Transient behavior of CO poisoning of the anode
catalyst layer of a PEM fuel cell
H.S. Chu
a, C.P. Wang
a, W.C. Liao
a, W.M. Yan
b,∗aDepartment of Mechanical Engineering, National Chiao Tung University, Hsinchu, Taiwan 300, ROC bDepartment of Mechatronic Engineering, Huafan University, Shih Ting, Taipei, Taiwan 223, ROC Received 4 October 2005; received in revised form 20 December 2005; accepted 20 December 2005
Available online 3 February 2006
Abstract
A one-dimensional transient mathematical model is applied to simulate the carbon monoxide poisoning effect on the performance of a PEM fuel cell. Based on the CO kinetic model developed by Springer et al. [T.E. Springer, T. Rockward, T.A. Zawodzinski, S. Gottesfeld, J. Electrochem. Soc. 148 (2001) A11–A23], the transient behaviors of the CO poisoning process across the anode catalyst layer is investigated. The results show that the hydrogen coverage,θH, decreases with the time due to CO adsorption on the catalyst sites. A higher CO concentration results in fewer
available catalyst sites for hydrogen electro-oxidation and a significant decrease in the response time to reach steady state, tss. Increasing the anode
overpotential and the gas porosity would result in an increase in the current density, especially at low levels of CO concentration. © 2006 Elsevier B.V. All rights reserved.
Keywords: PEM fuel cell; Unsteady state; CO poisoning; Anode catalyst layer
1. Introduction
Fuel cell performance and durability are strongly influenced by impurities in the hydrogen fuel gas. Reforming of methanol or gasoline fuels is the most widely used method to generate the hydrogen fuel which contains 45% hydrogen, 10 ppm CO,
15% CO2, and 1% CH4[1]. Nevertheless, 5–10 ppm CO would
reduce the hydrogen electro-oxidation effectively by occupying the Pt reacting surface sites which results in a decrease in the cell performance[2]. To keep maintain long and stable opera-tion, how to reduce the CO concentration effectively from the reformer and enhance the CO tolerance of the fuel cell will become a significant topic.
Many researchers have focused their interest on the effects of the impurities ions from both hydrogen fuel and air on the
cell performance. Okada et al.[3–7]have examined the water
transport in the membrane of the fuel cell with the effects of the impurity ions. Water content, electricity, transference coef-ficient, and water permeability decrease with an increase in the impurity ions. Halseid et al.[8]indicated that the cell
perfor-∗Corresponding author. Tel.: +886 2 26632102; fax: +886 2 26632143.
E-mail address: wmyan@huafan.hfu.edu.tw (W.M. Yan).
mance was also influenced the ammonia from the reformer. A higher concentration of ammonia ions caused a lower membrane conductivity. To reduce the decay of the cell performance, Uribe et al.[9]used a Pt–Ru alloy to slow down the performance decay. The CO tolerance of a carbon-supported platinum–ruthenium catalyst at elevated temperatures and atmospheric pressure in a PEM fuel cell was investigated by Si et al.[10]. They indicated that the CO became dissociated from the platinum reaction sites at high operation temperatures which possess a high CO toler-ance and can provide more catalyst sites for hydrogen.
Zhang et al. [11]found that the CO poisoning process can
be accelerated at a high mass flow rate. However, increasing cathode pressure can raise the oxygen diffusion from cathode to anode side and assist in CO oxidation on the anode catalyst. In the other studies[12,13], the promotion of CO tolerance can be achieved by adopting different cell structure designs. Yu et al.
[12]used the Pt–Ru alloy as a filter to reduce the CO poison-ing. In their study, the Pt–Ru alloy was placed between the gas diffusion layer and the catalyst layer. In a similar way, Santiago et al. [13]used Ru metal to fabricate the gas diffusion layer. Before the CO entered into the catalyst layer, the high oxidation ability of Ru oxidized the CO to CO2. It is concluded from Refs.
