oscillations of pem fuel cells at low cathode humidification

Journal of Electroanalytical Chemistry 649 (2010) 219–231

Contents lists available at ScienceDirect

Journal of Electroanalytical Chemistry

journal homepage: www.elsevier .com/locate / je lechem

Oscillations of PEM fuel cells at low cathode humidification

Daniel G. Sanchez a, Domingo Guinea Diaz a, Renate Hiesgen c, Ines Wehl c, K. Andreas Friedrich b,*

a Spanish National Research Council (CSIC), Industrial Automation Institute (IAI), Carretera de Campo Real, km 200, La Poveda, Arganda del Rey, Madrid E-28500, Spainb Deutsches Zentrum für Luft und Raumfahrt (DLR) Institut für Technische Thermodynamik, Pfaffenwaldring 38-40, 70569 Stuttgart, Germanyc Faculty of Basic Sciences, University of Applied Sciences Esslingen, Kanalstr. 33, 73728 Esslingen, Germany

a r t i c l e i n f o

Article history:Received 16 December 2009Received in revised form 29 March 2010Accepted 6 April 2010Available online 10 April 2010

Dedicated to Prof. Jacek Lipkowski inrecognition of his achievements and on theoccasion of his 65th birthday

Keywords:Polymer electrolyteFuel cellWater managementOscillations

1572-6657/$ - see front matter � 2010 Elsevier B.V. Adoi:10.1016/j.jelechem.2010.04.005

* Corresponding author. Tel.: +49 711 6862278; faxE-mail address: [email protected] (K.A. Frie

a b s t r a c t

Oscillatory fluctuations of a single polymer electrolyte fuel cell appear upon operation with a dry cathodeair supply and a fully humidified anode stream. Periodic transitions between a low and a high currentoperation point of the oscillating state are observed. The transition time of 20–25 s for the change fromthe low to the high operation is fast and does not depend on the operating parameters. Contrasting withthis behavior, the downward transition depends strongly on the operating conditions. Impedance mea-surements indicate a high ionic resistance with low water content for the low current operation and alow ionic resistance of the membrane with high water content for the high current operation. An insightinto the transitions is obtained by current density distributions at distinct times indicating a propagatingactive area with defined boundaries. The observations are in agreement with assuming a liquid water res-ervoir at the anode with a downward transition period depending on the operation conditions. The highcurrent operation possesses a high electro-osmotic drag and a high permeation rate (corresponding toliquid–vapor permeation) leading to a large water flux to the cathode. Subsequently, the liquid reservoirat the anode is consumed leading to an anode drying. The system establishes a new quasi-stable opera-tion point associated with a low current, low electro-osmotic drag coefficient, and a low water perme-ation (corresponding to vapor–vapor permeation). When liquid water is formed at the anode interfaceafter some time the fast transition to the high current operation occurs. This interpretation is supportedby conductive atomic force microscopy current images of the membrane showing a strong dependence ofthe ionic conductivity on the activation procedures with or without liquid water and also showing oscil-latory behavior after the membrane is activated. Specifically, activation with liquid water yields a highconductivity with currents larger by three orders of magnitude.

� 2010 Elsevier B.V. All rights reserved.

1. Introduction

Polymer electrolyte fuel cells (PEFC) are expected to be an inter-esting alternative as power supply in portable, automotive and sta-tionary applications because of their high power density and lowenvironmental impact. Proper water management is of paramountimportance for PEFCs since performance losses are known to resultdue to cathode flooding, gas dilution or membrane dehydration. Asa consequence, water management is one of the most important is-sues for successful operation, high performance and durability ofPEFC. Excess of humidified inlet gases produces an accumulationof liquid water in the porous electrodes and gas diffusion media,thus decreasing the electrochemical activity and performance (ef-fect known as flooding). On the other hand, an insufficient levelof humidification lowers the ionic conductivity in the membraneand results also in a performance reduction. From the system pointof view the necessity to ascertain a well-humidified membrane and

ll rights reserved.

: +49 711 68621278.drich).

concurrently to avoid condensation of liquid water adds tosignificant complexity. In general, dedicated components forhumidification of the gases and an intricate control system are re-quired. The complex water management also limits the dynamicresponse of the system as fast changes of humidity are to beavoided. In particular the swelling of the membrane and corre-sponding dimensional changes are problematic in this respect.Therefore, numerous studies have investigated the operation ofPEFC under dry conditions to simplify operation. Early work todemonstrate stable performance for PEFC using dry or slightlyhumidified gas has been reported by Büchi and Srinivasan [1],and Watanabe et al. [2]. They showed that cell performance canbe achieved without humidifying the gas streams through use ofproduct water produced by the electrochemical cell reaction, orin the case of Watanabe et al., by direct humidification of the mem-brane by wicking from an auxiliary water supply. Strategies foroperating polymer electrolyte fuel cells include also the reductionof humidification of both reactant gases [3–6] or the dry operationof cathode [7–9] or anode sides [10]. The studies are not onlyconcerned with demonstrating stable operation under reduced

http://dx.doi.org/10.1016/j.jelechem.2010.04.005

mailto:[email protected]

http://www.sciencedirect.com/science/journal/15726657

http://www.elsevier.com/locate/jelechem

220 D.G. Sanchez et al. / Journal of Electroanalytical Chemistry 649 (2010) 219–231

humidification conditions. A major focus lies also in the determina-tion of water transport processes, namely the water transport coef-ficient and the electro-osmotic drag coefficient through themembrane. This is the aim of the work of Colinart et al. that sum-marizes the literature on water management and in addition deter-mines the net water transport properties by operating the fuel cellwith completely dry anode feeds [10]. With this approach it is pos-sible to derive the water transport from cathode to anode as a func-tion of operation parameters like flow rates, degree of cathodehumidification, and temperature of cells. Furthermore, dedicatedexperiments have been performed to determine under full humid-ification the electro-osmotic drag coefficient, providing experi-mental values from 0.3 up to 5 [11–15]. In the literature thevalue b is often used which is the ratio of the net water flux to pro-ton flux. If b is positive the direction of net water flux is to the cath-ode and if negative the net flux is to the anode. For Nafion� 105under wet, dry and differential pressure conditions Janssen andOvervelde observed that b is positive for almost all operating con-ditions and only negative when the anode is dry [16].

An understanding of the operating membrane conductivity dur-ing fuel cell operation is one of the endeavoring tasks in fuel cellresearch as several processes affect the water level of the compo-nents. From a technical point of view, particularly dry air feed isdesirable as the amount of water needed for the humidificationof air dominates which is commonly provided with high stoichi-ometry. In consequence, more studies have looked into the possi-bility of dry operation on the fuel cell cathode side [7–9].

In general the following mechanisms for water transportincluding two-phase phenomena are important and have to beconsidered in polymer electrolyte fuel cells and are schematicallyillustrated in Fig. 1:

� Diffusion of water vapor from humid to dryer regions.� Electro-osmotic drag from the anode to the cathode.� Hydraulic permeation due to pressure gradients.� Capillary flow of liquid water within the gas diffusion layer

(GDL).� Water droplet formation and detachment at the GDL-gas chan-

nel interface.� Liquid film flow on the channel walls.

