A transient model of PEM fuel cells based on a spherical thin film-agglomerate approach

(1)

A transient model of PEM fuel cells based on a spherical

thin film-agglomerate approach

Shih-Ming Chang, Hsin-Sen Chu

∗

Department of Mechanical Engineering, National Chiao Tung University, HsinChu, Taiwan 300, ROC

Received 18 March 2007; received in revised form 7 May 2007; accepted 8 May 2007 Available online 18 May 2007

Abstract

The objective of this study is to investigate the transient behavior of a PEM fuel cell by using a one-dimensional, two-phase mathematical model. This model treats the catalyst layer as a spherical thin film-agglomerate. Effects of various transport parameter as well as other factors such as catalyst loading, gas diffusion layer thickness and liquid water permeability on the transient evolution of major model properties and cell performance are investigated in detail. Numerical results show that the evolution of ionic potential drop and oxygen consumption experience several steps before they reach steady state. The same situation can also be seen for the evolution of water saturation and current density. A detail inspection of these phenomena shows a close relation between these transport variables and cell performance. Parametric studies of other design factors’ effects reveal there optimum values which lead to a greater current output during its evolution period.

Keywords: PEMFC; Transient analysis; Two-phase; Thin film-agglomerate model

1. Introduction

It is of great important to increase the performance of fuel cells (FCs) for modern consumer application such as portable and mobile devices. A number of studies have shown that the cathode is the most important fuel cell component, and its flood-ing problem is the major cause of performance deterioration in proton exchange membrane (PEM) fuel cells. One way to alleviate this phenomenon is by developing a better water man-agement strategy. However, this requires a better understanding of electrochemical reactions and liquid water transport in the cat-alyst layers, which are the most complex components of the fuel cell.

Due to the extreme small geometry of the electrodes, it is difficult to directly measure the liquid water distribution in the gas diffusion layers and catalyst layers. Many researchers have developed mathematical models to investigate how liquid water transport affects the PEM fuel cell performance. Springer et

al.[1] first developed an isothermal, one-dimensional,

steady-∗_{Corresponding author. Tel.: +886 3 571 2121x55115; fax: +886 3 572 7930.}

E-mail address:[email protected](H.-S. Chu).

state model for a complete polymer electrolyte fuel cell model. The model predicted an increase in membrane resistance with increased current density and demonstrated the great advantage of a thinner membrane in alleviating the resistance problem.

Bernardi and Verbrugge[2,3]proposed a one-dimensional

math-ematical model of the solid-polymer-electrolyte (SPE) fuel cell for liquid water transport in porous electrodes assuming a con-stant liquid water volume fraction and no interactions between liquid and gas flows. A numbers of catalyst layer models had

been developed, include the interface model[1–3], the thin film

model [4], the agglomerate model [5–11], and the thin

film-agglomerate model[12–14]. Using the agglomerate model, the

researchers at National Research Council of Canada studied the optimum performance of PEM fuel cell for a number of optimization parameters: type of agglomerate, catalyst layer thickness, catalyst layer porosity, distribution of Nafion

con-tent, Pt loading, etc.[15–21]. Besides these models, the works

of Wang and Wang[22,23]treated the catalyst layer as an

indi-vidual zone with various conservation equations employed in the modeling of transient study and various time constants for the transient transport phenomena were proposed.

As for the investigation of two-phase flow, Wang et al.[24]

pioneered the research on this issue through analytic and

(2)

Nomenclature

a surface area per unit volume

b Tafel slope

Cj concentration of species j (mol cm−3)

Dj diffusion coefficient of species j (cm2s−1)

F Faraday constant

H Henry constant (atm cm3mol−1)

i current density (A cm−2)

kc condensation rate

kT oxygen reaction rate constant (s−1)

kv evaporation rate

K permeability (cm2₎

M molecular weight

Nj molar flux of species j (mol cm−2s−1)

P pressure (atm)

r radius (cm)

s liquid water saturation level in porous medium

T temperature (K)

