Catalysis for Low Temperature Fuel Cells PART II: THE ANODE CHALLENGES

By T. R. Ralph and M. P. Hogarth Johnson Matthey Technology Centre, Blounts Court, Sonning Common, Reading RG4 9NH. U.K

In thefirst part of this three partpaper; published in January 2002, we dealt with enhancing the performance of low temperature proton exchange membrane fuel cells by improvements to the platinum-based cathode materials and to the cathode design. In this second part of the paper we shall discuss the improvements in the reformate (CO and COJ tolerance at the anode and in extending the MEA durability. Improvements have been achieved by advances in platinum/ntthenium electrode design, in the application of bilayer anodes for durable air bleed operation, and in the addition oj‘a water electrolysis electrocatalyst to the anode for cell reversal tolerance. In the third part of the paper the particular challenges presented by the direct methanolfuel cell will be discussed.

The proton exchange membrane fuel cell (PEMFC) operating on air (or oxygen for space applications) and on selected fuels, is being devel- oped as an electrochemical power source for a variety of applications (1). While at the cathode, the choice of oxidant is either air or oxygen, the selection of the fuel is not so stmghtforward and a number of options are being pursued.

Fuel Selection The simplest and highest performing PEMFC

systems employ pure hydrogen as the fuel. For example, DaimlerChrysler and Ballard Power Systems are manufactwing transit buses with a 250 kW PEMFC engine that can provide a range of 250 miles using compressed hydrogen stored in the roof space, with overnight refuelling at the depot. This hydrogen system is envisaged as an early mar- ket entry for fuel cells. For the automotive market, using the currently available storage technology, there is not sufficient space in a car to store hydro- gen as a gas, while for stationary power generation the hydrogen supply infrastructure does not exist. In an automobile requiring 75 kW of fuel cell power, a liquid methanol feed is normally employed, although gasoline is now being exam- ined. For stationary power generation in the home 3 to 5 kW fuel cell systems are required, while 250 kW power plants are being built for larger indus- trial or residential sites. In this case natural gas is

used as the fuel since it is readily available and there are significant worldwide reserves.

The methanol, gasoline or natural gas is reformed using steam, partial oxidation or autothermal reforming to produce the reformate - a hydrogen-rich gas stream ( H z , @70”/.) contain- ing carbon dioxide (COz, 15-25%), carbon monoxide (CO, 1-2Y0) and small quantities of inert gases, such as nitrogen and water vapour. However, at current PEhfFC stack operating tem- peratures of around 80°C the membrane electrode assemblies (MEAs) within the stack cannot toler- ate such hgh CO levels. Therefore, the reformate must be passed from the reformer to a shift reac- tor and then to a catalytic preferential oxidation (CPOq reactor to reduce the CO content to less than 100 ppm, and in some cases down to a few ppm. The reformate can then enter the stack and react in the anode electrocatalyst layer of the MFA The additional reformer, shift and CPOX reactors significantly complicate and add extra cost to the PEMFC system (see (1) and (19)).

Hydrogen Electrooxidation The electrochemical oxidation of hydrogen is

extremely facile on a wide range of metals (20). The platinum group metals (pgms) such as plat- inum (Pt) and palladium (Pd) show a key advantage: they have the required stability to oper- ate in the acidic environment inside the MEA. The

Pkdnnrn Met& Ray., 2002,46, (3), 117-135 117

Fig. 9 Progressive poisoning jrom IO, 40 and I00 ppm CO on pure Pt and Pto.jRuo J alloy anodes. Increased CO tolerance is shown by the Pto sRuo J alloy anodes. The anodes are preparedjbm 20 wt.% Pt/Vulcan XC72R or 20 wt.% PI. 10 wt.% RdVulcan XC72R at a loading of’0.25 mg Pt cm-.’. The cathode uses 40 wt.% Pt/Vulcan XC72R at a loading of0.6 mg Pt cm”. The MEAs are based on catalysed substrates bonded to Nafion NE-115 membrane. The Ballard Mark 5E single cell is operated at S O T , 308/308 kPu, 1.3/2 stoichiometry with full internal membrane humidif cation

hgh activity of Pt for hydrogen electrooxidation is shown by an exchange current density of A cm-’ Pt at ambient temperature. This is some lo7 to lo9 times more facile than oxygen reduction at the cathode. As a consequence, when operating with pure hydrogen, at practical current densities, the anode potential is typically less than + 0.1 V (vs. RHE (reversible hydrogen electrode)), which is only some 0.1 V from the standard reversible potential for hydrogen electrooxidation (Reaction

6)):

Hz = 2H’ + 2e- E0z50c = 0.00 V (vs. NHE) 6)

where NHE is the normal hydrogen electrode. Under such operating conditions the cell potential is only shghtly lower than the cathode potential and the MEA performance essentially reflects the cathode operation.

CO Poisoning on Pt The mechanism of hydrogen electrooxidation

on Pt in acid electrolytes is thought to proceed by rate-determining dissociative electrosorption of molecular hydrogen (Reaction (ii)) followed by facile electron transfer (Reaction CI)) (21):

Hz + 2Pt = 2Pt-HA 2Pt-Hd. = 2Pt + 2H’ i 2e- P)

At the typical MEA operating temperature of 80°C, the CO in the reformate stream binds very

strongly to the Pt electrocatalyst sites in the anode layer. Even at ppm levels of CO, the CO coverage is above 0.98 (22). The adsorbed CO prevents the dissociative electrosorption of hydrogen (Reaction (ii)) and dramatically lowers the cell potential pro- duced by the MEA since a much hgher anode potential is required to sustain the rate of hydro- gen electrooxidation (23). This is demonstrated in Figure 9 by the effect of pprn levels of CO on an MEA with a Pt anode. The anode contains 0.25 mg Pt cm-’, and is based on a 20 wt.% Pt/Vulcan XC72R carbon-supported electrocatalyst.

As the CO content is increased from 10 to 100 ppm there is a progressive lowering in the cell potential from that achieved with pure hydrogen. At such low Pt loadmgs on the anode, the cell potential losses are s@cant even at low current densities below 200 mA cm”. But at %her cur- rent densities the cell potential losses are more dramatic. Eventually the cell potential is reduced to - 0.3 V, at which the electrooxidation of CO to C02 can occur (Reaction (iv)).

CO + HzO = COZ + 2H+ + 2e- E0z5”C = 4.10 V (VS. NHE) (iv)

Under these circumstances the anode potential reaches - +0.45 V (vs. RHE), some 0.55 V anod- ic of the standard reversible potential for Reaction (iv). This is sufficient to drive CO electrooxidation and some of the poisoned Pt sites are now free to

118

Table 111

Physical Characterisation of PtRu Alloy Electrocatalysts

Electrocatalyst

67 wt.% Pt, 33 wt.% R U (unsupported)