Important for the discussion of the results obtained here is thework by Adachi et al. that has studied the permeation through Naf-ion� NRE 211 proton exchange membranes [15]. This work studiedthe difference of the fluxes for liquid–liquid permeation (LLP),liquid–vapor permeation (LVP), and vapor–vapor permeation(VVP) for dispersion cast, thin NRE 211 membrane at 70 �C. From

Fig. 1. Schematic overview of wate

an analysis of water fluxes and the corresponding chemicalpotential gradients across the membrane, the study concludes thatthe water flux through this thin Nafion� membrane is largest whenthe membrane is exposed to liquid on one side and vapor on theother. It is found that LVP water transport is largely responsiblefor regulating water balance within the operating MEA. This leadsto a conclusion that formation of a chemical potential gradient forwater and good hydration at the interface including the formationof liquid water on one side of the membrane is important for theefficient operation of the membrane electrode assembly (MEA).As discussed later the formation of LVP at our single cell is animportant requirement for inducing a transition in the system.

The study of the current oscillations under low cathode hydra-tion has been the focus of several research groups during the lastyears, with the aim at understanding the variables that controlthe phenomenon and its origin [3,17–19]. Studied oscillations inPEFC can result from either poisoning effects of the catalysts [19]or complex water management [3,17,18] which is of special inter-est here. Benziger et al. studied auto-humidified stirred tank reac-tor PEFC and found a sudden increase of current when themembrane water content is greater than a critical level of �1.6H2O/SO3; and in addition reported that the high performing oper-ation is ‘‘extinguished” when the initial membrane water contentis below this critical level [17,18]. These complex dynamics ofPEM fuel cell operation are associated with the membrane wateruptake. The authors hypothesized that water produced leads toswelling of the membrane thereby altering the interfacial mem-brane–electrode contact. Atkins et al. investigated PEFC at lowerhumidity levels and found strong periodic oscillations in currentand cell resistance as the feed stream humidification is decreased[3]. A cyclic mechanism was proposed that may offer a qualitativeexplanation for these unusual oscillations. In particular, a positivefeedback between proton conduction in the membrane and waterproduction from the fuel cell reaction has been studied.

This contribution investigates the oscillatory fluctuations ob-served when operating a single polymer electrolyte fuel cell withdry air. This oscillation state consists of two distinct operation per-formances of the cell which are investigated as a function of oper-ation conditions recording the local current density distributions.Using impedance spectroscopy it is clearly demonstrated that thedistinct cell performances are associated with varying membraneconductivities and water content. We discuss in this contributionthe possible sequence of the processes leading to the oscillatoryfluctuation. The transition time from low to high performance ismostly operation independent; in contrast to the transition fromhigh to low current which depends strongly on the experimentalconditions. Although degradation effects associated with cell oper-

r transport processes in PEFC.

D.G. Sanchez et al. / Journal of Electroanalytical Chemistry 649 (2010) 219–231 221

ation under these unconventional operating conditions have beenobserved these are not the focus of this contribution. Here wemainly describe the experimental findings for explaining the pro-cesses leading to oscillations. Quantitative understanding of thecomplex non-linear dynamics has to be based on modeling takinginto account two-phase flow. The modeling work is in progress andwill be presented in a forthcoming paper. Since membrane interfa-cial behavior is important for the cell behavior we include conduc-tive AFM measurements on the Nafion� membrane surface todemonstrate the importance of the presence of liquid water forthe membrane conductivity. The measurements are performed un-der comparable conditions with one dry side and a wet anode. Inthis case also current oscillations can be observed. In addition,the importance of liquid water for activating a high conductivityof a Nafion� membrane is shown. The work presented here is ofsignificance since various techniques including local resolved char-acterization are used in combination to study the complex watermanagement behavior. The determined sensitivity of the mem-brane|electrode interface to the presence of liquid water is impor-tant for general polymer fuel cell operation.

2. Experimental details

The investigation of the dynamic response of polymer electro-lyte fuel cells at reduced temperatures was performed on singlecells with an electrode area of 25 cm2 at the testing stands ofDLR (German Aerospace Center). The test bench is home-madeand is controlled by programmable logic controllers (PLC), whichallow an automatic control of the input and output conditions,such as pressure, temperature, flow rate of gases, humidity of reac-tants. The relative humidity of the inlet gases was controlledthrough water-filled heated sparger bottles. Lines between humid-ifiers and the fuel cell were heated in order to avoid water conden-sation. The gas mass flow rates were regulated at the fuel cell inletswhereas the pressure was controlled at the outlets. The pressurewas fixed for the experiment at constant value of 1500 mbar. Theflow of gases (air and H2) is controlled through the test stationand can be varied between 0 and 500 ml min�1 on the anode sideand between 0 and 2000 ml min�1 on the cathode side. The refer-ence operating conditions are summarized in Table 1.

In this study a dry operation of the cell and the resulting non-linear response is of special interest. These testing stands havethe possibility to by-pass the bubblers and introduce dry gassesinto the cell. We use the normal way of defining the relativehumidity (RH). A relative humidity of ca. 0% is applied to the cath-ode side which is accomplished by introducing ambient gas usingthe bubbler by-pass. In order to assure that liquid water does notenter the cell which could introduce disturbances in the experi-ments, pre-heaters were implemented. These pre-heaters are tanksthan can be heated to guarantee dry gases at the cell inlet. Com-mercial electronic loads are used in the test stands and for the

Table 1Reference operating conditions in this study.

Cell temperature T = 70 �CTemperature of anode bubbler 80 �C, nominally RH = 152%, varying in

measurements in Figs. 2b and 6Pressure P = 1.5 barHydrogen stoichiometric ratio kH2 = 1.4, Varying in measurements

in Figs. 10 and 11Air stoichiometric ratio kair = 2, Varying in measurements

in Figs. 10 and 11Humidification Dry air, hydrogen humidified

at 100% RH or under condensingconditions

Flow rates Varying

measurements here the potentiostatic mode is used meaning thatthe cell voltage is controlled and the corresponding current canchange with time.

A commercial MEA was used in this study produced by IonPower Inc. company (Nafion� and catalyst coated membrane).The active area of the cell is 25 cm2 (5 � 5 cm) and the membraneelectrode assembly (MEA) used is a commercial Nafion�111-IPwith anode and cathode Pt loading of 0.3 mgPt cm�2. As GDL aSGL Group Sigracet� 35 BC is used for all measurements.