Vs cell potential (V)

y mole fraction

Greek

δ thickness (cm)

ε porosity or volumetric fraction

φ ionic potential (V) ϕ Thiele modulus κ conductivity (−1cm−1) λ water content μ viscosity (g cm−1s−1) ρ density (g cm−1) τ tortuosity ξ effectiveness factor

Subscripts and superscripts

agg agglomerate

CL catalyst layer

eff effective

g gas phase

GDL gas diffusion layer

MEM membrane N Nafion phase O2 oxygen p proton Pt platinum sat saturation v vapor water w liquid water

ical methods. A threshold current density was proposed to distinguish the scenarios between single- and two-phase regimes of water distribution and transport. In the subsequent works of

their group, sophisticated models[25–28] were developed to

simulate the flooding and liquid water distribution in PEM fuel

cells. The mature multiphase mixture (M2) formulation with

single set of conservation equations was employed to mimic the two-phase transport process. The results showed that it is important to consider the gas-phase pressure field in the GDL as a new mechanism of reactant transport. Another analytic model

performed by Pasaogullari et al. [29], focusing on the liquid

water transport in hydrophobic electrode, indicated that capillary transport dominates the water remove from flooded GDL. The

model of Natarajan and Van Nguyen[30,31], which considered

the evaporation or condensation of liquid water, also demon-strated the importance of its transport on cell performance. Moreover, comprehensive review summarizing the current stat-ues of the fuel cell modeling as well as point the development

direction of this emerging field is presented by Wang[32].

With the consideration that the catalyst layer is the most important component in the electrochemical reaction of a fuel cell, also the experimental results of TEM images by Siegel et

al.[7]clearly showed that carbon-supported, spherical Pt pellets

exist in the catalyst layers. To mimic the actual morphology of the catalyst layers, a spherical thin film-agglomerate expression along with a one-dimensional, two-phase model is developed in this paper to investigate the transient evolutions of various model properties as well as cell performance. The boundary conditions at the CL/membrane interface also take into account the effects of water content on certain design parameters such as GDL permeability, CL permeability, catalyst loading, and GDL thickness.

2. Mathematical modeling

This study presents a transient analysis of a PEM fuel cell cathode. The model domain consists of a gas diffusion layer (GDL), a catalyst layer (CL), and a membrane, arranged as

shown inFig. 1. Our previously published cylindrical thin

film-agglomerate model [13] is extended to a spherical model in

this work. The electrochemical kinetics and transport of oxy-gen, vapor water, and liquid water are also analyzed in the GDL, CL, and proton exchange membrane (PEM). When the PEM fuel cell is turned on, air diffuses through the GDL from the channel to the catalyst layer. Meanwhile, oxygen dissolves into the Nafion film and eventually reaches the pellet surface, where liquid water is generated. The following assumptions are made: (1) The gas-phase obeys the ideal gas law.

(2) Electronic resistance is negligible in the GDL and CL. (3) CL and GDL are assumed hydrophilic.

(4) The catalyst pellets are spherical, consist of carbon-supported platinum and Nafion, and are covered by a Nafion film.

(5) All the catalyst pellets in the CL have the same radius and Nafion film thickness.

(6) The catalyst pellets are homogeneous, and oxygen diffuses into them via the Nafion film.

(7) The oxygen reduction reaction generates only liquid-phase water.

(8) The ionic potential between the anode catalyst layer and the membrane is approximately zero, because the hydrogen oxidation reaction is fast.

(3)

Fig. 1. Schematic of the model[12].

Our previous paper [13]provides details on the derivation

of the model. The conservation equations applied are listed in

Table 1, for the following five variables:

(1) The concentration of oxygen in the gas-phase, [C_O2g ].

(2) The concentration of water vapor in the gas-phase, [Cgv].

(3) The liquid water saturation level, [s] (the ratio between liq-uid water volume and void volume in the porous medium).

