Vous êtes sur la page 1sur 11

DOI: 10.1002/fuce.

201000092

Hydrogen Mass Transport in Fuel Cell

ORIGINAL RESEARCH PAPER


Gas Diffusion Electrodes
J. St-Pierre1a*
1
Department of Chemical Engineering, University of South Carolina, 301 Main Street, Columbia, SC 29208, USA
a
Present address: University of Hawaii – Manoa, School of Ocean and Earth Science and Technology, Hawaii Natural Energy Institute,
1680 East-West Road, Honolulu, HI 96822.

Received May 30, 2010; accepted November 25, 2010

Abstract
An oxygen transport model derived for limiting current concentration difference models. The proposed model is
density operation was employed to explore its use for preferable because it is more general and derived from a
hydrogen transport. Limiting current data obtained from a specific physical mechanism. Experimental precautions
variety of fuel cell designs (alkaline, proton exchange mem- needed to ensure well-defined limiting currents without the
brane, sulphuric acid) demonstrated the model’s validity presence of artefacts were reemphasized and expanded.
and allowed determination of the hydrogen mass transport Model use for predictive purposes is also briefly discussed.
coefficient, a key cell performance parameter. Under specific
operating conditions, the proposed model is consistent with Keywords: Fuel Cell, Hydrogen Mass Transport Coefficient,
the continuous stirred tank reactor and the logarithmic mean Hydrogen Mixtures, Limiting Current, Model, Validation

1 Introduction arises in stacks with healthy cells forcing the current through
abnormal cells (the stack voltage is positive). Other failures
Fuel cells are actively being developed as alternative such as membrane holes and frozen valves can also lead to
power systems [1]. Focus areas of improvement include cost, fuel starvation. Owing to its irreversibility, carbon oxidation
performance, degradation and freezing. Hydrogen mass takes place above 0.207 V versus SHE [4], with significant cur-
transport impacting cell performance has received relatively rents occurring only at much higher potentials. Carbon oxida-
little attention [2]. Fuel starvation [3,4] and recirculation [5] tion also creates irreversible damage to the catalyst carbon
are two examples of reactant stream dilution needing a mass support and cell voltage [8–12]. Several approaches are pre-
transport coefficient to assess the overall impact on cell per- sently being explored to: (i) increase carbon resistance to oxi-
formance. dation (graphitisation, surface oxidation [13,14]), (ii) find
Figure 1 schematically shows oxygen reduction and alternate materials (conductive oxides and carbides [14,15])
hydrogen oxidation reactions (ORR and HOR) polarisation or (iii) eliminate the catalyst support [16,17]. Figure 1 shows
curves for two different cases. Under normal conditions, the that an increased hydrogen mass transport coefficient and
current demand originating from the system control is hydrogen oxidation limiting current increases cell resistance
located below both oxygen- and hydrogen-limiting currents. to fuel starvation, extends life and decreases cost. Therefore,
The net cell voltage (a positive value) corresponds to the dif- the hydrogen mass transport coefficient represents a key cell
ference between the polarisation curves corrected by subtract- performance parameter. Fuel recirculation, a system design
ing the ohmic loss. Under abnormal conditions occurring as a concept that increases hydrogen utilisation, also contributes
result of liquid water blockages or channel tolerance differ- to the fuel starvation risk because nitrogen diffuses through
ences between cells (uneven reactant flow distribution) that the membrane and accumulates in the anode compartment
lead to fuel starvation [8,9], the hydrogen concentration- [5]. As well, a larger hydrogen mass transport coefficient
dependent limiting current is located below the current increases cell robustness.
demand. This situation leads to an increase in anode potential
until other reactions occur to bridge the gap in current that –
reverse the cell voltage (a negative value). This situation [*] Corresponding author, jsp7@hawaii.edu

FUEL CELLS 11, 2011, No. 2, 263–273 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 263
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes
ORIGINAL RESEARCH PAPER