Electrochemical impedance spectroscopy (EIS), recorded withZahner instruments, is used to obtain a more detailed understand-ing of the processes inside a fuel cell. It allows correlating losses inperformance to individual components, namely the loss contribu-tions of the anode, cathode, or membrane. Especially, it can be usedto analyze the resistance of membrane to detect the drying out ofthe membrane (flooding does not affect significantly the mem-brane resistance). A restriction of EIS is that it can be applied tothe cell only under quasi-stationary conditions. Therefore in ourcase, the frequency had to be limited to the high-frequency rangemeasuring frequencies >500 Hz since the transient behavior ofthe cell is investigated.

For the AFM measurements a Nanoscope III (Digital Instru-ments/Veeco, Santa Barbara, CA) equipped with a fluid cell and avoltage source was placed into an environmental chamber withconstant humidity and temperature. All AFM experiments wereperformed on a single side coated Nafion� 112 membrane. Beforeelectrode preparation the membranes have been fully protonated.The electrode consists of commercial Pt catalyst (1.0 mg cm�2), ap-plied by a dry layer spraying preparation method at the GermanAerospace Center (DLR, Stuttgart) [20]. This back electrode servedas anode. Measurements on the uncoated membrane surface areobtained by moving the Pt-coated AFM probe ((NanoWorld�,25 nm platinum layer) in contact mode. It served as a counter elec-trode (cathode) necessary for conductivity measurements and as aprobe of surface morphology. Details of the set-up were given pre-viously [21–23]. The contact force between the AFM tip and thesample was in the range of 10 nN. The atmosphere was humidifiedambient air on both sides. The back electrode is soaked with waterbefore use. At an applied voltage at the extended anode the elec-trode reaction is oxygen evolution while the cathode reaction atthe cantilever tip is oxygen reduction to water, identical to thereaction in a polymer electrolyte fuel cell on the air/oxygen side.The current detected by the conductive AFM tip correspondsto the local proton flow through the membrane which is drivenby the electrochemical reactions and the applied voltage. Allimages were analyzed using the WSxM Scanning Probe Microscopysoftware version 4.0 [24].

Most in situ methods are integral and therefore do not allow alocally resolved analysis of cells. It is important as inhomogeneousdistribution of reactants and products often occur depending onthe operating conditions. Locally resolved current density mea-surements in single cells have been developed at DLR and areimplemented to gain insight into the nature of the oscillating fluc-tuations. A segmented bipolar plate based on printed circuit board(PCB) technology with integrated temperature sensors is used inthe single cell to analyze the locally resolved current density distri-butions [20,25]. The temperature sensor is needed for resistancecalibration and for ensuring an isothermal measurement; the PCBboard has been introduced at the anode side. Other groups havealso developed methods and tools for current density distributionmeasurements, e.g. in [26–28].

An important aspect of the measurement is the flow field usedand the configuration of the flow of reactants: For this study, a sin-gle serpentine channel for the humidified anode and for the nonhumidified cathode is used. Complexities of the two-phase watertransport can be very important in the overall behavior of fuel cells,

1.0

0.8

0.6

0.4

0.2

0.0

Cel

l Vol

tage

/ V

1.21.00.80.60.40.20.0Current Density / A cm-2

(A) (B)

Fig. 2. (A) Segment distribution of a flow field arrangement with temperature sensors and single serpentine channel flow field. (B) Initial U(i) curve of Nafion�111-IP atreference conditions with full humidification.


as e.g. discussed by Owejan et al. [29]. By choosing a singleserpentine channel at the humidified anode we ensure that no par-tial blocking of the flow field by liquid water is possible. As thecathode is operated with dry gases the flow field configuration isless critical. As mentioned, the flow field is partitioned intosegments for the current density distribution which are classifiedin-house as depicted in Fig. 2A. Essentially, measurements at twodifferent flow configurations are presented which provide differentresults:

� Configuration A (co-flow).Anode: input G1 output A7.Cathode: input G1 output A7.� Configuration B (counter flow).

Anode: input A1 output G7.Cathode: input G1 output A7.

Fig. 2B shows initial performance of Ion Power Nafion�111-IP atreference conditions with full humidification (100% RH).

0.6High Current Density (H)

3. Results and discussion

Oscillating fluctuations of PEFC single cells were observed un-der dry operation at the cathode and high humidification at the an-ode and were further investigated to understand this phenomenon.The cell was first operated in the flow configuration A (co-flow) un-der full humidification at anode and cathode giving an adequatepower density, see Fig. 2B. Then the humidifier was by-passed atthe cathode reducing instantaneously the cathode humidity todry conditions and the humidity at the anode was increased to

Fig. 3. Instabilities of a single cell operated with a dry cathode feed (RH = 0%) and awet anode at nominal RH = 152%, cell voltage 500 mV; FH2 = 180 ml min�1,Fair = 560 ml min�1, co-flow configuration.

condensing conditions in the cell compartment. The current is re-corded versus time (ca. 2.7 h) in Fig. 3 at a constant cell voltageof 500 mV under dry cathode operation and a nominal humidityof 152% at the anode. This relative humidity >100% indicates con-densing conditions with liquid water present at the anode. At firstthe current density decreases somewhat from 650 mA cm�2 toabout 610 mA cm�2 within 1000 s, then a fluctuation with increas-ing amplitude is observed which is followed by a continuous oscil-lating fluctuation between ca. 610 mA cm�2 and ca. 60 mA cm�2

with a gradual decrease of current due to degradation of cell per-formance. A periodicity of the fluctuation is evident although someirregularities in the oscillation are present. The studies presentedherein aim at understanding of this phenomenon.

What is apparent from this measurement is that the fluctuationconsists of two operation performance points that are stable forsome time. The system is oscillating between them. For these oper-ating conditions the low current regime has a longer residencetime of 300 s than the high current point which is constant forabout 50 s. As will be shown the amplitude and also the frequencyof the oscillations depend on the operating conditions and the RHvalue at the anode. It is important to emphasize that no oscillatingfluctuations are observed when both gas feeds at anode and cath-ode are dry. This observation already points toward the importanceof the cell water balance for the oscillatory response of the cell.

Fig. 4 shows one of the oscillations in more detail. It can be ob-served that the transition between the low current operation point

0.5

0.4

0.3

0.2

0.1

0.0

Cur

rent

Den

sity

/ A

cm-2

200180160140120100806040200Time / s

Low Current Density (L)

Fig. 4. Detail of one typical oscillation between a low current state (50 mA cm�2)and high current state (550 mA cm�2), conditions as in Fig. 3; symbols aremeasurement data and broken line a fit with a sigmoid equation.


and the high currents takes about 20 s. The high current point ofthe oscillation is stable for ca. 50 s with a small change of currentand then a transition to the lower performance operation is ob-served. The transition from high to low currents is normally slowercompared to the upward conversion. The transition to the highstate is relatively fast and is similar to the observations by Benzigeret al. who have denominated this an ‘‘ignited” state [18]. In the fol-lowing for clarity the high current density point is denominated Hand the low performance operating point is denominated L. Thetransition from L to H is well fitted by a general sigmoid functionof the form:

jðtÞ ¼ Aþ ðB� AÞ � tD

CD þ tDð1Þ

The fit to the sigmoid equation (specifically a Hill equation) isshown by broken line and will be discussed below.