(4) The concentration of liquid water in the Nafion phase, [CN

w].

(5) The ionic potential, [φ] (in the Nafion phase).

Experimental studies [5,7] have shown that the

carbon-supported Pt catalysts in the CL are approximately spherical. As mentioned above, this investigation treats them as perfect spheres. The derivation of spherical thin film-agglomerate model described here is based on the cylindrical models of Lin et al.

[12]and Chang and Chu[13]. The principal difference between

the cylindrical model and the spherical model is the outer sur-face area per unit volume of the agglomerate. The sursur-face area per unit volume of the agglomerate is defined as,

ar=

Agglomerate’s surface area

Geometric volume (1)

For spherical thin film-agglomerate model, aris:

ar = 4π(rp+ δN)2 (4/3)π(rp+ δN)3/(1 − εCL0 ) = 3 (rp+ δN) (1− εCL0 ) (2)

The initial values of all variables are zero, except for the liq-uid water concentration in the Nafion phase which is assumed to be in equilibrium with the water vapor activity in the

gas-phase. All boundary conditions are listed in Table 2. The

CL/membrane interface boundary condition for the

concen-tration of liquid water in the Nafion phase (Eq. (3)) is taken

from the model of Lin and Nguyen[14], where the water

con-tent below 14 mol H2O (mol SO3−)−1was determined by water

vapor activity. When the gas stream is saturated, however, and liquid water exists in the pores of the CL (between 14 and

16.8 mol H2O (mol SO3−)−1), the water content varies linearly

with the liquid water saturation level[1]. Since liquid water is

generated when a fuel cell starts up, any liquid water in the CL/MEM interface pores can dissolve into the Nafion phase. This leads to the boundary condition:

CN,eq

w = (0.043 + 17.81α − 39.8α2+ 36.0α3)Cf+ 2.8Cfs,

(3)

whereα is the water activity in the gas-phase in the CL

cath-ode and Cf is the concentration of fixed charged sites in the

membrane.

All the correlations used in this model are listed inTable 3.

The governing equations and boundary conditions are dis-cretized by the finite difference method. The convergence criteria for a steady state are:

Vnew i − Viold Vold i ≤ 1 × 10−4 (4)

where Viis an arbitrary variable.

3. Results and discussion

Water management plays an important role on the perfor-mance of PEM fuel cells. Understanding their starting process and performance evolution are also crucial in the actual operat-ing applications. Simulation results for this issue by employoperat-ing the PEM fuel cell cathode model is presented in the follow-ing sections. Investigations on how the structural parameters of the cell such as GDL permeability, CL permeability, cata-lyst loading, and GDL thickness affect its performance are also demonstrated.

To validate the numerical model, the comparison of calcula-tion results and Navessin et al.’s experimental results are carried

out. The comparison results are shown inFig. 2. The

experi-mental operating conditions and parameters of Navessin et al.

are listed inTable 4, and those of the model are listed inTable 5.

The I–V polarization curve represents the fuel cell’s steady state

at an operating temperature of 25◦C. The PEM fuel cell

oper-ates under uniform conditions from the time it is turned on until a steady state is achieved. It is found that current model results agree well with the experimental data.

(4)

