Ifl

AF3

a

ARR2AA

KPFNWP

1

tohlitamrfltcap

ica

w

0d

Electrochimica Acta 55 (2010) 2734–2745

Contents lists available at ScienceDirect

Electrochimica Acta

journa l homepage: www.e lsev ier .com/ locate /e lec tac ta

mpact of channel wall hydrophobicity on through-plane water distribution andooding behavior in a polymer electrolyte fuel cell

hmet Turhan, Soowhan Kim1, Marta Hatzell, Matthew M. Mench ∗

uel Cell Dynamics and Diagnostics Laboratory, Department of Mechanical and Nuclear Engineering, The Pennsylvania State University,27 Reber Building, University Park, PA 16802, United States

r t i c l e i n f o

rticle history:eceived 2 September 2009eceived in revised form5 November 2009ccepted 28 November 2009vailable online 16 December 2009

a b s t r a c t

Through-plane liquid accumulation, distribution and transport inside polymer electrolyte fuel cell (PEFC)components were analyzed as a function of channel wall hydrophobicity with the use of high-resolutionneutron imaging. Neutron images were taken with polytetrafluoroethylene (PTFE) coated and uncoatedflow channel walls. Anode to cathode liquid distribution was analyzed for each case at low and highcurrent conditions over 20 min of operation. The form and amount of liquid water inside the channels

eywords:olymer electrolyte fuel celllow channeleutron imagingater storage

and diffusion media (DM) were compared for hydrophobically coated channels and hydrophilic channels,and a primary liquid transport-flooding mechanism is suggested for each case. The location and valueof maximum water storage in DM at low and high current operation were analyzed and slopes of watermass versus distance curve were calculated to compare the significance of capillary liquid flow and phase-change-induced flow within the diffusion media. A significant effect of CL|MPL and MPL|DM interfaceson liquid transport and flooding is found through the analysis of micro-porous layer (MPL) water content

ng th

hase change and saturation profile alo

. Introduction

In polymer electrolyte fuel cells (PEFCs), liquid water accumula-ion, transport and removal are critical, due to the significant impactn fuel cell operation and durability [1–3]. Experimental studiesave shown that flooding inside a fuel cell may exist in the catalyst

ayer (CL), diffusion media (DM), flow channels or along the CL|DMnterface [4–10]. In order to estimate the level of flooding in each ofhese components, the liquid transport inside the fuel cell at oper-ting and after shutdown conditions has also been investigated byany modelers [11–25]. Pasaogullari et al. [14] modeled the satu-

ation distribution in DM based on a capillary pressure dominatedow regime, and proposed a tree-like structure for liquid waterransport in DM. Gurau et al. [16] included catalyst layer and gashannels in their multi-phase transport model and estimated themount of water in DM and catalyst layer based on pore size andermeability.

Although most available two-phase transport models are usefuln terms of estimating the water saturation and removal in fuel cellomponents, the use of certain assumptions are questionable suchs isothermal conditions or capillary flow only inside DM pores,

∗ Corresponding author. Tel.: +1 814 865 0060; fax: +1 814 863 4848.E-mail addresses: mmm124@psu.edu, mmench@psu.edu (M.M. Mench).

1 Current address: Energy and Efficiency Division/Energy Materials, Pacific North-est National Laboratory, United States.

013-4686/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.oi:10.1016/j.electacta.2009.11.095

e CL|MPL and MPL|DM interface region.© 2009 Elsevier Ltd. All rights reserved.

which may not be suitable for real operating conditions. Exper-imental visualization techniques such as direct imaging with aconventional charge coupled device (CCD) camera, magnetic reso-nance imaging [26–28] and X-ray tomography [29–31] can be usedto measure the real-time liquid water accumulation in a speciallydesigned experimental fuel cell. A recent excellent comprehensivereview of the visualization techniques in fuel cells is given by Bazy-lak [32]. The only technique that is capable to measure the nearreal-time liquid water accumulation in an unmodified fuel cell isneutron radiography (NR). NR has captured the attention of manyresearchers [33–47]. Hickner et al. [37] showed a decreasing waterstorage in the cell at high current conditions due to high evapora-tion rates and increased vapor capacity of the gas flow even thoughthere is more water production in the cell at higher current. Turhanet al. [38] also found a similar behavior in their work. Owejan etal. [39] tested rectangular and triangular cross section, serpentineflow field designs and found that smaller droplets were formed, andstored water amount in the cell is lowered with triangular cross sec-tion. Hussey et al. [41] increased the pixel resolution limit of NR to∼16 �m by using a micro-channel plate based detector. Hickner etal. [42] analyzed cross-sectional water profile using this techniqueand found significant liquid water residue in the anode gas chan-

nels. At current densities above 0.75 A/cm2, with anode and cathodeflow rates set to stoichiometry of two, the residual water in theanode gas channels was found to be 1.2–1.8 times higher than thecathode gas channels, indicating anode liquid accumulation couldbe a problem at low anode flow rates. Weber et al. [43] developed

mica A

atafcwccti

ofhlcodtawbfhamwb

A. Turhan et al. / Electrochi

computational model based on these results and estimated thehrough-plane saturation in fuel cell components at different oper-ting conditions. In their high-resolution testing, Boillat et al. [44]ound that water tends to accumulate under the ribs more than thehannels for under-humidified cathode flow, and this is reversedhen cathode flow is fully-humidified. All these studies had major

ontributions to the understanding of water management in fuelells, however, the flooding mechanism and through-plane liquidransport in terms of capillary flow and phase-change-induced flows not yet completely understood.