supported on XC72R

supported on XC72R

40 wt.% Pt, 20 wt.% R U

20 wt.% Pt, 10 wt.% RU

~~~

XRD crystallite size, nm

2.9

2.5

1.9

take part in hydrogen electrosorption and electro- oxidation (Reactions (ii) and (iii)). The cell potendal then levels out.

PtRu Alloy Electrocatalysts It is well documented (24-27) that in the

PEMFC at 80°C platinum-ruthenium (PtRu) alloys are much more tolerant to CO poisoning than pure Pt electrocatalysts. This has made them the electrocatalysts of choice for reformate opera- tion.

For example, Figuie 9 shows the large improve- ment in CO tolerance for an MEA with an anode based on a PbdtQ5 alloy electrocatalyst prepared as 20 wt.% Pt, 10 wt% Ru/Vulcan XC72R The cell potentials are sgdicantly hgher at all current densities compared with the MEA based on pure Pt, for operation with hydrogen containing 10,40 and 100 ppm CO. This improved CO tolerance has resulted in much effort to optimise the PtRu electrocatalyst for reformate operation.

For industrial-scale manufacture of PtRu alloy electrocatalysts Johnson Matthey use an aqueous slurry route with chemical reduction to form the metal alloy particles (3). Table III shows physical characterisation data of an unsupported Pb 5Rw alloy black electrocatalyst and two hgh metal- loaded Pto5R~5 alloys supported on Vulcan XC72R carbon black at 40 wt% Pt, 20 wt.% Ru and at 20 wt.% Pt, 10 wt% Ru.

Particularly interesting is the hlgh metal dis- persion of all of the electrocatalysts, especially at

Calculated metal area, m2 g-' PtRu

114

131

150

CO chemisorption metal area, m2 g-' PtRu

77

104

139

XRD lattice parameter,

nm

0.388

0.388

0.388

high metal lo-. The X-ray difftacton (XRD) crystallite sizes are 2.9 nm (PtRu black), 2.5 nm (40 wt.% Pt, 20 wt.% Ru) and 1.9 nm (20 wt% Pt, 10 wt% Ru). These are much lower than the corre- spondmg values of 5.8 nm (Pt black), 4.5 nm (60 wt.% Pt) and 2.8 nm (30 wt% Pt) for the corre- spondmg pure Pt electrocatalysts. The hlgh degree of PtRu dispersion is most probably a reflection of the surface characteristics of the materials. Although the Pb.5Ru0.5 electfocatalysts were wd alloyed the contraction in the lattice parameter shown in Table III suggests all were shghtly Pt rich (that is Pb.5pRuo.41). The lack of any crystalhe Ru- rich phases in the XRDs suggests that the unalloyed Ru is present as an amorphous phase. Cyclic voltammetty and X-ray photoelectron spec- troscopy (XPS) studies performed at Johnson Matthey indicate that the surface of the electrocat- alysts were rich in amorphous Ru oxide. This amorphous Ru oxide may play an important role in the codeposition of the PtRu particles and in reducing the degree of sintering during the electro- catalyst manufacturing process.

It is also interesting to compare the measured CO chemisorption metal areas with the calculated metal areas from the XRD crystallite sizes (assum- ing spherical particles of density 18.2 g Table llI shows the measured CO chemisorption metal areas of 77 m2 g-' PtRu (PtRu black), 104 mz g-' PtRu (40 wt.% Pt, 20 wt% Ru) and 139 m2 g-' PtRu (20 wt.% Pt, 10 wt% Ru) are 48,26 and 8% lower, respectively, than the calculated values. The

Pkatinnm Metah Rm, 2002,46, (3) 119

Fig. 10 TEMs o f20 wl.% Pt,IO wt.% Ru/Vulcan XC72R and unsupported PtRu alloy black electrocatalysts. Le j : Highly dispersed PtRu alloy crystallites on the surface ofthe primary carbon black particles. Right: The PtRu black particles clump together to form small aggregates and larger agglomerate structures

measured values become progressively lower than catalyst, the effect of alloy formulation on CO tol- the calculated metal areas as the metal loadmg on erance was examined. Electrocatalysts were the electrocatalyst increases. This probably reflects prepared, both as fully alloyed and as completely the increased clustering and fusing of the PtRu unalloyed materials, with a nominal composition alloy crystallites as the metal 10- increases. of pb 5Ru0 5.

Figure 10 shows the transmission electron The unalloyed Pb5Ruo5 was prepared by micrographs QTMs) of the 20 wt.% Pt, 10 wt.% depositing Ru onto a pre-reduced Pt electrocata- Ru/Vulcan XC72R and of the unsupported PtRu lyst XRD confirmed that there was no alloying black electrocatalysts. For the carbon-supported from the Pt lattice spacing of 0.392 nm, that is electrocatalyst, the hghly dispersed PtRu alloy equivalent to pure Pt. In addition there was no sig- crystallites are clearly evident on the primary car- nificant crystalline Ru, suggesting the Ru was bon black particles that are fused together to form present as amorphous Ru oxide. Half-cell studies the aggregate and the larger agglomerate structure were performed in sulfuric acid & S o d ) at 80°C typical of fuel cell supports. This is in contrast to using typical PEMFC anode structures containing the unsupported electrocatalyst where a large Nafion perfluorinated sulfonic acid. Cyclic number of small PtRu alloy particles form an voltammetry provided insight into the nature of extended array of PtRu aggregates and agglomer- the electrocatalyst surface. Figure 11 shows typical ates. The morphology of the unsupported cyclic voltammograms for a PbsRu05 alloy and a electrocatalyst is very different from the supported pure Pt electrocatalyst recorded at a lower temper- materials. ature of 25°C.

The alloyed Pb 5Ru0 5 had features that are pre- Effect of PtRu Composition dominantly Ru in character with no resolvable Pt

For the volume manufacture of PtRu electro- hydtide peaks. The unalloyed P~&.Ios showed catalysts it is important to understand the effect of predominately Pt character with resolvable Pt the composition on the MEA performance. Using hydride peaks, although there was strong evidence a 20 wt.% Pt, 10 wt.% Ru/Vulcan XC72R electro- of surface interaction between the Pt and the Ru

Phtnum Metah Rev., 2002,46, (3) 120

oxide. The peak potential for CO electrooxidation (Reaction (iv)) occurred at +0.44 V (vs. RHE), dose to the +0.37 (vs. RHE) for the P ~ & I O S alloy and much lower than the +0.58 V (vs. RHE) for the pure Pt electrocatalyst. The peak potentials are much lower than in Figure 11, due to the much hgher electrolyte temperature employed, although the relative peak positions are comparable.

With hydrogen containing 100 pprn CO as reac- tant the specific activity for hydrogen electrooxidation was measured. At +0.010 V (vs. RHE) the specific activity increased from 8 mA an-’ Pt for pure Pt to 25 mA an-’ PtRu for the unalloyed Pb&u05 and to 60 mA an-’ PtRu for the Pb&u05 alloy. This suggests that the alloying process which incorporates the Ru into the Pt lat- tice (and reduces the Pt lattice spacing) is important for improving the CO tolerance. Besides enhancing the intimate contact between the Pt and the Ru, extended X-ray absorption fine structure (EXAFS) analysis has shown that alloy- ing removes electron density from the Pt (28). Both these effects are capable of promoting a lower CO coverage on Pt and increasing the hydrogen electrooxidation activity of the anode.

A possible consequence of these fin- is that low Ru contents mght not show such htgh CO tolerance in the PEMFC. As an interesting exercise the Ru content in the PtRu alloy was reduced to match the output ratio of the pgms fiom the refin- ery. Thus the Ru content in a fully alloyed 20 wt.%

Pt, Ru/Vulcan XC72R electrocatalyst was lowered from Pb 5Ruo 5 to the mined ratio of Pb 7Ruo 3, and then examined in the PEMPC. The XRD lattice spacing of 0.390 nm confirmed a Pb7Ru03 d o y was formed. For hydrogen containing 10, 40 and 100 pprn CO MEAs prepared with Pb7Ru03 matched the performances in Figure 9 for the Pbau05 electrocadyst at an identical anode load- ing of 0.25 mg Pt cm?. This coniirmed that using the Johnson Matthey manufacturing route at an atomic ratio of Pt,&u03 there is sufficient Ru incorporated into the Pt lattice to adequately mod- lfy the CO tolerance of Pt.

These fin- were in agreement with the fun- damental kinetic work of Gasteiger et al. (29) who used 1000 pprn CO in hydrogen and bulk, planar

Fig. I 1 Cyclic voltammograms recorded in I mol dm-’ H ~ S O Q at 25”Cf’or Pto.75Moo.zs, Pto JRUO 5 and pure Pt electrodes. For the Pto.7~Moo.z~ alloy# CO electrooxidation activity starts at +O. 15 V (vs. RHE). For the PIO JRUO 5 alloy, the peak potential f’or CO electrooxidation is at the lowest electrode potential. The electrodes are prepared.from 35 wt. % Pt, 5 wt. % Mo/Vulcan XC72R. 40 wt.% Pt, 20 wt.% Ru/Vulcan XC72R or 40 wt.% Pt/Vulcan XC72R at a loading of’ 0.25 mg PI em-’. The electrodes were placed in a 3-compartment cell with a palladium/carbon electrode purged with hydrogen to act as the RHE and a Pt counter electrode. The electrolyte was first purged with pure CO to poison the electrode at +0. I5 V (vs. RHE) and then with nitrogen to remove any solution hase CO. The electrode potential was swept at 10 m V s- from i0.15 to i 1 .0 V (vs. RHE) then back to 0 V (vs. RHE). The second (dashed lines) and subsequent potential sweeps are between 0 V and +I.0 V (vs. RHE)