In order to understand the nature of the transition from L to H,EIS measurements corresponding to the quasi-stable operation ofthe cell have been performed. The high-frequency resistance ob-tained by EIS measurements gives integral values for the mem-brane resistance. The measurements corresponding to the samecell at H and L of the oscillating state are given in Fig. 5. The fre-quency had to be restricted to the high-frequency region in orderto ascertain the stability criterion for EIS measurements corre-sponding to the specific states of the cell. The Nyquist representa-tion in Fig. 5 at 600 mV cell voltage shows that the two operationpoints are characterized by a large change of the high frequencyohmic resistance of the cell. Since we have no indication that thecontact resistances due to roughness can change so much duringoperation of the cell, it is plausible to associate the change from28 mX to 5.3 mX to a change of membrane or interfacial ionicresistance. However, it is noted that we cannot distinguish if thechange is taking place at the bulk of membrane or just at the inter-face electrode|membrane.

The change of current from 3.2 to 17 A going from L to H is alsoassociated with a charge transfer resistance change which, how-ever, is consistent with the measured change of the resistance.Therefore, it can be concluded that the ionic resistance is the deter-mining factor for the observed dynamics of the cell under theseconditions. The membrane ionic resistance is associated with thewater content of the membrane. Due to the limited time for therecording of the EIS we cannot distinguish at present if the changein ionic resistance is taking place only in the membrane or also in

5 10 15 20 25 30 35

0

-20

-10

-15

-5

10

5

Z' / mΩ

Z'' / mΩ

599 mV - 17.088 A

599 mV - 3.199 A

1.2 kHz

500 Hz

30 kHz

500 Hz

R=5.3 mOhm R=28 mOhm

Fig. 5. Nyquist representation of impedance spectra of the high (red) and low (blue)current state indicating the large change in membrane resistance. Measurement at600 mV for frequencies higher than 500 Hz. Flow rates, area: 25 cm2 co-flowconfiguration. (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

the catalytic layer of the electrode. The semi-circle in Fig. 5 showssome distortion which, however, is not pronounced enough for fur-ther analysis of the properties of the catalyst layer.

It is interesting to determine if the operation conditions affectthe residence time in L and H points. For the conditions describedabove H exhibits a residence time of only some 50 s whereas theresidence time in L is around (300 s). In order to investigate if thisis a general trend the RH of the anode feed was sequentially variedunder water condensing conditions from nominally 237% to 100%always with a dry cathode. Introducing a large amount of liquidwater into the anode side of the cell is accomplished by changingthe temperature of the cell because cell temperature can be chan-ged faster compared to the humidifier which possesses a consider-able thermal inertia. As can be seen in Fig. 6 at high liquid waterlevels of the anode no fluctuations are discerned. However, aftersome time a small periodic change of performance indicates thatthe cell is approaching a situation where unstable oscillatorybehavior is possible. When the cell temperature is increased from60 �C to 65 �C the fluctuations become pronounced and H persistsfor much longer times compared to L. The downward transition isslower compared to the always fast upward transition. Withincreasing cell temperature and less liquid water at the anodethe oscillating effect becomes more pronounced with a time-dom-inating L point. At the highest cell temperature the upward transi-tion ceases and only L is stable. The measurement shows thatliquid water at the anode plays an important role in the observedtransitions and that the stability of the high current operation de-pends on the amount of liquid water present at the anode.

The relatively fast change during the upward transition is ofinterest as it is clearly related to liquid water in our case. It isnormally assumed that the slowest process at the cell level ismembrane hydration with time constants in the range of at least20–25 s. Therefore, when the membrane undergoes water contentchanges, the water accumulation term is dominating the transientbehavior of the cell. The observed transition times here are in thelower range associated with hydration effects and it is interestingto understand how a macroscopic area of 25 cm2 is affected. There-fore, the current density distribution with 49 segments for thetransition at co-flow and counter flow are investigated. The associ-ated measurements are displayed in Figs. 8 and 9 during thedownward and upward transition for the two distinct flow config-urations, respectively. For all locally resolved measurements a col-or representation of the current density is used for easiercomprehension. The color coding for the current density measure-ments of Figs. 8 and 9 is given in Fig. 7.

Fig. 6. Time dependence of current (red) and cell temperature (blue) yieldingdifferent RH on the anode (cathode RH 0%) with nominal RH 237%, RH 187%, RH152%, and RH 100%. Other conditions as in Fig. 3. (For interpretation of thereferences to color in this figure legend, the reader is referred to the web version ofthis article.)

0.8

0.6

0.4

0.2

0.0

Cur

rent

Den

sity

/ A

cm-2

250200150100500Time / s


High Current Density (H)

0.8

0.6

0.4

0.2

0.0

Cur

rent

Den

sity

/ A

cm-2

250200150100500Time / s


High Current Density (H)

Fig. 7. Transition from low to high current state and consecutive decay for counter flow configuration B at cell voltage of 500 mV, Fair = 660 ml min�1, FH2 = 200 ml min�1, celltemperature Tcell = 70 �C. Symbols are measurement data and broken line a fit with a sigmoid equation. Right hand side: Color representation of current densities formeasurements in Figs. 7–9. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)


Fig. 7 shows a fluctuation in a counter flow configuration. Withthis configuration the stability of H is favored as will be discussedbelow. The higher stability is evident from a much slower decay ofH–L in the order of hundreds of seconds. This compares to about20 s rise at the upward transition. In addition the amplitudechange between L and H is significantly less than in the co-flowcase. Interestingly, the upward transition shows the same charac-teristic time as in the co-flow configuration (again the broken lineindicates the Hill equation fit). The associated current density colorplots are provided in the insets of the image showing a homoge-neous current density distribution at H associated with a low ionicresistance, high humidification and presence of liquid water at theanode. For L the current density distribution is inhomogeneouswith some high current densities remaining at the anode inlet (cor-responding to cathode outlet location) and low current densities atthe cathode inlet.

Fig. 8 exhibits the current density distributions of the transi-tions between H and L for (A) the investigated co-flow configura-tion and (B) the counter flow configuration. As was mentionedbefore, configuration (A) leads to a faster downward transition;the low current density is initiated at both anode and cathode inlet.After 100 s the only area with higher current density remains inthe left corner where the anode and cathode inlet is located atthe bottom. When the transition is completed the only inhomoge-neity is left directly at the outlet of the reactant flows. In contrast,the counter flow configuration (B) leads to a stabilization of H andthe transition is slower and more gradual. The decrease in currentdensity starts at the segments corresponding to the cathode inletand the anode outlet region, specifically at the right hand side ofthe images. The transition is not as complete as in co-flow config-uration (A) and an inhomogeneous distribution still persists in L. Inthe anode inlet region which is where the humidity enters the cellthe current density remains high.