of the hydrogen mass transport coefficient from limiting cur-


rent densities requires modification to take this effect into
account and to minimise errors. Measurements show that the
current distribution is highly non-uniform under limiting
conditions irrespective of the active area (~2 cm2 [22],
289 cm2 [6]). The presence of convection in the gas diffusion
layer [23], which is exacerbated by the use of inter-digitated
flow fields [24], increases the hydrogen flux towards the cata-
lyst and intensifies the hydrogen concentration non-unifor-
mity along the channel length. Subsequently, the hydrogen
reaches the ionomer layer covering the catalyst. Hydrogen
transport within the ionomer layer is a diffusion-controlled
Fig. 1 Oxygen reduction and hydrogen oxidation reactions polarisations process [25].
curves for normal and abnormal fuel flow distribution cases.  1 The oxygen
reduction plot was generated with i ˆ 1=il ‡ eF=RT …E 0:85† where i rep- Gas diffusion electrodes (GDEs) were studied by circulat-
resents the current density (A m–2), il the limiting current density (A m–2), ing different fuel compositions on one of their side while the
–1
F the Faraday constant (96,500 C mol ), R the ideal gas constant catalysed side was in contact with an aqueous sulphuric acid
(8.3143 J mol–1 K–1), T the temperature (K) and E the electrode
solution [26,27]. The method used to extract the mass trans-
potential (V vs. SHE), il = 1.4 A cm–2, and T = 80 °C [6]. The hydrogen
oxidation plots were generated with i ˆ il 1 e 2F=RT …E E † where E0
0
port coefficient was derived under the continuous stirred
0
represents the standard electrode potential (V vs. SHE), E = 0 V vs. SHE, tank reactor (CSTR) assumption. This assumption is not valid
il = 0.3 and 2.8 A cm–2, and T = 80 °C [7]. The standard potential for operation under limiting conditions as the current density
for the carbon oxidation reaction (E0 = 0.207 V vs. SHE,
and reactant concentration distributions are strongly depen-
C + 2H2O → CO2 + 4H+ + 4e–) is also indicated.
dent on location along the flow field length [6, 22]. A similar
study was completed with an aqueous alkaline solution [28]
Figure 2 illustrates the hydrogen flow path from the fuel
but a hydrogen mass transport coefficient was not computed.
cell inlet to the catalyst surface. The hydrogen circulates first
In contrast, the most recent study was completed with a pro-
in the flow field channels and is characterised by a reactor
ton exchange membrane fuel cell (PEMFC) membrane/elec-
Péclet number [18] Pé = UL/D = 6,300–62,700 where U rep-
trode assembly [29]. A logarithmic mean concentration differ-
resents the average reactant velocity, L the flow field channel
ence was used to derive a global mass transport coefficient,
length and D the diffusion coefficient. The Pé values were cal-
but the validity of this approach is also unclear because it is a
culated using U = 63–627 cm s–1 based on a 0.06–0.6 s oxi-
generic method to treat reactor behaviour and minimise
dant transit time [19], the following channel length and cor-
aspect ratio effects [21]. Thus, it is not derived from a specifi-
rection factors of 1.2:1.8, 0.21 and 4 for, respectively, a lower
cally applicable model. Only three other PEMFC limiting cur-
fuel stoichiometry than air, a higher fuel purity than air and a
rent models have appeared in the literature. The Zhukovsky
smaller fuel channel cross section than air, L = 67 cm [20],
and Pozio [30] serpentine flow field channel model is not
and D = 0.67 cm2 s–1 for an H2/N2 mixture at 0 K and 1 at.
applicable because it does not take into account the signifi-
[21]. Therefore, the channel flow is governed by convection.
cant effect of transport in the ionomer (Figure 2). The concen-
Subsequently, the hydrogen penetrates the gas diffusion
tration profile is also assumed to be linear between the cell
layer. By contrast, the fluid Pé number [18] Ut/D = 2.1–20.6
inlet and outlet. Yet another model was derived by assuming
based on the gas diffusion layer thickness t (t = 0.022 cm
that the flow velocity remains constant along the flow field
[19]) shows predominant convection with significant diffu-
length [31], which equates to a dilute reactant stream. As a
sion impact. In other words, the diffusive flux through the
result, their model cannot be generalised as presented
gas diffusion layer is not negligible in comparison to the con-
because the concentration is not allowed to change along the
vective flux along the channel. The hydrogen concentration
channel length due to the flow of diluent gas within the
along the channel is therefore not uniform. The computation
cell (Eq. (3)). Finally, the Kazim et al. [32] model, cannot be
considered because the inter-digitated flow field does not
represent the prevalent design.
Previous studies’ data [26–29] are reanalysed to: (i) fulfil
the need for hydrogen mass transport coefficients, (ii) high-
light shortcomings of previous treatment methods, including
error estimates due to the use of a model either neglecting or
improperly taking into account reactant consumption along
the channel length, and (iii) emphasize an already proven
analysis alternative [6]. Necessary precautions to ensure
experimental data quality are stated. Furthermore, the inlet
Fig. 2 Schematic representation of the two stage hydrogen transport pro- hydrogen stream composition is predicted to meet pre-deter-
cess from the flow field channel to the catalyst surface. mined operating conditions.

264 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.fuelcells.wiley-vch.de FUEL CELLS 11, 2011, No. 2, 263–273
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes

2 Theory

ORIGINAL RESEARCH PAPER


c = pr/RT remains constant along the flow field channel and
is consistent with constant pressure and temperature assump-
2.1 General Case tions (the flow decreases along the channel length owing to
The main equation used to analyse prior experimental data consumption). The hydrogen mass balance in the flow field
was derived using the following assumptions: (i) water channel obeys:
vapour saturated inlet reactants (to avoid electrolyte de- dN il
hydration), (ii) reactant mass transport limited regime ˆ (4)
dy nF
obtained by controlling the cell potential to a low value, (iii)
relatively constant operating conditions along the flow field where y represents the cartesian coordinate along the flow
channel length (temperature, pressure and cell potential), and field channel (m). Replacement of Eqs. (1) and (3) in Eq. (4)
(iv) a reactant flow displaying both plug flow reactor and leads to:
ideal gas law behaviour [6]. It is noted that controlling the cell dN pr N
potential to a low value is not necessarily sufficient to achieve ˆ k (5)
dy RT N ‡ Nd
the desired limiting current regime. As illustrated in Figure 1,
the oxygen reduction limiting current needs to be higher than Equation (5) is scaled using the following dimensionless
the hydrogen oxidation limiting current. This is achieved by variables:
combining a higher reactant concentration such as pure oxy-
y ^ ˆ N
gen with a higher stoichiometry. Although only applicable ^
yˆ ; N (6)
L N…0†
for small mass transport rates and mole fractions, the local
mass transport coefficient is defined by: where y^ represents the dimensionless cartesian coordinate
^ the dimensionless hydro-
along the flow field channel and N
il ˆ nFkc (1) gen gas flux. The dimensionalised Eq. (5) is:

where il represents the limiting current density (A m–2), n the ^


dN kL pr N ^ nFk pr N ^ ^
N
ˆ ˆ ˆ ^i ^
; N…0† ˆ1
number of electrons exchanged in the HOR, F the Faraday d^
y ^ ‡f
N…0† RT N ^ ‡f
ie RT N ^ ‡f
N
constant (96,500 C mol–1), k the overall hydrogen mass trans- (7)
port coefficient (m s–1) and c the local dry fuel hydrogen con-
centration in the flow field channel (mol m–3). This choice where ^i represents the ratio of the limiting current density il
simplifies the model while taking into account the major with Nd = 0 (f = 0) to the inlet hydrogen flow rate equivalent
effect of reactant consumption along the flow field length current density ie. The solution to Eq. (7) yields the dimen-
[21]. Equation (1) is similar to another commonly used limit- sionless outlet reactant flux:
ing current expression [32–35] but is more general because  
^
N…1† ^
‡ f ln N…1† ˆ1 ^i (8)
the mass transport coefficient k takes into account convection
in addition to diffusion:
c
il ˆ nFD (2) 2.2 Dilute Reactant Stream
d
A reduction was derived for the case of a dilute reactant
where d represents the boundary layer thickness (m). The stream (f >> 1) [6]:
flow field channel hydrogen concentration is defined with
the following expressions: ^ ^i=f
N…1† ˆe (9)
pr N
cˆ ; pr ˆ p psat ;
RT N ‡ Nd Equations (8) and (9) are displayed in Figure 3 for several
(3) values of ^i. Equations (8) and (9) converge to the same f val-
f N 1 fH ie L
Nd ˆ fN …0†; fˆ ˆ ; N…0† ˆ
fH fH nF ^
ues for large N…1† values (low reactant conversion) and any ^i
value, thus confirming the validity of the derived approxima-
where pr represents the dry inlet reactant stream pressure tion. The dilute reactant stream approximation applicability
(Pa), R the ideal gas constant (8.3143 J mol–1 K–1), T the tem- range increases with the ^i value (for instance, Eqs. (8) and (9)
perature (K), N the hydrogen gas flux (mol m–1 s–1), Nd the overlap for ^i ˆ 100). For plotting convenience, Eq. (9) is used
diluent gas flux (mol m–1 s–1), p the inlet reactant stream in conjunction with Eq. (7) to derive relationships between
pressure (Pa), psat the water vapour saturation pressure (Pa), the average limiting current density and (i) the inlet hydrogen
f the inert gas to hydrogen fraction in the dry inlet reactant flow rate equivalent current density or (ii) the fraction of
stream, fN the nitrogen fraction in the dry inlet reactant hydrogen in the dry inlet stream. Equation (9) leads to (ief can
stream, fH the hydrogen fraction in the dry inlet reactant experimentally be set as a constant):
stream and ie the inlet hydrogen flow rate equivalent current    
iave ˆ ie 1 ^
N…1† ˆ ie 1 e nFkpr =RTie f
(10)
density (A m–2). For Nd = 0, the dry hydrogen concentration