P.

pb&u05 and Pb&uOl alloys. The h d q s also

agree with the more applied results in the PEMFC of Iwase and Kawatsu (26) for alloys of composi- tion Pbs5Rua15 to Pbl&uOs5 all based on 20 wt.% Pt, 10 wt.% Ru/Vulcan XC72R electrocatalysts. They found that the CO tolerance was sqpticantly lower only at Ru contents below Pb BRuo 2 or above P b 2 R ~ a . However, there have also been a few reports of lower CO tolerance at lower Ru levels

PLdinnnz Mefa.lr Rcv., 2002,46, (3) 121

including P t d h 3 (27). This highhghts the impor- tance of the manufacturing route on the degree of PtRu alloying and on the nature of the electrocat- alyst surface. This variability probably accounts for the significant spread in CO tolerances reported for PtRu anode electrocatalysts (24).

Mechanism of CO Tolerance on PtRu While it is dear that PtRu provides considerably

enhanced CO tolerance, an understandmg of the mechanism operating at the anode is needed for the development of improved electrocatalysts. Based on early work by Watanabe and Motoo (30) (methanol electrooxidation) a bifunctional mecha- nism involving water activation by Ru (Reaction (v)) and subsequent CO electrooxidation on a neghbouring Pt atom (Reaction (vi)) has been pos- tulated by a number of groups (26,27,29,31-33).

(v) Ru + H20 = Ru-OH + H+ + e-

Pt-CO + Ru-H = COz + Hf + e- + Pt + Ru (vi)

To investigate the mechanism of CO tolerance of PtRu, half-cell studies in H2S04 at 80°C with PEMFC anode structures containing Nafion were performed using pure CO as the reactant. The half-cell outlet gas was analysed using in-line mass spectrometry for evidence of COZ - to examine PtRu for CO electrooxidation activity (Reaction (vi)). The results indicated that CO electrooxida- tion was occurring on PtRu at above +0.25 V (vs. ME). This is some 0.2 V less anodic than required for pure Pt electrocatalysts and reflects the greater ability of the Ru to adsorb water (Reaction (v)) compared with Pt. Figure 11 shows that, in the cyclic voltammograms measured in &So4 at 25"C, the promotiod effect from PtRu is also reflected in a lower peak potential for CO electrooxidation at the electrodes: PtRu (+0.5 V (vs. RHE)) compared with Pt (+0.8 V (vs. RHE)). Both the half-cell and the cyclic voltammetry stud- ies confirmed that at anode potentials above +0.25 V (vs. RHE) the improved CO tolerance of PtRu is due to the enhanced electrocatalysis of CO elec- trooxidation to COz @actions (v) and (vi)).

At lower anode potentials, where the more effi- cient anode electrocatalysts function during PEMFC operation 6.e. < +0.2 V (vs. ME)), the

situation is open to debate. In half-cell/mass spec- trometry studies at low anode potentials of 0.0 to +0.2 V (vs. RHE), there was no measurable CO electrooxidation current or detection of COz prod- uct in the half-cell outlet gas. While it has been calculated that currents as low as A cm-' Pt can remove sufficient CO from the electrocatalyst to give effective CO tolerance (34), it is reasonable to argue that CO removal is more likely due to a weakening of the P t X O bond strength. As dis- cussed, EXAFS has shown (28) that alloying does reduce the Pt electron density which should weak- en the Pt-CO bond strength and decrease the CO coverage on PtRu. This will increase the number of electrocatalyst sites available for hydrogen elec- trosorption and electrooxidation. The rate of Reactions (iiii and (ii) would then increase. This mechanism would also be consistent with the observed importance of PtRu alloying on the CO tolerance discussed earlier. Particularly relevant is the observation that the unalloyed PtRu material shows comparable CO electrooxidation activity in the cyclic voltammograms, although it is not as CO tolerant. Thus, it seems more probable that the mechanism of CO tolerance at practical anode potentials is due to a weakening in the Pt-CO bond strength rather than to the promotion of CO elemooxidation to C02.

CO, Poisoning in the PEMFC While it has been known for many years that

CO poisons Pt-based electrocatalysts, it has only recently been discovered that COZ also acts as a poison in the PEMFC (25, 35). As shown in Figure 12, the effect on a Pt anode of a hgh COz concentration of 25%, typical of many reformate streams, is modest. Only at the hghest current densities do cell potential losses become signifi- cant In fact, comparing the performance of an h4EA with an identical Pt anode in the presence of 10 ppm CO (Figure 9) and 25% COz (Figure 12) shows that the cell potentials are much lower with 10 ppm CO under the particular smgle-cell operat- ing conditions examined.

COZ poisoning can come from two sources both of which generate CO which then poisons the Pt electrocadyst sites for hydrogen electroox-

Pbdnwm Meta.lr Rev., 2002,46, (3) 122

Fig. 12 Poisoning effect of 25% C02 on Pt and Ptn 5 R U O alloy anodes. The plots show the increased C02 tolerance of the Pto sRun 5 alloy anodes. The anodes and Pt loading are as described in Fig. 9. The cathode and MEA are as described in Fig. 9. The Ballard Mark 5E single cell operating conditions are as described in Fig. 9. except that here it is using hydrogen or (hydrogen with 25% CO2)/air

idation. The CO sources are: the reverse water gas shift reaction (RWGSR, Reaction (vii)) and the electrochemical equivalent of the RWGSR (Reaction (viii)).

COz + Hz = H20 + CO 2Pt + 2H' + 2e- = 2Pt-HA 2Pt-Hds + COz = Pt-CO + H20 + Pt

(4

EO25-c = 4.10 V (a. NHE) (a) From thermodynamics it has been calculated

that at 80°C for refonnate containing 25% COZ in hydrogen, 100 to 200 ppm CO could be generated by the RWGSR (36). The exact amount of CO produced is however hlghly dependent on the water concentration. In practice, as shown by com- pacing Figures 9 and 12, with the Johnson Matthey anodes much less than 10 ppm CO is produced. This suggested that with these anode designs, the RWGSR may not be the major mode of poisoning.

C 0 2 Electroreduction Consequently, at a Pt and a PbdtuoS alloy elec-

trode in HzS04 at 70°C, an examination of the electrochemical poisoning by CO2 was performed. By holding the electrode potential constant and saturahg the &so4 with pure COZ it was shown that both the pure Pt and the Pb&uo5 alloy elec- trocatalysts form a 'Pt-CO'-like poison at electrode potentials below + 0.4 V (vs. RHE). In th is potential range there is sufficient Pt-Hdd, avd-

able to promote COZ electroreduction. Indeed, the 'Pt-CO' coverage correlates well with the Pt-Hd, coverage.

The 'Pt-CO' poison was removed by sweeping the electrode to &her potentials. The characteris- tic CO electrooxidation peak was evident in the cyclic voltammograms at peak potentials of - +0.4 V (vs. RHE) for the Pb ~RUO 5 d o y electrode, and - +0.6 V (vs. RHE) for the pure Pt electrode. It was therefore concluded that at operating anode potentials in the PEMFC, COz poisoning occurs predominantly by the electrochemical reduction reaction (Reaction (viii)), with the 'Pt-CO' poison removed at much hgher anode potentials.

Most significantly, as shown in Figure 13, the potential-dependent coverage by the 'Pt-CO' poi- son, calculated from the CO electrmxidation current in the cyclic voltammograms, is much %her for the Pt than for the Pbam5 d o y elec- trocatalyst For example, the maximum coverage recorded at +0.05 V (vs. RHE) is - 0.9 on the pure Pt electrode and - 0.5 on the Pb &Q 5 doy. The lower 'Pt-CO' coverages on the Pbauos d o y electrocatalyst suggested that it should provide an improved CO2 tolerance in the PEMFC.

COz Performance at PtRu Anodes Figure 12 shows that for single-cell operation

on 25% COZ in hydrogen, the Pb &uo 5 alloy has improved the COZ tolerance compared to the pure

Pbtinum Metah h., 2002,46, (3) 123

1 ,

In- : 0 . 2 5

0.1 0.2 0:3 0 4 ELECTRODE POTENTIAL FOR COz

POISONING, V vs. RHE

At 538 rnAcrr-’

Fig. 13 Lower ‘Pt-CO ’ coverage due to CO2 electroreduction at a Pt0.sRu0.5 alloy electrode compared to pure Pt. as a function of the electrode potential. The electrodes and Pt loading are as described in Fig. 9 for the anodes. The electrode was placed in a 3-compartment cell containing I mol dm-’ H2S04 at 70°C with a palladium/ carbon electrode purged with hydrogen. to act as the RHE and a Pt counter electrode. The electrolyte was purged with pure C02 to poison the electrode at the selected rest potential. then with nitrogen to remove any solution phase C02. The electrode potential was swept at 10 m V s-’from the rest potential to t1.0 V (vs. RHE) then back to 0 V (vs. M E ) . The ‘Pt-CO ’ electrooxidation current was used to determine the ‘Pt-CO ’ coverage