The liquid accumulation inside the fuel cell also might dependn surface hydrophobicity of the channel walls. There are aew studies in literature investigating flooding behavior withydrophobic channel surface [39,48,49]. Owejan et al. [39] ana-

yzed the liquid distribution in flow channels with and without PTFEoating using in-plane neutron radiographs and found large slugsf water inside the channels without PTFE coating and discreteroplets with PTFE coating. Zhu et al. [48] used microcomputedomography to look in droplet formation in hydrophobic channelsnd found that droplets did not detach from DM before removalhereas in hydrophilic channels, a thin water layer formed at the

ottom of the channel away from the DM. Bazylak et al. [49] per-ormed experiments on a model fuel cell to determine effects of

ydrophobic land surface on droplet removal and found that withhydrophobic land surface, droplets experience minimal entrap-ent in the DM|land interface. The findings of all these studiesere important in terms of understanding the effect of hydropho-

ic surface on channel level liquid accumulation but they did not

Fig. 1. Schematic of test cell and im

cta 55 (2010) 2734–2745 2735

describe the through-plane liquid transport from DM to the chan-nels due to either in-plane neutron imaging resolution limitationsor lack of capturing actual fuel cell operation. In this study, thethrough-plane water distribution and liquid removal were inves-tigated in an actual operating fuel cell for different flow channelsurface properties, using high-resolution neutron radiography.

2. Experimental setup

The neutron imaging facility at the National Institute of Stan-dard and Technology (NIST) was used to perform through-planeneutron radiography tests. The details of the test facility were givenin Hussey et al. [41]. Each neutron image has 54 s of total integra-tion time and 15.4 �m × 15.4 �m pixel resolution. An alignmentprocedure was followed prior to each experiment. The fuel cell wasrotated by increments of 0.1◦ and an image was taken at each orien-tation. Then the total thickness of anode DM-MEA-cathode DM wasmeasured from each neutron image and all the measured thicknessvalues were plotted with respect to the rotation angle. The mini-mum of the resulting parabola is selected as the correct alignmentposition which enabled an accurate representation of the MPL|DMand DM|channel interfaces.

The fuel cell used in all of the tests has an active area of 17.2 cm2

and was machined out of 6061 aluminum with a gold plated, sin-gle serpentine design flow channels having 1 mm width and 1 mmdepth. The width of the landings was also 1 mm. The cell was ori-ented vertically, with a co-flow design where gases flow from topto bottom direction. A Gore Primea 5710 MEA, with 0.4 mg/cm2 Pt

aging area used in all tests.

2736 A. Turhan et al. / Electrochimica Acta 55 (2010) 2734–2745

F ophilih

lupw

escapdrectatep

3

3d

c

ig. 2. Neutron images of tests with 0.2 A/cm2 operation and (a) anode channel hydrydrophilic.

oading on each electrode, and two SGL 10BB diffusion media weresed in all experiments at a cell temperature of 65 ◦C, and a cellressure of 1 atm. Anode and cathode flows were fully humidifiedith stoichiometric ratios of 1.5 and 2.0, respectively.

Two different channel wall surface conditions were used in thexperiments. In the first set of experiments, an imaging area con-isting of 4 channels and 3 landings at the bottom of flow fieldenterline, as shown in Fig. 1, was selected and a PTFE coating waspplied to the cathode flow channel walls only. The uncoated gold-lated walls are naturally hydrophilic. The through-plane wateristribution data were obtained for 0.2 A/cm2 and 1.0 A/cm2 cur-ent density conditions at various times. For the second set ofxperiments, a PTFE coating was applied only to the anode flowhannel walls of an imaging area of 4 channels and 3 landings athe top of flow field centerline. The neutron images were takent the same current conditions as the first set of data to comparehe effect of PTFE coating on channel wall surface. For both sets ofxperiments, the fuel cell was purged with dry nitrogen for 5 minrior to applying current, to ensure similar initial conditions.

. Results and discussion

.1. Effect of channel wall surface hydrophobicity on liquid wateristribution

The neutron images of hydrophilic-anode/PTFE-coated-athode and PTFE-coated-anode/hydrophilic-cathode channel

c/cathode channel PTFE coated and (b) anode channel PTFE coated/cathode channel

walls at 0.2 A/cm2 operation are shown in Fig. 2, to illustratethe water build-up mechanism in the flow channels based onsurface treatment. In all of the images, liquid water thickness ateach pixel is represented by different colors (color online) wherewater thickness increases from dark blue to red, as shown inthe color scale. In PTFE coated channels, it is observed that theliquid water formed as droplets on the side walls, whereas onthe uncoated channels, water spreads from DM to the entire wallsurface forming a liquid film layer. This observation is consistentwith the planar neutron imaging results of Owejan et al. [39] wherethey found large slugs of water without PTFE coating and discretedroplets with PTFE coating. To better understand the characteristicthrough-plane water profile with and without PTFE coating, thethrough-plane distribution of liquid water is plotted with respectto time in Fig. 3. In Fig. 3a, the water content inside the anodechannel peaks around the through-plane position of zero, whichcorresponds to anode channel back wall. This indicates a thin filmformation formed at this location, which is consistent since theanode channel is uncoated. Note that initially the liquid mass atthe anode back wall is at a minimum due to low liquid generation(0.2 A/cm2) and significant liquid condensation inside the cathode

DM. As liquid builds-up in the cathode DM with time, the backdiffusion of liquid into anode also increases, as seen from thefigure, and a peak in liquid water mass at anode channel back walloccurs. For the cathode channel, peak water content is observednear the center of cathode channel, at a through-plane position

