You are on page 1of 14

Journal of The Electrochemical

Society

Model for Polymer Electrolyte Fuel Cell Operation on Reformate Feed:


Effects of CO, H;2 Dilution, and High Fuel Utilization
To cite this article: T. E. Springer et al 2001 J. Electrochem. Soc. 148 A11

View the article online for updates and enhancements.

This content was downloaded from IP address 109.171.234.170 on 12/02/2023 at 17:43


Journal of The Electrochemical Society, 148 (1) A11-A23 (2001) A11
S0013-4651/2001/148(1)/A11/13/$7.00 © The Electrochemical Society, Inc.

Model for Polymer Electrolyte Fuel Cell Operation on


Reformate Feed
Effects of CO, H2 Dilution, and High Fuel Utilization
T. E. Springer, T. Rockward, T. A. Zawodzinski,* and S. Gottesfeld**,z
Los Alamos National Laboratory, Electronic and Electrochemical Materials and Devices Group, Los Alamos,
New Mexico 87545, USA

We describe a polymer electrolyte fuel cell model emphasizing operation on hydrocarbon reformate, i.e., the anode feed stream con-
sists of dry H2 concentrations as low as 40%, inlet CO levels of 10-100 ppm, and hydrogen fuel utilization as high as 90%. Refine-
ments of interfacial kinetics equations used in our previous work on CO effects in H2 anodes have yielded a better quantitative fit to
the measured dependence of voltage loss on inlet CO level [in Electrode Materials and Processes for Energy Conversion and Stor-
age, J. McBreen, S. Mukerjee, and S. Srinivasan, Editors, PV 97-13, pp. 15-24, The Electrochemical Society Proceedings Series, Pen-
nington, NJ (1997)]. We calculate anode potential losses by coupling such interfacial kinetic processes to reactant diffusion limita-
tions and ionic resistance in the catalyst layer, and by accounting for the drop in local hydrogen concentration along the flow chan-
nel due to significant fuel utilization. As a result of internal readjustment of cell overpotentials when hydrogen concentration drops
along the flow channel, we show that loss of current, or power, under the realistic condition of constant cell voltage is smaller than
loss of current at constant anode potential. We show that voltage losses associated with CO poisoning are significantly amplified with
diluted hydrogen feed streams and particularly so under high fuel utilization. We make projections on improvements required, qual-
itative and quantitative, in the physical parameters of the anode catalyst surface chemistry to significantly improve “CO tolerance.”
© 2000 The Electrochemical Society. S0013-4651(00)03-119-0. All rights reserved.

Manuscript submitted March 27, 2000; revised manuscript received August 29, 2000.

Recently developed fuel cell power systems for transportation ap- guide the proper design and operation of a fuel cell power system. It
plications have relied on processing of liquid fuels on board the vehi- can be used to study the effects of total anode pressure, temperature,
cle to generate a hydrogen-rich gas mixture as the feed stream for a hydrogen dilution in N2 and/or CO2, and fuel utilization along the
polymer electrolyte fuel cell (PEFC) stack. Steam reforming of meth- anode flow field. These effects can also be studied under realistic
anol and autothermal reforming of gasoline-like fuels have been the reformate anode conditions of 10-100 ppm CO levels and under low
two most commonly considered and investigated systems of on-board or high power demand. We compare model results to data measured
fuel processing. Methanol has the advantage of a relatively low-tem- in our laboratory to determine certain parameter values and to show
perature steam-reforming process, yielding a gas mixture with hydro- the ability of the model to reproduce experimental results.
gen at levels as high as 75% of dry gas composition. Gasoline, on the Some recent experimental work focused on surface catalyst deac-
other hand, has an advantage of relying on an existing fuel infra- tivation by low levels of CO. In their studies of hydrogen oxidation on
structure but requires significantly higher temperatures for auto- Pt and Pt alloy (Pt/Sn, Pt/Ru) rotating disk electrodes exposed to
thermal processing. The latter process yields typical compositions of H2/CO mixtures, Gasteiger et al 3,4 showed that a small hydrogen oxi-
35-40% hydrogen in the dry gas mixture (see, for example, Ref. 1) In dation current is observed well before the onset of major CO oxidative
both of these fuel-processing systems, water-shift and fuel cleanup stripping (ca. 0.4 V) on Pt/Ru. However, they concluded that such cur-
steps follow the steam or autothermal reforming step. These steps rent observed at low anode overpotential was “too low to be of practi-
lower CO levels in the product gas mixture from several percent down cal value.” Nonetheless, both others and we5-7 have found experimen-
to 10-100 ppm as required for use as an anode feed stream in a PEFC tally that it is possible to operate a PEFC with a Pt/Ru anode with CO
operating at 80⬚C. The anode feed stream in such systems is thus typ- levels in the range 10-100 ppm at CO voltage losses of the order of
ified by hydrogen mole fractions (before humidification) ranging 10–100 mV. Such past5-7 fuel cell measurements were not performed
between 0.4 and 0.75 and by the presence of 10-100 ppm CO, a seri- with diluted hydrogen feed streams under high fuel-utilization condi-
ous poison of Pt based anode catalysts operating at 80⬚C. tions, where the effects of a given CO level are amplified, as shown in
To quantitatively evaluate PEFC stack losses originating from the this paper. These past results suggested that PEFC operation at signif-
combined effects of hydrogen dilution and CO surface poisoning, a icant current densities may be possible in the presence of 10-100 ppm
comprehensive model for a PEFC anode is required. Such a model CO, even without resorting to air bleeding into the anode feed stream.8
should elucidate the effects of H2 dilution alone, particularly under The latter approach has been shown to be highly effective in elimina-
high fuel-utilization conditions, and then treat the implications of tion of Pt anode catalyst poisoning effects at CO levels as high as 500
10-100 ppm levels of CO in diluted hydrogen under high-utilization ppm, in cells operating at 80⬚C with low Pt catalyst loading, at the cost
conditions. We described in an earlier contribution2 modeling of of 2-4% in cell fuel efficiency. Air bleeding lowers CO partial pressure
anode losses incurred by addition of CO at 10-100 ppm levels to a to extremely low levels in the anode plenum by catalytic (chemical)
neat hydrogen anode feed stream. In this paper, we refine parameters oxidation of CO by dioxygen at the anode catalyst. In this modeling
used for these interfacial kinetic equations to achieve better simulta- work we do not include, however, any description of oxygen bleeding
neous fitting of experimental data obtained with a range of CO inlet effects and concentrate on the behavior of the anode with feed streams
levels. We next model effects of dilution of hydrogen in inert dilu- of limited flow rate (i.e., under high fuel utilization), consisting of H2
ents (N2 and CO2) under high fuel-utilization conditions, and final- diluted in inerts, with or without CO.
ly, we model the effects of CO poisoning under such hydrogen dilu-
tion and high-utilization conditions. Theoretical Anode Model
The anode model described in this paper is based on physical prin- We initially address the kinetic equations for the catalyst/ionomer
ciples whose input parameters have physical meaning. The model can interface that provides local current generation as a function of local
** Electrochemical Society Active Member.
hydrogen and CO concentrations and the local overpotential. Subse-
** Electrochemical Society Fellow. quently, we address the mass-transport equations that set the local
*z E-mail: gottesfeld@lanl.gov conditions at the catalyst surface when a cell is operated with diluted
A12 Journal of The Electrochemical Society, 148 (1) A11-A23 (2001)
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