Pt anode. The clear benefit from the Pb.5Ruo.5 alloy becomes progressively greater as the current den- sity increases - because the COZ losses on pure Pt increase. Indeed, at a current density of 1 A cm-’ the cell potential of the Pb.5Ru0.5 alloy-based MEA is some 0.2 V higher. This is a significant benefit.

The better CO2 tolerance of the PtRu anode is probably due to a weakening in the ‘Pt-CO’ bond strength that is believed to be the source of the improved CO tolerance of PtRu below +0.2 V (vs.

WE). While the precise nature of the adsorbed CO intermediate(s) from COz electroreduction are not clear (37-39) the cyclic voltammograms for the electrooxidation of the ‘Pt-CO poison in&- cate that they are very similar to those produced by pure CO in this potential range.

Anode Design for Reformate Tolerance As with the development of improved cathodes

for the PEMFC discussed in Part I of this paper (40), the optimum anode for reformate operation requires not only a PtRu electrocatalyst but also an optimised anode electrocatalyst layer.

The importance of the anode design in pro- moting the CO tolerance of some Pb 5Ruo 5 anodes is hhlighted in Figure 14. This shows for single- cell operation at 538 m A cm”, the cell potential loss (relative to pure hydrogen) with fuel contair- ing from 10 to 100 ppm CO. While anode loadmgs of 1.0 mg Pt are not economicd for the majority of PEMFC applications, the performance of unsupported PtRu black layers at this hgh load- ing shows the importance of the anode layer thickness and the in sitrr anode electrode Pt surface area (EPSA), measured using cyclic voltammetry (4). The cell potential losses in Figure 14 are only 5 mV (at 10 ppm CO), 30 mV (at 40 pprn CO) and 40 mV (at 100 pprn CO). This is because the anode layer thickness is less than 10 pm and the measured in ritrr anode EPSA is high at close to 770 cmz PtRu ad.

It is important to maximise the anode EPSA as this measures the maximum PtRu surface area available for hydrogen electrooxidation in the

PtRu I X C 7 2 R

PtRu/ XC72R Medium EPSA

High EPSA

2 0 60 100 140 CO CONCENTRATION, ppm

Fig. 14 Reduction in the cell potential loss at 538 mA c K 2 (relative to pure hydrogen operation) due to 10. 40 and I00 ppm CO as the anode EPSA is increased.for Pto.sRuo.s alloy anodes. The anodes are prepared from 40 wt. % Pt, 20 wt. % Ru/Vulcan XC72R or PtRu black at a loading of 0.35 mg Pt cm , respectively. The cathode and MEA are as described in Fig. 9. The Ballard Mark 5E single cell operating conditions are as described in Fig. 9 except that here it is operating at 1.512 stoichiometry

and 1.0 mg Pt -2

Phtiinum Metah Rev., 2002, 46, (3) 124

Fig. 15 The reduction in cell potential loss at 754 mA cm-2 (relative to pure hydrogen opera- tion) due to 25% CO2 as the anode EPSA is increased for Pto 5Run J alloy anodes. The anodes and Pt loadings are as described in Fig. 14. The cathode is as described in Fig. 9. The MEAs are based on catalysed substrates bonded to Dow XUS13204.10 membrane. The Ballard Mark SE single cell operating conditions are as described in Fig. 9, except that here it is operating at 1.5/2 stoichiometry

0.07 - 0.06-

0.05 - 0.04.

0.03-

0.02-

0.01 -

At 754mA cm-2

100 200 so0 400 500 600 760

absence of CO and COZ poisoning. As the anode layer is thin, the whole anode layer thickness is active and all of the EPSA can therefore be utilised during PEMFC operation. Proton conductivity and hydrogen permeability do not limit the active layer thickness of the anode. This is particularly important at hlgh current densities (41).

In contrast, even using a hlgh loaded 40 wt.% Pt, 20 wt.% Ru/Vulcan XC72R-supported electro- catalyst at 1.0 mg Pt cm-’, the anode layers are much too thick. As a consequence, much of the depth of the anode layer is redundant and the cell potential losses are much higher than shown in Figure 14 for the unsupported PtRu black anode layer at 1.0 mg Pt m-’.

Significantly, however, at more economical electrode Pt loadmgs of 0.35 mg Pt cm-’ the anode layer is sufficiently thin for the carbon-supported electrocatalysts to ensure all of the anode thickness is used during PEMFC operation. In this case, the PtRu black anode cannot achieve the same degree of CO tolerance as a well designed anode layer using 40 wt.% Pt, 20 wt% Ru/Vulcan XC72R This is because the CO metal area is lower for the PtRu black (see Table m) and the anode EPSA is reduced.

For carbon-supported electrocatalysts to

achieve a lugh anode EPSA it is necessary to pay attention to the interaction between the PtRu sites on the carbon support (shown in the TEM in Figure 10) and the protons in the perfluorinated sulfonic acid solution used to provide proton con- duction within the anode layer. With the traditional aqueous-based electrocatalyst inks employed at Johnson Matthey to deposit the anode layer, the

EPSA, cm2 P ~ R U cm?

initial in situ anode EPSAs for the 40 wt.% Pt, 20 wt.% Ru/Vulcan XC72R layers were low. Initial measurements at an anode loachg of 0.35 mg Pt cm-2 indicated that the anode EPSA was below 200 cm’ PtRu m-’. This was much lower than the 364 cm’ PtRu cm” potentially available.

Therefore, improved aqueous-based PtRu inks were manufactured which enhanced the interac- tion between the electrocadyst and the Nafion proton conducting polymer solution. With the improved inks, the in situ anode EPSA of the resulting anode layers was raised much closer to the maximum 364 cm’ PtRu cm-’. This produced the significantly lower cell potential losses in Figure 14 for the 40 wt.% Pt, 20 wt.% Ru/Vulcan XC72R anodes. The increasing anode EPSA pro- duces cell potential losses at 538 mA m-’ that are reduced by 40 mV (at 10 ppm CO), 70 mV (at 40 pprn CO) and 60 mV (at 100 ppm CO), respec-

Comparable improvements in the COZ toler- ance of the h4EA are also observed for the better PtRu anode designs. Figure 15 shows, for smgle cell operation at 754 mA cm-’, the relationship between the cell potential loss (relative to pure hydrogen) with 25% CO, in hydrogen and the in situ anode EPSA. As the EPSA is increased from 160 to 600 m’ PtRu cm? for the 40 wt.% Pt, 20 wt.% Ru/Vulcan XC72R and the unsupported PtRu black-Nafion ink layers, the CO2 tolerance is improved. In Figure 15, the cell potential loss is reduced from 65 mV to as little as 10 mV.

Further work has shown the importance of ensuring that the in sih anode EPSA does not fall below the values in F w e 15. For an anode layer

tively.

Phtinnm Metab Rnt., 2002,46, (3) 125

containing 0.25 mg Pt cm-', and based on 20 wt.%

Pt, 10 wt.% Ru/Vulcan XC72R, removing the Nafion from the catalyst ink (by using an altema- tive binder) dramatically lowers the COz tolerance. For single-cell operation at 538 mA cm-' with 25% C02 in hydrogen at the operating conditions detailed in Figure 12, the cell potential was nearly 150 mV lower. The reduction in COz tolerance is due to a very low in J-&+ anode EPSA below 100 cmz PtRu ~ n - ~ . It seems that the Pt not in contact with Nafion can catalyse the RWGSR (Reaction (vii)) to generate levels of CO approachmg 100 PPm.

It is therefore clear that to achieve a hgh degree of reformate tolerance it is important to maximise the in siru anode EPSA by increasing the Pt dispersion of the electrocatalyst and by max- imising the proton contact to the PtRu sites from the proton conducting polymer in the anode. Equally important to the in situ anode EPSA is the production of thin electrocatalyst layers. This reduces the possibility of the proton conductivity and the hydrogen permeability in the anode layer from limiting the performance of the MEA, par- ticularly at htgher current densities (41).

Alternative Reformate Tolerant Electrocatalysts

There has been a significant effort over several decades to mitigate the effects of CO poisoning at the anode by developing electrocatalysts that are superior to PtRu (36,42,43). The electrocatalysts generally fall into two categories.

(i) Intrinsic Mechanism The fist category is the wide range of Pt alloys

that have been examined in an attempt to modify the CO and hydrogen electrosorption properties of Pt - to reduce the CO coverage and increase the rate of hydrogen electrooxidation. Much study has focused on PtRu, PtRh and PtIr alloys. These were ori@y examined in the 1960s by General Electric (e.g. see 44-46) as unsupported metal blacks at high anode loadmgs of 30 mg Pt ern-'. More recently (47) they have been prepared on carbon supports and examined at lower anode loacllngs of 0.5 mg Pt cm-'. It was shown that at

below 100°C none of the Pt alloys was superior to PtRu but that the CO tolerance of PtRh approached that of PtRu. Iwase and Kawatsu (26) also found that at 80°C a large number of Pt alloys supported on Vulcan XC72R (PtRh, PtIr, PtPd, PtV, PtCr, PtCo, PtNi, PtFe, PtMn) a l l gave infe- rior CO tolerance compared to PtRu.

