The influence of hydrogen- and cation-underpotential deposition on oxide-mediated Pt dissolution in proton-exchange membrane fuel cells

Electrochimica Acta 56 (2011) 8387–8393

Contents lists available at ScienceDirect

Electrochimica Acta journal homepage: www.elsevier.com/locate/electacta

The influence of hydrogen- and cation-underpotential deposition on oxide-mediated Pt dissolution in proton-exchange membrane fuel cells Seokkoo Kim, Jeremy P. Meyers ∗ Texas Materials Institute and The Center for Electrochemistry, The University of Texas at Austin, University Station, C2200 Austin, TX 78712-0292, United States

a r t i c l e

i n f o

Article history: Received 4 April 2011 Received in revised form 5 July 2011 Accepted 8 July 2011 Available online 19 July 2011 Keywords: Pt-oxide Pt dissolution ICP-MS EQCN RRDE

a b s t r a c t In order to fully understand the influence of a lower potential limit on platinum dissolution and the likely mechanism for mass and surface-area loss under potential cycling conditions, the dissolution of a Pt catalyst in a N2 -saturated 0.5 M H2 SO4 solution was examined using an electrochemical quartz nanobalance (EQCN) flow cell, a rotating ring-disk electrode (RRDE) and inductively coupled plasma mass spectroscopy (ICP-MS). Due to the observation that cycling to a lower potential limit, which coincides with the hydrogen under-potential (HUPD ) region, results in a decrease in the dissolution rate, cations capable of interfering with the hydrogen UPD process (Zn2+ , Li+ , Na+ , K+ , and Cd2+ ) were introduced to the solution. Larger rates of mass loss were observed in the presence of these cations during the cycling process in the UPD region, despite apparently negligible effects on the behavior with more positive lower potential limits or on oxide formation and stripping. It was found that the quantity of soluble Pt species produced during the electrochemical reduction of PtO2 was proportional to the charge associated with oxide stripping at the disk electrode during the RRDE experiment. © 2011 Elsevier Ltd. All rights reserved.

1. Introduction Proton-exchange membrane fuel cells (PEFCs) have attracted attention as energy conversion devices for electric vehicles, smallscale stationary power generators, and portable electronics devices [1]. One of the key challenges to PEFC commercialization is the gradual loss of cell performance over time, which limits the lifetime of the fuel cell at the required power and efficiency. In particular, the deterioration of the PEFC’s performance by degradation of the electrocatalysts is a major problem. The catalyst degradation in the cathode has been studied extensively and is generally thought to be related to three major mechanisms: Ostwald ripening (Pt dissolution and redeposition onto lower surface energy sites) [2,3], coalescence of Pt nanoparticles via migration and surface diffusion on carbon supports [2,4,5], and carbon corrosion-induced Pt detachment or agglomeration [6]. Although the kinetic properties of electrodes are frequently measured on pristine catalyst surfaces, PEFC cathode electrocatalysts are commonly covered by oxide films that may have a profound influence on the chemistry, stability and kinetics of the oxygen reduction reaction in the potential range where the

∗ Corresponding author. Tel.: +1 512 964 4288; fax: +1 512 471 7681. E-mail addresses: [email protected], [email protected] (J.P. Meyers). 0013-4686/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.electacta.2011.07.031

electrodes normally operate. While experimentation and discussion continues, a consensus has emerged that suggests a strong physisorption of water on the surface of platinum in the range of 0.27–0.85 V, followed by the formation of a PtO layer in the range of 0.85–1.15 V, and a formation of a PtO2 layer at higher potentials [7–10]. Mitsushima [11,12] and Guilminot et al. [13] reported the formation of PtO2 at potentials greater than 1.5 V vs. RHE and noted that PtO2 is formed by further oxidation of an existing thin layer of PtO, resulting in a two-layer film. Sun et al. [14] also proposed that a two-layer oxide structure exists with the inner layer (PtO) in the potentials region of 1.0 V < EH < 1.35 V vs. RHE and an outer layer (containing PtO2 ) at potential greater than EH > 1.35 V vs. RHE. Furthermore, it has been noted that potential cycling accelerates the rate of surface area loss during PEFC operation; in fact, potential cycling has been standardized as an accelerated testing protocol for membrane-electrode assembly durability by the Department of Energy [15]. There are several experimental studies that show that cycling the potential of platinum electrodes to a range in which oxides are formed and subsequently stripped results in a loss of electrochemically active surface area (ECSA) and is more rapid than holding the potential at either the upper or lower potential limit [16–19]. Therefore, it has been proposed that the presence of oxides can, in effect, passivate the surface, and protect platinum from the dissolution rates that would be expected at more positive potentials [3,20]. If platinum dissolution occurs through an electrochemical