hydrogen and at high fuel utilization. We stress again that this model hydrogen electrode (RHE)]. An initial step of “water discharge” is a
is intended to study effects of CO in operation on diluted, highly uti- likely rate-determining step in the electro-oxidation of adsorbed CO
lized hydrogen fuel when air bleeding is not used. at such low anode potentials, forming an active surface oxygen inter-
Kinetic equations for hydrogen anode.—We would like to use a mediate, e.g., OHads, that subsequently reacts readily with adsorbed
single set of kinetic equations for all cases of anode feeds. In either CO. As further discussed later, different rates for this initial step of
the presence or absence of CO in the anode gas stream, the same water discharge at a given low anode potential is one possible expla-
kinetic equations should still be valid. We would also like the kinetic nation for different degrees of “CO tolerance” exhibited by various
equations to predict measured cell performance for a variety of CO anode catalysts (e.g., PtRu vs. Pt).
concentrations in the feed stream without changing any model para- Equations 5 and 6 represent the steady-state catalyst surface bal-
meters, other than xco. The kinetic equations are a small fraction of the ance of adsorption, desorption, and charge-transfer fluxes of CO and
modeling code but, in the presence of CO, have a major effect. Where hydrogen. If the intermediate hydrogen step is second order in catalyst
we have observed some discrepancies between simpler model predic- sites, then n ⫽ 2. If it were first order, then n ⫽ 1. The second-order
tion and experiment, we have invoked modified assumptions that solution is given here; the first-order solution, considered by Watan-
apparently better describe the actual physical processes occurring. For abes group,10 is presented in Appendix A. These equations determine
example, we have now made some interfacial rate constants be func- ␪CO, the fraction of catalyst sites with adsorbed CO, and ␪h, the frac-
tions of fractional CO coverage on the catalyst sites. We assume that tion with adsorbed H at steady state, at anode potential Va. The fluxes
isothermal conditions exist, that the feed stream is always saturated are also all in units of equivalent current density. The quantity ␳ repre-
with water vapor, and that the kinetic equations are not affected by the sents the molar area density of catalyst sites times the Faraday constant
presence of N2 and CO2 except for their effect on the partial pressures ␳␪˙ CO ⫽ kfc x CO PA (1 ⫺ ␪ CO ⫺ ␪ h )
of H2 and CO.
We assume the four processes expressed in Eq. 1-4 determine the Va ⫹VNernst
interfacial kinetics. By itself M corresponds to a vacant catalyst site, ⫺ bfc kfc␪ CO ⫺ kec␪ CO e bc
⫽ 0 [5]
otherwise H or CO is attached to the site
␳␪˙ h ⫽ kfh x h PA (1 ⫺ ␪ CO ⫺ ␪ h ) n ⫺ bfh kfh ␪ nh
kfc
CO ⫹ M o (M-CO) [1]  V ⫹ VNernst 
bfc(␪CO)kfc ⫺ 2␪ h keh sinh  a  ⫽ 0 [6]
 bh 
kfh(␪CO)
Va ⫹VNernst
H2 ⫹ 2M o 2(M-H) [2]  V ⫹ VNernst 
bfhkfh(␪CO) jh ⫽ 2 keh␪h sinh  a  , jCO ⫽ 2 kec␪CO e bc [7]
 bh 
keh Current density expressions for hydrogen and CO oxidation, jh and jCO,
(M-H) r H⫹ ⫹ e⫺ ⫹ M [3] are given in Eq. 7 as a function of anode potential Va referenced to a
kec standard hydrogen electrode (SHE) corrected for local Nernstian po-
H2O ⫹ (M-CO) r M ⫹ CO2 ⫹ 2H⫹ ⫹ 2e⫺ [4] tential shift associated with local values of PH2 different than unity. Va
can be negative if the hydrogen partial pressure is greater than 1 atm.
The first two equations represent the competing processes of CO ad- However, Va ⫹ VNernst will never be negative. We assume a transfer
sorption and dissociative chemisorption of H2. The forward rate con- coefficient of 0.5 for the electro-oxidation of hydrogen, reducing the
stants for hydrogen and CO adsorption, kfh and kfc, are expressed in Butler-Volmer equation to a hyperbolic sine function that allows the
A/cm2 (geometric) atm. The desorption rates are implicit in bfc and current to approach zero at zero overpotential. Note that the same over-
bfh, the ratio of backward to forward rate constants with units of at- potential is assumed at a given anodic potential for both Hads and COads
mospheres, i.e., the (inverse) equilibrium constants for processes 1 electro-oxidation. This means that both kfh and kfc describe the rate of
and 2, respectively. As indicated, we assume that the rate constant kfh the anodic process at 0 V RHE (at full surface coverage by the corre-
and the backward/forward ratio bfc are functions of CO site occupa- sponding surface species). A Tafel relation can be used for the poten-
tion. In other words, the desorption rate for CO and the adsorption rate tial dependence of current density for CO electro-oxidation (Eq. 7).
for hydrogen are assumed to be functions of ␪CO, whereas it was con- If all the rate constants are independent of ␪CO, i.e., kfh and bfc are
venient to hold the adsorption/desorption rate ratio for hydrogen and constant, then an analytic solution for ␪h and ␪CO is given by Eq. 9
the adsorption rate for CO constant with CO coverage. The hydrogen with intermediate variables defined in Eq. 8. From the expressions
and CO electro-oxidation rate constants keh and kec are expressed in for jh and jCO in Eq. 7, the kinetic current densities of hydrogen and
A/cm2 in Eq. 3-4. The third process (Eq. 3) generates the current den- of CO electro-oxidation, can now be evaluated
sity corresponding to anodic hydrogen oxidation, i.e., the electro-
chemical oxidation of adsorbed hydrogen atoms. The fourth process  V ⫹ VNernst 
s1 ⫽ keh sinh  a 
(Eq. 4), corresponding to the electrochemical oxidation of CO to CO2,  bh 
reaches significant rates particularly at high anode overpotential, gen-
erating at such high overpotential a larger number of CO-free catalyst Va
sites under steady-state conditions. However, we show that even very a1 ⫽ e bc kec ⫹ kfc (bfc (b fc ⫹ PA x CO )
low rates of process (Eq. 4), that would apply at low anode potentials,
can be sufficient to significantly increase the rate of hydrogen electro-  Va 
oxidation in the presence of CO. Note that in Eq. 4, the oxygen atom a2 ⫽  e bc kec ⫹ bfc kfc  a3 ⫽ bfh kfh ⫹ 2 s1 [8]
 
required for complete electro-oxidation of CO is assumed to originate
from a water molecule. Recent analysis by Cairns’ group9 showed that
the overall stoichiometry implied by Eq. 4, i.e., two electrons per  a a a k P x ⫹ a2s2 ⫺ a2s ⫺ a k P x 
 1 2 3 fh A H 1 1 1 1 2 fh A H 
adsorbed CO molecule, indeed appears to apply at Pt catalysts under ␪H ⫽
the relevant low anode overpotentials [e.g., <200 mV vs. the reference kfh (bfh a12 ⫺ a2 PA x H )

kfc PA x CO  a1b fh kfh ⫹ a1s1 ⫺ 2


a2 bfh kfh PA x H ⫹ 2 a2 kfh PA x H s1 ⫹ a12 s12 
 
␪ CO ⫽ [9]
kfh (bfh a12 ⫺ a2 PA x H )
Journal of The Electrochemical Society, 148 (1) A11-A23 (2001) A13
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

(A solution for first-order kinetics at uncovered sites is given in


Appendix A.)
As indicated in Eq. 1, bfc is next made a function of ␪CO. By using
a Temkin isotherm11 Eq. 10 allows the free energy of CO adsorption
to decrease linearly with ␪CO. Use of a Temkin isotherm was advo-
cated originally by Bellows and Marcuchi-Soos12 as a means to
achieve a better fit to the experimentally observed variation of the
critical current with CO level. ␦(⌬GCO) is the difference in adsorption
free energy assumed between ␪CO ⫽ 1 and ␪CO ⫽ 0. Conrad et al.13
have found the binding energy for CO on single-crystal Pd decreased
from 35 kcal/mol at zero coverage to 5 kcal/mol at saturation cover-
age of 0.8. For Pt a decrease from 30 kcal/mol at a surface coverage
of ␪CO ⫽ 0.12-10 kcal/mol at ␪CO ⫽ 0.69 has been found. Other
researchers14 have found the heat of adsorption of CO on Pt 111 to
decline with increasing coverage from 32 to 12 kcal/mol. Thus in
Eq. 10 we express the CO adsorption-to-desorption rates ratio as
␦( ⌬Gco )
␪ co [10]
bfc ⫽ bfc0 e RT

Figure 1. Plot of bfc and kfh as functions of ␪CO (see Eq. 10 and B-2) using
Because the model has been found relatively insensitive to the actu- ␦(⌬GCO)/RT ⫽ 6.8, ␦(⌬EH)/RT ⫽ 4.6 and p ⫽ 5.
al forward rate kfc, we consider it constant. The variation of bfc with
␪CO in Eq. 10, using a ␦(⌬GCO) value of 20 kcal/mol, is needed to
reproduce the proper dependence of the “critical current density” calculates a local current density and anode potential along the
(current at which the anode potential begins to increase steeply) on anode flow field, where the local concentrations in the flow channel
CO partial pressure. have been determined by inlet flow conditions and cell current.
In Eq. 2 we indicate that we also want the H2 dissociative chemi- We show in the next section how the current distribution down the
sorption rate constant to be a function of ␪CO. This proposed relation flow channel and the average current density can be calculated as a
is needed to explain, with the same model, anode polarization results function of fuel utilization using this 1D model. (Obviously, the 1D
obtained with hydrogen greatly diluted in inerts (e.g., nitrogen) and model also describes the current density of the total cell for the case of
anode polarization results obtained with neat hydrogen containing “zero” fuel utilization or very high anode flow rate). Under significant
10-100 ppm CO. The results with diluted hydrogen imply very high fuel utilization conditions, we apply the 1D model repeatedly for each
rates of hydrogen dissociative chemisorption on a CO-free anode of the various dashed 1D sections in Fig. 2. In each section, the back-
catalyst surface. We found experimentally (to be discussed) that with ing inlet composition, the current density, and the anode overpotential
only 6% hydrogen in the flow channel, the kinetic limiting rate assume different values, even though the cell voltage remains constant
derived from the dissociative chemisorption step still significantly along the flow channel. To avoid confusion, we let J represent current
exceeds 1 A/cm2, i.e., it must be of the order of 10 A/cm2, at least. density in the 1D model with the local value for J varying along the
Consideration of the combined effects of hydrogen dilution by more flow channel, and I represents the average cell current density. As
than an order of magnitude (6/100) vs. the standard state of 1 atm, shown in Fig. 2, the remaining (non-anode) part of the cell is includ-
and the limited anode catalyst utilization calculated with such a ed here without detail. We include in this 1D PEFC model a “standard
diluted hydrogen feed stream (see the following discussion), leads to cathode-membrane voltage,” shown in Fig. 3, which is the cell voltage
the conclusion that kfh must significantly exceed 100 A/cm2, i.e., be vs. current relationship expected at zero anode loss. This standard
of the order of 1000 A/cm2 under the standard condition PH2 ⫽ curve is used to determine the anode potential at a fixed overall cell
1 atm. However, based on the recorded effects of CO, kfh apparently voltage, based on the local anode conditions and all other cell voltage
falls sharply in the presence of partial catalyst surface coverage by
CO, otherwise, limiting currents of less than 1 A/cm2 would not be
expected from Eq. 6 (assuming n ⫽ 2) even at high surface coverage
by CO, contrary to experimental observation of “critical currents.”
To generate an effect of ␪CO on kfh, as required to explain the exper-
imental results, we propose a surface catalyst model described in
Appendix B. Figure 1 shows a plot of bfc and kfh as functions of ␪CO,
using values of ␦(⌬GCO), ␦(⌬EH), and p (see Eq. 10 and B-2) re-
quired to explain our experimental results both without and with CO.

