMFC GDLs.
It was also found that the porosity determinations in Chapter 5 were sensitive to thresholding.
Though Otsu’s technique has been utilized in the literature for binarizing GDL tomograms, the
method should be validated. To validate this thresholding method, the calculated porosity should
be compared to the porosity of the same GDL sample obtained from another non-destructive
technique such as gas pyncometry. Additionally, the intrusion of the GDL into the channel was
58
not fully studied in Chapter 5. The accurate description of GDL channel intrusion may be
informative to modellers studying the effect of GDL compression on PEMFC performance.
59
References
[1] P. Beckhaus, M. Dokupil, A. Heinzel, S. Souzani, C. Spitta, J.Power Sources, 145 (2005)
639-643.
[2] N. Demirdoven, J. Deutch, Science, 127 (2004) 8-15.
[3] Y. Wang, K.S. Chen, J. Mishler, S.C. Cho, X.C. Adroher, Appl.Energy, 88 (2011) 981-1007.
[4] C.K. Dyer, Journal of Power Sources, 44 (2003) 83.
[5] F. Urbani, G. Squadrito, O. Barbera, G. Giacoppo, E. Passalacqua, O. Zerbinati, J.Power
Sources, 169 (2007) 334-337.
[6] J. Larminie, A. Dicks, Fuel cell systems explained, 2nd ed., J. Wiley, New York; 2003, pp.
406.
[7] M. Mathias, J. Roth, J. Fleming, W.L. Lehnert, Diffusion media materials and
characterisation, In: Vielstich W, Lamm A, Gasteiger HA, (Eds), Handbook of fuel cells :
fundamentals, technology, and applications, Wiley, Chichester, England ; Hoboken, N.J., 2003,
pp. 1-21.
[8] G. Lin, T. Van Nguyen, J.Electrochem.Soc., 152 (2005) A1942-A1948.
[9] L. Cindrella, A.M. Kannan, J.F. Lin, K. Saminathan, Y. Ho, C.W. Lin, J. Wertz, J.Power
Sources, 194 (2009) 146-160.
[10] S. Park, B.N. Popov, Fuel, 88 (2009) 2068-2073.
[11] S. Park, S. Kim, Y. Park, M. Oh, Journal of Physics: Conference Series, 165 (2009) 012046.
[12] S. Escribano, J. Blachot, J. Ethève, A. Morin, R. Mosdale, J.Power Sources, 156 (2006) 8-
13.
[13] K.G. Gallagher, R.M. Darling, T.W. Patterson, M.L. Perry, J.Electrochem.Soc., 155 (2008)
B1225-B1231.
60
[14] R. Flückiger, S.A. Freunberger, D. Kramer, A. Wokaun, G.G. Scherer, F.N. Büchi,
Electrochim.Acta, 54 (2008) 551-559.
[15] Y. Chou, Z. Siao, Y. Chen, L. Sung, W. Yang, C. Wang, J.Power Sources, 195 (2010) 536-
540.
[16] J.T. Gostick, M.W. Fowler, M.D. Pritzker, M.A. Ioannidis, L.M. Behra, J.Power Sources,
162 (2006) 228-238.
[17] J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, Electrochemistry
Communications, 10 (2008) 1520-1523.
[18] J.T. Gostick, M.W. Fowler, M.A. Ioannidis, M.D. Pritzker, Y.M. Volfkovich, A. Sakars,
J.Power Sources, 156 (2006) 375-387.
[19] A.A. Adedeji, M. Ngadi, International Journal of Food Science and Technology, 46 (2011)
1266-1275.
[20] D. Mandal, D. Sen, S. Mazumder, M.R.K. Shenoi, S. Ramnathan, D. Sathiyamoorthy,
Ceramic Engineering and Science Proceedings, 32 (2011) 165-170.
[21] V. Berejnov, D. Sinton, N. Djilali, J.Power Sources, 195 (2010) 1936-1939.
[22] B. Gao, T.S. Steenhuis, Y. Zevi, J.Y. Parlange, R.N. Carter, T.A. Trabold, J.Power Sources,
190 (2009) 493-498.