However, there is a considerable drive to raise the operating temperatures of the PEMFC to above 100°C. This would raise the system efficien- cy and dramatically improve the CO tolerance of Pt-based electrocatalysts. For example, the CO coverage on Pt drops from in excess of 0.98 at 80°C to - 0.5 at 160°C (43). At this temperature a Pt-based anode can function in the presence of 1 to 2% CO rather than with ppm levels of CO, eliminating the need for the CPOX reactor. If the temperature can be raised to 160"C, studies in the phosphoric acid fuel cell (PAFC) with 1 to 2% CO indicate that PtRh and PtNi may offer superior CO tolerance (48, 49).

Indeed, at 200°C pure Pt is the favoured elec- trocatalyst in the PAFC. At hlgher temperatures PtRu may not be the electrocatalyst of choice in the PEMFC. There is currently much research (50) aimed at developing membranes capable of proton conduction at 120 to 200°C.

(ii) Promoted Mechanism The second category of electrocatalysts

involves modlfymg Pt with metal oxides to catal- yse the electrooxidation of CO to COZ (Reaction (iv)) at operating anode potentials. This would reduce the CO coverage and increase the number of Pt electrocatalyst sites available for hydrogen electrooxidation. Agam the most promising group of materials date from the early General Elecmc work (44,51). Based on these studies, carbon-sup- ported platinum-molybdenum (PtMo), PtCoMo, PtWOt and PtCoWO3 have recently been investi- gated. The electrocatalysts were shown to have enhanced CO tolerance properties compared to carbon-supported PtRu electrocatalysts (47,52).

In particular, it was recently reported (53) that a carbon-supported Pb Moo alloy electrocatalyst produced, in a small PEMFC smgle cell operating at 85"C, only a 50 mV loss in cell potential for

Phtinum Metals Rev., 2002, 46, (3) 126

Fig. 16 Speci/ic activity as m./unction of anode potential ,for hvdrogen electroo.ridation in the presence of 100 ppm CO and .for pure CO electroosidation at pure Pt, Pto jRirn 5 alloy arid Pto. 7 . 4 f o g . 2 . ~ alloy electrodes. At low anode potentials the Pto. 7jMO0.25

allov shows a higher specijic activiy,for hj'drogen and CO electrooxidation. The electrodes and Pt loading are as described in Fig. 11. The electrodes were located in a hulf'cell with a hvdrogen-putged platinised Pt electrode as the RHE and a PI counter electrode. The halfcell was placed in I ntol dnii3 H ~ S O J at 80°C and the electrolyte deo.iygenated with nitrogen. The stea(i).-state current was measured at selected electrode potentials from 0 V (vs. RHE) using hydrogen with 100 ppm CO und then pure CO. The electrode potential was returned to 0 V (vs. W E ) . Cyclic voltammograms were recorded at 10 ni V s-' between 0 and t1.0 V Ivs. RHE) to obtain the CO electroosidation currents used to calculate the EPSAs and the speci/ic activities

operation with hydrogen containing 100 pprn CO. In contrast, the carbon-supported Pb.sRu0.~ alloy electrocatalyst showed a cell potential loss of 160 mV. The promising PtMo system has been exam- ined at Johnson Matthey.

PtMo Alloy Electrocatalysts A Pto 7&00 25 alloy was prepared as 20 wt.% Pt,

3 wt.% Mo by sequentially depositing Pt oxide and Mo oxide followed by thermal reduction at high temperature. Half-cell measurements in HZS04 at 80°C using Nation-containing electrodes were per- formed to determine the specific activity for hydrogen electrooxidation in the presence of 100 pprn CO. Figure 16 shows the resulting plots of anode potential versus the specific activity for the Pb 75M00.25 alloy electrode compared to the corre- sponding PtosRuo.5 alloy electrode and a pure Pt electrode. As expected, due to the greater level of CO poisoning it is only at the pure Pt electrode that anode potentials in excess of +0.05 V (vs. WE) are required to sustain hydrogen electrooxi- dation. Most significantly, at anode potentials below +0.025 V (vs. W E ) a higher specific activ-

ity for hydrogen electrooxidation is measured for the Pb.75Mo0.25 alloy electrode compared to the Pto.sRu0.5 alloy electrode.

The source of the improved hydrogen elec- trooxidation activity was confirmed by using pure CO as the reactant and monitoring the half-cell exhaust for product COZ using mass spectrometry (Reaction (iv)). Figure 16 shows that, for the Pb.7&fo0.25 alloy electrode with pure CO, there is considerable specific activity for CO electrooxida- tion starting fi-om anode potentials of +0.05 V (vs. WE). Mass spectrometry of the exhaust gas con- firmed that the electrooxidation current was due to product COZ formation.

This indicated that at anode potentials below +0.2 V (vs. RHE) (needed for efficient PEMFK operation), the CO tolerance of the PtMo electrocatalyst comes from promoting CO electrooxidation (Reaction (iv)). The low electrode potentials required for CO electrooxidation were also evident in the cyclic v o l t a m m o p s . Figure 11 shows for the Pb,&fo0.~ alloy electrode a broad oxidation current starting at +0.15 V (vs. RHE) that is complicated by Mo surface redox reactions (54).

Plarinum Met& Rm, 2002,46, (3) 127

Fig. I7 The ability of Pto 7Moo 25 and Pto 5& 5 alloy anodes to tolerate 46 ppm CO in hydrogen and 25% C02 in hydrogen. The Ptn 75MOo 25 alloy is more CO tolerant but much less C02 tolerant. The PtMo and PtRu anodes and their Pt loading are as described in Fig. 11. The cathode and MEAs are as described in Fig. 9. The Ballard Mark 5E single cell operating conditions are as described in Fig. 9, except that here it is operating at 1.512 stoichiometry

These results are in direct contrast to those from both the pure Pt and the Pb.&w.5 alloy elec- trodes. There is no detectable specific activity for CO electrooxidation below +0.2 V (vs. RHE) in Figure 16 and no product COZ was detected in the haf-cell exhaust gas. Figure 11 confirmed that there was no electrooxidation activity measured at these low electrode potentials for either the Pb.&u~.~ alloy or the pure Pt electrodes in the cyclic voltammograms.

CO Tolerance of PtMo MEAs Based on the promise offered by the half-cell

results, the Pb7~o025 alloy electrocatalyst was examined in the PEMFC. The ME4 performances confirmed the much improved CO tolerance for hydrogen containing from 40 to 100 ppm CO. For example, Figure 17 shows that for operation with hydrogen containing 46 pprn CO, at an electrode loadmg of 0.25 mg Pt cm?, the P b 7 & h 0 2 5 d o y provides much higher performance than the Pb &Q 5 anode. The performance benefit increas- es progressively from 400 mA cm-’ and at 1.0 A cm-’ the Pb7@0025 alloy is operating at some 0.2 V above the Pb 5Rw 5 alloy.

However, as the CO concentration is reduced the benefit due to the Pb 7@00 25 d o y electrocata- lyst is reduced, until at 10 pprn CO, the Pb5RuO5 alloy electrocatalyst is the more CO tolerant. The difference in reaction order for the Pb 75Moo 25 and

the Pb.5R~.5 alloys is perhaps indicative of the dis- tinct mechanisms of CO-tolerance operating at the electrocatalysts.

COz Tolerance of PtMo MEAs However, a much larger issue than the MEA

performance at low ppm levels of CO was found for the Pb&i0025 alloy. When operating on hydrogen with 25% C02, the pb7&k1025 alloy shows poor CO2 tolerance. As shown in Figure 17 the Pb7&f0025 alloy shows a similar performance to hydrogen containing 46 pprn CO. This con- trasts with the Pb5Rw5 alloy where the single-cell performance in Figure 17 with hydrogen c o n k - ing 25% CO2 is much &her than with hydrogen containing 46 ppm CO. Indeed, comparison with Figure 9 shows the cell potential loss with 25% COz is less than with 10 pprn CO. As a conse- quence, for full reformate operation with hydrogen con- both 25% COZ and 40 pprn CO, the Pb &a, alloy is much more tolerant than the Pb7&f0025 alloy. Figure 18 indicates that at 538 mA cm ’, compared with the hydrogen perfor- mance, the Pk&w5 alloy shows a cell potential loss of 100 mV but the Pb7@0025 alloy shows a much higher cell potential loss of 215 mV. As expected, the pure Pt-based anode provides the lowest performance and is operating at 390 mV below its performance with pure hydrogen. Thus, today, with typical reformate feeds that contain

Pkatinum Metah REV., 2002,46, (3) 128

C02, PtRu remains the electrocatalyst of choice for PEMFC operation below 100°C.