1D anode model and consideration of fuel utilization.—We turn


now to a description of the transport processes in PEFC anodes oper-
ated on reformate at high fuel utilization. These transport processes
determine the local concentration and local interfacial potential dif-
ference at an anode catalyst site located within the catalyst layer.
Limited rates of such transport processes could be significant
sources of voltage loss for operation with diluted hydrogen. Figure
2 is a schematic diagram of a general PEFC model that is the subject
of this paper. Depicted along the left and right sides of the diagram
are anode and cathode flow channels through which reactant gases
flow vertically downward. Lying horizontally across the cell
between the two flow channels is a series of differential one-dimen-
sional fuel cell sections. A typical 1D section model is depicted at
the bottom. This 1D model has specified anode gas composition at
its left side, an anode gas-diffusion-backing region, a catalyst layer
region, and a “standard cathode-membrane” region. The 1D model Figure 2. PEFC anode model schematic.
A14 Journal of The Electrochemical Society, 148 (1) A11-A23 (2001)
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

would be measured at constant anode potential. This is a result of the


“overpotential trade-off” described previously.
As shown at the bottom of Fig. 2, the 1D model consists of three
sections; gas diffusion backing, catalyst layer, and the “standard
cathode-membrane.” We now describe the assumptions and equa-
tions for these sections.

Anode backing layer.—The backing usually consists of carbon


cloth filled with wetproof carbon powder. We assume the gas mole-
cules transported through the backing collide with other gas mole-
cules at much higher frequency than with the surrounding backing
materials. Then the total pressure remains constant and the Stefan-
Maxwell equations (Eq. 11) describe the shift in component mole
fractions xi as a function of steady-state component fluxes ␾i
through the backing. Indexes 1-5 correspond to H2, CO, N2, CO2,
and H2O respectively. If water vapor remained unsaturated with no
liquid water in the backing, the five linear first-order differential
equations can be solved analytically. With the constraint on the fifth
equation, there must be a flux of water vapor ␾5 (Eq. 13) through a
hydrophobic component of the backing that is a function of the mole
Figure 3. Standard cathode-membrane polarization curve, Vstd, used in this
paper to convert the “anode model” to a “cell model.” Vstd ⫽ 0.8177 ⫺
fractions. A corresponding return flux of liquid must flow through
0.1897Jtot ⫺ 0.0176 ⫻ ln[Jtot] ⫺ 1.0077 ⫻ 10⫺5 exp[5.3687]. Vstd is equiv- part of the cross-sectional area. The treatment of saturated water
alent to cell polarization with no anode losses. vapor in a cathode model has been described previously.15 Eliminat-
ing ␾5 from the first four equations makes them nonlinear. Howev-
er, a linear solution with an error of only about 1% is introduced by
losses (represented by the standard curve) at a given current density. holding ␾5 constant at the inlet mole fractions. Resolving the ana-
Such consideration of current continuity through each segment of the lytic solution at the exit conditions for ␾5 verifies this, and averag-
cell converts the PEFC anode model to a PEFC cell model. ing the two solutions eliminates most of the error
Addressing the changing anode conditions as a cell, rather than
∂x i xi ⌽ j ⫺ x j⌽ i

anode, problem reveals an interesting “trade-off” that takes place in V ␶T
reality between anode and other voltage losses. As the H2 mole frac- 1 ⱕ i ⱕ 4: ⫽ S
∂z ⑀TS DSij
tion decreases, the constant cell voltage condition for the 1D model j⫽1
forces an increase in local anode overpotential at the expense of low-
∑ ␣ (x ⌽ ) VS ␶T
ering in combined cathode-membrane overpotential, resulting in a ⫽ ij i j ⫺ x j ⌽ i ; ␣ ij ⫽ [11]
relatively smaller decrease in local current density than that expected ⑀TS DSij
j⫽1
at constant anode potential. This is explained by the example given in
Fig. 4. This figure describes solutions for base-case parameters JH J
(Table I) of the 1D model (discussed in detail later) for a range of ⌽1 ⫽ ; ⌽ 2 ⫽ CO ; ⌽ 3 ⫽ 0; ⌽ 4 ⫽ 0 [12]
2F 2F
concentrations of hydrogen below 40%, representative of flow chan-
∂x 5
∑␣
nel concentrations. Two current density vs. mole fraction curves are
⫽0⫽ 5 j ( x5 ⌽ j ⫺ x j ⌽ 5 );
shown, one at constant Vcell and one at constant Vanode. As seen in ∂z
Fig. 4, current and power losses incurred from hydrogen dilution j⬆1
under conditions of constant overall cell voltage (that always apply to
the segments of a cell in a stack) are seen to be smaller than those that ⌽ 5 ⫽ x 5 ( ␣ 51⌽1 ⫹ ␣ 52 ⌽ 2 ) / ∑␣
j⬆5
5jx j [13]

Anode catalyst layer.—We assume the catalyst layer has a uni-


form effective ionic conductivity ␴ and uniform effective diffusion
coefficients DH and DCO for hydrogen and CO. Local generated
kinetic ionic current densities, jH and jCO from hydrogen and CO
conversion, respectively, are expressed as equivalent A/cm2 per geo-
metric electrode area. Both jH and jCO are functions of CH and CCO
and local overpotential ␩

Table I. The base-case parameters used for the anode model.

bc 0.06 V ᐉcl 5 ⫻ 10⫺4 cm


bh 0.032 V ␴ 0.03 S cm⫺1
bfc0 1.510⫺8 atm kec 1 ⫻ 10⫺8 A cm⫺2
bfh 0.5 keh 4.0 A cm⫺2
DCO 5.34 ⫻ 10⫺5 cm2 s⫺1 kfc 10 A cm⫺2 atm⫺1
DH 2 ⫻ 10⫺4 cm2 s⫺1 kfh 4000 A cm⫺2 atm⫺1
Figure 4. 1D model calculations of the variations of anode potential and cell ⑀/␶ 0.26 Tcell 80⬚C
current density with dry hydrogen mole fraction. Current density is shown for K1 0.8177 V ␦(⌬GCO)/RT 6.8
two cases of (---) constant cell voltage of 0.6 V and (—) constant anode K2 0.1897 ⍀ cm2 Pa 1-3 atm
potential. The constant anode potential corresponds to a H2 mole fraction of K3 0.0176 ⑀/␶ 0.26
0.4. With fall in hydrogen concentration, cell current is seen to fall to a less- K4 1.0077 ⫻ 10⫺5 V ␦(⌬EH)/RT 4.6
er extent under the (realistic) conditions of constant cell voltage, as compared K5 5.3687 cm2 A⫺1
with conditions of constant anode potential. ᐉB 0.03 cm
Journal of The Electrochemical Society, 148 (1) A11-A23 (2001) A15
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

tions of cell voltage and gas composition at the anode flow field/back-
∂ 2 CH jH ᐉ cl 2 FPATS xHF/ B DH
⫽ ; I Dh ⫽ ; ing interface. We next develop a method for obtaining integrated cell
∂y 2 IDh VST ᐉ cl current generated along the flow field where the backing entrance mole
fractions vary as hydrogen and CO are utilized along the flow channel.
xHB/ CL ∂CH Let w be the normalized position variable along the flow field.
CH (0) ⫽ ; (1) ⫽ 0 [14]
xHF/ B ∂y Along the flow field the H2 mole fraction changes so that the rate of
change of H2 flow equals the equivalent H2 current produced per unit
∂ 2 CCO jCO ᐉ cl F/B
2 FPATS xCO DCO length of w. A similar relation for CO exists. We consider as CO cur-
2 ⫽ I ; IDCO ⫽ ; rent that which arises from Eq. 4. We express as f the molar flow of
∂y DCO VS T ᐉ cl
the feed stream normalized to that of a unit area of the fuel cell and
B/ CL
xCO ∂CCO expressed in terms of equivalent current density for hydrogen, or CO
CCO (0) ⫽ ; (1) ⫽ 0 [15] (i.e., multiplied by two times the Faraday constant F). Utilization u is
F/B
xCO ∂y the fraction of fuel gas consumed producing cell current
∂ 2 Va ( jH ⫹ jCO )ᐉcl ∂Va u ⫽ Iave/(xHIN ⫹ x CO
IN
)f IN ⫽ 1 ⫺ [(xH ⫹ xCO)f ]/[(xHIN ⫹ x CO
IN
)f IN] [20]
⫽ ; (0) ⫽ 0; Va (1) ⫽ Vanode [16]
∂y 2 ␴ ∂y The mole fractions xH and xCO correspond to the location where the
Accumulated ionic current densities JH and JCO range from zero average current density reaches Iave, which can be any value of u be-
at the backing interface to the total current density at the membrane tween 0 and 1.
interface. The total current density from H2 and CO electro-oxida- Because only H2 and CO are consumed along the flow field, the
tion at the membrane interface is equal to the respective fluxes times total molar flow rate everywhere is
2F at the backing interface 1 ⫺ x HIN ⫺ x CO
IN
f ⫽ fIN [21]
tot ∂CH tot 1 ⫺ x H ⫺ x CO
J H ⫽ 2 F⌽1 ⫽ ⫺I Dh (0); J CO ⫽ 2 F⌽ 2
∂y and the combined mole fraction of the reacting species is
∂CCO ( x HIN ⫺ x CO
IN
)(1 ⫺ u)
⫽ ⫺I DCO x H ⫹ x CO ⫽ [22]
∂y y⫽0 1 ⫺ ( x HIN ⫺ x CO
IN
)u
[17]
We wish to specify utilization and then determine the total current
␴ ∂␩ density Jtot. Because we do not know the feed rate fIN to achieve this
J tot ⫽ J Htot ⫹ J CO
tot
⫽ [17]
ᐉcl ∂y y⫽1
utilization, it is convenient to switch the integration variable from w,
the position down the flow channel, to u⬘(w), the local utilization
At the membrane interface a relationship for the “standard cathode- down the flow channel
membrane” voltage Vstd as a function of current density is supplied w