Fig. 9 shows the current density distributions during the up-ward condition for both studied configurations. This is the ‘‘igni-tion” process which is quite fast with 20 s transition timeobserved for the cell for co-flow (A) and counter flow (B) configu-rations. Whereas in the counter flow configuration the area of highcurrent density remains, the co-flow configuration yields a com-pletely ‘‘extinguished” current operation point. In both cases thecurrent density increases from the high humidity anode inlet side,namely from the right side in case (A) and from the left hand sidefor case (B). The dry inlet of the cathode is in both cases the rightside. It is further noted that in the parallel case a lower currentdensity remains at cathode and anode inlet at the end of the tran-sition. The time resolution of the scanner for the current density

distribution does not allow further resolution of intermediatestages.

From these observations it is clear that the area with high cur-rent density changes during the transitions and the current densitydistribution depends on the flow configuration. The counter flowconfiguration is of advantage regarding performance (high currentoperation stability) and exhibits a propagating active area with aboundary during the transitions. The downward transition startsat dry side of the cell area (cathode inlet) and the upward transi-tion starts at the wet side (anode inlet). Since the change of currentis associated with increase of ionic resistance we can derive the fol-lowing equation for the upward transition which is similar in allinvestigated cases:

1RionicðtÞ

¼ rHAHðtÞ

lþ rL

ðAtotal � AHðtÞÞl

ð2Þ

Rionic(t) is the average ionic resistance, rH and rL are the specificconductivities of the high and low current points, AH is the fractionof high current density area, and l the thickness of the membranewhich is considered to be constant. This description is better suitedfor the counter flow transition where the high current area is bet-ter defined. The time area behavior during the upward transitioncan be fitted very well with a sigmoidal Hill equation (see Eq.(1)) giving a very steep Hill slopes of 8.7–11.8 (±0.3) (D in Eq.(1)) for the upward transitions in the different flow configurations.Sigmoidal behavior is well known for cooperative systems in bio-technology where transitions are characterized by a sharp transi-tion, indicating that a large variation of the output takes place ina very small interval of the independent variable. Examples ofthese systems are e.g. the binding of four oxygen molecules toone haemoglobin molecule; the thermal transition from gel to solphase of artificial and natural membranes or the unfolding of mac-romolecules like DNA. When D = 1, the transition is non-coopera-tive, while for a very large D the transition is all or none and thetotal change takes place in a very narrow interval of time. This isconsistent with a propagating front with relatively well-definedboundaries during the transition. In contrast, the downward tran-sition is characterized by a lower cooperative factor, e.g. in Fig. 7the broken line indicates a fit with D = �4.7 (±0.3). It has to benoted that since this sigmoidal fitting gives just one characteristicparameter D of the transition the different processes which areimportant for the transition are not distinguished.

In order to understand the significance of the water input andwater removal it is necessary to investigate the dependence ofthe transitions on the flow rate of the gases. In Fig. 10 the oscilla-tory fluctuations upon changing the flow rate at the anode are

(A)°co-flow configuration

(B)° counter flow configuration

Fig. 8. Current density during the downward transition from the high current to the low current state. cell voltage Ucell = 500 mV, (A) co-flow configuration with total time of30 s from first to last image, (B) counter flow configuration with total time of 150 s from first to last image. Other operating conditions as in Fig. 7.


shown. Three different anode feed flows have been investigated,namely 200 ml min�1, 300 ml min�1 and 400 ml min�1. Thehydrogen is again humidified at 153% RH which corresponds to58 mgH2O min�1, 87 mgH2O min�1, and 116 mgH2O min�1 introducedinto the cell, respectively. Not all of this water flux is used formembrane humidification due to the water removal in the anodecirculation, but it is noteworthy that the cathode flow has thecapacity to remove a large amount of water, specifically 131mgH2O min�1 with 660 ml min�1 dry air. Therefore the cathodecan generally be considered free of liquid water in agreement withfindings in the literature [16]. In Fig. 10 it is evident that withincreasing water influx at the anode the current differences be-tween L and H become smaller. The L operation shows compara-tively higher total currents with higher humidity at the anode.This can be interpreted that the high ionic resistance is coincidentwith desiccation from the anode side. It is also noted that withincreasing anode flux the H operation shows a lower maximumcurrent although the hydrogen stoichiometry increases with theflow rate. The upward transition time is mostly unchanged and

also no significant changes in repetition frequency are observed.It is expected that the inhomogeneity of the current density distri-butions should be reduced when increasing anode flow and waterinflux leading to less dry conditions in L. This expectation is con-firmed in the current density distribution measurements at L inFig. 11 for the counter flow configuration (B). With increasing totalcurrent of L the current density distribution becomes less inhomo-geneous. Again the current density is low at the anode outlet andhigh at the humid influx area.

Similarly also the flow at the cathode is investigated with aconstant anode feed flow of 261 ml min�1 corresponding to awater influx of 76 mgH2O min�1. Increasing the dry flux of airfrom 832 to 1662 ml min�1 has a reverse effect compared tothe anode flow as can be seen in Fig. 12. Increasing the cathodeair flow increases the differences between high and low currentstates. Higher cathode flow increases the water removal fromthe cathode side, and water removal capacity is 165 mgH2O min�1

at 832 ml min�1 and 330 mgH2O min�1 at 1662 ml min�1 (seeFig. 12).

Fig. 10. Time dependence of oscillating fluctuations on anode flow rate atUcell = 500 mV, with a dry cathode flow Fair = 660 ml min�1, counter flow configu-ration, Tcell = 70 �C. Different anode flows are shown (a) FH2 = 200 ml min�1 (blue,diamonds), (b) FH2 = 300 ml min�1 (red, squares) c) FH2 = 400 ml min�1 (green,diamonds). (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

(A)° co-flow configuration

(B) ° counter flow configuration

Fig. 9. Current density distribution during the upward transition from the low current to the high current state. Cell voltage Ucell = 500 mV, (A) co-flow configuration, (B)counter flow configuration, the total transition is around 20 s. Other operating conditions as in Fig. 7.


Summarizing the observed effects it is evident that the ob-served operation points and the transitions between them are aconsequence of the water balance of the membrane or the inter-face area. This can be directly determined by EIS. However, thetransition in the ionic resistance leading the distinct performancepoints is not an integral effect but correlated to inhomogeneousconductivities over the area. This transition is strongly affectedby the flow configuration (counter flow versus co-flow) and bythe flow rates at anode and cathode. Other influencing parameterslike the relative humidity at anode and cathode as well as the influ-ence of flow field geometry and gas diffusion layers is also signifi-cant but will be discussed in a forthcoming paper including amodel for the cell behavior. Also the observed degradation effectswill be discussed later.