A. Turhan et al. / Electrochimica Acta 55 (2010) 2734–2745 2737

Fac

oPwcPwcmicafl

amb0cu2aam1u

ig. 3. Through-plane liquid water distribution for 0.2 A/cm2 operation with (a)node channel hydrophilic/cathode channel PTFE coated and (b) anode channel PTFEoated/cathode channel hydrophilic.

f ∼2.1 mm, a result of droplet formation on the side walls due toTFE coated channels. A similar trend is also observed in Fig. 3b,here water mass peaks around center for the coated anode

hannel and near the end for the uncoated cathode channel. TheTFE coating also has an effect on water removal from the channelsith respect to time. On the hydrophobically coated cathode

hannel, the accumulated droplets on the walls are purged at 11thinute of operation whereas, without the coating, no sudden drop

n channel water content is observed during testing. The PTFEoating on the walls seems to retain more water in the channels onverage, but at the same time, the sporadic droplet purge is morerequent due to discrete droplet formation rather than liquid filmayer.

The total normalized water mass in the cathode side for 0.2nd 1.0 A/cm2 cases are shown in Fig. 4. The total liquid waterass measured in each component, channels and DM, is divided

y the total volume of that component to normalize the results. At.2 A/cm2, the total water mass on average inside the PTFE-coated-athode channels were 100% greater compared to water in thencoated cathode channels at the same condition over the entire0 min of operation. As current density increased to 1.0 A/cm2, thisverage difference dropped to 70%. When the transient behavior is

nalyzed, the water mass values in the PTFE coated channels showore frequent fluctuation, especially at low current density after

0 min of operation, indicating a high frequency of water build-p and sudden removal compared to uncoated channels. When

Fig. 4. At every minute of operation, total liquid water mass values inside the cath-ode DM and the cathode channel are plotted for (a) 0.2 A/cm2 and (b) 1.0 A/cm2.

the total liquid inside the cathode DM is considered, at the initialminute of operation, the coated cathode channel case has ∼150%more DM water for 0.2 A/cm2 and ∼100% more for 1.0 A/cm2. Thisinitial difference decreases with time and at the end of 20 min oper-ation, for 0.2 A/cm2, there was only ∼7% more water in the cathodeDM with the hydrophobically coated channel walls. At 1.0 A/cm2,this difference was found to be ∼15%, with the coated channelhaving more liquid water. These total water mass results suggestthat surface energy of channel walls effects the water accumula-tion, back diffusion and removal. This will have an impact on nettransport coefficient which should be included in water transportmodels. The effect of surface energy of the channel walls seems tobe more significant at high current operation, however, the smallerimpact at low current is still important especially at idling con-ditions of cold-start operation. The higher DM water content inhydrophobically coated channels may not be beneficial during ini-tial start-up at sub-zero temperatures.

The through-plane neutron imaging results also reveal signif-icant evidence that water discharge mechanism from DM in tothe flow channels is coupled to the land|channel interface, andits hydrophilic–hydrophobic nature might have important impli-cations on performance loss due to DM flooding. In Fig. 5 neutron

images of both hydrophobically coated and uncoated cathode chan-nel conditions are shown for 0.2 A/cm2, from the start of operationto the point where significant channel water accumulation wasobserved.

2738 A. Turhan et al. / Electrochimica Acta 55 (2010) 2734–2745

F s at 0.a

aw[mm(iiua9wotDptwiaa

ig. 5. Neutron images of water build-up in DM and water discharge into channelnd (b) anode channel PTFE coated/cathode channel hydrophilic.

For the first two images in both Fig. 5a and b, the liquid waterccumulation was observed mainly in the DM, under the landings,hich was expected based on previous neutron imaging studies

50,51]. Then, until significant liquid accumulates in the cell (8thinute in Fig. 5a and 18th minute in Fig. 5b), increasing liquid waterass under the landings moves in two directions; (1) in vertical

through-plane) direction such that from DM-under-the-landingnto the flow channel through the DM-channel wall intersection, (2)n lateral (in-plane) direction within the DM such that from DM-nder-the-landing to DM-under-the channel location. The waterccumulation under the landing and removal of water through the0◦ interface is common for both cases. The major difference is,ith a hydrophobic surface water tends to form large droplets

n the wall, which relatively inhibits the suction of water intohe channels. This enhances the lateral movement of water insideM, and can lead to a connectivity under the channels which canromote flooding losses. In Fig. 5a, the DM-under-the-channel loca-

ions of the hydrophobic land|channel interface filled with liquidater much faster compared to hydrophilic land|channel interface

n Fig. 5b. Also, after significant water built-up in the channels,s shown in the last images of both Fig. 5a and b, more liquidccumulation was observed in DM-under-the-channel locations

2 A/cm2 operation for (a) anode channel hydrophilic/cathode channel PTFE coated

with hydrophobic channels. Both of these observations suggestthat with the hydrophobically coated channels, more connectiv-ity of liquid in DM-under-the-channel locations occurs comparedto uncoated channels. This also results in more water storage insideDM and a peak water mass near the DM|channel interface. Thethrough-plane liquid distributions in Figs. 6 and 7 for both currentconditions and channel PTFE coatings show this difference moreclearly.