FUEL CELLS 11, 2011, No. 2, 263–273 www.fuelcells.wiley-vch.de © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 265
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes
ORIGINAL RESEARCH PAPER

Equations (14) and (15) also show that the local limiting
current density (Eq. (1)) is equal to the average current den-
sity. In general (Eq. (8)), the hydrogen concentration is situ-
ated between the extreme cases given by Eqs. (11) and (15).
Equations (8) and (13) are displayed in Figure 3 for several
values of ^i . Equations (8) and (13) also converge to the same f
^
values for large N…1† values (low reactant conversion) and
^
any i value, thus confirming the validity of the derived
approximation. The high stoichiometry approximation
applicability range increases with a decrease in ^i (for instance,
Eqs. (8) and (13) overlap for ^i ˆ 0:1), corresponding to the
reverse direction that was observed with Eqs. (8) and (9).
The relative difference between approximate Eq. (13) and
the general Eq. (8) is expressed by:
Fig. 3 Plots of outlet hydrogen flux Eqs. (8), (9), and (13) for several ^i pa-
2  3
fgen fapprox ^
ln N…1†
rameter values.
ˆ1 4 5 (16)
fgen ^
N…1† 1
where iave represents the average limiting current density where fapprox represents the f parameter computed from the gen-
(A m–2). Equation (10) implies that the hydrogen concentra- eral reactant flux equation approximation (Eq. (13)). The differ-
tion varies exponentially along the channel length (f >> 1): ence between the high stoichiometry approximation and the
pr N p N p N p f ^ pr fH general equation is independent of ^i . The relative difference
^i^
y=f
cˆ ˆ r ˆ r ˆ r HN ˆ e shrinks between the high stoichiometry approximation and the
RT N ‡ Nd RT Nd RT fN…0† RT RT
(11) ^
general equation as N…1† increases (Figure 4), partly reflecting
the observed behaviour of Eqs. (8) and (13) in Figure 3.
The relative difference between approximate Eq. (9) and
the general Eq. (8) is expressed by:
fgen fapprox 1 ^
N…1† 3 Discussion
ˆ (12)
fgen 1 ^
N…1† ^i 3.1 Model Validation
where fgen represents the f parameter computed from the gen- Data presented in Figure 4 [28] for a GDE in contact with
eral reactant flux Eq. (8) and fapprox the f parameter computed an aqueous alkaline electrolyte and Eq. (14) are used to create
from the general reactant flux equation approximation Figure 5. The use of Eq. (14) is justified as reported data are a
(Eq. (9)). The relative difference shrinks between the dilute ^
function of fH and, considering N…1† > 0:93 leads to a relative
reactant stream approximation and the general equation as ^i difference lower than 0.037 (Eq. (16), Figure 4). For a high
increases (Figure 4), partly reflecting the observed behaviour 6 M KOH concentration, data follow Eq. (14). Deviations are
of Eqs. (8) and (9) in Figure 3. however noted for a lower 1 M KOH concentration. Never-
theless, a slope was computed to derive a hydrogen mass
2.3 High Stoichiometry transport coefficient by assuming that the data point corre-
sponding to fH = 0.1 was not affected by the mechanism caus-
Another reduction is derived for a high stoichiometry ing these deviations. Computed hydrogen mass transport
^
(N…1† ≅ 1 achievable with stoichiometries of the order of 10). coefficients are summarised in Table 1. The larger mass trans-
Expansion of the logarithm term around 1 in Eq. (8) and
keeping only the first term leads to:
  ^i
^
N…1† ^
‡ f N…1† 1 ˆ1 ^i; ^
N…1† ˆ1 (13)
1‡f
Equation (13) leads to:
  i ^i nFkpr nFkpr fH
iave ˆ ie 1 ^
N…1† ˆ e ˆ ˆ (14)
1 ‡ f RT…1 ‡ f† RT
Equation (14) implies that the hydrogen concentration is
^
constant along the channel length (N…1†≅1):

pr N p N p 1 pr 1 pf
cˆ ˆ r ˆ r ˆ ˆ rH
RT N ‡ Nd RT N ‡ fN…0† RT 1 ‡ f RT 1 ‡ f RT Fig. 4 Relative difference between the general outlet hydrogen flux
^
N (Eq. (8)) and two approximations (Eqs. (9) and (13)) computed using
(15) Eqs. (12) and (16).

266 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.fuelcells.wiley-vch.de FUEL CELLS 11, 2011, No. 2, 263–273
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes

ORIGINAL RESEARCH PAPER


Fig. 6 Comparison between a model approximation for a dilute reactant
Fig. 5 Comparison between a model approximation for a high stoichiome- stream (Eq. (10)) and experimental data for a GDE in contact with a sul-
try (Eq. (14)) and experimental data for a GDE in contact with an alkaline phuric acid solution [26]. T = 30 °C, pr = 97.0 kPa, ief = 5.02, 10.9 and
solution [28]. T = 25 °C, pr = 98.1 kPa. 17.8 A cm–2.