[12,13]that the CO poisoning effect can be reduced by a simple structure. But, the fabrication cost was increased. To avoid the
0378-7753/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jpowsour.2005.12.045
Nomenclature
a contact area of Pt catalyst (cm2cm−3)
bfCO ratio of forward to backward of CO adsorption
(atm)
bfH ratio of forward to backward of hydrogen
adsorption (atm)
C concentration of reactant gas (mol cm−3)
D diffusion coefficient (cm2s−1)
EH activation energy change for hydrogen
dissociative adsorption near CO occupied sites (J mol−1)
Gf variation of free energy of CO adsorption between
zero and full coverage (J mol−1)
i current density (A cm−2)
keCO CO electro-oxidation rate constant (A cm−2)
keH hydrogen electro-oxidation rate constant
(A cm−2)
kfCO CO adsorption rate constant (A cm−2atm−1)
kfH hydrogen adsorption rate constant
(A cm−2atm−1)
n number of electrons
P total pressure (atm)
R universal gas constant (J mol−1K−1)
s stoichiometric coefficient t time (s) T temperature (K) X molar fraction z distance (m) Greek letters ε porosity η overpotential (V)
θCO coverage ratio of CO on Pt catalyst site
θH coverage ratio of hydrogen on Pt catalyst site
ξ molar area density of Pt catalyst sites (C cm−2)
φ potential (V) Subscripts CO carbon monoxide H2 hydrogen gas ss steady state Superscripts
in inlet at catalyst layer (z = 0)
0 initial state
use of a precious metal, some researchers adopted some meth-ods to restore the cell performance during the operation process, including oxidant-bleeding[14–16], self-oxidation[2,17], and current-pulsing[17,18].
In theoretical studies, Springer et al.[19]derived a mathe-matical model to describe CO adsorption, desorption and charge fluxes on the Pt catalyst site. The adsorption phenomena of hydrogen and carbon monoxide on the Pt catalyst are the main mechanisms which cause the deterioration in the cell
perfor-mance. In addition, the relationship between the rate constant and the coverage ratio was proposed by Springer et al. Chan et al.[20]combined the theoretical models developed by Springer et al.[19]and Bernardi et al.[21,22]to examine the CO kinetics. The steady-state fuel gas and coverage profiles were presented
for different CO concentrations. Bhatia and Wang[23]treated
the anode catalyst layer as a boundary condition to analyze the transient CO poisoning behaviors at different CO levels. A simple relationship between rate constants and coverage was
used. Baschuk and Li[24–27] studied the CO poisoning and
O2bleeding in a PEM fuel cell to reduce the poisoning effect
of CO. Although the transient variations of the fuel cell per-formance under different CO concentrations were observed, the actual transient coverage profiles and the reactant gas distribu-tions across the anode catalyst layer were not presented.
In the present study, a transient one-dimensional mathemat-ical model of the CO poisoning behavior in a PEM fuel cell was studied. In order to improve the cell performance, several physical parameters are considered to promote CO tolerance and analyze the influence on the response time, tss.
1.1. Theoretical model
Consider an anode catalyst layer of thickness Lc, as
schemat-ically shown in Fig. 1. Transient analysis of CO poisoning
behavior across the anode catalyst layer is investigated. The the-oretical model of CO kinetics proposed by Springer et al.[19]
was adopted in this work. For hydrogen, CO, and Pt interfacial kinetics, four expressions are described as follows:
H2+ 2Pt ↔ 2(H − Pt) (1)
CO+ Pt ↔ (CO − Pt) (2)
(CO− Pt) + H2O→ Pt + CO2+ 2H++ 2e− (3)
H− Pt → Pt + H++ e− (4)
Hydrogen dissociative chemisorption and the CO adsorption on the Pt catalyst sites are described by Eqs.(1)and(2), respec-tively. In the above equations, Eqs.(3) and (4)represent the CO and hydrogen electro-oxidation, respectively. Under the time-dependent conditions, Eqs.(5) and (6)describe the first order transient process of adsorption, desorption and charge fluxes:
ξdθH dt = kfH+ XH2P(1 − θH− θCO)− bfHkfHθH − 2keHθHsin nH2Fη 2RT (5) ξdθCO
dt = kfCO+ XCOP(1 − θH− θCO)− bfCOkfCOθCO − 2keCOθCOsin nCOFη 2RT (6) whereθHandθCOdenote the fraction of catalyst site coverage by hydrogen and CO, respectively. The forward rate constant of hydrogen and CO adsorption-to-desorption rate ratios are expressed as kFHand bfc, which are functions of CO coverage
ratio and expressed as follows:
kfH= kfH0exp −δ(GCO) RT 1− exp λθCO θCO− 1 (7)
bfCO= bfCO0exp
δ(GCO)
RT θCO
(8)
Unsteady transport equations for H2and CO across the anode
catalyst layer can be written as
εc∂CH2∂t = εcDH2 ∂2CH2 ∂z2 − diH2 dz sH2 nH2F (9) εc∂CCO ∂t = εcDCO ∂2CCO ∂z2 − diCO dz sCO nCOF (10) whereεcstands for the gas porosity,DH2 and DCOdenote the
diffusion coefficient of hydrogen and CO, respectively. s is the stoichiometric coefficient, n the number of electrons, F the Fara-day constant, andddzi is the electro-chemical reactions which are described by Eqs.(11)and(12). Subscripts H2and CO represent
the hydrogen and carbon monoxide, respectively. The operating current density is then:
di
dz =
diH2
dz +
diCO
dz = 2akeHθHsin
nH2Fη
2RT
+ 4akeCOθCOsinh
nCOFη 2RT
(11) To investigate the transient behaviors of the reactant gases distri-butions and coverage ratio distridistri-butions across the anode catalyst layer, the initial conditions are all set from zero which expressed as follows: CH2(z, 0) = C 0 H2(z, 0) (12) CCO(z, 0) = C0CO(z, 0) (13) θH(z, 0) = θH0(z, 0) (14) θCO(z, 0) = θ0CO(z, 0) (15)
At the boundary z = 0, the hydrogen and CO are given a fixed amount of concentration. At the interface between the anode catalyst layer and the membrane (z = Lc), the flux of reactant
gases equal to zero. The corresponding boundary conditions are illustrated as follows: CH2 = Cin H2, z = 0 (16) CCO= Cin CO, z = 0 (17) DH2 ∂CH2 ∂z = 0 (18) DCO∂CCO ∂z = 0 (19)
Because no electro-oxidation occurs at the interface between the anode gas diffusion layer and the anode catalyst layer, the current density is set to be zero.