The hydrophobic coating on the walls slightly increased peakwater mass in 0.2 A/cm2, as shown in Fig. 6, and significantlyincreased (∼15%) peak water mass in 1.0 A/cm2, as shown in Fig. 7.The location of cathode DM peak water mass is also differentbetween the two cases. In the 1.0 A/cm2 operation of uncoatedcathode case, the peak water mass was observed at 0.524 mmwhereas in the coated channel case, the peak shifted towardsDM|channel interface, to a distance of 0.554 mm. In the anodeside, it was also observed (Fig. 7b) that the hydrophobically coated

anode channel walls caused a peak in DM water storage near theDM|channel interface (∼0.04 mm) having ∼30% more water stor-age than hydrophilic channel case, as shown in Fig. 7a. These resultsindicate the land|channel interface is critical in removal of waterfrom DM. It is suggested that the hydrophilic interface enhances

A. Turhan et al. / Electrochimica Acta 55 (2010) 2734–2745 2739

F(P

twdrnc

ltfriswfflipbdaftHti

ig. 6. Through-plane liquid water distribution inside DM and MEA at 0.2 A/cm2 fora) anode channel hydrophilic/cathode channel PTFE coated and (b) anode channelTFE coated/cathode channel hydrophilic.

he capillary suction from DM and results in lower water storagehereas, the PTFE coated interface suppresses the liquid in DMue to the water repellent nature of the surface. Based on theseesults, a schematic of water accumulation and removal mecha-ism is suggested for both hydrophobic and hydrophilic channelases and shown in Fig. 8.

At this point, it is important to note that for the discharge ofiquid water from diffusion media into the flow channel, many ofhe studies consider droplet emergence on diffusion media sur-ace as the main discharge mode, without consideration of theole of the channel|land interface. In our results, however, theres almost no droplet accumulation observed directly on the DMurface. Instead, liquid directly accumulated on the channel wallshich strongly indicates that continuous discharge of liquid water

rom DM occurs mainly from under-the-landing location into theow channel through the land|channel interface. This does not

mply that droplet emergence directly on to DM surface is notossible. In fact, there are studies in literature that visualized thisehavior [52,53]. However, our results imply that the liquid waterischarged from DM-under-the-channel location is easily remov-ble, and may not be accumulated on the DM surface long enough

or detection by the 54 s exposure of the neutron image, comparedo liquid discharge onto channel walls from land|channel interface.ence, it can be effective to engineer the water storage in the cell by

ailoring the geometry and the surface energy of the land|channelnterface.

Fig. 7. Through-plane liquid water distribution inside DM and MEA at 1.0 A/cm2 for(a) anode channel hydrophilic/cathode channel PTFE coated and (b) anode channelPTFE coated/cathode channel hydrophilic.

3.2. Generic through-plane liquid distribution and significance ofcapillary versus phase-change induced flow inside the DM

In this section, the common trends observed in the through-plane water distribution for both of the channel wall surface energyconditions are analyzed to understand the generic water transportmechanisms inside the DM. The through-plane water mass pro-files of only the hydrophobically coated cathode tests are shownfor the sake of simplicity, however, the same trends are also validfor the hydrophobically coated anode case. The liquid distributionof 0.2 A/cm2 and 1.0 A/cm2 are represented in two different for-mats. In the first format, normalized liquid water mass in the entirefuel cell (total water mass) and in each of the fuel cell components(flow channel, DM and MEA) are plotted over 20 min of operation,starting at open circuit voltage condition. Note that the total liquidwater mass values in the entire fuel cell are shown for unit surfacearea of the cell. This format enables comparison of total water massbetween components at transient and steady conditions, as well asat low and high current. In the second format, water mass valuesat each pixel location along the through-plane direction are shownto analyze the water transport mechanisms inside DM and MPL.

In both Fig. 9a and b, a significant difference in liquid water

accumulation between anode and cathode DM was observed atconstant current operation. At low current density (0.2 A/cm2) thewater mass per unit cathode DM volume was 84% higher than thatof anode DM during the first minute of operation. After steady-statewas reached, this difference increased to 120%. For the 1.0 A/cm2

2740 A. Turhan et al. / Electrochimica Acta 55 (2010) 2734–2745

F ncoatel

cttaccacowacswttw

ig. 8. Schematic of liquid water transport and removal with coated channels and uiquid build-up and (d) real-time neutron images for each case.

ondition, however, cathode DM water was only 86% higher thanhat of anode DM at steady-state. Based on this observation, a 1:1.8o 1:2.2 for the anode to cathode DM liquid water ratio may begood range to assume for basic calculations depending on the

urrent condition. At 1.0 A/cm2, the liquid water built-up in thehannels was removed by the gas flow, and the water mass fluctu-ted only ∼10% from an average (steady-state) value. In 0.2 A/cm2

ase, however, the first water removal was observed after 10 minf operation and even at 20 min of operation, a true steady-stateas not observed. A similar trend in steady-state behavior was

lso observed in MEA and DM water content between the twoonditions. At 1.0 A/cm2, the steady-state is reached in DM water

2

torage within 4 min of operation, whereas at 0.2 A/cm , 10 minas required for steady-state. The five times greater water genera-

ion in 1.0 A/cm2 case as well as the higher channel shear explainshis difference, but in order to better compare the differences inater mass accumulation between two current conditions, the per-

d channels at (a) start-up, (b) after couple minutes of operation, (c) after significant

cent increase of water mass in each component from 0.2 A/cm2 to1.0 A/cm2 is shown in Fig. 10. Here, a negative value representsa lower water amount in 1.0 A/cm2 case compared to 0.2 A/cm2.The first difference was observed in channel water accumulationbetween the two conditions. There was consistently more chan-nel water in 0.2 A/cm2 throughout the entire operation, except thefirst couple minutes. This is consistent with observed low-powerinstability in stacks, and is a result of low channel shear flow. Thelow stoichiometric flow rate, and therefore decreased channel flowvelocity is the main reason of the low channel shear at this current.The average channel water accumulation at 0.2 A/cm2 was found tobe 57% higher in the cathode side and 47% higher in the anode side

compared to high current density operation. It is also important tonote that there was significant anode channel water in both currentconditions.