∫ (I
to determine Vanode for a given cell voltage, Vcell dw
( x HIN ⫹ x CO
IN
) fIN u ′( w ) ⫽ H ⫹ ICO )dw or
Vanode ⫽ Vstd(Jtot) ⫺ Vcell [18] du ′
0
As explained previously, the purpose of introducing this polarization ( x HIN ⫹ x CO
IN
) fIN
relation for the cathode and membrane in an anode model is to eval- ⫽ [23]
I H ⫹ ICO
uate the anode under the valid constant-cell-voltage condition. To
generate Vstd as a function of current density (Fig. 3), we used an Equating the loss of hydrogen and CO along the flow field to the cur-
experimental PEFC voltage vs. current density relationship using rent produced, we obtain
neat hydrogen feed and an air cathode, corrected for the small anode df
overpotential associated with neat hydrogen operation. The resulting ⫽ ⫺ I H ⫺ ICO [24]
“standard cathode-membrane” voltage vs. current curve was then dw
represented mathematically, choosing a form suggested by Kim dx H f df dx dx
⫽ ⫺ IH ⫽ x H ⫹ f H ⫽ ⫺( I H ⫹ ICO ) x H ⫹ f H
et al.16 that can be fit well with five parameters dw dw dw dw
Vstd(Jtot) ⫽ K1 ⫺ K2Jtot ⫺ K3 ln(Jtot) ⫺ K4 exp [K5Jtot] [19] [25]
dx CO f df dx CO
Equation 19 is used in conjunction with Eq. 18 to determine the local ⫽ ⫺ ICO ⫽ x CO ⫹ f
dw dw dw
value of Vanode and Vstd that generates equal current densities at the
anode and the cathode at a specified cell voltage and local H2 con- dx
⫽ ⫺( I H ⫹ ICO ) x CO ⫹ f CO [26]
centration. dw
To solve the previous nine equations a banded matrix solution By substituting Eq. 21, 23, and 24 into 25 and 26, we obtain deriva-
similar to Newman’s BAND routine was used.17 The catalyst layer tives of w, xH, and xCO with respect to u⬘. However, fIN is not known
region (Eq. 14-16) was divided into a number of nodes, typically 21, so we divide it into w and use the relation that w ⫽ 1 when u ⫽ u⬘
and implicit difference equations were programmed at each node for
each variable. Equations 14-16 plus Eq. 11-13 and 18 for backing  
and cathode technically define the problem. However, the solution d  1 ⫽ 1
[27]
stability was greatly improved by explicitly integrating the current
sources jH and jCO backward through the catalyst layer to get JHtot and ( IN
)
du ′  x H ⫹ x CO fIN 
IN

I H ( x H , x CO ) ⫹ ICO ( x H , x CO )
tot. Passing the total current density from backing to membrane end
JCO
dx H (1 ⫺ x H ⫺ x CO )[ x H ICO ⫺ (1 ⫺ x H ) I H ]( x IN ⫹ x CO
IN
)
of the catalyst layer forced the same current density to be used in the ⫽ [28]
backing and “standard cathode-membrane” and to be consistent with du ′ (1 ⫺ x HIN ⫺ x COIN
)( I H ⫹ ICO )
the current generation. This required introducing three more vari-
ables or “concentrations” to the banded matrix. dx C (1 ⫺ x H ⫺ x CO )[ x CO I H ⫺ (1 ⫺ x CO ) ICO ]( x HIN ⫹ x CO
IN
)

du ′ (1 ⫺ x H ⫺ x CO )( I H ⫹ ICO )
IN IN
Calculation of cell current under high hydrogen utilization.—The
1D anode model gives us hydrogen and CO current densities as func- [29]
A16 Journal of The Electrochemical Society, 148 (1) A11-A23 (2001)
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

Equations 27-29 may be integrated using the fourth-order Runge with (i) highly preferential adsorption of CO from a CO/H2 mixture
Kutta method. The 1D model determines IH and ICO as functions of and (ii) a much higher specific rate per catalyst site at low anodic
Vcell, xH, and xCO. The integration begins at u ⫽ 0 (where fIN is infi- potentials for hydrogen electro-oxidation vs. CO electro-oxidation.
nite) with xHIN and xCO
IN
and ends at the maximum value of u desired. These two key features determine the behavior of Pt-based PEFC
We could eliminate Eq. 28 by taking xH from Eq. 22 using xCO anode catalysts when exposed to a mixture of hydrogen with CO in
obtained by integrating Eq. 29. The integral of Eq. 27 allows us to the 10-100 ppm range.
determine the fIN as a function of u. The total current density as a In the polarization curves of Fig. 5, we refer to the highest cur-
function of u is then rent density available without a major loss in cell voltage (i.e., a
sharp increase in anode overpotential) as the “limiting” or “critical”
Iave(u) ⫽ (xHIN ⫹ xCO
IN
)fIN(u)u [30] current. This critical current indicates control of the overall rate of
the anode process at the Pt catalyst by the limited rate of the disso-
Results and Discussion ciative chemisorption of hydrogen (Eq. 2) occurring on a limited
Interfacial kinetics at a Pt anode catalyst in the presence of number of CO-free Pt surface sites. To obtain a simple analytical
CO.—Critical current density and fit for a range of xCO.—We start expression for this limiting current density, we consider a somewhat
the discussion of our model results with an examination of the inter- extreme case of the rate of CO electro-oxidation, kec ⫽ 0 (kec is very
facial reactivity predicted for a hydrogen anode in the presence CO. small in reality at a Pt catalyst). When kec ⫽ 0, the limiting, or crit-
We discuss here the general features of this interfacial kinetics ical current density, jhl, from Eq. 5-7 when ␪h approaches zero, is
model, while highlighting the effects of the refinement made here vs. given by
the first version presented in Ref. 1. The model results shown in the
relevant figures are all quantified for a Pt anode catalyst, although n
 bfc 
they may also apply for Pt alloy catalysts by making appropriate jhl ⫽ kfh x h PA (1 ⫺ ␪ CO ) n ⫽ kfh x h PA   [31]
changes to the values of one or two of the surface chemistry/electro-  fc
b ⫹ x CO A 
P
chemistry parameters. We refer to the possible nature of such
changes in several of the examples shown. This represents, at given CO partial pressure and cell tempera-
A set of IR-corrected cell polarization curves, measured for a ture, a limit on the current density that can be generated by the anode
H2/air fuel cell employing a Pt anode catalyst, at various added con- with acceptable voltage losses. An interesting conclusion from
centrations of CO into the neat H2, is shown in Fig. 5. There is a Eq. 31 is that for a fixed CO mole fraction, if the order n for free cat-
region of low current density for which little voltage loss is observed alyst sites is greater than unity, a lower overall anode pressure PA
as a result of addition of CO to hydrogen. This is followed by a much increases the limiting current density jhl. This assumes that the CO
steeper decline in cell voltage and eventual leveling off at an almost partial pressure is equal to or greater than bfc, a condition necessary
constant, low cell voltage. The range of little polarization loss at low for serious poisoning effects. The finding that a limiting current den-
current densities is dependent on CO concentration. The lower the sity of hydrogen oxidation could arise at low anode overpotentials in
level of CO, the higher is the current density achievable at minimal the presence of CO was described in the 1975 work by Vogel et al.18
voltage loss. for phosphoric acid fuel cell anodes. In that case the fuel cell tem-
Table I lists the set of representative interfacial-kinetics-model perature was significantly higher (190⬚C) and the relevant CO con-
parameters selected to match such experimental data. Note that the centrations were much higher (about 1%).
values chosen reflect basic documented characteristics of this inter- Figure 6 shows a computed set of cell polarization curves using
facial system, e.g., equilibrium coverage by CO is substantial (at the system parameter values in Table I. The dashed curves represent a
80⬚C) at partial pressures of CO as low as 10-100 ppm (see value constant value of bfc, and the solid curves represent the case of bfc in-
chosen for bfc). The choice of parameters describes a catalyst surface creasing with ␪CO to correspond to a decrease of CO adsorption ener-
gy with increasing CO coverage, as described in Fig. 1. Comparing
Fig. 6 to Fig. 5, we see that with constant bfc the performance at high
(250 ppm) CO concentration is predicted to be significantly poorer
than that observed. Assuming bfc varies with ␪CO according to Eq. 10