8388

S. Kim, J.P. Meyers / Electrochimica Acta 56 (2011) 8387–8393

mechanism, such as the electrochemical oxidation of Pt, as shown in Eq. (1): Pt → Ptz+ + ze−

(1)

one expects a higher equilibrium concentration of the mobile Pt ions in solution at more positive potentials. Wang et al. [3] conducted a series of experiments in which platinum was held at fixed potentials until an equilibrium concentration in the solution was reached, and they observed that the platinum concentration in the solution increases with increasing electrode potential up to approximately 1.1 V. In addition, a marked decrease in solution-phase Pt concentration at more positive potentials was observed, which suggest a protection of the surface from the direct electrochemical dissolution mechanism at potentials where the surface is fully covered by the oxides. While the literature suggests that the presence of oxides can help protect the platinum surface from anodic dissolution at higher potentials, there are unanswered questions about what role the oxide film plays and the mechanism of Pt dissolution; specifically, whether the oxide-mediated dissolution occurs primarily during the oxidation or reduction of the film, and which mobile platinum species are formed in the dissolution process. In this study, a variety of electrochemical techniques were used in combination with ICP-MS to clarify the surface state of Pt during electrochemical experiments to understand the Pt dissolution rate and to the likely mechanism(s) of Pt dissolution as well as to confirm that all methods give equivalent results under identical potential cycling or potential sweep conditions. In particular, we examined the influence of cycling potential limits in an attempt to corroborate the measurements made by Mitsushima et al. [11], and we showed what we believe to be very good agreement between the EQCN and ICP-MS experiments. This agreement allowed us to assess mass loss over a relatively small number of cycles and revealed that mass loss of the sample is matched by dissolution of platinum ions from the electrode into the solution. When confronted with a rather surprising result that cycling to lower potentials somehow protects the surface in the subsequent cycle, we examined the role of cation adsorption at these lower potentials. It has been established that Zn under-potential deposition (UPD) changes the nature of the HUPD region by surface reordering and blocking in the UPD potential region [21,22]. Thus, we sought to examine whether this surface reordering would have an effect on the results reported by Mitsushima et al. that cycling in the HUPD region resulted in less Pt loss than when the lower potential limit was maintained above the HUPD region. We discuss these results to clarify the deterioration mechanism and the influence of the cation UPD on the Pt catalysts in acidic media. To more precisely detect the nature of the mobile species involved in the dissolution process and to determine which potential conditions result in the detachment of the mobile species from the platinum surface, we performed a series of experiments on the formation and stripping of Pt oxide by using RRDE experiments in combination with ICP-MS analysis. 2. Experimental 2.1. Preparation of a thin film Pt catalyst layer on the EQCN and RRDE working electrode The catalyst layer of the EQCN and RRDE working electrode was fabricated by dispersing 32 mg of Pt-black catalyst (E-Tek) and 182.4 mg of a 5 wt% Nafion solution in lower aliphatic alcohols and water (Aldrich) in a mixture of 2 ml of isopropyl alcohol and 4.8 ml of distilled pure water. The mixture was ultrasonicated until a dark homogeneous dispersion was formed (∼1 h). Subsequently, 8 ␮l of the dispersed ink solution was placed on a glassy carbon (GC) disk

electrode for RRDE experiments, or on an AT-cut Au quartz crystal for EQCN experiments, to produce a catalyst loading of 35,800 ng. The loadings were 188.5 ␮g cm−2 for the EQCN experiments and 144.6 ␮g/cm2 for the RRDE experiments, due to a difference in the size of the electrodes. The electrodes were subsequently dried in an atmosphere of ultra-pure N2 gas (99.999%).