For the DM and MEA water content, the commonly expectedresult would be a greater water mass accumulation at high current

A. Turhan et al. / Electrochimica Acta 55 (2010) 2734–2745 2741

Fc0

dt1c

F0h

ig. 9. At every minute of operation, the total liquid water mass values in the cellomponents for anode channel hydrophilic/cathode channel PTFE coated case at (a).2 A/cm2 and (b) 1.0 A/cm2.

ue to more water generation. This behavior is seen in Fig. 10 forhe first 10 min of operation where steady-state was reached for.0 A/cm2 but not for 0.2 A/cm2 case. After 10 min, where bothonditions reached steady-state, there was ∼20% more water in

ig. 10. The difference of liquid water mass values in each cell component between.2 A/cm2 and 1.0 A/cm2 at every minute of operation for the anode channelydrophilic/cathode channel PTFE coated case.

Fig. 11. The liquid water mass distribution with respect to through-plane distance at(a) 0.2 A/cm2 and (b) 1.0 A/cm2 for the anode channel hydrophilic/cathode channelPTFE coated case.

anode DM and ∼10% more water in MEA at 1.0 A/cm2 compared to0.2 A/cm2. On the other hand, the water accumulation differencein cathode DM was almost zero (±3% fluctuation). In other words,even though there is five times more water generation at 1.0 A/cm2

condition, the water storage at cathode DM is almost same as0.2 A/cm2, while the anode DM water does increase.

In order to understand this interesting result, the liquid watermass distribution with respect to the through-plane position for0.2 and 1.0 A/cm2 current density conditions were shown in Fig. 11.In the graph, initial value of x-axis corresponds to anode DM|flowchannel interface and final data point corresponds to cathodeDM|flow channel interface. The water mass distribution is similarin both conditions, with a peak value near the middle of cathodeDM. At 0.2 A/cm2 condition, increasing liquid accumulation in cath-ode DM from 77 mg/cm3 to 115 mg/cm3 for the first 10 min can beclearly seen. After this point, the water mass values do not fluctuate,indicating a pseudo steady-state is reached. For the 1.0 A/cm2, thereis almost no fluctuation in the water mass values and steady-stateis reached in the first few minutes. The peak water mass value in theanode DM for 1.0 A/cm2 is ∼20% greater than that of 0.2 A/cm2 case,as expected due to higher water generation. However, the graphclearly shows that, at steady-state, peak water mass values in cath-

ode DM are the same (115 mg/cm3) for both current conditions,despite the significant difference in water generation.

This result might indicate an important water transport mech-anism inside the DM. It is shown in literature that liquid water

2 mica A

tdD[cttl

742 A. Turhan et al. / Electrochi

ransports from hot to cold side by vapor diffusion and con-ensation in the existence of a temperature gradient across theM; a transport mode named as phase-change-induced (PCI) flow

54–56]. It is also known that a temperature gradient occurs in theathode DM during operation due to heat generation at the reac-ion sites. The through-plane results presented in Fig. 11 show thathere is water transport within DM via PCI-flow. For a solely capil-ary flow, the distribution should be monotonically decreasing from

Fig. 12. (a) Isolated holes and connected cracks on MPL surface [62] a

cta 55 (2010) 2734–2745

CL|DM to DM|channel interface following a Darcy law relationshipfor one-dimensional flow in porous media, as shown in Eq. (1).

( )

nx

w = Kw(s)�w

∂Pc

∂s

∂s

∂x(1)

where nxw is the liquid flux, Kw(s) is the permeability depending on

liquid saturation (s), �w is the liquid viscosity, and capillary pres-

nd (b) liquid water mass peaks around MPL|DM interface [63].

mica A

sMbvlflclDwhtfitagicctt

tcflTttflf0ttau1ncdtlattiufsuoaitmsdb

auiIMmm

connected and elongated cracks at the interface. Significant liquidflow may occur inside the connected cracks via capillary action aswell as condensation of water vapor may cause liquid accumulationinside the isolated holes. Both of these cases will result in significant

A. Turhan et al. / Electrochi

ure (Pc) is defined in terms of saturation via the Leverett function.any two-phase models assuming isothermal conditions show this

ehavior in their saturation plots where condensation of waterapor is neglected. However, the existence of a source term foriquid water inside DM, which is the vapor condensation due to PCIow, is seen in Fig. 11 with a peak of liquid water mass near theenter of DM where temperature is lower than the cathode cata-yst layer. The balance of PCI flow compared to capillary action inM, however, is not known. Considering only PCI flow, the peakater mass value should be higher in the 1.0 A/cm2 case due toigher water generation (such that more water source is availableo condense), and its location should be closer to DM|channel inter-ace since the temperature gradient across cathode DM is highern 1.0 A/cm2 compared to 0.2 A/cm2. Our results, however, showhat both the location and the magnitude of the peak water massre same for both current densities. Based on this, it may be sug-ested that once a certain level of liquid water storage (saturation)s reached in the cathode DM due to condensation, the rate ofapillary liquid back flow and flow into the channels is dominantompared to water transport due to PCI flow. Under these condi-ions, even though there is more water generated in catalyst layer,he stored liquid water content in DM reaches a plateau.