Figure 5. Measured iR-corrected cell voltage on Pt anode with H2/air fuel


cell at 80⬚C and high stoichiometric flow of hydrogen with various concen- Figure 6. Cell polarization as function of current density calculated for base
trations of CO. Experimental conditions: cell active area 5 cm2; temperature: case parameters in Table I for different CO concentrations in the anode feed
80⬚C (cell)/105⬚C (anode)/90⬚C (cathode); Pt catalyst loading 0.2 mg/cm2 on stream: (---) constant bfc ⫽ 4.5 ⫻ 10⫺6 atm; (—) use the Fig. 1 relation for
each side; Nafion 117 membrane. ␪CO-dependent bfc. Kinetics order (Eq. 6), n ⫽ 2.
Journal of The Electrochemical Society, 148 (1) A11-A23 (2001) A17
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

raises the critical current density at higher CO concentration and also


lowers it relatively at lower concentrations. This feature of ␪CO depen-
dence on bfc (that we added here to our model, see Ref. 2 for com-
parison), is supported by the documented variation of the heat of
adsorption of CO with CO coverage on Pt and is seen (Fig. 6) to pro-
vide a better fit to the actual experimental behavior. Interestingly, a
similar improvement in the fit to variation of the critical current den-
sity with xCO was achieved if first-order kinetics in free catalyst sites
was assumed in the H2 dissociative chemisorption step, as shown in
Fig. 7. Indeed, with bfc held constant (at a smaller value), the match
to Fig. 5 data appears to be better with n ⫽ 1 than with n ⫽ 2. Exper-
imentally, we have observed, however, poorer performance at higher
overall anode pressure, as seen below in Fig. 7, and (as explained pre-
viously in this section) Eq. 31 indicates that n should therefore be
greater than unity. Hence, use of n ⫽ 2 with variable bfc for modeling
appears to be preferable, although we chose to include Fig. 7 as a pos-
sible indication that reality may lie somewhere between n ⫽ 1 and 2
for the effective order of free catalyst sites in the kinetics of Eq. 2.
[Note: Differential pressure effects on oxygen crossover could not
contribute significantly to the trends seen in Fig. 8, considering the Figure 8. Measured performance with Pt anode at high stoichiometric feed
very low rates of oxygen crossover through the Nafion 1135 mem- rate of 40% H2 with 10 and 20 ppm CO at three anode gauge pressures (air
brane-based membrane electrode assembly (MEA)]. cathode pressure constant at 60 psig). Decreased performance seen with
increasing total anode pressure indicates kinetic order n in free catalyst sites
Interfacial kinetics at a Pt anode catalyst with CO.—Calculated (Eq. 6) is greater than 1. Experimental conditions: cell area 5 cm2; H2 160
effects of variations in key interfacial parameters.—As explained pre- sccm; air 550 sscm; temperature 80⬚C (cell)/105⬚C (anode/90⬚C (cathode); Pt
viously, the current obtained in the low-polarization region is appar- catalyst loading 0.2 mg/cm2 on each side; Nafion 1135 membrane used for
ently generated by hydrogen oxidation at a small fraction of uncovered MEA preparation.
(by CO) catalyst sites. The availability of such uncovered sites is relat-
ed primarily to the equilibrium coverage of the catalyst by CO at the
relevant temperature (i.e., by CO partial pressure and bfc) and by any age as a function of anode potential. The rate of CO electro-oxidation
marginal rate of CO electro-oxidation at low anode overpotentials (i.e., as a function of applied potential is kept at the base-case level of 3 ⫻
by kec). To quantify the possible effects of such factors, we now exam- 10⫺10 A/cm2 at 0 V RHE in all cases, and under such conditions the
ine calculated variations in the polarization curve brought about by rate of CO stripping is calculated to be well below ␮A/cm2 (geom.) in
variations in these model parameters using second-order kinetics at the potential domain of interest (<0.2 V). An increase of bfc, i.e., a drop
free catalyst sites. We specifically explore effects of variations in base- of the equilibrium constant for CO adsorption even to the limited
case values of these parameters for a 40% H2, 50 ppm CO anode feed extent shown in Fig. 9, could substantially improve the performance at
stream at 1 atm, assuming zero fuel utilization. Figure 9 shows the the 50 ppm CO level, as indicated by the fivefold increase of the cal-
effect of varying bfc, the (inverse the) equilibrium constant for CO culated limiting kinetic current density of hydrogen electro-oxidation.
adsorption when bfc is constant (——) and when bfc assumes the ␪CO-
dependent form of Fig. 1 (----). We use the base-case value of bfc0 and
a factor of ten higher. The constant bfc values are chosen to have the
same value as the ␪CO-dependent values at the broad maxima in the
␪CO vs. Vanode curve. Consequently, the constant bfc values exhibit a
ratio of only 2.8. The upper half of Fig. 9 shows the hydrogen and CO
current density and the lower half shows the hydrogen and CO cover-

Figure 9. Effect of increase in base case bfc on fractional CO and Hads cov-
erage (bottom) and hydrogen current density (top). CO concentration is 50
ppm in 40% H2, 35% N2, 25% CO2 at 1 atm. The constant bfc value for the
base case is 4.8 ⫻ 10⫺6 and bfc0 is 1.5 ⫻ 10⫺8 for the ␪-dependent curve.
Figure 7. Modeled performance prediction as in Fig. 6 but using first-order These are increased by a factor of 10 in the “⫻10 case.” Other model para-
kinetics in free catalyst sites (n ⫽ 1 in Eq. 6). bfc is fixed here and lowered meters have values as in Table I. The abscissa shows anode overpotential vs.
to 5 ⫻ 10⫺7 atm. RHE.
A18 Journal of The Electrochemical Society, 148 (1) A11-A23 (2001)
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

This increase takes place, as seen from the figure, as a result of lower- tion through the catalyst layer. We use the model to illustrate this
ing of steady-state CO coverage from 0.83 to 0.63 at Vanode near 0.1 V point by setting base-case parameters for the model and then observe
vs. RHE. This change in bfc might be achieved experimentally by rais- the implications of varying some of them. Table I lists the base-case
ing the temperature of the PEFC with a given anode catalyst, or, alter- parameters we have selected. Certain of the parameters, such as cat-
natively, by developing novel anode catalysts of lower affinity to CO alyst layer resistance, Rcl, effective catalyst layer diffusion coeffi-
at 80⬚C, while maintaining high hydrogen electro-oxidation activity. cients DH and DCO, and gas diffusion backing porosity/tortuosity fac-
Figure 10 shows calculated effects of raising the rate constant for tor e/t could be inferred from results of our fuel cell cathode model19
electrochemical stripping of CO, kec, (process described by Eq. 4) by modified for the different gases.
one or two orders of magnitude above the very low level of Figure 11 shows a comparison of the anode potential calculated
10⫺8 A/cm2 (geom.) at 0 V, assumed in the base-case. The figure as a function of current density for neat hydrogen and for 10%
describes a calculation for 100 ppm CO in H2. The top part of Fig. 10 hydrogen in nitrogen at two different values of catalyst layer specif-
shows calculated CO stripping currents, the middle part shows the ic protonic conductivity, 0.1 and 0.01 S/cm. Note that at 1 A/cm2 the
corresponding steady-state CO coverage as a function of anode anode overpotential penalty for hydrogen dilution by a factor of ten
potential, and the bottom part shows the resulting hydrogen oxida- increases from 75 to 100 mV between the higher and lower conduc-
tion currents as a function of anode potential. Other parameters are tivity catalyst layers. In Fig. 12 we see the catalyst layer profiles for
kept at base-case levels (see Table I). It is evident from these curves the four cases in Fig. 11 depicted at a current density of 1 A/cm2. The
that low CO stripping current density of less than 100 ␮A/cm2 overall anode potential (top plot) is higher for the 10% dry H2 cases,
(geom.) (i.e., sub ␮A/cm2 Pt), can translate into quite significant and the increase in Va across the catalyst layer is more pronounced
effects on steady-state CO coverage in the relevant anode potential in the lower conductivity cases. This is caused by the faster drop in
range and, consequently, into significant hydrogen oxidation current reactant concentration in the 10% H2 cases that moves peak current
densities. For example, for the case of kec ⫽ 10⫺8, jCO at an anode generation toward the backing side of the catalyst layer, consequent-
potential of 0.28 V is approximately 100 ␮A/cm2 (geom). This re- ly passing the protonic current through greater ionic resistance. A
sults in a lowering of CO steady-state coverage to a level sufficient diluted feed stream is thus seen to generate an increased anode polar-
to enable a hydrogen oxidation current density of 1 A/cm2 at the ization originating from combined concentration and resistance
same anode potential. The strong effect, of what is still a very low polarization.
CO electro-oxidation rate, is a consequence of the high rate of the We next compare our 1D model calculations with experimental
hydrogen electro-oxidation process per catalyst site available. anode polarization data obtained with various dilutions of hydrogen
Calculated effects of hydrogen dilution in the complete absence in nitrogen in the range of 30-6% hydrogen under very low fuel uti-
of CO.—We next address the effects of hydrogen dilution (in inerts) lization. An air cathode was used and anode dilution loss (“anode
as projected by our model calculations. Losses associated with lim- overvoltage”) was defined as the lowering of cell voltage vs. that of
ited protonic conductivity and/or limited hydrogen diffusivity in the the same cell with a neat hydrogen anode. Figure 13 shows the ex-
anode catalyst layer generally increase during operation with diluted perimental data. The indicated mole fractions are for dry gas, but the
hydrogen feed streams because of a redistribution of current genera- gas was then fully saturated with water vapor at the cell temperature
of 80⬚C. At low hydrogen concentrations, we see a sharp rise in
anode potential expected to be caused by the limited rate of diffusion
in the backing layer. Figure 14 shows the model predictions for
anode polarization corresponding to the conditions of Fig. 13. A
good match is obtained with the base-case parameters for the inter-
facial process and the anode catalyst layer. Whereas the limiting-cur-
rent behavior could be fully accounted for with the physical backing
properties as shown, it is interesting to note that interfacial kinetics
can also generate, in principle, a limit on jh of kfhxhPA. This follows
from Eq. 6 and 7 for large values of ␩. A high value of kfh0 of 4000 A
cm⫺2 atm⫺1, invoked in our model for the process of hydrogen dis-
sociative chemisorption on CO-free Pt, was dictated by lack of indi-

Figure 10. Effect of increase by factor 10, or 100, in rate constant for CO
electro-oxidation on CO electro-oxidation current (top), CO surface coverage
(middle), and hydrogen electro-oxidation current (bottom). Other model Figure 11. Anode polarization calculated for neat and 10% H2, at higher and
parameters have values as in Table I. The abscissa shows anode overpotential lower catalyst layer resistance. Assumed conditions are dry gas pressure of
vs. RHE. 2 atm, 80⬚C; Vanode is vs. SHE.
Journal of The Electrochemical Society, 148 (1) A11-A23 (2001) A19
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

Figure 14. Model prediction for results in Fig. 12 using ⑀/␶ ⫽ 0.09 for the
anode backing. Other parameters are from Table I.

ation on a diluted hydrogen inlet stream, still in the absence of CO.