T able 1 Go v erning equations V ariables GDL CL MEM C g O2 ∂ ∂t((1 − s) ε GDL 0 C g O2 ) = DO 2 ε GDL τ 0 [(1 − s)] τ∇ 2C g O2 +∇ ·[(1 − s)] τ∇· C g O2 ∂ ∂t((1 − s) ε GL 0 C g O2 ) =− RO 2 + DO 2 ε GL τ 0 [(1 − s)] τ∇ 2C g O2 +∇ ·[(1 − s)] τ∇· C g O2 – C g v ∂ ∂t((1 − s) ε GDL 0 C g )v =− Rw + Dv ε GDL τ 0 [(1 − s)] τ∇ 2C g +∇v ·[(1 − s)] τ∇· C g v ∂ ∂t((1 − s) ε GL 0 C g )v =− Rw + Dv ε GL τ 0 [(1 − s)] τ∇ 2C g +∇v ·[(1 − s)] τ∇· C g v – s ε GDL 0 ρw Mw ∂s ∂t= Rw + ρw Kw ,0 Mw μw − dP c ds (s∇ 2s+ (∇· s) 2) ε GL 0 ρw Mw ∂s ∂t= (4 n CL d + 2) RO 2 + Rw + ρw Kw ,0 Mw μw − dP c ds (s∇ 2s+ (∇· s) 2)– C N w – ε CL m ∂C N W _∂t = D N w∇ 2C N w εm ∂C N W _∂t = D N w∇ 2C N w + nd KN F ∇ 2V + φ – KN ,ef f ∇ 2φ − 4FR O2 = 0 ∇ 2φ = 0

Fig. 2. A comparison of the model to experimental data.

Fig. 3displays the evolution of the total ionic potential drop

(IPD) in the catalyst layer, for three different cell potentials Vs.

IPD=

x=CL x=GDL

(φp|x=CL− φp(x)). (5)

This plot can be divided into five sections: (1) a rapid rise, which lasts from startup to 0.01 s; (2) from 0.01 to 0.1 s, the potential plateaus (remains approximately constant); (3) at 0.1 s the poten-tial rises rapidly again to its maximum value; (4) the potenpoten-tial drops again; (5) a second plateau is reached representing the steady state.

The total dimensionless oxygen consumption (TOC) in the

catalyst layer also evolves over time, as shown inFig. 4. This

quantity is defined as:

TOC= x=CL x=GDL Cg O2|x=GDL− C g O2(x) Cg,inlet O2 . (6)

The plot can be divided into six sections: (1) a rapid rise from

startup to 5× 10−3s; (2) the first plateau, which lasts from

Fig. 3. Evolution of the total ionic potential drop in the catalyst layer under various cell voltages.

(5)

Correlations used in this study

Diffusivity of oxygen in Nafion membrane[3](cm2_s−1₎ _DN

O2= 0.0031 exp

−2768

T

Diffusivity of oxygen in liquid water (cm2_s−1₎ _DW

O2= 2.14 × 10 −5μTr μT 1.026_T Tr

Diffusivity of oxygen in gas stream[33](cm2s−1) Dg_O

2= 0.1775

_T

273.15

1.823

Diffusivity of vapor water in gas stream[33](cm2_s−1₎ _Dg v= 0.256

_T

307.15r

2.334

Henry’s constant for oxygen between Nafion and air[3](cm3_mol−1₎ _HN

O2= 1.33 × 10

6_exp−498

T

Henry’s constant for oxygen between liquid water and air[34](cm3_mol−1₎ _HN

O2= 1.33 × 10 6_exp−498 T Water content λ =CNw Cf

Net electro-osmotic drag coefficient[1] nd= 222.5λ

Conductivity of Nafion membrane[1](l ( cm)−1) κN= exp

1268 1 303− 1 T (0.005139λ − 0.00326) Diffusivity of liquid water in Nafion membrane[1](cm2_s−1₎ _CN

w(2>λ≥3)= 10−6exp 2416 1 303− 1 T (−3.1 + 2.0λ) CN w(3>λ≥4)= 10−6exp 2416 1 303− 1 T (6.89 − 1.33λ) CN w(λ4)= 10−6exp 2416₃₀₃1 −_T1(2.563 − 0.33λ + 0.0246λ2_{− 0.000671λ}3₎

5× 10−3to 5× 10−2s (consumption increases slightly during

this phase); (3) another rise from 5× 10−2to 0.5 s; (4) a

sec-ond plateau from 0.5 to 2 s (where the consumption decreases slightly); (5) a rapid rise to the peak value from 2 s onward; finally (6) a steady-state plateau. Oxygen consumption increases less than 15% in the first plateau, and decreases by only 2% in the second plateau. As the IPD increases, the fuel cell must over-come higher and higher activation energies to maintain a higher rate of electrochemical reaction. After the fuel cell overcomes