It is worth considering the reason for the poor COZ tolerance of the PtMo electrocata- lyst and the impact on the anode design for the PEMFC. The poor C02 tolerance of the PtMo system reflects an increased ability of the electrocatalyst to promote the electrore- duction of COa (Reaction (viii)). This produces much higher coverages of the 'PtXO' poison. The source of the CO toler- ance shown by the PtMo electrocatalyst, namely its ability to electrooxidise CO (Reaction (iv)) also results in an ability to catalyse the reverse electroreduction reaction (Reaction (viiii). The relative magnitude of the electrooxidation of CO and the elec- troreduction of COz will depend strongly on the concentrations of each reactant and on the PEMFC operating conditions. At present, for operation below lOO"C, the reformate streams usu- ally contain much more COz than CO. Figure 18 indicates in t h i s situation the rate of C02 electrore- duction is dominant. This implies that the search for alternative electrocatalysts to provide improved CO tolerance by electrooxidising CO at low anode potentials may be fundamentally flawed for current PEMFC stack operating conditions.

The work to develop new membranes to allow operation of the stack at lugher temperatures, above 100°C could change the situation. For exam- ple at 160°C much greater CO concentrations of 1 to 2% can be fed to the MEA. Alternatively, mem- brane purification (55) could be adopted to reduce C02 to ppm levels. Using membrane purification would allow PtMo to show a clear benefit over PtRu electrocatalysts at CO concentrations greater than 10 ppm.

The Air Bleed Technique At present, for PEMFC operation below lOO"C,

PtRu is st i l l the favoured electrocatalyst for full reformate operation. However, even with PtRu in a thin anode layer designed to have a hlgh EPSA there are sti l l losses in cell potential with full reformate operation that lower the MEA's performance. The losses can be largely mitigated

Fig. 18 The reformate tolerance ofpure Pt, Pto 7sMoo.z~ alloy and Pto 5RUO 5 alloy anodes. The change in cell potential with time at 538 mA cm-2 on switching the anode fuel from hydrogen to hydrogen with 40ppm CO and 25% C02 shows the much higher tolerance of the Pto 5Ruo s alloy anode. The anodes and Pt loading are as described in Fig. 11. The cathode and MEAs are as described in Fig. 9. The Ballard Mark 5E single cell operating conditions are as described in Fig. 9, except that here it is operating at 1.5/2 stoichiometry

by injecting a small air bleed (I 2%) just ahead of the PEMFC stack (23) or within the stack itself (56). This air removes the CO from the PtRu elec- trocatalyst sites by gas phase dissociative chemisorption of oxygen on the CO-free Pt cata- lyst sites, followed by bimolecular surface reaction between Pt-CO and Pt-0 to form C02 (Reaction CW (57, 58):

CO + %OZ = COZ AH = -283 kJ mol-' (is)

The gas-phase catalytic oxidation reduces the CO coverage on the PtRu electrocatalyst and increases the hydrogen electrooxidation activity. Normally temperatures in excess of 100°C are required for catalytic oxidation of CO, but since the CO levels are low (10-100 ppm) there are suf- ficient free Pt sites to promote Reaction (ix) at 80°C. It seems that the presence of hydrogen and water in the PEMFC has little impact on the cat- alytic oxidation. In fact, complete tolerance to CO levels of up to 100 ppm can be achieved in w d - designed PtRu anode structures.

Typically, however, with PtRu anodes the small loss in cell potential due to COZ poisoning, shown in Figure 12, is not completely recovered. This is probably because there are some PtRu electrocata- lyst sites not in contact with the proton conducting polymer that are free to promote the RWGSR

PMnnm Mexklr Rm, 2002,46, (3) 129

Fig. 19 Durabilit>j of a standard arid a bilaver anode (se1o.v layer) structuse f o s lydrogen/ais and refosmaie (70% Iiydrogen. 5% nitrogen, 25% CO., 40 ppni CO)/uis operution at 754 mA em-- with an air. bleed. The standard anode ,fails after 1000 h0ur.s hut the hilayer anode is siiN,functioning ufter 2000 1iour.s of continuous operation with a 3% air bleed. The anodes use based on 20 wi.% Pi. 10 wt.% Ru/Vulcan XC72R at a loading of 0.25 mg Pt cni ’. The bilayes anode cont$ns a catal\.iic o.riduiiori IaJjes (selox laver) psepared,froni 20 wt.% Pi/Shawinigan at a loading of0. I mg Pt cnii‘. The cathode is as descrihed in Fig. 9. The MEAs use as de.scrihed in Fig. 15. The Ballasd Mark 5E single cell operating condiiions use as described in Fig. 9, except that here it is operating ai 70°C at stoichiometsies of l .5 /2

(Reaction (vii)). There have been a few recent reports (24) that the COZ losses can be completely mitigated in very thin anode layers at electrode loadings below 0.2 mg Pt crn-’. In this case, all of the PtRu electrocatalyst is probably in contact with the proton conducting electrolyte.

Durability Issues While the air bleed technique is capable of

recovering the cell potential losses due to CO poi- soning and most of the losses due to CO, poisoning there are st i l l issues to be addressed. For instance, in addition to the catalytic oxidation of CO to CO2 (Reaction (ix)) the air can recombine with the hydrogen fuel in the presence of Pt-based catalysts (Reaction (x)):

H2 + %02 = H 2 0 AH = -251 kJ mol ’ (x)

Recombination removes a little of the hydrogen fuel but the major concern is that both the CO oxidation and the recombination reactions are highly exothermic. This increases the local temper- ature within the MEA, and indeed, temperatures

exceeding 100°C have been recorded in MEAs that are functioning in PEMFC stacks operating at 80°C.

The increased temperature can result in sinter- ing of the PtRu electrocatalyst and eventually to pin-holing of the membrane and failure of the MEL4 and the stack. Figure 19 shows, for example, the performance with time of a standard Johnson Matthey anode based on a Pb&h5 alloy electro- catalyst. The MEA is operating at 754 mA cm? and 70°C on reformate containing 25% COZ and 40 ppm CO. Initially a 3% air bleed is sufficient to sustain the cell potential but this has to be increased progressively with time to 4, 5 and then 7% to maintain the MEA’s performance. Cyclic voltammetry in situ in the PEMFC indicated that the anode EPSA was decreasing with time. XRD confirmed that this was due to sintering of the P~.sRuo.s d o y electrocadyst.

The resulting reduction in the number of active electrocatalyst sites was balanced by increasing the air bleed, to sustain the number of CO-free PtRu electrocatalyst sites. However, Figure 19 shows

Platinum Metah Rev., 2002,46, (3) 130

that the pure hydrogen performance was not affected as only a few electrocatalyst sites are required in this case. Half-cell studies have shown that in the absence of reformate poisoning, as little as 0.025 mg Pt cm-’ of a 20 wt.% Pt/Vulcan XC72R electrocatalyst is sufficient to sustain hydrogen electrooxidation below +0.05 V (vs. M E ) . Eventually at a very hlgh air bleed level of 7’4 applied after 1000 hours of continuous opera- tion, the MFA based on the standard anode failed due to pin-holing of the membrane.

Bilayer Anodes

Johnson Matthey and Ballard Power Systems have significantly extended anode lifetimes by introducing a bilayer anode structure. In this struc- ture, a carbon-supported Pt-based catalyst layer is inserted between the PtRu electrocatalyst layer and the anode subsixate (59). As the function of the additional layer is to promote gas-phase CO oxida- tion, it does not contain a proton conducting ionomer - in contrast to the electrocatalyst layer. The ionomer can hinder the distribution of the air bleed.

Figure 19 shows that the bilayer anode structure (that is the selox layer) achieved a comparable MEA performance to the standard anode when operating both on reformate and on pure hydro- gen. More importantly, the bilayer anode did not require an increase in the level of the air bleed beyond 3% to sustain the MEA performance, even after 2000 hours of continuous operation. The heat generated by Reactions (ix) and (x) has been largely removed from the electrocatalyst layer and the electrocatalyst layer-membrane interface. This minimises the degree of electrocatalyst sintering and significantly reduces the possibility of mem- brane pin-holing. The anode gases and the stack cooling can more effectively remove the heat gen- erated in the gas-phase selox layer.

Using the bilayer anode has resulted in stable MEA performances over many thousands ofhours of continuous operation. For example, for an MEA with a bilayer anode containing 0.35 mg Pt cm-’ the reformate performance with a 2% air bleed was stable after in excess of 8000 hours of continuous operation. This bilayer approach may

produce the required MEA lifetimes for PEMFC applications which need an air bleed to sustain reformate operation.

Cell Reversal Tolerant Anodes

The use of an air bleed is not the only durabil- ty issue evident at the anode. During operation it is possible that one or more MEAs in a stack, or even a complete stack in a multi-stack system, will show a reverse in polarity. Such cell reversal incidents reduce the power output from the system and can irreversibly damage the stack. There can be sevetal sources of cell reversal but most damaging is reactant starvation of the MEA, especially at the anode. Fuel and oxidant starvation can arise from a lack of system control during a sudden change in reactant demand. Alternatively, a flow field chan- nel may be blocked due to stack material debris or by water flooding and inert gas blanketing of the MEA.