We have selected operating conditions of 1.5 atm pressure, 80⬚C
temperature, 0.6 V cell voltage, and a reformate mixture of 40% H2,
25% CO2, 0% CO, and 35% N2. Basic cell parameters are given in
Table I. Figure 15 describes the predicted performance of this cell as
the hydrogen utilization is varied from 0 to a maximum of 0.9. The
top curve plots the feed rate in equivalent current density for the case
where u⬘ is the utilization at the anode outlet (u ⫽ u⬘). The middle
curve plots the H2 mole fraction at the specified utilization. For the
Figure 12. Distribution in anode catalyst layer of potential, H2 concentration, bottom curve of Fig. 15, because there is a distribution of anode
and current production at 1 A/cm2 for the four cases in Fig. 11. The backing potentials along the feed path, we describe an effective anode loss,
interface is at y ⫽ 0 and the membrane interface is at y ⫽ 1. analogous to Eq. 18 for the 1D model, by means of an effective
anode potential Van,eff which at some average cell current Iave, is
cation of such kinetic rate limits in operation with the highly diluted defined as
H2 feed streams. Current limits resulting from hydrogen adsorption
were evident in our model calculations if a significantly lower value Van,eff ⫽ Vstd(Iav) ⫺ Vcell [32]
of kfh0 were used. (Mass transport in the catalyst layer exacerbates
the effects of dilutions as high as employed in the experiment depict- In the bottom plot we show the values of Jlocal and Van,local, the local
ed in Fig. 13 and increases the tendency to generate limiting currents current density and anode potential when utilization has reached u⬘,
at lower kfh values.) together with the values of Iave and Van,eff. Jlocal and Van,local, in the
lower part of Fig. 15, correspond to solutions of 1D model sections
Cell losses with realistic feed streams.—As a first step we illus- depicted in Fig. 2. Each solution uses the overall cell voltage of
trate in this section the consequences of high fuel utilization in oper- 0.6 V and an H2 mole fraction xH (top part of Fig. 15) for the corre-
sponding value of u⬘. We see the local current density drops about
32% as the dry mole fraction drops from 0.4 to about 0.053. This rel-
atively small change is sustained because of the relatively large
increase in anode potential from 34 to 102 mV [and the low Tafel
slope for the hydrogen oxidation reaction (HOR)]. This increase in
anode overpotential occurs together with an equivalent decrease in
overvoltage across the cathode membrane, as required to sustain the
constant Vcell of 0.6 V. Such interplay between anode and cathode
overpotentials down the flow channel reduces the penalty for higher
utilization. Although examination of expected local current and
potential variations is instructive, in operation we observe only Iave
and Van,eff. The loss of average cell current and, hence, cell power
density, as a result of raising (40%) H2 utilization from 0 to 90%, is
seen to be only 9.3% at cell voltage of 0.6 V. This is a remarkably
small penalty for high fuel utilization with such a diluted (40% H2)
inlet feed stream, explained basically by the very high interfacial
kinetics of the HOR that enables us to achieve high currents in spite
of very low local concentrations of hydrogen.
To examine more closely the combined implications of hydrogen
dilution and high fuel utilization, the effective anode potential Van,eff,
Figure 13. Measured anode loss caused by hydrogen dilution at various
as defined in Eq. 32, is plotted as a function of current density in
dilutions of hydrogen in nitrogen at 3 atm, 80⬚C (5 cm2 cell). Anode catalyst
Pt/C at 0.2 mg Pt cm⫺2. Anode feed humidified at 105⬚C. Hydrogen flow rate Fig. 16 and 17 for a range of fuel dilution and fuel utilization. (In
very high. Fig. 16, Vloss becomes negative at 3 atm pressure, because the standard
A20 Journal of The Electrochemical Society, 148 (1) A11-A23 (2001)
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

Figure 17. Same as Fig. 16 but at 3 atm anode pressure.

10% larger volume than that required for operation on neat hydrogen,
it would not amount to a technology barrier of any real significance.
However, having reached such an optimistic conclusion, we note that
both from ourselves and from reports by others to date, it appears that
reaching 90% fuel utilization with such diluted hydrogen feed streams
has not been so straightforward in reality. In the context of stack rather
than single-cell operation, this may reflect nonuniformity of anode
feed flow between cells in the stack under high overall fuel utilization,
Figure 15. Local H2 mole fraction, current density, and anode potential along bringing about high polarization in some “fuel-starved” cells. We view
the anode flow field plotted as a function of local utilization. Operating con- the results of the model as indication that a loss of only 10% in max-
dition is 1.5 atm, 40% H2, 25% CO2, 0% CO, and 35% N2 at a cell voltage imum power between using neat hydrogen and 40% H2 at 90% fuel
of 0.6 V and maximum fuel utilization u ⫽ 0.9. Other parameters are in utilization should be reachable. However, this apparently requires (i)
Table I. sufficiently humidified anode catalyst layers of good transport charac-
teristics, (ii) properly designed anode backing layers, (iii) properly
anode is defined for PH2 ⫽ 1 atm). Figure 16 conditions are replotted designed anode manifolds and flow fields, and (iv) that the CO level
in more familiar terms of cell voltage vs. current density in Fig. 18. in the feed stream is practically zero.
The results of Fig. 18 reveal a cell voltage loss from operation on dilut- We move now to calculations that consider all causes for anode
ed hydrogen feed streams at high fuel utilization that can be described loss with realistic anode feed streams: a dilute inlet stream containing
as significant but not overwhelming. Loss near the maximum power CO and operation under high fuel utilization. Cell voltage vs. current
point, relative to operation on neat hydrogen, of either current or volt- density plots, calculated for diluted feed streams containing CO, are
age with the other held constant is about 10% in the most demanding shown in Fig. 19 through 21. Whereas dilution and high utilization do
case of 40% H2 at 90% utilization. The corresponding loss encoun- not penalize cell performance very much in the absence of CO, as can
tered with 75% H2 inlet is smaller than 5%. If the implication of oper- be seen in Fig. 18, the voltage loss under similar conditions could be
ation with gasoline reformate were only the need of a PEFC stack of large when some CO is present. This amplifying effect that hydrogen

Figure 16. Calculated effective anode loss in the absence of CO for three Figure 18. Calculated cell polarization curves for three anode feed streams
anode feed streams of neat H2, 75% H2, and 40% H2 (no CO) at utilization of neat H2, 75% H2, and 40% H2 (no CO), under fuel utilization of 0%, 50%,
of 0, 0.5, and 0.9. and 90%.
Journal of The Electrochemical Society, 148 (1) A11-A23 (2001) A21
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

dilution and high fuel utilization could have on the extent of surface The interfacial parameters we have used in this model up to this
poisoning by CO can be understood by examining Eq. 31. This equa- point have been set to match experimental data for (unalloyed) Pt
tion shows that the H2 electro-oxidation “limiting” or “critical” cur- catalysts. The question remains for improved anode catalysts, known
rent density is directly proportional to H2 mole fraction and is not for better “CO tolerance,” whether the same strong amplification of
affected by anode potential. For operation at high fuel utilization, the CO-related loss is also observed with dilute feed streams at high fuel
hydrogen mole fraction drops along the flow channel to well under utilization. Figure 20 shows the polarization curves for the identical
the inlet level, which could itself be as low as 0.4. However, the CO conditions as the bottom, right plot in Fig. 19, except that kec has
mole fraction typically remains the same or actually rises, because been increased from 1 ⫻ 10⫺8 to 1 ⫻ 10⫺6 A/cm2. This amounts to
CO is not consumed to a significant extent (very low rates of the an assumption that the advantage of the improved catalyst over Pt is
process in Eq. 4 at relatively low anode potentials). Figure 15 shows a significantly higher rate of CO electro-oxidation at low anodic po-
that the mechanism to correct for the local fall in hydrogen concen- tentials. While still low, such somewhat higher kec values can have a
tration in the absence of CO is a rise in anode potential along the flow significant effect on steady-state ␪CO and, hence, on the rate of hy-
channel. The local current density consequently recovers significant- drogen oxidation at the CO-covered catalyst surface at a given low
ly by boosting the rate of the reaction described by Eq. 3. However, anode potential (see Fig. 10). Figure 20 exhibits milder drops of po-
in the presence of CO, the local “critical” current density at some seg- tential within each polarization curve and smaller penalties with
ment along the flow channel would be determined by the reaction rate increased fuel utilization. This is because the assumed increase in
of Eq. 2, rather than of Eq. 3. The local current in the presence of CO rate of CO electro-oxidation has removed the exclusive control of
is therefore proportional to the local low mole fraction of H2 and can- the rate of hydrogen electro-oxidation by a chemical process (Eq. 2),
not be increased above the critical value by raising the local anode making hydrogen electro-oxidation now under the mixed control of
potential. Figure 19 (top, left) shows that at just 10 ppm CO, the crit- processes Eq. 2 and 3. Polarization curves with such qualitative pat-
ical current density at a Pt anode catalyst at 80⬚C is 0.8 and 0.4 A/cm2 terns of softer drops of cell voltage with current density in the pres-
in operation on 75% hydrogen inlet at 0 and 90% fuel utilization, ence of CO have been reported for PEFCs with PtRu or PtMo anode
respectively. The same figure (bottom, left) shows, for the more dilute catalysts.20 Further possible improvements can be invoked, beyond
40% hydrogen inlet with just 10 ppm CO, critical current densities of what is presented by Fig. 20, with a stronger increase of kec as well
0.5 and 0.2 A/cm2 at 0 and 90% fuel utilization, respectively. Effects as an increase of bfc. It has been our practical experience, however,
of a very low inlet CO level of just 10 ppm are thus seen to be sig- that at 50 ppm CO, the only effective way to avoid excessive voltage
nificantly amplified when the inlet feed is diluted and the fuel uti- losses under operation with dilute hydrogen at significant fuel uti-
lization is high. Under similar conditions, the situation in cells with lization is air bleeding. To put the issue of “CO tolerance” in practi-
Pt anode catalysts is projected to be much worse at an inlet level of cal perspective, we finally present in Fig. 21 a recent result for the
50 ppm, as seen in the two panels on the right in Fig. 19. Our model performance of a PEFC (single cell) with an inlet of 40% H2 with
projections depicted in Fig. 19, of strong effects of fuel utilization on 100 ppm CO at 70% fuel utilization. This performance was achieved
cell performance in the presence of CO, still await full comparison with an air bleed into the anode inlet at a level of 4%. In terms of the
with experiment. However, trends seen with increased fuel utilization key parameters of the model described here, the effect of air bleed is
in the presence of CO are along the lines projected, as can be seen, to lower xCO to practically zero by chemical oxidation of CO
for example, from the exceptionally low cell currents measured under adsorbed on anode catalyst sites according to 2COads ⫹ O2 ⫽ 2CO2.
such conditions with 100 ppm CO (see curve on left side of Fig. 21). Conclusions
Operation with “critical” anode current densities as low as 0.5 A/cm2,
Model for a PEFC with an anode operating on reformate.—We
or below, is quite impractical and, consequently, a catalyst with much
described here a model for a PEFC with a hydrogen anode that oper-
better “CO tolerance” is required to provide sufficient performance
ates at high fuel utilization on diluted hydrogen feed streams contain-
under such realistic operation conditions, particularly in the case of
ing CO. The components of the feed stream, other than hydrogen, can
40% hydrogen at high utilization. Alternatively, the air bleeding8
be carbon dioxide, nitrogen, and carbon monoxide. We assume the
which is apparently employed by most PEFC stack developers to
gas mixture is saturated with water vapor at the specified temperature.
maintain high performance on reformate feed streams provides a rad-
The model incorporates diffusion losses in the anode gas diffusion
ical solution by lowering xCO in the anode to practically zero, using
layer, and diffusion and ohmic losses in the anode catalyst layer, to
chemical oxidation of CO by dioxygen within the anode.
define the distribution of hydrogen reactant as a function of location