2.2. EQCN flow cell system and electrochemical measurements The EQCN flow cell system is composed of a quartz crystal analyzer (Maxtek RQCM, INFICON Inc.), a gear pump and a potentiostat (CHI760C, CH Instruments Inc.) controlled by a computer through a general-purpose interface bus. During all measurements, the mass change resulting from the oxidation and reduction of Pt was measured by EQCN. The EQCN cell consists of a working electrode (7.995 MHz AT-cut Au quartz crystal, 0.196 cm2 of geometric area, ICM Inc.), a hydrogen reference electrode (HydroFlex, Gaskatel GmbH Inc.), and a Pt wire counter electrode. All the potential values were measured and reported relative to the reversible hydrogen reference electrode (RHE). The electrolyte solution was delivered to the EQCN cell from a reservoir at a volumetric rate of 122.5 mm3 /s using a gear pump (Cole Palmer) with a retention time of 0.27 s. The 0.5 M H2 SO4 electrolyte solution was de-aerated by bubbling high purity N2 gas for 1 h prior to beginning the electrochemical test, and the solution was continuously purged with high-purity nitrogen during the experiment. Prior to the measurements, the working electrode was pretreated by a potential cycle between 0.05 V and 1.0 V vs. RHE at a rate of 100 mV/s for 20 cycles. The triangle wave voltammetry (TW) and square wave voltammetry (SW) of the EQCN working electrode were performed at a scan rate of 500 mV/s and a frequency of 1 Hz for 500 cycles with respect to the lower (EL ) and upper (EH ) limits on the potential. Before each cycling test, an LSV (linear sweep voltammogram) was conducted at a scan rate of 100 mV/s immediately before the experiment to measure the electrochemically active area (ECSA). For the experiments that examined the effects of other dissolved cations, a 0.5 M H2 SO4 electrolyte solution containing 5 mM of fully dissolved cations was prepared using analytical grade Aldrich reagents, such as ZnSO4 , Li2 SO4 , Na2 SO4 , K2 SO4 , and CdSO4 .

2.3. Quartz crystal calibration The sensitivity factor for the EQCN was determined using a method described by Vatankhah et al. [23]. This method enables facile calibration by measuring the silver electrodeposition–electrodissolution processes using cyclic voltammetry. All the electrochemical experiments were conducted in a 0.5 M H2 SO4 solution containing 1.5 mM of analytical grade Ag2 SO4 (Fisher Scientific) at T ∼ = 298 K. The cyclic voltammetry profiles for silver deposition and stripping on a thin film Pt catalyst layer were recorded in the potential range 0.425–0.775 V vs. RHE at a scan rate of 1 mV/s. The calibration constant of the EQCN determined by this method was 10.40 ± 0.53 ng Hz−1 cm−2 , which is larger than the theoretical value of 6.957 ng Hz−1 cm−2 , due to the thickness (∼100 ␮m) and porosity of the Pt catalyst layer on the EQCN working electrode crystal. The presence of Nafion in the electrode structure might preclude a completely rigid structure that we assume during calibration and could obscure the interpretation of the results within a single cycle; however, good agreement between the EQCN and ICP analyses (discussed below) suggests that mass loss over multiple cycles can be quantified using these techniques.

S. Kim, J.P. Meyers / Electrochimica Acta 56 (2011) 8387–8393 70

The reaction mechanism of Pt dissolution was investigated using an RRDE. A GC disk – Au ring electrode (E7R9, PINE Inc.) was used with a geometric surface area of 0.2475 cm2 and 0.1866 cm2 for the disk and ring electrodes, respectively, and a collection efficiency of N = 0.37 (N = |iring |/|idisk |). When turned on, the electrode was rotated at 1600 rpm, using a speed controller (modulated speed rotator (MSR), PINE Inc.). Approximately 100 ml of a 0.5 M H2 SO4 electrolyte solution was de-aerated with vigorous bubbling of high-purity N2 gas for 1 h prior to beginning the electrochemical test and was continuously purged during the experiment. Before performing any measurements, the working electrode was pretreated by potential cycling between 0.05 V and 1.0 V vs. RHE, at a rate of 100 mV/s for 20 cycles, while bubbling N2 in the solution. The hydrogen electrode (HydroFlex, Gaskatel GmbH Inc.) and the Pt wire electrode were used as the reference and counter electrodes, respectively.