Air Starvation If the MEA is starved of oxidant, then rather

than oxygen reduction sustaining the current flow- ing through the cathode, hydrogen evolution takes place (Reaction (xi)):

2Hf + 2e- = Hz E2yc = 0.00 V (vs. NHE) (xi)

The anode reaction and the anode potential remain unchanged and the PEMFC acts as a hydrogen pump. The cell potential at 500 m A cm-’ is typically close to -0.1 V since both hydrogen electrooxidation and hydrogen evolution are facile on Pt-based electrocatalysts. Such a condition does raise reliability concerns since hydrogen is pro- duced in the cathode chamber and significant heat is generated in the MEA. With a lack of water pro- duced at the cathode, this can damage the membrane.

Fuel Starvation However, more dramatic effects arise if the

MEA is starved of fuel. Studies at Ballard Power Systems (60) have shown that the electrochemical processes in this case can be monitored by follow- ing the cell potential with time and by gas chromatographic analysis of the anode outlet gas for oxygen and CO2 production. Figure 20 shows

Pkztinnn Me& Rcu., 2002,46, (3) 131

with nitrogen at the anode (replirating,fiiel starvation)/aic 200/200 kPa, 1 3 2 stoicliionietn. The o.:vgen and C'O? were detected in the anode outlet gas using gus chromatogruphv

the typical response from an MEA with a carbon- supported PtRu anode. In this case the cathode reaction and the cathode potential remain unchanged and the change in cell potential reflects the anode potential. As the anode is starved of fuel the anode potential increases until water electroly- sis occurs (Reaction (xii)):

H20 = % 0 2 + 2H' + 2 4 Ezyc = +1.23 V (VS. NHE) (s)

Figure 20 shows that at 40 mA cm-' this corre- sponds to a cell potential of 4 . 6 V. After falling quickly to 4 . 6 V, the cell potential gradually decays to -0.9 V correspondtng to an anode potential of +1.4 to +1.7 V (vs. WE). That the resulting water electrolysis plateau in Figure 20 is due to Reaction (xii) was confirmed by the hlgh current efficiency for oxygen evolution. At such a high anode potential, a small degree of carbon cor- rosion (Reaction (xiii)) accompanies water electrolysis as shown in Figure 20 by the small quantity of COZ detected in the anode outlet gas:

C + 2Hz0 = COz + 4H' + 4e E"2yc = -0.21 V (VS. NHE) (xiii)

If the supply of water to the anode electrocata- lyst runs out or if the Pt-based anode electro- catalyst is deactivated for water electrolysis, the cell potential moves to more negative potentials. Figure 20 shows a carbon corrosion plateau as the

Fip. 20 The chanpe i n the ceil potential 111 41) niA cm ' arid the rate ofoxygen arid CO2 produced at un MEA during cell reversal due to ,file1 starvation. The anode is prepared,frvm 2 0 $!'I.%

Pt. 10 wt. !% Ric/Shuwiriiguri at u,loading 01'0.25 mg Pt cm--. The cathode uses Pt black at (I loading of'4.0 nig Pi c K 2 . The MEAs are based on cuta!vsed sub- strates bonded to Nufiori NE- I 13.5 membrane. The Ballard Murk S E single cell operating conditions are us described iri Fig. 9. except that here it i s operuting

cell potential reaches -1.4 V, corresponding to an anode potential of +2.1 V (vs. WE). At this stage the current is sustained increasingly by carbon cor- rosion rather than by water electrolysis, as shown in Figure 20 by the increase in the rate of COZ pro- duction and the decrease in the rate of oxygen production in the anode chamber. Significant irre- versible damage to the MEA occurs since the anode electrocatalyst carbon support, the anode gas diffusion substrate and the anode flow field plate (if carbon-based) all corrode. After a few minutes in this condition the h4EA is normally electrically shorted due to the significant amount of heat generated in the membrane.

Water Electrolysis Electrocatalysts The effects of cell reversal can be mitigated by

using diodes to carry the stack cutrent past each MEA or by monitoring the cell potential of each MEA and shutting the stack down if a cell poten- tial is low. However, such measures are complex and expensive to implement. A more practical approach to protect the carbon-based compo- nents at the anode is to incorporate an additional electrocatalyst in the anode catalyst layer to sustain the rate of water electrolysis. This approach has been jointly investigated by Johnson Matthey and Ballard Power Systems (60).

F w e 21 shows the performance in Ballard stack

hardware of three different cell-reversal-tolerant

Phtinrm Met& Rm, 2002,46, (3) 132

(i.e. water electrolysis) electrocatalysts prepared by Johnson Matthey. In this case the MEAS have received significant prior cell-reversal periods to drive the anodes to the limit of their tolerance. The impact that inclusion of a water electrolysis elec- trocatalyst has on the ability of the anode to sustain water electrolysis is evident. At the standard 40 wt.% Pt, 20 wt.% Ru/Shawinigan carbon black- based anode the water electrolysis plateau is so short it is difficult to detect. Only by using a cell- reversal-tolerant electrocatalyst in the anode are water electrolysis plateaux evident in Figure 21.

As a result, the degree of carbon corrosion is significantly reduced by the cell-reversal-tolerant electrocatalysts. The typical PtRu anode electrocat- alyst layer shows carbon corrosion after only 15 seconds of operation under the final cell-reversal conditions of Figure 21. This is extended to 4.5 minutes by adding RuOZ to the electrocatalyst layer, to 24 minutes using Ru02/Ti02 and to 48 minutes with RuOz/Ir0~ in the anode layer. While the carbon corrosion plateaux are not clear in Figure 21, the relative rate of production of COz and oxygen conhrmed the time-scale required for significant levels of carbon corrosion.

Fig. 21 The change in cell potential at 200 mA em-’ during cell reversal due to fuel starvation at a PIO 5RUQ J alloy anode containing different water electrolysis electrocatalysts. The MEAs were

first subjected to cell reversal at 200 mA cm ’for 5 minutes then for 3 sets of 30pulses (10 seconds on, 10 seconds om with I5 minute recovery Deriods on

After this extended cell-reversal tesang, return- ing the more cell-reversal-tolerant anodes (optimised for PEMPC operation) to normal fuel cell operation indicated that there was negligible loss in the MEA power output This contrasts with MEA failure in the absence of the cell-reversal- tolerant electrocatalyst. In addition, incorporation of the water electrolysis electrocatalyst in the favoured anode structures adds very little extra pgm to the MEA (< 0.1 mg cm-3. Thus, this safe- guard to the stack durability comes at little additional MEA cost.

Conclusions At the current MEA operating temperature of

80°C there are sgmficant reformate poisoning issues at the anode. The most tolerant electrocata- lyst is still PtRu alloy. While unalloyed PtRu does promote CO electrooxidation above +0.25 V (vs. WE) the alloying process is important for max- imising the CO tolerance below +0.20 V (vs. RHE). This is probably because it weakens the Pt-CO bond srrength. With the Johnson Matthey manufacturing process the Ru content can be reduced from Pt0.5Ruo.~ to Pb ,Rue without any loss

hydrogedair and ajinal overnight recovery period. The anodes are as described in Fig. 20 but at a loading of 0. I mg Pt The water electrolysis electrocatalysts were prepared from 20 wt. % Ru02 /Shawinigan. 20 wt. % Ru02, no2 (Run PEO t)/Shawinigan or 20 wl.% RuO2, Ir02 (Run.9lro.t)/Shawinigan and added at an anode loading of 0.1 mg Ru em ’. The cathode is as described in Fig. 9. The MEAs are as described in Fig. 20. The Ballard Mark 5E single cell operating conditions are as described in Fig. 9, except that here it is operating at 75°C using nitrogen at the anode (replicating fuel starvation)/ai,: 160/160 kPa. with stoichiometries of 1.311.5

P h f j n m Mefuh Rey., 2002,46, (3) 133

in anode performance, presumably because there is sufficient Ru incorporated into the Pt lattice.

To realise the highest reformate tolerance from PtRu it is important to maximise the in situ anode EPSA and to produce thin anode layers. The high in sitI4 EPSA increases the number of poison-free PtRu sites and minimises CO production from the RWGSR Thin layers are produced by using high loadings of 40 wt.% Pt, 20 wt.% Ru with carbon- supported materials, and high in situ anode EPSAs from good PtRu ink preparations. Unsupported PtRu black functions well in PEMFC applications that employ high anode Pt loadings.