port coefficient resulting from the dilute solution is ascribed Operating conditions used to derive Eq. (14) imply that
to a larger hydrogen permeability value (Table 2, [28]). Addi- the hydrogen concentration is uniform along the flow field
tionally, it is hypothesised that other factors play a role, con- channel (Eq. (15)). Therefore, the average limiting current
sidering that the permeability increase from 6 to 1 M KOH corresponds to the local value. Use of Eq. (15) in Eq. (14) cor-
(approximately a factor of 10, Table 2) only leads to an responds to Eq. (1). Therefore, this simple model is applicable
approximate tripling of the hydrogen mass transport coeffi- ^
with N…1†∼1.
cient (Table 1). Potential factors include partial mass trans- Data presented in Figures 4–7 [26] for a GDE in contact
port control in the gas phase (Section 3.2) and contact angle with an aqueous acid electrolyte and Eqs. (8)–(10) are used to
change between the alkaline solution and the GDE [28]. create Figures 6 to 8. For Figure 6, Eq. (10) is used to display
Deviations at larger fH values and 1 M KOH concentration data because f and ^i values were not kept constant (Eq. (8)
are attributed to a lower OH– permeability at lower KOH con- could not be used) whereas the product ief was constant.
centrations and exacerbated by local solution dilution due to Furthermore, Eq. (14) could not be used because 0:35 < N…1† ^
water production [28]. Additional data for alkaline solutions < 0:76 leading to a relative difference varying from 0.14
are desirable to better define model limitations. to 0.62 (Eq. (16), Figure 4). By contrast, the use of
Eq. (10) led to a relative difference
^
< 0.07 (N…1† ˆ 0:44, ^i ˆ 8:8, Eq. (12),
Table 1 Hydrogen mass transport coefficients.
Figure 4). For Figure 7, Eq. (10) is
Electrolyte(a) Operating conditions Hydrogen mass transport Reference
–1 also used for the same reasons with a
coefficient / m s
relative difference <0.04 (N…1†^ ˆ 0:68,
Present model Other models
–4 (b) ^i ˆ 8:9, Eq. (12), Figure 4). Data in
Aqueous KOH 1 M KOH, N2, 25 °C 6.7 × 10 – [28]
6 M KOH, N2, 25 °C 2.0 × 10–4 – Figures 6 and 7 follow Eq. (10). One
Aqueous 0.5 M H2SO4 2.33 cm3 s–1 N2, 30 °C 5.7 × 10–3 – [26](c) set of data was characterised by a
5.08 cm3 s–1 N2, 30 °C 6.4 × 10–3 – smaller hydrogen mass transport
8.25 cm3 s–1 N2, 30 °C 6.7 × 10–3 –
coefficient (Table 1). Some curves
2.33 cm3 s–1 N2, 30 °C 1.5 × 10–2 –
overlap in Figure 7 (30 and 50 °C, 20
5.08 cm3 s–1 N2, 20 °C 5.5 × 10–3 6.3 × 10–3
5.08 cm3 s–1 N2, 30 °C 6.4 × 10–3 6.9 × 10–3
and 60 °C). For Figure 8, Eq. (9) is
3 –1
5.08 cm s N2, 40 °C 7.1 × 10 –3
7.3 × 10 –3 first used because most data points
3 –1
5.08 cm s N2, 50 °C 7.4 × 10 –3
7.1 × 10 –3
correspond to a dilute stream and
5.08 cm3 s–1 N2, 60 °C 7.6 × 10–3 6.4 × 10–3 ie was kept constant, while f is
0.058 cm3 s–1 H2, 30 °C 7.0 × 10–3 – varied (relative difference <0.02,
Aqueous H2SO4 solution 5.08 cm3 s–1 Ar, 20 °C 5.7 × 10–3 6.9 × 10–3 [27](c) ^
N…1† ˆ 0:52, ^i ˆ 40:5, Eq. (12), Fig-
5.08 cm3 s–1 N2, 20 °C 5.6 × 10–3 7.3 × 10–3
ure 4). Because ^i and k are constant
5.08 cm3 s–1 He, 20 °C 9.3 × 10–3 1.17 × 10–2
Ionomer (W. L. Gore & CO2, 41 °C 7.3 × 10 –3
8.0 × 10 –3
[29] (d) for high f values (Table 1), data were
Associates) N2, 41 °C 1.20 × 10 –2
1.35 × 10 –2 replotted in the inset using Eq. (8). It
He, 41 °C 1.52 × 10–2 1.57 × 10–2 is observed that two data points are
(a)
In contact with a GDE. still outliers. This situation is attribut-
(b)
Hydrogen mass transport coefficient estimates were not provided. Thus, Eq. (14) was used for their ed to the very low volumetric N2 gas
estimation. flow rates (Table 1) which are more
(c)
A CSTR model was used to calculate the hydrogen mass transport coefficient (Eq. (17)).
(d)
A logarithmic mean concentration model was used to calculate the hydrogen mass transport coefficient susceptible to measurement errors
[29]. Values were converted to the proper units by multiplication with RT. (presence of leaks, tubing permeabil-

FUEL CELLS 11, 2011, No. 2, 263–273 www.fuelcells.wiley-vch.de © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 267
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes
ORIGINAL RESEARCH PAPER

Fig. 9 Comparison between a model approximation for a dilute reactant


stream (Eq. (10)) and experimental data for a GDE in contact with a sul-
Fig. 7 Comparison between a model approximation for a dilute reactant phuric acid solution [27]. T = 20 °C, pr = 98.9 kPa, ief = 10.9 A cm–2.
stream (Eq. (10)) and experimental data for a GDE in contact with a sul-
phuric acid solution [26]. T = 20, 30, 40, 50 and 60 °C, pr = 81.4, 89.0,
93.9, 97.0 and 98.9 kPa, ief = 5.01 and 10.9 A cm–2. tration is underestimated along the flow channel, Eq. (1)
shows that at every channel length location the mass trans-
port coefficient is significantly overestimated. Thus, the
ity, etc). However, the model may also be responsible for
hydrogen mass transport coefficients from Ref. [26,27] are
these differences because ^i values are only constant for high f
generally larger than values derived from the present model
values. Thus, Figure 8 data follow both Eqs. (8) and (9) in the
(Table 1, Figure 10). Figure 10 kCSTR values are calculated
dilute H2 stream region (larger gas flow rate).
using a CSTR model [26,27] and all data pertaining to Fig-
Data presented in Figure 3 [27] for a GDE in contact with
ures 6–9:
an aqueous acid electrolyte and Eqs. (8) and (10) are used " #
to create Figure 9. Equation (10) is used to display data CSTR iave V_ …AVM iave =nF†
k ˆ (17)
because the product ief is constant (relative difference < 0.02, _
nF Vc…0† …Aiave =nF†
^
N…1† ˆ 0:68, ^i ˆ 17:4, Eq. (12), Figure 4). Equation (8) is also
where kCSTR represents the overall hydrogen mass transport
used because all data are characterised by only three f values
coefficient calculated using a CSTR model (m s–1), V_ the volu-
(Figure 9 inset). Both Eqs. (8) and (10) lead to a reasonable fit
metric inlet gas flow rate (m3 s–1), A the geometric electrode
and hydrogen mass transport coefficients derived from
surface area (m2) and VM the gas molar volume (m3 mol–1). A
Eq. (10) are given in Table 1.
relationship between kCSTR and kPEMFC mass transport coeffi-
Equation (11) derived under the same operating condi-
cients is derived by redefining the average current density
tions as Eq. (10) indicates that the hydrogen concentration
and using Eq. (11):
steeply varies along the flow field length. Therefore, a CSTR
model [26,27] is not adequate to derive a mass transport coef- p r fH ^i=f
iave ˆ nFkCSTR c…L† ˆ nFkCSTR e (18)
ficient because it assumes that the hydrogen concentration is RT
constant and fixed to the outlet value (^y ˆ 1). As the concen-
Equations (7) and (18) lead to:
RTiave …1 ‡ f† nFpr kPEMFC =RTie f
kCSTR ˆ e (19)
nFpr