Figure 19. Fuel utilization effect on cell polarization in presence of CO, cal-
culated using base case parameters at 1 atm, 75% (top) or 40% H2 (bottom) Figure 20. Fuel utilization effect on cell polarization in the case of 40% H2
inlet, with 10 ppm (left), or 50 ppm CO (right). In each case the polarization feed stream containing 50 ppm CO, calculated for a catalyst which is associ-
curve is calculated for four fuel utilization levels of 0, 0.5, 0.8, and 0.9. ated with 100 times higher kec.
A22 Journal of The Electrochemical Society, 148 (1) A11-A23 (2001)
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

along the catalyst layer at given composition in the flow channel adja- long as the anode feed stream is completely free of CO. One inter-
cent to the backing. Having defined local hydrogen and CO concen- esting reason for this rather “forgiving” behavior of the cell is that
trations, the model provides detailed equations for interfacial kinetics under constant voltage conditions, the cell tends to readjust anode
that apply for hydrogen electro-oxidation at Pt in the presence of CO. losses vs. losses in the remainder of the cell as the hydrogen con-
Some refinements of the interfacial kinetic equations employed centration drops along the anode flow channel at high fuel usage.
by us previously1 to describe anode deactivation in presence of CO In contrast, when the level of CO is even as low as 10 ppm, the
were explored. For the dissociative chemisorption of H2, we first losses incurred by CO are amplified significantly as a result of lower-
assumed in Eq. 1 simultaneous involvement of two adjacent Pt cata- ing the local concentration of hydrogen and therefore become worse
lyst sites (Tafel mechanism).1 When comparing to measured polar- with a more diluted feed stream and with higher fuel usage levels.
ization losses, the “critical current” projected by the model dropped Replacement of Pt by PtRu, or similar, as anode catalyst, is likely
too sharply with a rise in inlet ppm CO. The fit was significantly to enable operation with 10 ppm CO in neat hydrogen, and possibly,
improved by introducing a Temkin isotherm to allow the equilibrium in reformate at low fuel utilization. It is less likely that such an anode
constant for CO adsorption to drop with ␪CO, i.e., allowing the model catalyst would be sufficient, however, to resolve the “tolerance” prob-
parameter bfc to be smaller at higher ␪CO, in line with solid/gas liter- lem even at 10 ppm CO (not to mention higher CO levels) with dilut-
ature reports. We also found that using a mechanism that assumes ed anode feed streams and under high fuel utilization. Air bleed there-
initial H2 adsorption on a single Pt site (Heyrovsky mechanism), the fore appears for the moment the most radical approach to minimize
“critical current” variation with CO concentration could be matched voltage losses in realistic PEFC operation on reformate, for reasons
by assuming bfc independent of ␪CO. The latter result may suggest that become clearer from our model calculations.
that both Tafel and Heyrovsky mechanisms could possibly be oper- Acknowledgments
ative. However, a key criterion for order higher than unity in free cat-
The authors are indebted to Dr. Mark Paffett for helpful discus-
alyst sites is met experimentally: measured voltage losses are exac-
sions on the effect of hydrogen adsorption rate in the presence of
erbated with increased total anode pressure at a given ppm CO. This
CO. We also appreciate suggestions from Dr. Richard Bellows that
suggests that the Tafel mechanism with coverage-dependent adsorp-
the model critical current density dependence on CO concentration
tion energy for CO is, most probably, the preferred description for
might better fit experiment if we used a Temkin rather than a Lang-
the HOR process in the presence of CO at a Pt catalyst. Along sim-
muir isotherm. The U.S. Department of Energy, Office of Advanced
ilar lines, to use the same kinetic equations of the model both with-
Automotive Technology, supported this work.
out and with CO, we had to invoke an activation energy for the dis-
sociative chemisorption of hydrogen on Pt that decreases substan- Los Alamos National Laboratory assisted in meeting the publication
tially over a group of surface sites around every site covered by CO. costs of this article.
Finally, we provided a method to calculate the cell voltage loss Appendix A
expected when the anode feed stream consists of diluted hydrogen
and the cell is operated under high fuel utilization. We pointed to the First-Order Kinetics of a CO Anode
redistribution of losses along the flow channel, as segments of the Because the second-order model decreases current more abruptly with
cell closer to the outlet would exhibit higher anode overpotential to Vanode than the measured data, we also calculated first-order kinetics for Eq. 5-
7. Equation 6 becomes Eq. A-1. The intermediate variables are different
correct, in part, for lower local hydrogen concentrations. Corre-
spondingly, cathode and membrane overpotentials were smaller near ␳␪˙ h ⫽ k fh x h PA (1 ⫺ ␪ CO ⫺ ␪ h ) ⫺ bfh k fh ␪ h
the outlet, maintaining the same overall cell voltage at each point
along the flow channel.  V ⫹ VNernst 
⫺ 2␪ h k eh sinh a  ⫽ 0 [A-1]
Conclusions  bh 
PEFC with anode operating on reformate.—The power loss from Va ⫹VNernst
PEFC operation on dilute hydrogen feed streams, as low as 40% in  V ⫹ VNernst  bc
j h ⫽ 2 k eh ␪ h sinh a  , jCO ⫽ 2 k ec ␪ CO e [A-2]
hydrogen, under high fuel usage, as high as 90%, should not exceed  bh 
10% of the stack power obtained in operation on neat hydrogen, as
V
 V ⫹ VNernst  a
s2 ⫽ 2 keh sinh a  , a4 ⫽ e bc kec ⫹ kfc bfc
 bh 

a5 ⫽ a4(bfh ⫹ xhPA) ⫹ bfhkfcxCOPA, a6 ⫽ a4 ⫹ kfcxCOPA,


a7 ⫽ kfha5 ⫹ s2a6 [A-3]

a3 k fh x h PA k x P (b k ⫹ s2 ) [A-4]
␪h ⫽ ; ␪ CO ⫽ fc co A fh fh
a7 a7

Appendix B
In deriving a relationship between kfh and ␪CO, we target a unified de-
scription of the hydrogen electro-oxidation rates observed in anode operation
on feed streams of (i) CO-free, very dilute hydrogen (e.g., 5% H2, 95% N2)
and (ii) hydrogen (or “reformate”) containing 10-200 ppm CO. Assuming, as
we did, that the “site-blocking” effect of CO on hydrogen adsorption is
described as (1 ⫺ ␪CO ⫺ ␪h)2 (see Eq. 6), further dependence of kfh on ␪CO in
Eq. 6 is dictated by measured anode activity on the two feed streams de-
scribed previously. If site blocking according to (1 ⫺ ␪CO ⫺ ␪h)2 were the
only effect on the rate of dissociative chemisorption, the very high value of
kfh, concluded from measured anode activity with CO-free diluted hydrogen,
Figure 21. Fuel tolerance to CO, demonstrated experimentally at 70% fuel is not compatible with the appearance of limiting currents significantly lower
utilization with inlet feed of 40% H2 containing 100 ppm CO. Bleeding air at than 1 A/cm2 in the presence of CO. Hence, we reach our conclusion that a
4% by volume into such anode feed stream with 100 ppm CO, is seen to significant drop of kfh with ␪CO must be added in Eq. 6. The catalyst surface
restore the performance back very close to the same level recorded with CO- model described in this Appendix targets a variation of kfh vs. ␪CO of the type
free, 40% H2 feed. Precious metal loading in the anode of this cell was seen in Fig. 1, i.e., a strong drop of kfh with ␪CO at lower CO coverage (␪CO <
0.6 mg/cm2. 0.3), following by leveling off at higher CO coverage. This pattern was found
Journal of The Electrochemical Society, 148 (1) A11-A23 (2001) A23
S0013-4651(00)03-119-0 CCC: $7.00 © The Electrochemical Society, Inc.