Efficient anodes operate at electrode potentials below +0.2 V (vs. RHE). At such anode potentials the PtRu alloy provides CO and C02 tolerance by weakening the Tt-CO’ bond strength. Only at much higher anode potentials does water adsorp- tion on Ru promote significant levels of CO electrooxidation to CO,. The search for electro- catalysts capable of electrooxidising CO at much lower anode potentials has led to PMo alloys that are capable of promotion below +0.1 V (vs. RHE). Consequently, in PEMFC testing above 10 ppm CO, carbon-supported Pf0.,&i0,,.2~ alloy shows an increased level of CO tolerance com- pared to PtRu.

However, the CO2 tolerance of the Pb.&f00.2~ alloy is very poor. The electrocatalyst also pro- motes the electroreduction of COz to a Tt-CO’ poison. The result is much poorer reformate toler- ance compared to PtRu using typical anode feeds containing CO and CO2. Application of a mem- brane purifier to significantly reduce the C02 content and raising the MEA operating tempera- ture to 160°C might allow the improved CO tolerance of the PfO.7&f0~.~~ alloy to be realised.

Even with PtRu in a well-designed anode layer there are losses in cell potential, with reformate operation, that are not recovered. The air bleed technique must be employed to introduce S 2% air into the fuel stream to catalyse the oxidation of CO to COZ. This recovers the cell potential losses at 100 ppm CO and most of the small loss due to 25% CO2. However, in standard anode designs the air bleed introduces durability issues, due to s@- icant local heat generation within the anode from

the catalytic oxidation of CO and the recombina- tion of hydrogen and oxygen. The PtRu electrocatalyst sinters and the membrane eventual- ly pin-holes. To extend the anode lifetime it is necessary to adopt a bilayer anode structure using a pgm-based gas-phase cataly& oxidation layer between the gas diffusion substrate and the PtRu electrocatalyst layer. This moves the heat genera- tion away from the the PtRu electrocatalyst layer and the membrane.

An additional major durability issue arises at the anode if the MEA is starved of fuel. The MEL4 changes polarity due to cell reversal. Initially water electrolysis sustains the power output but after some 20 minutes at 200 mA cm-2 the PtRu elec- trocatalyst is deactivated for water electrolysis and significant rates of carbon corrosion are detected. This quickly destroys the carbon-supported PtRu electrocatalyst, the gas-diffusion substrate and the flow field plate in contact with the anode. However, adding a RuO2/1102 water electrolysis electrocatalyst into the anode layer can provide a hgh degree of cell reversal tolerance. This sustains the rate of water electrolysis for an additional 40 minutes. Rather than MEA failure after 20 minutes the MEA can be returned to normal PEMFC operation after experiencing 60 minutes under the conditions of cell reversal due to fuel starvation.

The approach of using a bilayer anode for air bleed operation and of including a water electroly- sis electrocatalyst for cell-reversal tolerance can be provided at little additional cost. With continued development of the concepts, the pgm levels used can probably be thtifted to below 0.05 mg cm ’.

Acknowledgement Part of this paper is based on a keynote lecture by T. R.

Ralph at the 17th North American Chemical Society Meeting held at the Westin Coast Harbour Castle Hotel, Toronto, Ontario, Canada, from the 3rd June to 8th June 2001.

References 19 D. L. Trimm and 2. I. Onsan, Catal. REV., 2001,43,

(1-2), 31 20 A. J. Appleby, G. Bronoel, M. Chemla and H. Eta,

in “Encyclopedia of Electrochemistry of the Elements”, ed. A. J. Bard, Dekker, New York, 1983, Vol. IX, Part A, p. 383

21 P. Stonehart and P. N. Ross, Catal. REV. Sci. Eng., 1975,12, 1

Platinrrm MetaLr Rev., 2002, 46, (3) 134

22 T. R Ralph, G. A. Hards, D. Thompsett and J. M. Gascoyne, Fuel Cell Seminar Extended Abstracts, San Diego, CA, 1994, p. 199

23 S. Gottesfeld and J. Pafford, J. Ekhbem. Soc., 1988, 135, (lo), 2651

24 S . Gottesfeld and T. A. Zawodzinski, Adv. Ekctmcbm. Sci. Ens, 1997,5,219

25 M. S. Wilson, C. R Derouin, J. A. Valerio and S. Gottesfeld, Proc. 28th Intersodety Energy Conversion Engmeering Conf., Atlanta, Georgia, 1993, Vol. 1, p. 1203

26 M. Iwase and S. Kawatsy in “Fiit Int. Symp. on Proton Conducting Membrane Fuel Cells I”, eds. A. R Landgrebe, S. Gottesfeld and G. Halpert, The

27 H. F. Oetjen, V. M. Schmidt, U. Stimming and F. Trila, J. Ekcfmbem. SOL, 1996,143, (12), 3838

28 J. McBreen and S. Mukerjee, J. Ekhcbem. Soc., 1995, 142, (lo), 3399

29 H. A. Gasteiger, N. M. Markovic and P. N. Ross,]. Pbys. Cbm., 1995, 99, (45), 16757

30 M. Watanabe and S. Motoo, J. Ekctivad Ch., 1975,60,275

31 H. A. Gasteiger, N. Markovic, P. N. Ross and E. J. Cairns, J. Pbys. Cbem., 1994,98, (2), 617

32 H. A. Gasteiger, N. M. Markovic and P. N. Ross,]. Pbys. Cbem., 1995, 99, (20), 8290

33 K. A. Friedrich, K-P. Geyzers, U. Linke, U. Stimming and J. Stumper, J. Ekctivand Ch., 1996, 402, (1-2), 123

34 T. E. Springer, T. A. Zawodzinski and S. Gottesfeld, in “Electrode Materials and Processes for Energy Conversion and Storage IV”, eds. J. McBreen, S. Mukerjee and S. Srinivasan, The Electrochemical Society, Pennington, NJ, 1997, PV 97-13, p. 15

35 S. Swathirajan, Fuel Cell Seminar Extended Abstracts, San Diego, CA, 1994, p. 204

36 R J. Bellows, E. P. Marucchi-Soos and D. T. Buckley, I d Eng. Cbem. Rcs., 1996,35, 1235

37 J. Sobkowski and A. Czerwinski, J. PLys. Ch., 1985, 89, (2), 365

38 C. U. Maier, A. Band and M. Specht, J. Ekctmcbm. Soc., 1994, 141, (l), LA

39 M. C. Arevalo, C. Gomis-Bas, F. Hahn, A. Arecalo and A. J. Arvia, Ekhcbm. Ah, 1994,39,793

40 T. R Ralph and M. P. Hogarth, Phiinam Me& Rcv., 2002, 46, (l), 3

41 T. Zawodzinski, T. Springer, J. Bauman, T. Rockward, F. Uribe and S. Gottesfeld, in “Second Int. Symp. on Proton Conducting Membrane Fuel Cells II”, eds. S. Gottesfeld, T. F. Fuller and G. Halpert, The Electrochemical SOC., Pennington, NJ,

42 A. J. Appleby and F. R. Foulkes, “Fuel Cell Handbook”, Krieger Publishmg Company, Malabar, Florida, 1993, Chapter 11

43 D. P. Wilkinson and D. Thompsett, in “Proc. of the Second Int. Symp. on New Materials for Fuel Cell and Modem Battery Systems”, eds. 0. Savadogo and P. R Roberge, Les Edition de 1’Ecole Polytechnique de Montreal, Montreal, Canada, 1997, p. 266

~ s o C , P e n n i o g t g N J , 1 9 9 5 , W 9 5 - 2 3 , p . 1 2

1998, PV 9 ~ 7 , p. 127

44 L. W. Niedrach and I. B. Weinstock, Ekhcbm. Tecbnol., 1965,3,270

45 L. W. Niedrach, D. W. McKee, J. Paynter and I. F. Danzig, Ekhcbm. Tecbnol., 1967, 5, 318 and iM., 41 9

46 D. W. McKee and A. J. Scarpellino, Ekhbem. Tecbnol., 1968,6, 101

47 S. J. Cooper, A. G. Gunner, G. Hoogers and D. Thompsett, op. dt., (Ref. 43), p. 286

48 P. N. Ross, K. Kinoshita, A. J. Scarpellino and P. Stonehart, J. E k h a n d Cbem., 1975,59,177

49 D. W. McKee and M. S. Pak, J. Ekhcbm. Soc., 1969,116,516

50 M. Hogarth and X. Glipa, “High Temperature Membranes for Solid Polymer Fuel Cells”, ETSU Ref: ETSU F/02/00189/REP (DTI/Pub URN 01/893), Johnson Matthey Technology Centre, 2001

51 L. W. Niedrich and H. I. Zeliger,]. Ekhbem. Soc., 1969,116,152

52 S. Mukerjee, S. J. Lee, E. A. Ticianelli, J. McBreen, B. N. Grgur, N. M. Markovic, P. N. Ross, J. R Giallombardo and E. S . De Castro, E k h b m . Solid SfufeLcft., 1999,2, (l), 12

53 B. N. Grgur, N. M. Markovic and P. N. Ross, J. Ekhbem. Soc., 1999,146, (5), 1613

54 B. N. Grgur, N. M. Markovic and P. N. Ross, op. cif., (Ref. 41), p. 176

55 T. S. Moss, N. M. Peachey, R C. Snow and R C. Dye, Znf. J. Hjdmgen E n m , 1998,23, (2), 99

56 D. P. Wilkinson, H. H. Voss, J. Dudley, G. J. Lamont and V. Basura, US. Pak?nt 5,482,680; 1996

57 T. Engel and G. Ed, in ‘The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis”, eds. D. A. King and D. P. Woodruff, Elsevier, Amsterdam, 1982, p. 73

58 P. J. Berlowitz, C. H. F. Peden and D. W. Goodman, J. P4y-r. Cbm., 1988, 92, (18), 5213

59 D. P. Wilkinson, H. H. Voss, K. B. Prater, G. A. Hards, T. R. Ralph and D. Thompsett, US. Put& 5,795,669; 1998

60 S. D. Knights, D. P. Wilkinson, S. A. Campbell, J. L. Taylor, J. M. Gascoyne and T. R Ralph, WorkiAppl. 01/15247; 2001

The Authors Tom Ralph is Product Development Manager at the Johnson Matthey Technology Centre, responsible for MEA design. He has been working with PEMFCs since 1991 and with developing all aspects of the MEA: electrocatalyst, electrocatalyst layer, gas diffusion substrate and solid polymer membrane, and in integrating MEAs into customer hardware.

Martin Hogarth is a Senior Scientist at the Johnson Matthey Technology Centre and has worked in the area of DMFCs since 1992. His main interests are the development of new catalyst materials and high-performance MEAs for DMFCs. More recently his interests have expanded into novel high-temperature and methanol-impermeable membranes for the PEMFC and DMFC, respectively.

Phnnntn Metab Rey., 2002,46, (3) 135