Fig. 10 Comparison between hydrogen mass transport coefficients calcu-


lated using a PEMFC model (Figures 6–9 and 12 data [26, 27, 29],
Fig. 8 Comparison between a model approximation for a dilute reactant Eqs. (8), (9) and (10)), and, CSTR (Figures 6–9 data [26,27], Eq. (17))
stream (Eq. (9)) and experimental data for a GDE in contact with a sulphu- and logarithmic mean concentration difference (Figure 12 data [29],
ric acid solution [26]. T = 30 °C, pr = 97.0 kPa, ie = 0.125 A cm–2. Eq. (20)) models.

268 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.fuelcells.wiley-vch.de FUEL CELLS 11, 2011, No. 2, 263–273
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes

data are characterised by ^i≥10. Therefore, the hydrogen con-

ORIGINAL RESEARCH PAPER


centration significantly changes along the flow field channel.
Use of a logarithmic mean concentration difference [29] is an
improvement over the CSTR model [26,27] as shown in
Figure 10. Figure 10 kln values are calculated using a logarith-
mic mean concentration difference [29] (negligible pressure
drop and leak rate) and all data pertaining to Figure 12:
iave ln‰c…0†=c…L†Š
k ln ˆ (20)
nF‰c…0† c…L†Š
where kln represents the overall hydrogen mass transport
coefficient calculated using a logarithmic mean concentration
Fig. 11 Correlation between hydrogen mass transport coefficients calcu- difference model (m s–1). The adequacy of the logarithmic
lated using a CSTR model (Figures 6–9 data [26,27], Eq. (17)) and esti- mean model is further determined by using Eq. (11) in
mates (Eq. (19)) based on a PEMFC model (Figures 6–9 data [26,27],
Eq. (20):
Eqs. (8), (9) and (10)).
c…0† c…L† p f f ^i=f

iave ˆ nFk ln ˆ nFk ln r H 1 e (21)
ln‰c…0†=c…L†Š RT^i
where kPEMFC represents the overall hydrogen mass trans-
port coefficient calculated using a PEMFC model Equations (7) and (21) lead to:
(Eqs. (8)–(10)), m s–1). Figure 10 kPEMFC values are used to
iave kPEMFC …1 ‡ f†
compute kCSTR values with Eq. (19) and are plotted in Fig- k ln ˆ PEMFC =RTi f
 (22)
ie f 1 e nFpr k e
ure 11. The accuracy of Eq. (19) (the difference in slopes is
less than 6%) further supports the proposed model (Eq. (8)). Figure 10 kPEMFC values are used to compute kln values
Therefore, a CSTR model is only appropriate if N…1†∼1 ^ with Eq. (22) and are plotted in Figure 13. The accuracy of
because the hydrogen concentration is uniform along the flow Eq. (22) (the difference in slopes is less than 3%) further sup-
field length (Eq. (14)). ports the proposed model (Eq. (8)). Therefore, a logarithmic
Data presented in Figure 2 [29] for a GDE in contact with a mean concentration difference model is appropriate for inter-
polymer acid electrolyte and Eq. (8) are used to create ^
mediate N…1† values because the hydrogen concentration is
Figure 12. In this case, f values are relatively constant (target not uniform along the flow field length (Eq. (11)). The present
values are used to plot model curves in Figure 12). A good fit model is considered more desirable in comparison to the
is obtained with Eq. (8). Average hydrogen mass transport use of a logarithmic mean concentration difference model
coefficients for each inert gas diluent are given in Table 1. because the former is more general (consistent with CSTR
A simple general hydrogen concentration expression can- and logarithmic mean concentration difference models) and
not be derived because Eq. (8) is implicit, N ^ cannot be iso- is supported by a physical mechanism.
lated and inserted in Eq. (3). Figure 3 indicates that Eq. (8)
behaviour varies with the ^i value. At high ^i values (^i ≥ 10),
3.2 Hydrogen Mass Transport Coefficient
Eq. (8) behaves as Eq. (9) and the hydrogen concentration
profile is exponential (Eq. (11)). At low ^i values (^i ≤ 0:1), Table 1 data reveals that the hydrogen mass transport
Eq. (8) behaves as Eq. (13) and the hydrogen concentration coefficient is significantly smaller in the case of a GDE in con-
profile is uniform (Eq. (15)). Most Figure 12 experimental tact with an alkaline solution. Differences in gas solubility

Fig. 13 Correlation between hydrogen mass transport coefficients calcu-


Fig. 12 Comparison between the general case model (Eq. (8)) and lated using a logarithmic mean concentration difference model (Figure 12
experimental data for a GDE in contact with a polymer acid electrolyte data [29], Eq. (20)) and estimates (Eq. (22)) based on a PEMFC model
[29]. T = 41 °C, pr = 93.5 kPa. (Figure 12 data [29], Eq. (8)).

FUEL CELLS 11, 2011, No. 2, 263–273 www.fuelcells.wiley-vch.de © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 269
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes

Table 2 Gas transport in electrolyte properties.