Table 4

Parameters used by Navessin et al.[15]

Gas diffusion layer properties

Porosity 0.6

Thickness (cm) 0.035

Catalyst layer properties

Porosity 0.55

Catalyst loading (mpt) (mg) 0.78

Area per unit volumea_(m−1₎ _727,664

Volumetric fraction of Naion in catalyst pellet (εP_N) 0.2 Raduis of catalyst pellet, (Ragg) (cm) 2.5× l0−5 Exchange current densitya(mA cm−2) 1.27, 1.66 Reference concentration of O2, (CO2,ref)

a₍_{␮mol cm}−3₎ _{2.34, 2.25} Membrane properties Porosity 0.35 Thicknessa_(cm) _{0.0062, 0.0061} Operation conditions Temperature (◦C) 25.0 Pressure (atm) 1.0

a_{List value for HMEA-2 and HMEA-3, respectively.}

Table 5

Parameters used in this simulation (base case) Gas diffusion layer properties

Porosity 0.3 Thickness (cm) 0.025 Permeability (cm2) 10−9 −dpc ds (Dyne cm−2) 284.2

Catalyst layer properties

Porosity 0.06 Thickness (cm) 0.0016 Permeability (cm2₎ ₃_{× l0}−11 −dpc ds (Dyne cm−2) 568.4 Catalyst loading (mPt) (mg) 0.4

Specific surface area of Pt (aPt) (cm2_{(mg Pt)}−1₎ ₁₀₀₀ Volumetric fraction of Nafion in catalyst pellet (εP

N) 0.393

Radius of catalyst pellet (Ragg) (cm) 10−5

Thickness of Nafion (cm) 10−6

Exchange current density (i0,ref0◦C) (A cm−2) 10−6 Membrane properties

Porosity 0.35

Fixed charge site concentration (mol cm−3) 1.2× 10−3 Operation conditions

Temperature (◦C) 60.0

Pressure (atm) 1.0

Mole fraction of O2in inlet 0.206

Humidity in inlet (%) 10

(6)

Fig. 4. Evolution of the total dimensionless oxygen consumption in the catalyst layer under various cell voltages.

the maximum activation energy, the IPD decreases. As shown inFig. 4, when Vs= 0.2 V and 0.4 V, after 2 s, the TOC increases

rapidly. Thus, after a fuel cell overcomes the maximum activa-tion energy, it retains a high electrochemical reacactiva-tion rate with a lower IPD.

Fig. 5shows the evolution of the liquid water saturation level, measured at various cell voltages. The liquid water saturation level plot can be divided into four sections. From startup to

10−2s, liquid water accumulates in the catalyst layer. Between

10−2and 0.5 s, liquid water begins diffusing into the gas

dif-fusion layer. During the next phase, from 0.5 to 2 s, the liquid water saturation level in the CL decreases as more and more liquid water occupies pores in the GDL. In the final phase, the liquid water saturation level increase in both CL and GDL until a steady state is reached. Although the liquid water saturation level decreases in the CL between 0.5 and 2 s, note that it will increase again once the electrochemical reaction rate generates

enough liquid water to fill the Nafion phase. When Vs= 0.6 V,

for example, the electrochemical reaction rate is moderate and less liquid water is generated. In this case, at the time after 2 s,

the liquid water saturation level decreases (Fig. 5c). In contrast,

the liquid water saturation level increases after 2 s for Vs= 0.2

and 0.4 V (Fig. 5a and b).