[23] F.N. Büchi, R. Flückiger, D. Tehlar, F. Marone, M. Stampanoni, ECS Trans., 16 (2008)
587-592.
[24] J. Becker, V.P. Schulz, A. Wiegmann, Journal of Fuel Cell Science and Technology, 5
(2008) 021006.
[25] J. Becker, R. Flückiger, M. Reum, F.N. Büchi, F. Marone, M. Stampanoni,
J.Electrochem.Soc., 156 (2009) B1175-B1181.
[26] Z. Fishman, A. Bazylak, J.Electrochem.Soc., 158 (2011) B846-B851.
61
[27] Z. Fishman, A. Bazylak, J.Electrochem.Soc., 158 (2011) B841-B845.
[28] Z. Fishman, J. Hinebaugh, A. Bazylak, J.Electrochem.Soc., 157 (2010) B1643-B1650.
[29] J. James, MASc. Dissertation, Queen's University, Canada, (2012).
[30] H. Jhong, F.R. Brushett, L. Yin, D.M. Stevenson, P.J.A. Kenis, J.Electrochem.Soc., 159
(2012) B292-B298.
[31] P. Krüger, H. Markötter, J. Haußmann, M. Klages, T. Arlt, J. Banhart, C. Hartnig, I. Manke,
J. Scholta, J.Power Sources, 196 (2011) 5250-5255.
[32] H. Ostadi, P. Rama, Y. Liu, R. Chen, X.X. Zhang, K. Jiang, Chemical Engineering Science,
65 (2010) 2213-2217.
[33] H. Ostadi, P. Rama, Y. Liu, R. Chen, X.X. Zhang, K. Jiang, J.Membr.Sci., 351 (2010) 69-
74.
[34] H. Ostadi, P. Rama, Y. Liu, R. Chen, X. Zhang, K. Jiang, Microelectron Eng, 87 (2010)
1640-1642.
[35] R. Flückiger, F. Marone, M. Stampanoni, A. Wokaun, F.N. Büchi, Electrochim.Acta, 56
(2011) 2254-2262.
[36] J. Eller, T. Rosen, F. Marone, M. Stampanoni, A. Wokaun, F.N. Buchi, J. Electrochem.
Soc., 158 (2011) B963-B970.
[37] Z. Fishman, A. Bazylak, J.Electrochem.Soc., 158 (2011) B247-B252.
[38] J. Hinebaugh, Z. Fishman, A. Bazylak, J.Electrochem.Soc., 157 (2010) B1651-B1657.
[39] J. Yablecki, A. Bazylak, J.Power Sources, 217 (2012) 470-478.
[40] A. Bazylak, D. Sinton, Z.-. Liu, N. Djilali, J.Power Sources, 163 (2007) 784-792.
[41] P. Yi, L. Peng, X. Lai, J. Ni, Journal of Fuel Cell Science and Technology, 8 (2011) 011011.
[42] W. Lee, C. Ho, J.W. Van Zee, M. Murthy, J.Power Sources, 84 (1999) 45-51.
62
[43] J. Ge, A. Higier, H. Liu, J.Power Sources, 159 (2006) 922-927.
[44] Z. Shi, X. Wang, L. Guessous, Journal of Fuel Cell Science and Technology, 7 (2010)
021012.
[45] J. Tan, Y.J. Chao, W. Lee, J.W. Van Zee, Journal of Pressure Equipment and Systems, 5
(2007) 1-7.
[46] P. Zhou, C.W. Wu, J.Power Sources, 170 (2007) 93-100.
[47] P. Zhou, C.W. Wu, G.J. Ma, J.Power Sources, 163 (2007) 874-881.
[48] R. Roshandel, B. Farhanieh, E. Saievar-Iranizad, Renewable Energy, 30 (2005) 1557-1572.
[49] Y. Wang, K.S. Chen, J.Electrochem.Soc., 158 (2011) B1292-B1299.
[50] A. Bazylak, Int J Hydrogen Energy, 34 (2009) 3845-3857.
[51] T. Sasabe, P. Deevanhxay, S. Tsushima, S. Hirai, J.Power Sources, 196 (2011) 8197-8206.