For understanding the behavior shown here it is necessary toconsider the specific properties of the membrane water transport.Recently, the membrane properties have been reviewed by Ge et al.[13]. They reported that the water transport processes for amembrane in contact with water vapor and liquid water are quite

different. The results show that, at temperatures in the range of303–353 K (30–80 �C), the electro-osmotic drag coefficients in-creases with increasing water content of the membrane. At lowwater content of the membrane with the ratio H2O/SO2�

3 < 4 thetemperature has almost no influence on the electro-osmotic drag.At high water content H2O/SO2�

3 > 4 the electro-osmotic drag coef-ficient a increases with increasing temperature. At a water activityof about 1.0 – which corresponds to 100% RH – the electro-osmoticdrag coefficient of water is about 1.1. For a membrane in contactwith liquid water, however, the electro-osmotic drag coefficient in-creases linearly from 1.8 at temperature of 288 K to 2.7 at 358 K.Within the statistical errors, the values of electro-osmotic dragcoefficient were reported to be almost the same at current densi-ties in the range of 0.05–1.0 A/cm2 at 353 K. In our case we havea wet anode and a very dry cathode environment. For the inlet re-gion of the anode we can expect to have also liquid water present.Taking the values discussed above is would be reasonable to as-sume very different electro-osmotic drag coefficients ranging froma < 1 for vapor phase to a � 2.7 for a membrane in contact with li-quid water.

The study of Adachi et al. systematically investigated the per-meation properties of Nafion� membranes with interfaces with li-quid water or water vapor [15]. It was found that the presence of aliquid-membrane interface is significant when comparing theabsolute permeability of water under LVP (liquid–vapor perme-ation) conditions relative to VVP (vapor–vapor permeation)conditions. The absolute permeability of water for the LVP mea-surements (1.5 � 10�11 mol2 cm�1 s�1 J�1) is much greater thanfor VVP measurements (1.6 � 10�12 mol2 cm�1 s�1 J�1), eventhough the gradient in chemical potential is similar for both. It isclearly stated that interfacial phenomenon i.e., a liquid/membraneinterface versus a vapor/membrane interface, is of great impor-tance to the permeability of water.

In our case at the initial condensing conditions at the anode thepresence of liquid water is certain and the H operation is stable. Un-der this conditions at least 650 mA cm�2 at 500 mV are observed.This value corresponds to a reaction water rate of rH2O =3.36 � 10�6 mol cm�2 s�1 and an electro-osmotic drag of water ofca. rH2O;EOD � 9 � 10�6 mol cm�2 s�1 in the presence of liquid water

Fig. 11. Distribution of the current density in the low current state of measurement in Fig. 10 at different anode flow rates, Ucell = 500 mV, Fair = 660 ml min�1 (A)FH2 = 200 ml min�1, Icell = 132 mA cm�2; (B) FH2 = 300 ml min�1, Icell = 225 mA cm�2; (C) FH2 = 400 ml min�1 Icell = 303 mA cm�2.

Fig. 12. Effect of cathode flow rate on the oscillating fluctuation at 600 mV cellvoltage with a anode flow of 261 ml min�1 at 100% RH, counter flow, Tcell = 70 �C.(A) Fair = 832 ml min�1 (red), (B) Fair = 1662 ml min�1 (blue). (For interpretation ofthe references to color in this figure legend, the reader is referred to the web versionof this article.)


(using an a = 2.7) and ca. rH2O;EOD � 3 � 10�6 mol cm�2 s�1 in thecase of water vapor (a = 0.9). The lowest dry cathode flow used inour measurements can remove rH2O = 1.2 � 10�4 mol cm�2 s�1

High ionic conductivityLiquid water resevoir at anodeLVP

Dehydration ano

Anode without liquid waHigh ionic resistance

Low ionic conductivityAnode without liquid water

Rehydration Anode~ 20 s

Low electro-osmotic drag and low water permeation

Form

atio

n of

liqu

id w

ater

High electro-osmand water perm

High ionic conductivityLiquid water resevoir at anodeLVP

Dehydration ano

Anode without liquid waHigh ionic resistance

Low ionic conductivityAnode without liquid water

Rehydration Anode~ 20 sRehydration Anode~ 20 s

Low electro-osmotic dragand low water permeationLow electro-osmotic drag and low water permeation

Form

atio

n of

liqu

id w

ater



Fig. 13. Left: scheme for the explanation of the transition between high current and low cstates.

water molecules. Therefore, it can be assumed that no liquid wateris present at the cathode for all conditions even if reaction waterand electro-osmotic drag are added. However, this is not the casefor the anode side which is exposed to high humidity. The observedtransitions in the fluctuating state are induced by a changing humid-ity at the anode associated with varying water transport properties(mainly the electro-osmotic drag coefficient and water permeationrate). The changes of water transport rates result in a strong positivefeedback mechanism represented in Fig. 13. In the H operation liquidwater is present at the anode, resulting in a well-humidified mem-brane interface which possesses a high electro-osmotic drag coeffi-cient and high water permeability. The high drag of water fromanode to cathode and the effective removal of water at the dry cath-ode feed reduce the liquid water present at the anode until a dryingof the membrane from the anode side starts. This is expected whenthe anode water reservoir at the interface is totally consumed start-ing from the dry anode outlet region. This leads to an increase of thelow current area from the anode outlet until the whole area is in thelow current point with dry anode and cathode conditions. With thelow current and low water content in the membrane the electro-os-motic drag coefficient drops below 1 and the interface are dry. Thewater permeation corresponds to VVP conditions. Low current,low electro-osmotic drag and low permeation decreases the watertransport from anode to cathode dramatically. Under these condi-

H2 + H2O

O2 + N2 O2 + N2 + H2O

H2 + H2O

αdrag≈2.7 αdrag< 1H+ H+

H2 + H2O

O2 + N2 O2 + N2 + H2O

H2 + H2O

αdrag≈2.7 αdrag< 1H+ H+

de

ter

otic drageation

de

ter

otic drageationotic drag

eation

urrent states; right: schematic for depicting the membrane processes in the two cell


tions the current density is just 50 mA cm�2 corresponding to L oper-ation. The electro-osmotic drag coefficient is in the order of 0.5–0.6leading to a value of ca. rH2O;EOD = 4 � 10�7 mol cm�2 s�1 transportedto the cathode and the VVP rate (rH2O � 1 � 10�6 mol cm�2 s�1) is anorder of magnitude lower compared to LVP permeation. Under theseconditions the water influx at the anode is not removed by currentflow and liquid water can start to accumulate. The liquid water firstforms at the anode inlet because the humidity is expected to dimin-ish along the anode channel due to the water uptake of the mem-brane. Therefore the first highly conductive area forms at theanode inlet for all flow configurations. As soon as liquid water formsLVP conditions lead to high water transport properties at the anodeinterface and concurrently the transition from low current to highcurrent is induced. The L to H transition time constant is in the rangeof 20 s. Liquid water accumulates at the cathode but the reservoir de-pends strongly on the operating conditions. The co-flow configura-tion favors somewhat the L operation because it is more difficult tobuild up a liquid water reservoir as the cathode flow has the highestwater removal capacity in the anode inlet region. In case of the coun-ter flow configuration a liquid reservoir can be formed at the humidanode inlet which faces the cathode outlet region with the lowestwater uptake capacity.