appropriate for accounting for the experimental results. Very strong varia- T temperature, K
tions of Si, the initial sticking probability for dissociative chemisorption of Ts standard temperature 273.15, K
diatomic molecules, have indeed been observed experimentally at low cover- u, u⬘(w) utilization, local utilization Itot /(xHIN ⫹ xIN
CO)/f
IN

age of site-masking agents,21 including the case of H2 on Pt. Monte Carlo Va local anode potential in catalyst layer, V
simulations21 accounted for such observations in terms of domains of sever- Vanode anode potential at membrane interface, V
al surface atoms affected by the “masking agent” and a domain of several sur- VNernst Nernst potential RT ln(PAxh), V
face atoms required for a diatomic molecule to stick effectively. Vs standard molar volume, cm3 mol⫺1
The model described here is based on a similar assumption of an ad- w fractional distance along flow field
sorbed CO molecule having an effect on the rate of dissociative adsorption of x mole fraction
hydrogen at a number of CO-free adjacent sites. The activation energy for y fractional distance through catalyst layer
dissociative chemisorption of H2 is locally increased from the zero CO cov- z distance through backing, cm
erage value by a constant amount ␦(⌬⌭⌯) on sites on and within some radius Superscripts
around each CO-containing site. Let there be a total number of n sites per unit
IN flow field gas inlet
area. Assume p contiguous sites (p ⬇ 4-6) plus the occupied site contribute
INA 1D model backing gas inlet along flow field
to the fraction of coverage ␪A that suffers the ␦(⌬EH) shift. There are n␪CO-
covered sites per unit area, leaving n(1-␪CO) sites available for each CO-cov- Subscripts
ered site to affect. The probability that any single CO site will affect any of H, CO, CO2, N, W components of hydrogen, CO, CO2, N2, and water,
the n(1-␪CO) unoccupied sites is p/n(1 ⫺ ␪CO). Then the probability that any respectively
available site is not affected by all CO occupied sites is h, co components of hydrogen, CO
n␪CO Greek
 p  ␣ij Stefan-Maxwell coefficients, cm s mol⫺1
1 ⫺ ␪ A ⫽ 1 ⫺
 n(1 ⫺ ␪ CO  ␦(⌬EH) change in activation energy for hydrogen dissociative ad-
sorption near CO occupied site
n(1⫺␪co ) p␪co
⋅ p␪co
␦(⌬GCO) variation of free energy of adsorption of CO between zero
 p  p (1⫺␪co ) ⫺
(1⫺␪co ) and full coverage of CO, J mol⫺1
⫽ lim 1 ⫺ ⫽e
n(1 ⫺ ␪ CO  flux of component i in backing, mol cm⫺2 s⫺1
[B-1] ⌽i
n→∞ 
␳ molar area density of catalyst sites, mol cm⫺2
The following function dependence of kfh on ␪CO results ␴ catalyst layer conductivity, ⍀⫺1 cm⫺1
␪CO, ␪h fractional coverage, CO, hydrogen
 ␦( ⌬EH ) 
k fh (␪ CO ) ⫽ k fh0 exp ⫺ ␪A  References
 RT  1. S. G. Chalk, J. M. Miller, and F. W. Wagner, J. Power Sources, 86, 45 (2000).
2. T. Springer, T. Zawodzinski, and S. Gottesfeld, in Electrode Materials and Process-
  p␪co  
δ ( ∆EH ) ⫺ es for Energy Conversion and Storage, J. McBreen, S. Mukherjee, and S. Srini-
⫽ kfh0 exp ⫺ 1 ⫺ e (1⫺␪co )   [B-2] vasan, Editors, PV 97-13, pp. 15-24, The Electrochemical Society Proceedings
 RT  
    Series, Pennington, NJ (1997).
3. H. A. Gasteiger, N. M. Markovic, and P. N. Ross, J. Phys. Chem., 99, 8290 (1995).
Equations 9, 10, and B-2 are solved implicitly. 4. H. A. Gasteiger, N. M. Markovic, and P. N. Ross, J. Phys. Chem., 99, 16757 (1995).
5. M. Iwase and S. Kawatsu, in Proton Conducting Membrane Fuel Cells, S. Gottes-
List of Symbols feld, G. Halpert and A. Landgrebe, Editors, PV 95-23, p. 12, The Electrochemical
a1, a3 intermediate variables Eq. 8, A s⫺1 Society Proceedings Series, Pennington, NJ (1995).
6. T. A. Zawodzinski, Jr., C. Karuppaiah, F. Uribe, and G. Gottesfeld, Abstract 781,
a2 intermediate variable Eq. 8, A2 s⫺2
p. 944, The Electrochemical Society Meeting Abstracts, Vol. 96-2, San Antonio,
bc natural Tafel slope for CO electro-oxidation, V TX, Oct 6-11, 1996.
bfc back-to-forward CO adsorption ratio (1/Kequilibrium), atm 7. C. Karuppaiah, F. Uribe, T. A. Zawodzinski, Jr., and S. Gottesfeld, Abstract 1081,
bfc0 bfc at ␪CO ⫽ 0, atm p. 1335, The Electrochemical Society Meeting Abstracts, Vol. 97-1, Montreal, Que-
bfh back-to-forward H2 adsorption ratio, atm bec, Canada, May 4-9, 1997.

C backing/catalyst layer concentration/ inlet conc. 8. (a) S. Gottesfeld and J. Pafford, J. Electrochem. Soc., 135, 2651 (1988); (b) M. S.
DCO CO diffusion coefficient in catalyst layer, cm2 s⫺1 Wilson, T. E. Springer, T. A. Zawodzinski, and S. Gottesfeld, in Proceedings of the
28th Intersociety Energy Conversion Engineering Conference, Atlanta, GA, Vol. 1,
DH H2 diffusion coefficient in catalyst layer, cm2 s⫺1
pp. 1203-1208 (1993); (c) T. Zawodzinski, T. Springer, J. Bauman, T. Rockward, F.
DSij binary diffusion coefficients in backing at standard condi- Uribe, and S. Gottesfeld, in Proton Conducting Membrane Fuel Cells, S. Gottes-
tions, cm2 s⫺1 feld and T. F. Fuller, Editors, PV 98-27, p. 127, The Electrochemical Society Pro-
F Faraday constant 96484, C mol⫺1 ceedings Series, Pennington, NJ (1999).
f molar flow of feed times 2F per electrode area, A cm⫺2 9. B. M. Rush, J. A. Reimer, and E. J. Cairns, Abstract 2098, The Electrochemical
IDco catalyst layer CO diffusion current density Eq. 15, A cm⫺2 Society and the Electrochemical Society of Japan Meeting Abstracts, Vol. 99-2,
IDh catalyst layer H2 diffusion current density Eq. 14, A cm⫺2 Honolulu, HI, Oct 17-22, 1999.
10. H. Igarashi, T. Fujino, and M. Watanabe, J. Electroanal. Chem., 391, 119 (1995).
I Htot, I CO
tot , I
total total (average) current density from H2, CO, and both, re-
11. A. W. Adamson, Physical Chemistry of Surfaces, p. 626, Interscience, New York
spectively, A cm⫺2 (1967).
JHtot, J CO
tot , J
tot 1D model total current density from H2, CO, and both, 12. R. J. Bellows and E. Marcuchi-Soos, in Proton Conducting Membrane Fuel Cells,
respectively, A cm⫺2 S. Gottesfeld and T. F. Fuller, Editors, PV 98-27, p. 218, The Electrochemical Soci-
jCO local kinetic current density from CO, A cm⫺2 ety Proceedings Series, Pennington, NJ (1999).
jH local kinetic current density from hydrogen, A cm⫺2 13. H. Conrad, G. Ertl, J. Koch, and E. Latta, Surf. Sci., 43, 462 (1974).
K1,2…,5 constants for V 14. G. A. Somorjai, Introduction to Surface Chemistry and Catalysis, John Wiley, New
York (1994).
kec pre-exponential CO Tafel electro-oxidation rate, A cm⫺2
15. T. E. Springer, T. A. Zawodzinski, M. S. Wilson, and S. Gottesfeld, J. Electrochem.
keh pre-exponential Had Tafel electro-oxidation rate, A cm⫺2 Soc., 143, 587 (1996).
kfc electrode forward CO adsorption rate times 2F, A cm⫺2 16. J. Kim, S. Lee, and S. Srinivasan, J. Electrochem. Soc., 142, 2670 (1995).
atm⫺1 17. J. Newman, Electrochemical Systems, pp. 417-425, Prentice Hall, Inc., Englewood
kfh electrode forward H2 adsorption rate times 2F, A cm⫺2 Cliffs, NJ (1973).
atm⫺1 18. W. Vogel, J. Lundquist, P. Ross, and P. Stonehart, Electrochim. Acta, 20, 79 (1975).
ᐉcl, ᐉB thickness of catalyst layer, backing, cm 19. T. E. Springer, M. S. Wilson, and G. Gottesfeld, J. Electrochem. Soc., 140, 3513
n order of H2 adsorption step (1993).
20. T. Zawodzinski, J. Bauman, T. Rockward, P. Haridos, F. Uribe, and S. Gottesfeld,
p number of affected sites per CO site for kfh
in Proton Conducting Membrane Fuel Cells, S. Gottesfeld and T. F. Fuller, Editors,
PA total pressure, atm PV 98-27, p. 200, The Electrochemical Society Proceedings Series, Pennington, NJ
Rcl catalyst layer resistance, ⍀ cm2 (1999).
R gas constant 8.3, J mol⫺1 K⫺1 21. C. T. Campbell, M. T. Paffett, and A. F. Voter, J. Vac. Sci. Technol., A4, 1342
s1 intermediate variable in Eq. 8 and 9, A cm⫺2 (1986).

You might also like