[52] T. Sasabe, S. Tsushima, S. Hirai, Int J Hydrogen Energy, 35 (2010) 11119-11128.
[53] P. Krüger, H. Markötter, M. Klages, J. Haußmann, T. Arlt, H. Riesemeier, C. Hartnig, J.
Banhart, I. Manke, J. Scholta, Materials Testing-Materials and Components Technology and
Application, 52 (2010) 712-717.
[54] H. Markötter, I. Manke, P. Krüger, T. Arlt, J. Haußmann, M. Klages, H. Riesemeier, C.
Hartnig, J. Scholta, J. Banhart, Electrochem.Commun., 13 (2011) 1001-1004.
[55] P.K. Sinha, P. Halleck, C. Wang, Electrochemical and Solid State Letters, 9 (2006) A344-
A348.
[56] C. Hartnig, I. Manke, Measurement methods | Structural Properties: Neutron and
Synchrotron imaging for in-situ water visualization, In: Editor-in-Chief: Jürgen Garche, (Ed),
Encyclopedia of Electrochemical Power Sources, Elsevier, Amsterdam, 2009, pp. 738-757.
[57] I. Manke, C. Hartnig, M. Grunerbel, J. Kaczerowski, W. Lehnert, N. Kardjilov, A. Hilger, J.
Banhart, W. Treimer, M. Strobl, Appl.Phys.Lett., 90 (2007).
63
[58] Y.S. Chen, H. Peng, D.S. Hussey, D.L. Jacobson, D.T. Tran, T. Abdel-Baset, M. Biernacki,
J.Power Sources, 170 (2007) 376-386.
[59] M.C. Hatzell, A. Turhan, S. Kim, D.S. Hussey, D.L. Jacobson, M.M. Mench,
J.Electrochem.Soc., 158 (2011) B717-B726.
[60] M.A. Hickner, N.P. Siegel, K.S. Chen, D.S. Hussey, D.L. Jacobson, M. Arif,
J.Electrochem.Soc., 155 (2008) B294-B302.
[61] D.S. Hussey, D.L. Jacobson, Journal of Fuel Cell Science and Technology, 7 (2010).
[62] T. Kim, J. Kim, C. Sim, S. Lee, M. Kaviany, S. Son, M. Kim, Nuclear Instruments and
Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated
Equipment, 600 (2009) 325-327.
[63] D. Kramer, J. Zhang, R. Shimoi, E. Lehmann, A. Wokaun, K. Shinohara, G.G. Scherer,
Electrochim.Acta, 50 (2005) 2603-2614.
[64] R. Mukundan, J.R. Davey, T. Rockward, J.S. Spendelow, B. Pivovar, D.S. Hussey, D.L.
Jacobson, M. Arif, R. Borup, ECS Trans., 11 (2007) 411-422.
[65] N. Pekula, K. Heller, P.A. Chuang, A. Turhan, M.M. Mench, J.S. Brenizer, K. Ünlü,
Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers,
Detectors and Associated Equipment, 542 (2005) 134-141.
[66] P. Quan, M.-. Lai, D.S. Hussey, D.L. Jacobson, A. Kumar, S. Hirano, Journal of Fuel Cell
Science and Technology, 7 (2010) 051009.
[67] R. Satija, D.L. Jacobson, M. Arif, S.A. Werner, J.Power Sources, 129 (2004) 238-245.
[68] R. Satija, D.L. Jacobson, M. Arif, S.A. Werner, J.Power Sources, 129 (2004) 238-245.
[69] J.G. Pharoah, K. Karan, W. Sun, J.Power Sources, 161 (2006) 214-224.
[70] H. Tang, A. Santamaria, J.W. Park, C. Lee, W. Hwang, J.Power Sources, 196 (2011) 9373-
9381.
64
[71] T.A. Trabold, J.P. Owejan, D.L. Jacobson, M. Arif, P.R. Huffman, Int.J.Heat Mass Transfer,
49 (2006) 4712-4720.
[72] A. Turhan, K. Heller, J.S. Brenizer, M.M. Mench, J.Power Sources, 160 (2006) 1195-1203.