In this respect, our observations are in part consistent with oneof the mechanism proposed by Atkins et al. to explain the observedfluctuations in current and resistance during operation of a PEMfuel cell under dry conditions involving dehydration of the MEAat the anode [3]. We have to add, however, that we need to accountin addition for the liquid reservoir at the anode to explain ourdependences on flow configurations and flow rates. The essentialpoint is that the current of L operation is strongly influenced bythe flow at the anode and by the flow at the cathode with antipodaleffects.

This interpretation is supported by conductive AFM measure-ments which characterizes directly the conductivity of the mem-brane surface; that means the interface essential for inducing thetransitions in the fluctuating state. Although specific operationconditions are different for the AFM measurement, the generaloperation conditions of the full cell are present in the AFM set-up, namely the dry condition on the cathode and humid environ-ment on the anode side. With this arrangement, the distributionof ionic conductive surface regions and the magnitude of the cur-rent are measured for differently treated membranes. The goal ofthe study is to identify the activation properties of the mem-brane|electrode interface. In Fig. 14 the measured current on anarea of 2.1 lm � 2.3 lm is recorded from a Nafion 112 membranestored in water after electrode preparation for several days withoutfurther treatment. In general, between measurements all mem-

Fig. 14. (A) AFM current image of Nafion 112 membrane after storage in water for 1 weline.

branes are stored in water. Under these conditions always a lowconductivity is found with AFM. The threshold voltage for currentflow is high (U > 2 V). A typical current image and line profile is gi-ven in Fig. 14.

The maximum current does not exceed 20 pA at a voltage of2.4 V visible in the current line scan. The magnitude of the currentis almost constant across the area. Embedded are lamellar-like re-gions without any current. A prolonged storage in water at roomtemperature does not change the conductivity of the membrane.An activation procedure has been applied thereafter. The mem-brane was assembled into a fuel cell/electrolysis cell (from Helio-centris). No additional humidification has been applied to the cellafter integration into the set-up. Fig. 15 shows an AFM image ofa cell after 2 h electrolysis at 2 V followed by 2 h fuel cell operation.After electrolysis/fuel cell operation the electrode was always dryat removal. No significant changes in the current image in Fig. 15compared to the unprocessed membrane can be observed, themagnitude of the current is comparable to Fig. 14, and again re-gions without current are frequently visible.

In a subsequent variant procedure the membrane is operated inthe electrolysis cell for 2.5 h, but soaked in water during the wholetime. Thereby the conductivity of the membrane can be changedsignificantly as shown in Fig. 16. The maximum current now mea-sures several nA and is about three orders of magnitude highercompared to Figs. 14 and 15. The distribution of current is also dif-ferent and typically small spots with high current are present. Inthe regions in between these current spots with high conductivity,still small currents with a distribution similar to Figs. 14 and 15 arefound. However, these small currents are not visible at this en-larged scale. In addition the threshold voltage for current flow isreduced from 2 V to 1.5 V, respectively, indicating a decrease inmembrane resistance. Only an activation procedure with the mem-brane soaked in liquid water has changed the conductivity of themembrane. This behavior is found with different membrane elec-trolyte materials, so far without any exception. We therefore con-clude that liquid water is necessary for gaining a state of highconductivity and improve water transport. In addition, an appliedvoltage with current flow is needed to induce water uptake, sincesimple storage does not change the conductivity.

In the conducting channels of Nafion�, at least three differentconduction mechanisms are described [30]. A high conductivitycan only be provided by the Grothuss mechanism which is possiblein the so-called bulk water inside the conductive channels of themembrane. The water attached to the sulfonic acid groups formingthe channels provides hopping of the protons along the surfacewith a lower conductivity. The internal pressure inside these con-ducting channels is rather high [14]. To fill these channels with

ek, size 2.1 lm � 2.3 lm, U = 2.1 V, RH = 61%, T = 23 �C; (B) current along the profile

Fig. 15. (A) AFM current image of Nafion 112 membrane after 4 h activation by electrolysis and fuel cell operation without liquid water, size 2.1 lm � 2.3 lm, U = 2.1 V,RH = 69%, T = 23 �C; (B) current along the profile line.

Fig. 16. (A) AFM current image of Nafion 112 membrane after 1 h activation by electrolysis in contact with liquid water, size 2.1 lm � 2.3 lm, U = 2.0 V, RH = 69%, T = 23 �C;(B) current along the profile line.


additional water work has to be done. The energy needed can beprovided by an external voltage, applied or provided by theelectrochemical reactions. Thereby it could also be explained,why a prolonged storage in water does not result in a highly con-ductive state.

The conditions of AFM measurements are to some extendcomparable to the fuel cell operated with a dry cathode andwet anode. An upper electrode is formed by a platinum-coatedAFM tip and provides a point contact to the membrane. Up to ahigh relative humidity of more than 80% no extended water filmcan be found by AFM on the bare Nafion surface and only a thinwater film is present at the hydrophilic areas [22]. Under theseconditions and using an activated Nafion� 112 membrane witha high conductivity, current oscillations can be measured byAFM. The current image of Fig. 17 shows the temporal evolutionof current with time, recorded line by line starting from the upperleft corner at t = 0 s down to the lower right corner at t = 67 s. Theoscillation frequency is 0.14 Hz in Fig. 17 with a high current last-ing for about 5 s and a low currents lasting 2.5 s and immediatechanges between high current and no current flow. The frequencydepends on the humidity of the electrode as well as the appliedvoltage to the membrane.

The water transport to the region where a current is flowingthough the membrane to the point tip on top is provided by diffu-sion in the extended bottom electrode which acts as a water reser-voir. If the current through the membrane is high enough atemporal drying out or lack of liquid water at the contact regionis probable. As long as there is enough liquid water stored in theback electrode the water can humidify this region again after sometime and the current flow starts again. The transition time fromhigh current to zero is extremely fast (ms regime) but relates hereto a small conductive area. This is indicative of an interruption ofthe current path. A small amount of water leaves the membraneby evaporation due to reaction heat, but the major part seems tobe soaked into the extended neighboring conductive network ofthe membrane at the surface. An indication for such an internalwater spreading out can be found by recording the current flowon a single scan line for prolonged times. In Fig. 18 the first scanline is displayed at top of the image and the subsequent scan linesbeneath the previous one down to the bottom of the image after52 s. The few initially conductive channels at the top are branchingduring scanning at an almost regular time scale and thereby theconductive region at the surface is widened. The water transportfrom the bottom obviously leads to a built-up of pressure and

Fig. 17. (A) Current evolution at a fixed position measured by AFM, U = 1.8 V, RH = 55%, T = 27 �C; (B) current along the profile line.

Fig. 18. Current evolution during 100 s on single scan line at Nafion 112, treated for20 min in 10% H2O2 at 121 �C 1.2 bar, RH = 75%.


the side branches of the hydrophilic network are filled with liquidwater thereby leading also to a higher conductivity.