ORIGINAL RESEARCH PAPER

cient including transport in a


Electroactive Electrolyte Temperature Solubility / Diffusion coeffi- Permeability / Reference N2 mixture (m s–1) and kg,N the
species / °C mol cm–3 cient / cm2 s–1 mol cm–1 s–1 hydrogen mass transport coeffi-
H2 1 M KOH 30 5.6 × 10–7 3.3 × 10–5 1.83 × 10–11 [36] cient in a N2 mixture (m s–1).
6 M KOH 30 1.3 × 10–7 1.1 × 10–5 1.43 × 10–12
Equations (23)–(24) combination
0.5 M H2SO4 25 5.9 × 10–7 3.8 × 10–5 2.24 × 10–11 [37]
Nafion(a) 20 5.1 × 10–7 7.6 × 10–6 3.9 × 10–12 [38]
leads to:
40 4.2 × 10–7 1.29 × 10–5 5.4 × 10–12 kN 1=ke
60 6.5 × 10–7 1.30 × 10–5 8.5 × 10–12 ˆ (25)
kHe …1=kg;N † ‡ …1=ke †
O2 Nafion(b) 50 3.6 × 10–6 1.0 × 10–6 3.6 × 10–12 [39]
(a)
Water saturated. Equation (25) provides an esti-
(b)
In contact with 100% relative humidity, 153 kPa. mate for the hydrogen mass trans-
port coefficient fraction responsi-
ble for transport in the electrolyte
phase. Table 1 data lead to values
and diffusivity between the alkaline solution and other elec- of 0.60 and 0.79 for, respectively, the sulphuric acid electro-
trolytes offer a likely explanation. However, permeability val- lyte and ionomer cases. Further support for this rationalisa-
ues (Table 2) are insufficient to account for these differences. tion is provided by plotting sulphuric acid results obtained at
For the 6 M KOH case, the hydrogen mass transport coeffi- different temperatures (Figure 14). There is evidence for two
cient is an order of magnitude smaller than for other electro- different processes with activation energies of 11.1 and
lytes but the permeability is a factor of ~2 smaller for com- 3.42 kJ mol–1 for a, respectively, low and high temperature
parable operating conditions. Therefore, electrode design range. A similar change in mass transport resistance with
parameters must also account for the difference (electrode temperature was also observed with a PEMFC [42]. Activa-
thickness, porosity, tortuosity, contact angle, surface tension, tion energies are also consistent with hydrogen permeation in
etc). Hydrogen mass transport coefficient values for a GDE in Nafion (15.7 kJ mol–1 [38]) and through porous gas phase sep-
contact with a sulphuric acid solution are self-consistent, with aration membranes (2.2 kJ mol–1 [43]).
the exception of the data series obtained at 30 °C with a A more detailed separation of transport contributions is
2.33 cm3 s–1 N2 flow. An explanation for this apparent discre- also possible [44] and has revealed that within the electrode
pancy was not found. (Toray carbon fibre paper), the largest loss originates from
The hydrogen mass transport coefficient taking account of gas phase bulk diffusion rather than in the sub-layer or the
transport in both gas and liquid/solid phases (Figure 2) is ionomer layer (all these contributions are lumped here into a
larger with a low molecular weight diluent gas, indicating at single mass transport coefficient). This result indicates that
least partial mass transport control in the gas phase (Table 1, electrode design significantly impacts mass transport because
[27] and [29] data). The gas diluent would not have any effect the largest transport loss was associated with the electrolyte
if transport were solely controlled by the liquid/solid phase. phase rather than the gas phase for an E-TEK ELAT based
Binary gas mixture diffusion coefficients follow the same electrode (Table 1, [29] data).
trend (Table 3). This effect is electrode design-dependent as The hydrogen mass transport coefficient obtained in the
the ratio of the hydrogen mass transport coefficients in N2 presence of an ionomer (1.20 × 10–2 m s–1, Table 1) is signifi-
and He is different for a sulphuric acid solution (Table 1, [27] cantly larger than for oxygen in air (3.0 × 10–3 m s–1 [6]). This
data) than for an ionomer electrolyte (Table 1, [29] data). This is due to higher binary hydrogen mixture diffusion coefficient
observation is explained by a series combination of mass
transport resistances (Figure 2, [21, 42]) assuming that with a
He diluent, the hydrogen mass transport resistance in the gas
phase is negligible (Table 3). Nitrogen was selected for dis-
cussion because in practice it accumulates in the H2 stream
by diffusion through the electrolyte:
1 1 1 1
ˆ ‡ ≈ (23)
kHe kg;He ke ke

1 1 1
ˆ ‡ (24)
kN kg;N ke
where kHe represents the overall hydrogen mass transport
coefficient including transport in a He mixture (m s–1), kg,He
the hydrogen mass transport coefficient in a He mixture
Fig. 14 Hydrogen mass transport coefficient Arrhenius plot using some
(m s–1), ke the hydrogen mass transport coefficient in the elec- Figure 7 data [26] for a GDE in contact with a sulphuric acid solution
trolyte (m s–1), kN the overall hydrogen mass transport coeffi- (Table 1, 5.08 cm3 s–1 N2 cases).

270 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.fuelcells.wiley-vch.de FUEL CELLS 11, 2011, No. 2, 263–273
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes

Table 3 Binary gas mixture diffusion coefficients.

ORIGINAL RESEARCH PAPER


Mixture components Temperature / °C Diffusion coefficient(a) / cm2 s–1
H2/N2 20 0.76
41 0.86
60 0.96
H2/He 20 1.75
41 1.98
H2/Ar 20 0.80
H2/CO2 41 0.72
O2/N2 60 0.25
(a)
Calculated at 101.3 kPa by neglecting the effect of water vapour using Fuller
et al’s method [40]. A different method to calculate H2 diffusivity was recently
proposed [41] but several required parameters are unknown for the specific
case considered here.
Fig. 15 Plots of inert gas to hydrogen fraction in the dry inlet reactant
stream (Eq. (8) modified with Eq. (26)) as a function of ^i for several
m parameter values. The f value for a specific case defined by m = 1.2 and
^i ˆ 13:2 is also predicted (Section 3.4).