[73] J. Zhang, D. Kramer, R. Shimoi, Y. Ono, E. Lehmann, A. Wokaun, K. Shinohara, G.G.
Scherer, Electrochim.Acta, 51 (2006) 2715-2727.
[74] C. Hartnig, I. Manke, R. Kuhn, N. Kardjilov, J. Banhart, W. Lehnert, Appl.Phys.Lett., 92
(2008).
[75] J. Hinebaugh, P.R. Challa, A. Bazylak, J. Synchrotron Rad., (Accepted on September 13,
2012; Submitted on May 31, 2012).
[76] R. Kuhn, J. Scholta, P. Krüger, C. Hartnig, W. Lehnert, T. Arlt, I. Manke, J.Power Sources,
196 (2011) 5231-5239.
[77] S. Lee, S. Kim, G. Park, C. Kim, Int J Hydrogen Energy, 35 (2010) 10457-10463.
[78] I. Manke, C. Hartnig, M. Gruenerbel, W. Lehnert, N. Kardjilov, A. Haibel, A. Hilger, J.
Banhart, H. Riesemeier, Appl.Phys.Lett., 90 (2007) 174105-174105.
[79] H. Markoetter, I. Manke, C. Hartnig, P. Krueger, K. Wippermann, T. Arlt, G. Choinka, H.
Riesemeier, J. Banhart, Materials Testing, 52 (2010) 698-704.
[80] T. Mukaide, S. Mogi, J. Yamamoto, A. Morita, S. Koji, K. Takada, K. Uesugi, K. Kajiwara,
T. Noma, J. Synchrotron Rad., 15 (2008) 329.
[81] A. Schneider, C. Wieser, J. Roth, L. Helfen, J.Power Sources, 195 (2010) 6349-6355.
[82] C. Hartnig, I. Manke, R. Kuhn, S. Kleinau, J. Goebbels, J. Banhart, J.Power Sources, 188
(2009) 468-474.
[83] P. Challa, J. Hinebaugh, A. Bazylak, ASME Conf. Proc., 2011 (2011) 121-126.
[84] P.K. Sinha, C. Wang, Electrochim.Acta, 52 (2007) 7936-7945.
[85] G. Maggio, V. Recupero, L. Pino, J.Power Sources, 101 (2001) 275-286.
65
[86] J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, J.Power Sources, 173 (2007)
277-290.
[87] A. Rofaiel, J.S. Ellis, P.R. Challa, A. Bazylak, J.Power Sources, 201 (2012) 219-225.
[88] P.P. Mukherjee, C.Y. Wang, Q.J. Kang, Electrochim Acta, 54 (2009) 6861-6875.
[89] J. Hinebaugh, A. Bazylak, Proceedings of ASME 2012 6th International Conference on
Energy Sustainability & 10th Fuel Cell Science, Engineering and Technology Conference,
(2012).
[90] J.H. Hubbell, S.M. Seltzer, Tables of X-ray mass attenuation coefficients and mass energy-
absorption coefficients, 1.4th ed., National Institute of Standards and Technology, Gaithersburg,
MD; 2004.
[91] I. Sobel, G.:. Feldman, A 3x3 isotropic gradient operator for image processing, 1973, pp.
271-272.
[92] J.M.S. Prewitt, Object enhancement and extractionIn: Lipkin B, Rosenfeld A, (Eds), Picture
Processing and Psychopictorics, Academic, New York, 1970, pp. 75-149.
[93] J. Canny, Pattern Analysis and Machine Intelligence, IEEE Transactions on, PAMI-8 (1986)
679-698.
[94] R. Lenormand, E. Touboul, C. Zarcone, J. Fluid Mech., 189 (1988) 165-187.
[95] M. Rebai, M. Prat, J.Power Sources, 192 (2009) 534-543.
[96] O. Chapuis, M. Prat, M. Quintard, E. Chane-Kane, O. Guillot, N. Mayer, J.Power Sources,
178 (2008) 258-268.
[97] V. Gurau, F. Barbir, H. Liu, J.Electrochem.Soc., 147 (2000) 2468-2477.
[98] N. Otsu, IEEE Trans.Syst.Man Cybern., SMC-9 (1979) 62-66.