i = 0 (20)
2. Results and discussion
To examine the transient behavior of the poisoning process, various CO concentrations are employed to simulate a wide range of hydrogen fuel from the reformer. Several physical parameters are considered to analyze the reactant gas distribu-tion, coverage, current density, and the time response needed to reach the steady state condition after start-up operation.Table 1
presents the parameters used in this work.
Transient evolutions of the hydrogen and CO distributions across the anode catalyst layer with 100 ppm CO are shown in
Fig. 2. Because of the fast kinetics of hydrogen, the current den-sity was provided by the hydrogen electro-oxidation resulting in much lower concentration profiles. In contrast, the concentration distribution of CO was only depleted slightly across the anode catalyst layer. As a result, both the hydrogen and the CO con-centrations take 541 s to reach the steady-state condition after start-up.
The transient distributions of the hydrogen coverage across the anode catalyst layer for 100 ppm CO are indicated inFig. 3.
Table 1
The parameters used in the present model
T (K) 353 P (atm) 3 α 0.5 DH2(cm 2s−1) 2.59× 10−6 kfH0(A cm−2atm) 100 bFh 0.5 kEh(A cm−2) 4 kECO(A cm−2atm) 10 bfCO(atm) 1.51× 10−9 keCO(A cm−2) 1× 10−8
Fig. 2. Hydrogen () and carbon monoxide () distributions at various time steps across the anode catalyst layer for 100 ppm CO,εc= 0.4,ηa= 0.01, and
Lc= 10m.
It is clearly seen that the hydrogen coverage,θH, decreases with time due to the CO adsorption on the catalyst sites. Owing to the high affinity between CO and the Pt catalyst, a large anode over-potential is needed to oxidize CO. As shown inFig. 4, the catalyst sites are adsorbed onto by the CO with reaction time. This causes the hydrogen to diffuse deeply into the catalyst layer, which in turn, seeks more catalyst sites. Without extremely high overpo-tentials to make the CO oxidation, the accumulation of CO on
Fig. 3. Distributions ofθHat various time steps across the anode catalyst layer for 100 ppm CO,εc= 0.4,ηa= 0.01, and Lc= 10m.
Fig. 4. Distributions ofθCOat various time steps across the anode catalyst layer for 100 ppm CO,εc= 0.4,ηa= 0.01, and Lc= 10m.
the catalyst sites are sustained and therefore reduce hydrogen oxidation.
Fig. 5shows the unsteady variations of the hydrogen oxida-tion current density across the anode catalyst layer. It is observed inFig. 5that the hydrogen oxidation increases sharply after the
start-up operation (within the first 2m). Comparison of the
corresponding hydrogen concentration distributions in Fig. 2
indicates that the fast kinetics of hydrogen results in a signifi-cant increase in the hydrogen oxidation current. InFig. 6, the
Fig. 5. Distributions of hydrogen oxidation current at various time steps across the anode catalyst layer for 100 ppm CO,εc= 0.4,ηa= 0.01, and Lc= 10m.
Fig. 6. Distributions of CO oxidation current at various time steps across the anode catalyst layer for 100 ppm CO,εc= 0.4,ηa= 0.01, and Lc= 10m.
CO oxidation current is of the order of 10−8A cm−2which is much smaller than the hydrogen oxidation current. This implies that the CO oxidation current can be neglected.