4. Conclusions

This paper shows results concerning the appearances of oscilla-tory fluctuation when the cathode is dry and the humidity level atthe anode high. It is observed that for specific operating conditionsa fluctuating state exists with two operation points with character-istic high and low currents. Periodic transitions between thesestates are observed. The upward transition between the low tothe high current operation is faster compared to the downwardtransition with a characteristic time of 20–25 s. The downwardtransition depends strongly on the operation parameters and thefollowing trends are observed which favor the high currentoperation:

� Higher humidity at anode (especially water condensingconditions).� Higher anode flow at 100% RH.� Lower cathode flow.� Counter flow configuration.

The transition time is shortened by the following parameters(concurrently with an acceleration of the transition frequency):

� Decreasing RH in the anode.� Increasing flow at the cathode.� Increasing the flow at the anode.� Co-flow configuration.

It is postulated that a liquid water reservoir plays an essentialrole in the dynamics of the cell as it determines strongly the cur-rent transition decay to the low current state. Contrasting to thisbehavior, the upward transition has always similar characteristictimes for all operating conditions as it is associated with theoccurrence of liquid water at the anode inlet. The change of theelectro-osmotic drag coefficient and the occurrence of liquid–va-por permeation properties at the interface of the membrane underliquid water contact is a major feedback mechanism for the oscil-latory behavior. A similar current oscillation occurs during AFMcurrent measurements at Nafion 112 membrane under to some ex-tent comparable conditions. In addition, the analysis of differentlytreated membranes leads to the conclusion that a high conductiv-ity of a membrane can only be reached after current flow in thepresence of liquid water.

Acknowledgements

The authors wish to acknowledge the Project CICYT ENE2005-09124-C04-02, MEC (Spanish Ministry of Science and Innovation),the DLR (Deutsches Zentrum für Luft und Raumfahrt, GermanAerospace Center), and the CSIC (Spanish National Research Coun-cil) Institute of Industrial Automatic for their support. Some of theauthors would like to acknowledge financial support by the ‘‘Bren-nstoffzellenallianz Baden-Württemberg (BzA-BW)” and by theHelmholtz-NRC-Project ‘‘Durability of PEM Fuel Cells”. The authorswould also like to thank H. Sander and N. Wagner at DLR for sug-gestions and discussions.

References

[1] F.N. Buchi, S. Srinivasan, Journal of the Electrochemical Society 144 (1997)2767.

[2] M. Watanabe, H. Uchida, M. Emori, Journal of the Electrochemical Society 145(1998) 1137.

[3] J.R. Atkins, S.C. Savett, S.E. Creager, Journal of Power Sources 128 (2004) 201.[4] S. Vengatesan, H.J. Kim, E.A. Cho, S.U. Jeong, H.Y. Ha, I.H. Oh, S.A. Hong, T.H.

Lim, Journal of Power Sources 156 (2006) 294.[5] S. Vengatesan, H.J. Kim, S.Y. Lee, E. Cho, H.Y. Ha, I.H. Oh, T.H. Lim, Journal of

Power Sources 167 (2007) 325.


[6] N.S.M. Hassan, W.R.W. Daud, K. Sopian, J. Sahari, Journal of Power Sources 193(2009) 249.

[7] X.G. Yang, N. Burke, C.Y. Wang, K. Tajiri, K. Shinohara, Journal of theElectrochemical Society 152 (2005) A759.

[8] H.T. Kim, K.Y. Song, T.V. Reshetenko, S.I. Han, T.Y. Kim, S.Y. Cho, M.K. Min, G.S.Chai, S.C. Shin, Journal of Power Sources 193 (2009) 515.

[9] Y. Lee, B. Kim, Y. Kim, International Journal of Hydrogen Energy 34 (2009)7768.

[10] T. Colinart, A. Chenu, S. Didierjean, O. Lottin, S. Besse, Journal of Power Sources190 (2009) 230.

[11] T.A. Zawodzinski, J. Davey, J. Valerio, S. Gottesfeld, Electrochimica Acta 40(1995) 297.

[12] Y.H. Park, J.A. Caton, International Journal of Hydrogen Energy 33 (2008) 7513.[13] S.H. Ge, B.L. Yi, P.W. Ming, Journal of the Electrochemical Society 153 (2006)

A1443.[14] K.D. Kreuer, S.J. Paddison, E. Spohr, M. Schuster, Chemical Reviews 104 (2004)

4637.[15] M. Adachi, T. Navessin, Z. Xie, B. Frisken, S. Holdcroft, Journal of the

Electrochemical Society 156 (2009) B782.[16] G.J.M. Janssen, M.L.J. Overvelde, Journal of Power Sources 101 (2001) 117.[17] J.F. Moxley, S. Tulyani, J.B. Benziger, Chemical Engineering Science 58 (2003)

4705.[18] J. Benziger, E. Chia, J.F. Moxley, I.G. Kevrekidis, Chemical Engineering Science

60 (2005) 1743.

[19] H. Fukumoto, H. Maeda, K. Mitsuda, Electrochemistry 68 (2000) 794.[20] E. Gulzow, M. Schulze, N. Wagner, T. Kaz, R. Reissner, G. Steinhilber, A.

Schneider, Journal of Power Sources 86 (2000) 352.[21] E. Aleksandrova, R. Hiesgen, D. Eberhard, K.A. Friedrich, T. Kaz, E. Roduner,

Chemphyschem 8 (2007) 519.[22] E. Aleksandrova, R. Hiesgen, K.A. Friedrich, E. Roduner, Physical Chemistry

Chemical Physics 9 (2007) 2735.[23] R. Hiesgen, E. Aleksandrova, G. Meichsner, I. Wehl, E. Roduner, K.A. Friedrich,

Electrochimica Acta 55 (2009) 423.[24] I. Horcas, R. Fernandez, J.M. Gomez-Rodriguez, J. Colchero, J. Gomez-Herrero,

A.M. Baro, Review of Scientific Instruments (2007) 78.[25] M. Schulze, E. Gulzow, S. Schonbauer, T. Knori, R. Reissner, Journal of Power

Sources 173 (2007) 19.[26] S.J.C. Cleghorn, C.R. Derouin, M.S. Wilson, S. Gottesfeld, Journal of Applied

Electrochemistry 28 (1998) 663.[27] D.J.L. Brett, S. Atkins, N.P. Brandon, V. Vesovic, N. Vasileiadis, A.R. Kucernak,

Electrochemistry Communications 3 (2001) 628.[28] J. Stumper, S.A. Campbell, D.P. Wilkinson, M.C. Johnson, M. Davis,

Electrochimica Acta 43 (1998) 3773.[29] J.P. Owejan, J.J. Gagliardo, J.M. Sergi, S.G. Kandlikar, T.A. Trabold, International

Journal of Hydrogen Energy 34 (2009) 3436.[30] P. Choi, N.H. Jalani, R. Datta, Journal of the Electrochemical Society 152 (2005)

E123.

oscillations of pem fuel cells at low cathode humidification

Documents