The slope of water mass curve between the peak value andhe cathode DM|channel interface is calculated for both currentonditions, to better explain this mechanism. Note that the liquidux due to capillary flow is directly related to saturation gradient.hrough the same DM material, the saturation gradient is exactlyhe same as liquid water mass gradient; therefore, by calculatinghe slope of water mass curve, the magnitude of capillary liquidow from cathode DM into cathode flow channels can be compared

or both current conditions. The slopes are found to be 0.670 and.924 for 0.2 A/cm2 and 1.0 A/cm2 cases, respectively. This showshat the increased rate of capillary removal at 1.0 A/cm2 overcomeshe increased rate of PCI flow due to greater temperature gradientnd water generation at this condition. Therefore, almost same liq-id storage in cathode DM was observed for both 0.2 A/cm2 and.0 A/cm2. Based on this, it is suggested that the PCI flow is notormally the dominant water transport mechanism compared toapillary flow in the cathode DM of an operating fuel cell. The fun-amental physical reason behind this fact can be explained throughransition from pendular to funicular regime in DM water accumu-ation [56]. For a short duration after start-up, the water generatedt the CL accumulates in DM as isolated droplets, which is called ashe pendular regime. In this regime, the liquid moves in DM due tohe temperature gradient. As the fuel cell operates and more waters generated, water storage in DM exceeds a certain level of sat-ration and the isolated droplets get connected. This is called theenicular regime and in our case, this corresponds to the time whereteady-state is reached. Once the transition from pendular to fenic-lar regime is established, the capillary flow becomes dominantver phase-change induced flow and therefore, for both 0.2 A/cm2

nd 1.0 A/cm2, the peak water mass value and stored water contentn DM do not change. The only difference occurs in the time to reachhat transition, which is seen from Fig. 11. This result is in agree-

ent with model predictions of Khandelwal et al. [56] where theyhowed that magnitude of the temperature gradient through DMoes not affect the steady-state liquid distribution after shutdownut it greatly changes the time required to reach steady-state.

Finally, it is important to mention that there is a significantmount of water in the MPL, and the water mass curve is contin-ous through the MPL|DM and MPL|CL interfaces such that there

s no abrupt change or jump from DM to MPL or from MPL to CL.n two-phase flow models predicting the saturation curve through

PL and DM, an increase in saturation at MPL|DM interface is com-only suggested based on capillary pressure equilibrium, with auch higher saturation on the DM side. The MPL saturation is nor-

cta 55 (2010) 2734–2745 2743

mally predicted to 40–60% smaller than the DM saturation at theinterface depending on the MPL porosity. These saturation predic-tions can be converted into water mass values using the DM andMPL porosities. For SGL 10BB DM, the uncompressed porosity isgiven as 0.84 by the manufacturer [57], however the MPL porosityis not clearly stated. Gostick et al. [58] calculated the MPL poros-ity of SGL 10BB as 0.72 whereas Nam et al. [59] used a value of0.5. For both cases, the MPL porosity is smaller than DM poros-ity. Therefore, using the definition of saturation, it can be shownthat models predict much less water storage in the MPL comparedto water storage in the DM at the interface. In other words, thereshould be a jump in the water mass curve at the MPL|DM interfacebased on many capillary flow based model predictions. However,the continuous water mass curve in our results suggests this is notthe case, and in fact, there is another effect to be considered otherthan capillary pressure equilibrium between the pores of MPL andDM. This might be the effect of surface morphology at this inter-face. It is shown that there are significant number of cracks (up to5–7% of surface area) at MPL|CL and MPL|DM interface which havemuch larger size compared to pores in MPL and DM [56,60–62].As shown in Fig. 12a, these can be in the form of isolated holes or

Fig. 13. Through-plane liquid saturation inside DM at 0.2 A/cm2 current density for(a) MPL porosity of 0.72 and (b) MPL porosity of 0.5. In both cases, there is a reversejump at MPL|DM.

2 mica A

luMmsmtea

icceie0A(aebfsa

4

floshwwTlmfePrcuro

wwemialrslc

mbfMfd

[[

[[[[[

[[[[[[

[[[[

[[[

[

[

[[

[

[

[

[

[

[

[

[

[

[

744 A. Turhan et al. / Electrochi

iquid water mass through the MPL layer which will explain the liq-id water in MPL and continuous water curve through MPL|CL andPL|DM interfaces obtained from the actual through-plane liquidass profile. Other published evidence from Hartnig et al. [63], as

hown in Fig. 12b, also indicates the existence of high water accu-ulation next to the MPL that can only be reasonably explained by

hese interfacial effects. Therefore, interface surface morphologyffects also have to be considered in two-phase flow and floodingnalysis.

The models predicting the saturation curve uses the poros-ty values and capillary pressure equilibrium by considering eachomponent separately and do not consider their combined andompressed structure in the fuel cell, which involves the interfacialffects. The liquid water mass values shown in Fig. 11 are convertednto saturation values using the same approach where all interfaceffects are neglected. The MPL porosities were taken as 0.72 and.5, as mentioned earlier, and the results were plotted in Fig. 13.s opposite to model predictions, the water saturation decreases

a reverse jump) from MPL to DM since MPL have smaller poresnd lower porosity value compared to DM when they are consid-red separately. The same is valid for MPL|CL since the CL porosity isetween 0.4 and 0.6, which is lower than MPL porosity [64]. This dif-erence also shows the interfacial effects plays an important role inaturation distribution and may have important effect on floodingt these locations.

. Conclusions

In this study, the through-plane liquid storage, transport andooding mechanism inside a PEFC was investigated as a functionf channel wall hydrophobicity. The PTFE coating on the channelurface resulted in discrete droplets on the channel walls with aigher water removal frequency. The uncoated channels have lessater on average, and liquid forms a film layer around the wallshich promotes steadier operation, but is more difficult to purge.

he liquid removal from under the lands into the channels viaand|channel interface is found to be a significant water transport

echanism and the hydrophilic–hydrophobic nature of this inter-ace effects the water storage in DM. Hydrophilic channel wallsnhance the liquid suction from under-the-land locations whereasTFE coated land|channel interface suppress the liquid inside DMesulting more water storage in DM, as high as 15% at high currentondition. The PTFE coated interface also found to increase the liq-id connectivity in DM-under-the-channel location. Based on theseesults, a DM-liquid transport mechanism is suggested dependingn surface PTFE coating.