Under various levels of the CO concentration from the reformer, the cell performance decreases with an increase in the
CO concentration.Fig. 7shows the steady-state hydrogen
cov-erage under various CO concentrations in the range between 10 and 100 ppm. It is seen that the fuel with a high CO level would reduce the hydrogen coverage on the catalyst sites, which can reduce the cell current density significantly. Otherwise, a
sig-Fig. 7. Distributions ofθH at steady state across the anode catalyst layer for different CO concentration withεc= 0.4,ηa= 0.01, and Lc= 10m.
Fig. 8. Distributions ofθCOat steady state across the anode catalyst layer for different CO concentrations withεc= 0.4,ηa= 0.01, and Lc= 10m.
nificant rise in theθCOfrom 0.5 to 0.94 at the range from 10 to 100 ppm CO is found inFig. 8.
Fig. 9 presents the total cell current density distributions across the anode catalyst layer under various ppm levels of CO. The predicted CO poisoning results are compared with the experimental data of Oetjen et al. [28]. A current density of nearly 1 A cm−2was obtained atηa= 0.01 in the present results without CO contained. The experimental data were subjected to different CO concentration polarization curves at a 0.6 V cell voltage which corresponds to a CO-free at current density of
Fig. 9. Distributions of current density at steady state across the anode catalyst layer for different CO concentrations withεc= 0.4,ηa= 0.01, and Lc= 10m.
Fig. 10. Effects of the ppm CO concentration on the response time interval for different anode overpotentialsηaand gas porositiesεcwith Lc= 10m.
1 A cm−2. As shown inFig. 9, the results show a good
agree-ment with experiagree-mental data. A careful examination ofFig. 9
discloses that the current density would decrease from 1 to 0.487,
0.365 or 0.263 A cm−2when the hydrogen is subjected to 25,
50 or 100 ppm CO, respectively. The predicted steady-state cur-rent density under diffecur-rent ppm levels of CO is consistent with Oetjen et al.[28].
As mentioned above, the hydrogen with various CO concen-tration not only drops the cell current density but also affects the response time interval tss.Fig. 10shows the effects of the
ppm levels of CO on the response time interval tssfor different
anode overpotentials and gas porosity. An overall inspection of
Fig. 10indicates that the ppm of CO has a significant impact on the response time interval. This is because the hydrogen with a high CO concentration would increase the CO adsorption on the catalyst sites, which in turn, causes a decrease in the response time interval tss. While the overpotential and gas porosity have a
slight influence on the response time interval. The rate of hydro-gen oxidation reactions increase exponentially with the anode overpotential which drastically consumed hydrogen fuel at the anode catalyst sites. This result decreases the hydrogen cover-age. Although, the rate of CO oxidation reaction increases at the same time, it also raises the possibility of adsorbed CO on the catalyst sites. Therefore, the response time tssshows slight
differences for different anode overpotentials. In addition, the response time interval decreases with an increase in CO.
Fig. 11presents the variations of current density with CO con-centration under different anode overpotential and gas porosity. It is seen that with a higher anode overpotential or gas poros-ity at the catalyst layer one can obtain a much greater current density, especially at low CO concentrations. This can be made plausible by noting the fact that the catalyst layer with a high porosity, the hydrogen fuel can be easily fed into the catalyst
Fig. 11. Effects of the ppm CO concentration on the current density for different anode overpotentialsηaand gas porositiesεcwith Lc= 10m.
layer and a high anode overpotential can free up the catalyst reacting sites for hydrogen oxidation by bringing about the CO oxidation. With a 0.01 V anode overpotential and 10 ppm CO,
the corresponding current density is 0.79 and 0.6 A cm−2 for
porosity of 0.5 and 0.3, respectively. But, when the CO concen-tration is increased, the effects of the gas porosity on the current density would become less significant. A similar trend can be obtained for the case of an overpotential of 0.005 V at the same gas porosity under various CO concentrations.