(Table 3) and hydrogen permeability in an ionomer (Table 2)


than for oxygen, although the influence of electrode design
parameters cannot be ruled out (these were not provided, also increases with m resulting in smaller ^i values (Eq. (7)) for
thus precluding their discussion). otherwise constant operating conditions. For a constant ^i
value, f increases with an increase in m. This is also expected
because the larger flow is accommodated with a more dilute
3.3 Conditions for Model Validity
stream to maintain constant ^i and ie values (Eq. (7)). An f
Many experimental precautions are necessary to achieve value is predicted for an application-specific case. For
well-defined limiting current conditions without the presence instance, potential operating condition targets may include a
of artefacts [6]. An anode kinetic control is achieved by the fuel stoichiometry of 1.2 and a cell current density of 1 A cm–
use of a high cathode stoichiometry, oxygen instead of air, 2
corresponding to an ie value of 1.2 A cm–2. Therefore ^i is
dilute hydrogen streams or low anode stoichiometry. Water equal to 13.2 (Eq. (7) with n = 2, F = 96,500 C mol–1,
management is required to avoid electrolyte dehydration k = 1.20 × 10–2 (N2 diluent, Table 1), pr = 200 kPa,
(use of saturated inlet gases), electrolyte dilution (short term R = 8.3143 J mol–1 K–1 and T = 353 K). This is indicated with
experiments) or liquid water flooding (use of a minimal pres- the modified Eq. (8) and Figure 15 (m = 1.2, ^i ˆ 13:2), f = 6.9
sure drop to drive liquid water droplets out of the cell [45, 46] (fH = 0.13). A relatively dilute hydrogen stream is therefore
and maintain a constant pressure or dilute hydrogen streams still able to fulfil the required operating condition targets. The
leading to little liquid water production). The cell tempera- investigation of other relevant cases is possible using a simi-
ture needs to be maintained at a constant level by the use of a lar approach. For example, rather than computing f, k could
high coolant flow rate, a small active surface area or a dilute be computed to achieve a pre-determined f value.
hydrogen stream (heat generation is reduced). A clearly
defined limiting current is obtained by using a sufficiently
low cell voltage or high anode potential (potentiostatic con- 4 Conclusion
trol preferable [6]). All these precautions are not necessarily
The validated model offers a relatively simple alternative
sufficient as indicated by Figures 5 and 8 results. For a low
to further explore different GDE and flow field designs and
KOH concentration, hydroxyl ion transport is presumably
separately optimise gas phase and electrolyte phase transport
controlling (Figure 5) whereas flow measurement errors are
properties. The model represents a significant development
likely responsible for discrepancies between experimental
as higher current density operation was identified as a poten-
data and model predictions (Figure 8).
tial solution to reduce fuel cell cost. Additional work is
ongoing to adapt the model to other cases. For example, in a
3.4 Acceptable Inlet Inert Gas to Hydrogen Fraction solid oxide fuel cell, the water vapour pressure cannot be
treated as a constant, and in a direct methanol fuel cell the
The relationship between the reactant stoichiometry m and
^ anode reactant is a dilute liquid solution suggesting a concen-
N…1† using Eq. (10) or (14) is:
tration definition change.
iave 1 ^
ˆ ˆ1 N…1† (26)
ie m
Acknowledgements
where m represents the reactant stoichiometry. Equation (8)
^
modified by replacing N…1† using Eq. (26) is plotted in Fig- The author is indebted to Dr Rune Halseid for providing
ure 15 using m as a parameter. For a constant f value, ^i additional experimental data that were not reported in the
decreases with an increase in m. This is expected because ie original publication [29].

FUEL CELLS 11, 2011, No. 2, 263–273 www.fuelcells.wiley-vch.de © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 271
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes

List of Symbols
ORIGINAL RESEARCH PAPER

t Gas diffusion layer thickness / m


T Temperature / K
Latin Letters U Average reactant velocity / m s–1
A Geometric electrode surface area / m2 V_ Volumetric inlet gas flow rate / m3 s–1
c Local dry fuel hydrogen concentration in the flow VM Gas molar volume / m3 mol–1
field channel / mol m–3 y Cartesian coordinate along the flow field channel / m
D Diffusion coefficient / m2 s–1 ^
y Dimensionless cartesian coordinate along the flow
E Electrode potential / V versus SHE field channel
E0 Standard electrode potential / V versus SHE
f Inert gas to hydrogen fraction in the dry inlet reac- Greek Letters
tant stream
fapprox f parameter computed from the general reactant flux d Boundary layer thickness / m
equation approximations (Eqs. (9) and (13)) m Reactant stoichiometry
fgen f parameter computed from the general reactant flux
Eq. (8)
fH Hydrogen fraction in the dry inlet reactant stream
References
fN Nitrogen fraction in the dry inlet reactant stream [1] J. Garche, C. K. Dyer, P. Moseley, Z. Ogumi, D. Rand,
F Faraday constant / 96,500 C mol–1 B. Scrosati (Eds.) in Encyclopedia of Electrochemical Power
i Current density / A m–2 Sources, Elsevier, Amsterdam, 2009.
^i Ratio of the limiting current density il with Nd = 0 [2] M. G. Santarelli, M. F. Torchio, P. Cochis, J. Power
(f = 0) to the inlet hydrogen flow rate equivalent cur- Sources 2006, 159, 824.
rent density ie (Eq. (7)) [3] B. Wetton, G.-S. Kim, K. Promislow, J. St-Pierre, Proc.
iave Average limiting current density / A m–2 4th Int. Conf. Fuel Cell Sci., Eng. Technol., Irvive, CA,
ie Inlet hydrogen flow rate equivalent current density / USA, 2006, paper FUELCELL2006-97027.
A m–2 [4] J. P. Meyers, R. M. Darling, J. Electrochem. Soc. 2006, 153,
il Limiting current density / A m–2 A1432.
k Overall hydrogen mass transport coefficient / m s–1 [5] S. S. Kocha, J. D. Yang, J. S. Yi, AIChE J. 2006, 52, 1916.
kCSTR Overall hydrogen mass transport coefficient calcu- [6] J. St-Pierre, B. Wetton, G.-S. Kim, K. Promislow, J. Elec-
lated using a CSTR model (Eq. (17)) / m s–1 trochem. Soc. 2007, 154, B186.
ke Hydrogen mass transport coefficient in the electro- [7] N. R. Elezovic, L. Gajic-Krstajic, V. Radmilovic, L. Vra-
lyte / m s–1 car, N. V. Krstajic, Electrochim. Acta 2009, 54, 1375.
kg,He Hydrogen mass transport coefficient in a He mixture / [8] R. Borup, J. Meyers, B. Pivovar, Y. S. Kim, R. Mukun-
m s–1 dan, N. Garland, D. Myers, M. Wilson, F. Garzon,
kg,N Hydrogen mass transport coefficient in a N2 mixture / D. Wood, P. Zelenay, K. More, K. Stroh, T. Zawodzins-
m s–1 ki, J. Boncella, J. E. McGrath, M. Inaba, K. Miyatake,
kHe Overall hydrogen mass transport coefficient includ- M. Hori, K. Ota, Z. Ogumi, S. Miyata, A. Nishikata,
ing transport in a He mixture / m s–1 Z. Siroma, Y. Uchimoto, K. Yasuda, K.-i. Kimijima,
kln Overall hydrogen mass transport coefficient calcu- N. Iwashita, Chem. Rev. 2007, 107, 3904.
lated using a logarithmic mean concentration differ- [9] F. A. de Bruijn, V. A. T. Dam, G. J. M. Janssen, Fuel Cells
ence model (Eq. (20)) / m s–1 2008, 8, 3.
kN Overall hydrogen mass transport coefficient includ- [10] S. D. Knights, K. M. Colbow, J. St-Pierre, D. P. Wilkin-
ing transport in a N2 mixture / m s–1 son, J. Power Sources 2004, 127, 127.
kPEMFC Overall hydrogen mass transport coefficient calcu- [11] A. Taniguchi, J. Power Sources 2004, 130, 42.
lated using a PEMFC model (Eqs. (8)–(10)) / m s–1 [12] T. W. Patterson, Electrochem. Solid-State Lett. 2005,
L Flow field channel length / m 8, A273.
n Number of electrons exchanged in the hydrogen oxi- [13] L. C. Colmenares, A. Wurth, Z. Jusys, R. J. Behm,
dation reaction J. Power Sources 2009, 190, 14.
N Hydrogen gas flux / mol m–1 s–1 [14] E. Antolini, E. R. Gonzalez, Solid State Ionics 2009,
^
N Dimensionless hydrogen gas flux 180, 746.
Nd Diluent gas flux / mol m–1 s–1 [15] H. Chhina, S. Campbell, O. Kesler, J. Electrochem. Soc.
p Inlet reactant stream pressure / Pa 2009, 156, B1232.
pr Dry inlet reactant stream pressure / Pa [16] M. K. Debe, A. K. Schmoeckel, G. D. Vernstrom, R. Ata-
psat Water vapour saturation pressure / Pa nasoski, J. Power Sources 2006, 161, 1002.
Pé Péclet number [17] L. Gancs, T. Kobayashi, M. K. Debe, R. Atanasoski,
R Ideal gas constant / 8.3143 J mol–1 K–1 A. Wieckowski, Chem. Mater. 2008, 20, 2444.