60

2.5. ICP-MS measurement At the end of each electrochemical experiment, the solutions were analyzed by ICP-MS (Optimass 8000, GBC Scientific Inc.) to determine the amount of Pt dissolved in the solution. The detected Pt ion concentration was in the range of 1–100 ppb, depending upon the nature of the potential cycling experiment. All glassware was cleaned by dipping in 4 M of nitric acid for at least 12 h, and rinsing with DI water prior to use. The ICP-MS system had a detection limit of 0.1 ng/l for Pt ions. 3. Results and discussion 3.1. Pt dissolution rate under potential cycling in terms of potential cycling limits We begin our discussion of experimental results with a summary of how cycling conditions affect the rate of platinum mass loss from the electrode. As mentioned in the introduction, Wang et al. [3] showed that the concentration of dissolved Pt increases monotonically from 0.65 to 1.1 V, and then decreases at potentials greater than 1.1 V due to the presence of protective PtO and PtO2 layers. In their experiments, they allowed the surface to remain at a specified potential for enough time to reach what is assumed to be true equilibrium in the solution. However, under transient conditions such as those introduced via potential cycling, the dissolution rate has been shown to be accelerated, both with respect to the scan rate and the potential limits [11]. Indeed, under transient conditions, the platinum might not be in an equilibrium state dictated by local potential conditions. As such, dissolution might proceed well beyond the point predicted by thermodynamic considerations alone, as the solution is equilibrating with a surface that is not at a thermodynamic equilibrium state as designated by the potential. Fig. 1 shows the Pt dissolution rate as a function of the lower potential limit for cycling, EL , wherein the upper limit, EH , is fixed at 1.6 V vs. RHE both for triangular-wave cycling (TW) with a scan rate of 500 mV/s, and for square-wave voltammetry (SW) at 1 Hz, over the course of 500 cycles. The dissolution rate obtained by both EQCN and ICP analyses is 45–55 ng Hz−1 cm−2 , with good agreement between the two methods. These values are larger than the results presented in other papers [11,24] that used a Pt planar working electrode electrodeposited onto a EQCN quartz crystal in a stagnant electrochemical cell. The difference can be explained by the high roughness factor of the Pt catalyst layer and the fresh electrolyte feed from the reservoir; the enhanced surface area and

EQCN TW EQCN SW ICP TW

50 -1

(ngcm cycle )

40

-2

Catalyst dissolution rate

2.4. Rotating ring-disk electrode (RRDE) measurement

8389

30 20 10 0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

EL (V vs. RHE) Fig. 1. Pt dissolution rate as a function of EL with EH = 1.6 V vs. RHE at 298 K in 0.5 M H2 SO4 for 500 mV/s triangular-wave cycling (TW) and 1 Hz square-wave cycling (SW) for 500 cycles.