3. Conclusions
The transient analysis of carbon monoxide poisoning at the anode catalyst layer is investigated. The CO kinetic model
devel-oped by Springer et al.[19]was adopted and extended to
sim-ulate thetransient characteristics of the CO poisoning process. In this theoretical model, the catalyst layer is treated as a thin film instead of an interface in order to investigate the response time interval required to reach the steady state under different conditions. Current density distribution, reactant gas distribu-tion, and the coverage across the anode catalyst layer are also investigated. The relevant parameters include various levels of CO concentration, anode overvoltage, and gas porosity lead-ing to the followlead-ing conclusions:When carbon monoxide exists in the hydrogen fuel, the cell performance decreases with an increase in the CO concentration. As a result, theθHdecreases with increasing CO concentration, which in turn, causes a low hydrogen electro-oxidation.The response time interval needed to reach the steady state condition, tss, is strongly influenced by
the CO concentrations. This is due to the fact that the CO electro-oxidation is insufficient to free up the catalyst sites. Therefore, it is easy for CO to accumulate on the Pt catalyst site with high lev-els of CO concentration which decrease the hydrogen oxidation,
which in turn, causes a decrease in the response time interval,
tss.A better cell performance can be obtained for a system with a higher overpotential or gas porosity, especially at low levels of CO. This is because for a case with high porosity, the hydrogen fuel can be easily fed into the catalyst layer and a high anode overpotential can free up the catalyst sites for hydrogen oxida-tion by bringing about greater CO oxidaoxida-tion.Effects of the CO levels have a significant impact on the response time interval, especially for low levels of CO.
Acknowledgement
The study was supported by the National Science Council, the Republic of China, through the grants NSC 93-2212-E-009-001.
References
[1] R.C. Urian, A.F. Gulla, S. Mukerjee, J. Electroanal. Chem. 554–555 (2003) 307–324.
[2] J. Zhang, R. Datta, J. Electrochem. Soc. 149 (2002) A1423–A1431. [3] T. Okada, Y. Ayato, M. Yuasa, I. Sekine, J. Phys. Chem. B 103 (1999)
3315–3322.
[4] T. Okada, H. Satou, M. Okuno, M. Yuasa, J. Phys. Chem. B 106 (2002) 1267–1273.
[5] T. Okada, Y. Ayato, J. Dale, M. Yuasa, I. Sekine, O. Andreas, PCCP 2 (2002) 3255–3261.
[6] T. Okada, J. Electroanal. Chem. 465 (1999) 1–17. [7] T. Okada, J. Electroanal. Chem. 465 (1999) 18–29.
[8] R. Halseid, J.S. Vie Preben, R. Tunold, J. Electrochem. Soc. 151 (2004) A381–A388.
[9] F.A. Uribe, S. Gottesfeld, T.A. Zawodzinski, J. Electrochem. Soc. 149 (2002) A293–A296.
[10] Y. Si, R. Jiang, J.C. Lin, H.R. Kunz, J.M. Fenton, J. Electrochem. Soc. 151 (2004) A1820–A1824.
[11] J. Zhang, T. Thampan, R. Datta, J. Electrochem. Soc. 149 (2002) A765–A772.
[12] H. Yu, Z. Hou, B. Yi, Z. Lin, J. Power Sources 105 (2002) 52–57. [13] E.I. Santiago, V.A. Paganin, M. do Carmo, E.R. Gonzalez, E.A. Ticianelli,
J. Electroanal. Chem. 575 (2005) 53–60.
[14] Z. Qi, C. He, A. Kaufman, J. Power Sources 111 (2002) 239–247. [15] Z. Jusys, J. Kaiser, R.J. Behm, J. Electroanal. Chem. 554–555 (2003)
427–437.
[16] P.A. Adock, S.V. Pacheco, K.M. Norman, F.A. Uribe, J. Electrochem. Soc. 152 (2005) A459–A466.
[17] A.H. Thomason, T.R. Lalk, A.J. Appleby, J. Power Sources 135 (2004) 204–211.
[18] L.P.L. Carrete, K.A. Friedrich, M. Hubel, U. Stimming, Phys. Chem. 3 (2001) 320–324.
[19] T.E. Springer, T. Rockward, T.A. Zawodzinski, S. Gottesfeld, J. Elec-trochem. Soc. 148 (2001) A11–A23.
[20] S.H. Chan, S.K. Goh, S.P. Jiang, Electrochim. Acta 48 (2003) 1905– 1919.
[21] D.M. Bernardi, M.W. Verbrugge, J. Electrochem. Soc. 37 (1991) 1151–1163.
[22] D.M. Bernardi, M.W. Verbrugge, J. Electrochem. Soc. 139 (1992) 2477–2490.
[23] K.K. Bhatia, C.Y. Wang, Electrochim. Acta 49 (2004) 2333–2341. [24] J.J. Baschuk, X. Li, Int. J. Energy Res. 25 (2001) 695–713. [25] J.J. Baschuk, X. Li, Int. J. Energy Res. 27 (2003) 1095–1116.
[26] J.J. Baschuk, A.M. Rowe, X. Li, ASME J. Energy Resour. Technol. 125 (2003) 94–100.
[27] J.J. Baschuk, X. Li, Int. J. Global Energy Iss. 20 (2003) 245–276. [28] H.F. Oetjen, V.M. Schmit, U. Stimming, F. Trila, J. Electrochem. Soc. 143