A highly non-linear behavior with a peak near the center of DMas found for the water mass distribution regardless of channelall surface energy, suggesting phase-change-induced (PCI) flow

xists in DM as a significant source term. The steady-state waterass storage inside anode DM was determined to increase with

ncreasing current; however, significant difference in water stor-ge was not observed inside the cathode DM. The peak water massocation in cathode DM also did not change with increasing cur-ent. Based on this and the slopes of the water mass curves, it isuggested that after a critical liquid saturation is reached in DM,iquid removal from the DM via capillary flow dominates phase-hange-induced flow.

Liquid water storage was observed in the MPL and the water

ass curve through MPL|CL and MPL|DM interfaces was found to

e continuous. The liquid saturation curves obtained using dif-erent MPL porosities suggest a reverse saturation jump at the

PL|DM interface which indicates MPL surface cracks and inter-ace morphology have an important impact on liquid storage andistribution.

[

[

[

[

cta 55 (2010) 2734–2745

Acknowledgements

The authors would like to thank to Professor Mike Hickner ofThe Pennsylvania State University, Daniel S. Hussey and David L.Jacobson of The National Institute of Standards and Technology fortheir contributions to this study. Partial support of this work wasgiven through National Science Foundation CAREER award CBET#0644811.

References

[1] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, J. Electrochem. Soc. 138 (8) (1991)2334.

[2] K.G. Gallagher, B.S. Pivovar, T.F. Fuller, J. Electrochem. Soc. 156 (3) (2009) B330.[3] Y. Devrim, S. Erkan, N. Bac, I. Eroglu, Int. J. Hydrogen Energy 34 (8) (2009) 3467.[4] H. Yamada, T. Hatanaka, H. Murata, Y. Morimoto, J. Electrochem. Soc. 153 (9)

(2006) A1748.[5] X. Li, I. Sabir, J. Park, J. Power Sources 163 (2) (2007) 933.[6] S. Litster, D. Sinton, N. Djilali, J. Power Sources 154 (1) (2006) 95.[7] E.C. Kumbur, K.V. Sharp, M.M. Mench, J. Electrochem. Soc. 154 (12) (2007)

B1295.[8] E.C. Kumbur, K.V. Sharp, M.M. Mench, J. Power Sources 168 (2) (2007) 356.[9] S. Shimpalee, U. Beuscher, J.W. Van Zee, Electrochim. Acta 52 (24) (2007) 6748.10] X. Liu, H. Guo, F. Ye, C.F. Ma, Electrochim. Acta 52 (11) (2007) 3607.11] W. Dai, H. Wang, X.Z. Yuan, J.J. Martin, Z. Luo, M. Pan, J. Power Sources 185 (2)

(2008) 1267.12] A.Z. Weber, J. Newman, J. Electrochem. Soc. 152 (4) (2005) A677.13] X. Wang, T.V. Nguyen, J. Electrochem. Soc. 155 (11) (2008) B1085.14] U. Pasaogullari, C.Y. Wang, J. Electrochem. Soc. 152 (2) (2005) A380.15] G. Lin, W. He, T.V. Nguyen, J. Electrochem. Soc. 151 (12) (2004) A1999.16] V. Gurau, J. Thomas, A. Zawodzinski, J.J. Adin Mann, J. Fuel Cell Sci. Technol. 5

(2) (2008) 021009.17] B. Carnes, N. Djilali, Electrochim. Acta 52 (3) (2006) 1038.18] S. Park, B.N. Popov, Electrochim. Acta 54 (12) (2009) 3473.19] R. Alink, D. Gerteisen, M. Oszcipok, J. Power Sources 182 (1) (2008) 175.20] S. He, J.H. Lee, M.M. Mench, J. Electrochem. Soc. 154 (12) (2007) B1227.21] S. He, S.H. Kim, M.M. Mench, J. Electrochem. Soc. 154 (10) (2007) B1024.22] C. Chacko, R. Ramasamy, S. Kim, M. Khandelwal, M. Mench, J. Electrochem. Soc.

155 (11) (2008) B1145.23] S. Kim, B.K. Ahn, M.M. Mench, J. Power Sources 179 (1) (2008) 140.24] S. Kim, M.M. Mench, J. Power Sources 174 (1) (2007) 206.25] K. Tajiri, C.Y. Wang, Y. Tabuchi, Electrochim. Acta 53 (22) (2008) 6337.26] K.R. Minard, V.V. Viswanathan, P.D. Majors, L.Q. Wang, P.C. Rieke, J. Power

Sources 161 (2) (2006) 856.27] S. Tsushima, T. Nanjo, K. Nishida, S. Hirai, ECS Trans. 1 (6) (2006) 199.28] Z.W. Dunbar, R.I. Masel, J. Power Sources 182 (1) (2008) 76.29] S. Griesser, G. Buchinger, T. Raab, D.P. Claassen, D. Meissner, J. Fuel Cell Sci.