The fuel cell cannot be turned on if the Nafion phase is com-pletely dry, so it is assumed that the concentration of liquid water in the Nafion phase is equal to the concentration of fixed charge

sites at the anode.Fig. 6shows the evolution of the CL water

content over time at three different cell voltages. These plots can be divided into three sections: from startup to 3 s, the water con-tent decreases; from 3 to 13 s, the water concon-tent increases; and in the final phase, the water content is constant. Initially there is less liquid water amount in the gas pores than in the Nafion phase, so at first liquid water flows from the Nafion phase to the gas pores. When more liquid water begins to be generated, it dissolves into the Nafion phase again and we see increasing water content in both the CL and the membrane.

The current density evolution for various cell voltages is

shown inFig. 7. These plots can be divided into five sections.

From startup to 5× 10−3s, the current density rises rapidly.

Fig. 5. Evolution of the liquid water saturation level under various cell voltages, (a) Vs= 0.2 V, (b) Vs= 0.4 V and (c) Vs= 0.6 V.

(7)

Fig. 6. Evolution of the water content under various cell voltages, (a) Vs= 0.2 V, (b) Vs= 0.4 V and (c) Vs= 0.6 V.

Fig. 7. Evolution of the current density under various cell voltages.

From 5× 10−3to 2× 10−2s, the current density plateaus. The

current density begins to rise rapidly again at 2× 10−2s,

even-tually reaching its maximum value, then drops briefly. Finally, it

reaches a steady-state plateau. As seen inFig. 5, liquid water

hin-ders oxygen transport as it diffuses from the CL to the GDL. The electrical conductivity is also reduced during this period, due to decreased water content, and the current density levels off as a result. Note that the total oxygen consumption increases rapidly

after 0.05 s (Fig. 4), and that the ionic potential drop increases

rapidly after 0.1 s (Fig. 3). Thus, the current density increases

rapidly after 0.1 s. Oxygen transport resistance increases after 0.1 s, however, because at this point the liquid water saturation

level is increasing in both GDL and CL (Fig. 5). Thus, the current

density levels off after 1 s.

The current density drops off after reaching its maximum value for the following reason. The higher the current density, the more liquid water is generated and the more gas pores are occu-pied by liquid water. When this happens, however, the catalyst pellets also have less oxygen on their exposed surface. This tends to reduce the electrochemical reaction rate and causes a drop in the current density. At lower cell voltages the electrochemical reaction rate is higher, so more liquid water is generated and we see a more extreme drop in the current density. For instance,

when Vs= 0.2 V the drop is more than 30%.

Fig. 8shows the effect of modifying GDL permeability for

Vs= 0.4 V. Cases 1 through 4 used permeabilities of 9× 10−10,

1× 10−9, 1× 10−8, and 1× 10−7cm2, respectively. The higher

the permeability, the more liquid water will be drained; this low-ers the amount of liquid waver available and reduces the current density. If the GDL permeability is too low, however, liquid water cannot drain effectively. This will reduce the available oxygen on the catalyst surface, decreasing cell performance and limiting mass transport.

Fig. 9 examines the effect of CL permeability when

Vs= 0.4 V; cases 1 through 4 use values of 8× 10−12, 3× 10−11,

3× 10−10, and 3× 10−9cm2, respectively. Cases 3 and 4 have

the highest permeabilities and effective water drainage. But when liquid water drains from the fuel cell quickly, it also reduces the water content of the Nafion phase. Thus, a lower

(8)

Fig. 8. The effect of GDL permeability on the current density (Vs= 0.4 V).

CL permeability means that there will be more liquid water in the Nafion phase. The maximum current densities achieved in cases 1 and 2 are similar, but the steady-state current density in case 2 is higher. This is because in case 1, liquid water is occupying more of the gas pores and effectively covering the catalyst pellet surface, causing a series concentration overpo-tential.

The current density evolution for Vs= 0.4 V is shown in

Fig. 10, under catalyst loadings of 0.2, 0.3, 0.4, and 0.5 mg for cases 1 through 4, respectively. A higher catalyst loading means a more rapid electrochemical reaction. The current densities in cases 1 and 2 are obviously lower than those in cases 3 and 4, because the catalyst loading is too low in the former to drive the reaction. The maximum current density in case 4 drops rapidly, on the other hand, because of fuel cell flooding. Oxygen cannot reach the surface of the catalyst pellets when there is too much liquid water. This suggests that excessive catalyst loading does not improve cell performance.