272 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.fuelcells.wiley-vch.de FUEL CELLS 11, 2011, No. 2, 263–273
St-Pierre: Hydrogen Mass Transport in Fuel Cell Gas Diffusion Electrodes

ORIGINAL RESEARCH PAPER


[18] H. S. Fogler, Elements of Chemical Reaction Engineering, [33] E. Ivers-Tiffée, A. V. Virkar in High Temperature Solid
4th edition, Prentice Hall, Upper Saddle River, NJ, 2006, Oxide Fuel Cells: Fundamentals, Design and Applications
pp. 958. (Eds. S. C. Singhal, K. Kendall), Elsevier, Oxford, UK,
[19] J. Diep, D. Kiel, J. St-Pierre, A. Wong, Chem. Eng. Sci. 2003, pp. 234.
2007, 62, 846. [34] A. A. Kulikovsky, Electrochem. Commun. 2004, 6, 969.
[20] P. Berg, K. Promislow, J. St-Pierre, J. Stumper, B. Wet- [35] F. Barbir, PEM Fuel Cells: Theory and Practice, Elsevier
ton, J. Electrochem. Soc. 2004, 151, A341. Academic Press, Burlington, MA, USA, 2005, pp. 46.
[21] R. B. Bird, W. E. Stewart, E. N. Lightfoot, Transport Phe- [36] P. Ruetschi, J. Electrochem. Soc. 1967, 114, 301.
nomena, revised 2nd edition, John Wiley & Sons, New [37] R. M. Q. Mello, E. A. Ticianelli, Electrochim. Acta 1997,
York, 2007, pp. 423, 517, 675 and 687–690. 42, 1031.
[22] A. A. Kulikovsky, A. Kucernak, A. A. Kornyshev, Elec- [38] J. Jiang, A. Kucernak, J. Electroanal. Chem. 2004, 567, 123.
trochim. Acta 2005, 50, 1323. [39] L. Zhang, C. Ma, S. Mukerjee, J. Electoanal. Chem. 2004,
[23] M. V. Williams, H. R. Kunz, J. M. Fenton, J. Electrochem. 568, 273.
Soc. 2004, 151, A1617. [40] B. E. Poling, J. M. Prausnitz, J. P. O’Connell, The Proper-
[24] T. V. Nguyen, J. Electrochem. Soc. 1996, 143, L103. ties of Gases and Liquids, 5th edition, McGraw-Hill, New
[25] K. Broka, P. Ekdunge, J. Appl. Electrochem. 1997, 27, 117. York, 2001, pp. 11.10.
[26] J. J. T. T. Vermeijlen, L. J. J. Janssen, J. Appl. Electrochem. [41] Y. Shi, J. Xiao, S. Quan, M. Pan, L. Zhang, Int. J. Hydro-
1993, 23, 26. gen Energy 2010, 35, 2863.
[27] J. J. T. T. Vermeijlen, L. J. J. Janssen, J. Appl. Electrochem. [42] U. Beuscher, J. Electrochem. Soc. 2006, 153, A1788.
1993, 23, 1237. [43] S. Gopalakrishnan, J. C. Diniz Da Costa, J. Membr. Sci.
[28] F. Alcaide, E. Brillas, P.-L. Cabot, J. Electrochem. Soc. 2008, 323, 144.
2005, 152, E319. [44] D. R. Baker, D. A. Caulk, K. C. Neyerlin, M. W. Murphy,
[29] R. Halseid, R. Tunold, J. Electrochem. Soc. 2006, 153, J. Electrochem. Soc. 2009, 156, B991.
A2319. [45] J. St-Pierre, D. P. Wilkinson, H. Voss, R. Pow, Proc. 2nd
[30] K. Zhukovsky, A. Pozio, J. Power Sources 2004, 130, 95. Int. Symp. New Mater. Fuel Cell Modern Battery Systems
[31] A. A. Kulikovsky, Electrochem. Commun. 2002, 4, 527. (Eds. O. Savadogo, P. R. Roberge), Montréal, QC, Cana-
[32] A. Kazim, H. T. Liu, P. Forges, J. Appl. Electrochem. 1999, da, 1997, pp. 318.
29, 1409. [46] D. P. Wilkinson, J. St-Pierre, J. Power Sources 2003,
113, 101.

______________________

FUEL CELLS 11, 2011, No. 2, 263–273 www.fuelcells.wiley-vch.de © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 273

Vous aimerez peut-être aussi