Technol. 4 (1) (2007) 84.30] F.Y. Zhang, S.G. Advani, A.K. Prasad, M.E. Boggs, S.P. Sullivan, T.P. Beebe Jr.,

Electrochim. Acta 54 (16) (2009) 4025.31] S.J. Lee, N.Y. Lim, S. Kim, G.G. Park, C.S. Kim, J. Power Sources 185 (2) (2008)

867.32] A. Bazylak, Int. J. Hydrogen Energy 34 (9) (2009) 3845.33] N. Pekula, K. Heller, P.A. Chuang, A. Turhan, M.M. Mench, J.S. Brenizer, K. Ünlü,

Nucl. Instrum. Methods Phys. Res. Sect. A: Accelerators, Spectrometers, Detec-tors and Associated Equipment 542 (1–3) (2005) 134.

34] A.B. Geiger, A. Tsukada, E. Lehmann, P. Vontobel, A. Wokaun, G.G. Scherer, FuelCells 2 (2) (2002) 92.

35] R.J. Bellows, M.Y. Lin, M. Arif, A.K. Thompson, D. Jacobson, J. Electrochem. Soc.146 (3) (1999) 1099.

36] R. Satija, D.L. Jacobson, M. Arif, S.A. Werner, J. Power Sources 129 (2) (2004)238.

37] M.A. Hickner, N.P. Siegel, K.S. Chen, D.N. McBrayer, D.S. Hussey, D.L. Jacobson,M. Arif, J. Electrochem. Soc. 153 (5) (2006) A902.

38] A. Turhan, K. Heller, J.S. Brenizer, M.M. Mench, J. Power Sources 160 (2) (2006)1195.

39] J.P. Owejan, T.A. Trabold, D.L. Jacobson, M. Arif, S.G. Kandlikar, Int. J. HydrogenEnergy 32 (17) (2007) 4489.

40] T.A. Trabold, J.P. Owejan, D.L. Jacobson, M. Arif, P.R. Huffman, Int. J. Heat MassTransfer 49 (25–26) (2006) 4712.

41] D.S. Hussey, D.L. Jacobson, M. Arif, J.P. Owejan, J.J. Gagliardo, T.A. Trabold, J.Power Sources 172 (1) (2007) 225.

42] M.A. Hickner, N.P. Siegel, K.S. Chen, D.S. Hussey, D.L. Jacobson, M. Arif, J. Elec-trochem. Soc. 155 (4) (2008) B427.

43] A.Z. Weber, M.A. Hickner, Electrochim. Acta 53 (26) (2008) 7668.44] P. Boillat, D. Kramer, B.C. Seyfang, G. Frei, E. Lehmann, G.G. Scherer, A. Wokaun,

Y. Ichikawa, Y. Tasaki, K. Shinohara, Electrochem. Commun. 10 (4) (2008) 546.

45] J. Park, X. Li, D. Tran, T. Abdel-Baset, D.S. Hussey, D.L. Jacobson, M. Arif, Int. J.

Hydrogen Energy 33 (13) (2008) 3373.46] R. Fu, U. Pasaogullari, D.S. Hussey, D.L. Jacobson, M. Arif, ECS Trans. 11 (1) (2007)

395.47] D. Kramer, J. Zhang, R. Shimoi, E. Lehmann, A. Wokaun, K. Shinohara, G.G.

Scherer, Electrochim. Acta 50 (13) (2005) 2603.

mica A

[[[

[

[[

[[[[

[

[

[

[

A. Turhan et al. / Electrochi

48] W. Zhu, Z.W. Dunbar, R.I. Masel, ECS Trans. 16 (2) (2008) 995.49] A. Bazylak, J. Heinrich, N. Djilali, D. Sinton, J. Power Sources 185 (2) (2008) 1147.50] 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 (1) (2007) 411.51] A. Turhan, K. Heller, J.S. Brenizer, M.M. Mench, J. Power Sources 180 (2) (2008)

773.52] F.Y. Zhang, X.G. Yang, C.Y. Wang, J. Electrochem. Soc. 153 (2) (2006) A225.53] B. Gao, T.S. Steenhuis, Y. Zevi, J.Y. Parlange, R.N. Carter, T.A. Trabold, J. Power

Sources 190 (2) (2009) 493.54] A.Z. Weber, J. Newman, J. Electrochem. Soc. 153 (12) (2006) A2205.55] S. Kim, M.M. Mench, J. Electrochem. Soc. 156 (3) (2009) B353.56] M. Khandelwal, S. Lee, M.M. Mench, J. Electrochem. Soc. 156 (6) (2009) B703.57] SIGRACET® diffusion media manufacturer data sheet. Available: http://www.

sglcarbon.com/sgl t/fuelcell/index.html.

[

[

[

cta 55 (2010) 2734–2745 2745

58] J.T. Gostick, M.W. Fowler, M.A. Ioannidis, M.D. Pritzker, Y.M. Volfkovich, A.Sakars, J. Power Sources 156 (2) (2006) 375.

59] J.H. Nam, K.J. Lee, G.S. Hwang, C.J. Kim, M. Kaviany, Int. J. Heat Mass Transfer 52(11–12) (2009) 2779.

60] H. Bajpai, M. Khandelwal, E.C. Kumbur, M.M. Mench, ECS Transactions 25 (1)(2009) 3–13.

61] T. Swamy, F.E. Hizir, M. Khandelwal, E.C. Kumbur, M.M. Mench, ECS Transac-tions 25 (1) (2009) 15–27.

62] F.E. Hizir, T. Swamy, S.O. Ural, E.C. Kumbur, M.M. Mench, ASME Proceedings ofthe 7th International Conference on Fuel Cell Science Engineering and Tech-nology, ASME, (2009) 85092.

63] C. Hartnig, I. Manke, R. Kuhn, N. Kardjilov, J. Banhart, W. Lehnert, Appl. Phys.Lett. 92 (13) (2008) 134106.

64] M.M. Mench, Fuel Cell Engines, 1st ed., John Wiley & Sons, New York, 2008.