Finally,Fig. 11demonstrates the evolution of current

den-sity for Vs= 0.4 V under various thicknesses of the gas diffusion

layer. Through case 1 to 6, the thicknesses are 220, 235, 240, 245,

250, and 265␮m, respectively. The current density increases to a

Fig. 9. The effect of CL permeability on the current density (Vs= 0.4 V).

Fig. 10. The effect of catalyst loading on the current density (Vs= 0.4 V).

Fig. 11. The effect of GDL thickness on the current density (Vs= 0.4 V).

constant (steady state) value after 2 s for GDL layers less than or

equal to 240␮m thick. For the remaining cases, the current

den-sity rises again after 2 s, reaches its maximum, then decreases in value and reaches a steady state. A thicker gas diffusion layer means a higher maximum current density, but this value may overshoot the sustainable current. Thicker GDL also imposes a greater resistance to liquid water transport, which promotes flooding of the fuel cell. Liquid water drains more quickly with a thinner gas diffusion layer, but the fuel cell cannot be too dry or its performance will deteriorate.

4. Conclusion

A transient, one-dimensional, two-phase model of a PEM fuel cell cathode is presented in this study with the catalyst layer treated as a spherical thin film-agglomerate. This model is used to investigate the transient transport of gaseous species, protons, and liquid water. The effects of GDL permeability, CL perme-ability, catalyst loading, and GDL thickness on cell performance are also investigated.

Our model results agree well with the experimental data. The observed IPD and TOC variations show that after the fuel cell

(9)

overcomes its maximum activation energy, a high electrochem-ical reaction rate with low IPD is maintained. Electrochemelectrochem-ical reactions affect the amount of liquid water in the cell; when less water is generated, the cell reaction rate is reduced. Neverthe-less, the higher liquid water content and saturation level tend to lead to the flooding catastrophe.

Parametric sensitivity analysis also shows that at higher per-meability, more liquid water is able to drain from the cell. Cell performance will decrease if there is too little liquid water, but may also be influenced if the water cannot drain effectively. Among the values adopted in this paper, the optimum GDL

permeability is K_wGDL_,0 = 1 × 10−9cm2 and the optimum CL

permeability isKGL_w_,0 = 3 × 10−11cm2.

Moreover, a cell with higher catalyst loading leads to more rapid electrochemical reactions, but this may also result in too much liquid water generation. This ultimately impedes oxygen transport, so excessive catalyst loading does not improve cell performance. Among the values considered in this study, the

optimum loading is mPt= 0.4 mg.

Finally, GDL thickness was found to play an important role in liquid water transport. A thinner layer offers less resistance to liquid water diffusion and drainage. Water content and cell performance therefore tend to remain constant. When the gas diffusion layer is thick, liquid water cannot drain effectively and the catalyst pellets will have less surface oxygen. According to

these results, the optimum thickness isδGDL= 245␮m.

Fuel cell performance and durability are also strongly influ-enced by impurities in the hydrogen gas used. A particular level of carbon monoxide yields a stable performance loss. There-fore, another research topic which is currently undergoing in our group is the investigation of the CO poisoning effect with the thin film-agglomerate model.

References

[1] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, J. Electrochem. Soc. 138 (1991) 2334–2342.

[2] D.M. Bernardi, M.W. Verbrugge, J. Electrochem. Soc. 139 (1992) 2477–2491.

[3] D.M. Bernardi, M.W. Verbrugge, AIChE J. 37 (1991) 1151–1163.

[4] Y.W. Rho, S. Srinivasan, Y.T. Kho, J. Electrochem. Soc. 141 (1994) 2089–2096.

[5] K. Broka, P. Ekdunge, J. Appl. Electrochem. 27 (1997) 281–289. [6] F. Jaouen, G. Lindbergh, G. Sundholm, J. Electrochem. Soc. 149 (2002)

A437–A447.

[7] N.P. Siegel, M.W. Ellis, D.J. Nelson, M.R. von Spakovsky, J. Power Sources 115 (2003) 81–89.

[8] N.P. Siegel, M.W. Ellis, D.J. Nelson, M.R. von Spakovsky, J. Power Sources 128 (2004) 173–184.

[9] W. Sun, B.A. Peppley, K. Karan, J. Power Sources 144 (2005) 42–53. [10] W. Sun, B.A. Peppley, K. Karan, Electrochim. Acta 50 (2005) 3359–3374. [11] K.M. Yin, J. Electrochem. Soc. 152 (2005) A583–a593.

[12] G.Y. Lin, W.S. He, T. Van Nguyen, J. Electrochem. Soc. 151 (2004) A1999–A2006.

[13] S.-M. Chang, H.-S. Chu, J. Power Sources 161 (2006) 1161–1168. [14] G. Lin, T. Van Nguyen, J. Electrochem. Soc. 153 (2006) A372–A382. [15] T. Navessin, S. Holdcroft, Q.P. Wang, D.T. Song, Z.S. Liu, M. Eikerling,

J. Horsfall, K.V. Lovell, J. Electroanal. Chem. 567 (2004) 111–122. [16] D.T. Song, Q.P. Wang, Z.S. Liu, M. Eikerling, Z. Xie, T. Navessin, S.

Holdcroft, Electrochim. Acta 50 (2005) 3347–3358.

[17] D.T. Song, Q.P. Wang, Z.S. Liu, T. Navessin, M. Eioerling, S. Holdcroft, J. Power Sources 126 (2004) 104–111.

[18] D.T. Song, Q.P. Wang, Z.S. Liu, T. Navessin, S. Holdcroft, Electrochim. Acta 50 (2004) 731–737.

[19] Q.P. Wang, M. Eikerling, D.T. Song, Z.S. Liu, J. Electroanal. Chem. 573 (2004) 61–69.

[20] Q.P. Wang, M. Eikerling, D.T. Song, Z.S. Liu, T. Navessin, Z. Xie, S. Holdcroft, J. Electrochem. Soc. 151 (2004) A950–a957.

[21] Q.P. Wang, D.T. Song, T. Navessin, S. Holdcroft, Z.S. Liu, Electrochim. Acta 50 (2004) 725–730.

[22] Y. Wang, C.Y. Wang, Electrochim. Acta 50 (2005) 1307–1315. [23] Y. Wang, C.-Y. Wang, Electrochim. Acta 51 (2006) 3924–3933. [24] Z.H. Wang, C.Y. Wang, K.S. Chen, J. Power Sources 94 (2001) 40–50. [25] U. Pasaogullari, C.Y. Wang, Electrochim. Acta 49 (2004) 4359–4369. [26] U. Pasaogullari, C.Y. Wang, J. Electrochem. Soc. 152 (2005) A380–A390. [27] H. Meng, C.-Y. Wang, J. Electrochem. Soc. 152 (2005) A1733–A1741. [28] Y. Wang, C.-Y. Wang, J. Electrochem. Soc. 153 (2006) A1193–A1200. [29] U. Pasaogullari, C.Y. Wang, K.S. Chen, J. Electrochem. Soc. 152 (2005)

A1574–A1582.

[30] D. Natarajan, T. Van Nguyen, J. Electrochem. Soc. 148 (2001) A1324–A1335.

[31] D. Natarajan, T. Van Nguyen, J. Power Sources 115 (2003) 66–80. [32] C.Y. Wang, Chem. Rev. 104 (2004) 4727–4765.

[33] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, second ed., Wiley, New York, 2002.

[34] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley, P.R. Roberge, T.J. Harris, J. Electrochem. Soc. 142 (1995) 1–8.