the inability of the solution to become saturated with mobile ionic platinum species lead to a higher rate of dissolution. The trend in the Pt dissolution rate confirms that cycling between two potentials that are both sufficiently high as to maintain oxide formation on the surface over the course of the entire cycle results in a rather modest effective dissolution rate. Consistent with other data in the literature, when the lower potential limit is low enough for the oxide to become stripped, the regeneration of a bare platinum surface occurs, which is more prone to subsequent dissolution during the intervals when the potential is raised to higher values, but before the protective oxide layer is formed. Previous analysis suggests that platinum solubility would increase with increasing potential but that the oxide equilibrates at a much lower concentration of dissolved platinum in solution than the platinum metal does at higher potentials [20]. As such, one would expect the effect of the lower potential limit on the overall dissolution to be quite modest compared to the influence of whether the protective layer is repeatedly removed before exposing the platinum to higher potentials. Nevertheless, cycling at a lower potential limit in the range of HUPD appears to lower the cyclic dissolution rate significantly below that which is found when the lower limit is maintained in the oxide- and HUPD -free region. This observation will be discussed in greater detail in Section 3.3. Furthermore, a broad maximum of loss with a lower potential limit in the range of 0.2–0.6 V vs. RHE was observed by both EQCN measurement and ex situ ICP-MS analysis. This is a somewhat surprising result, although it does corroborate measurements made by Mitsushima et al. [11]. As the solubility of platinum is predicted to be quite low at the low potential values where HUPD occurs, one would not expect the adsorbed hydrogen to play as major of a role in protecting the surface as does the oxide, because the potentials at which oxides are present are much higher and are therefore expected to prevent the surface from equilibrating with a much higher concentration of dissolved mobile platinum species. We note that the trend works for both the triangular wave (TW) and the square wave cycling (SW), indicating that the total time spent at lower potentials or traversing the oxide-free region is not the controlling factor. Fig. 2 shows Pt catalyst mass loss resulting from the triangularwave potential cycling as a function of the upper potential limit EH , wherein the EL is fixed at 0.5 V vs. RHE, for scan rates of 500 mV/s. We note that an increase in the rotation speed results in an increase in the mass loss rate of the Pt catalyst at the upper potential limits above 1.2 V vs. RHE, whereas negligible differences are observed

8390

S. Kim, J.P. Meyers / Electrochimica Acta 56 (2011) 8387–8393 40

(a)

rpm0 rpm1600

35

Mass loss (wt%)

30 25 20 15 10 5

(b)

0 0.6

0.8

1.0

1.2

1.4

1.6

1.8

EH (V vs. RHE) Fig. 2. Pt mass loss (wt%) analyzed by ICP-MS with a changing upper limit potential (EH ), with a lower limit potential fixed at 0.5 V vs. RHE, for scan rates of 500 mV/s.

when the upper potential limit is kept below 1.2 V vs. RHE. It should be noted that when the upper potential limit is sufficiently high to form PtO2 on the surface of the electrode (generally considered to be at potentials more positive than 1.5 V vs. RHE), there is a mechanism by which the oxide can be reduced electrochemically to form a mobile species, namely: PtO2 + 4H+ + 2e− → Pt2+ + 2H2 O

E ◦ = 0.84 − 0.12 pH

− 0.03 log [Pt2+ ]

(2)

This mechanism might contribute significantly to mass loss during the cycling experiment where the upper potential limit is sufficiently high to generate PtO2 , but it does not likely play a role at upper potential limits below the formation of PtO2 . While not conclusive, the difference in dissolution rate at a more positive upper potential limit suggests the generation of mobile species that might redeposit in the case of a stagnant solution but which are removed from the point of generation by convection at higher rotational speeds. From these experiments, we observe the effect of potential cycling on the overall dissolution rate and obtain consistent results using different detection techniques. Again, there is evidence to suggest the protective nature of the oxide film; however, it is also likely that cycling the potential to sufficiently positive values wherein PtO2 is generated results in the direct dissolution of the oxide film, thereby contributing to Pt loss, though this mechanism is unlikely in the range of EH < 1.4 V. 3.2. Pt dissolution during reduction of Pt oxide We also wanted to investigate whether we could determine the nature of the species that are liberated and contribute to the mass loss. As such, we conducted RRDE experiments to detect the nature of species that are liberated from the platinum electrode upon cycling. Fig. 3 shows the result from the RRDE experiments in a 0.5 M H2 SO4 solution at a scan rate of 50 mV/s and an electrode rotation speed of 3000 rpm. Fig. 3(a) shows the current generated on the disk electrode as a function of potential, Idisk –Edisk , while Fig. 3(b) reveals the current at the ring electrode, Iring –Edisk , when the potential of the ring electrode is fixed at 1.4 V vs. RHE. The main effect of increasing the upper potential limit of the disk electrode is an increase in the oxidation charge at the ring electrode during the cathodic sweep of Edisk and a shift in the oxidation peak potential of the ring electrode toward less positive values, in accor-

Fig. 3. RRDE measurement in a N2 -saturated 0.5 M H2 SO4 solution at a scan rate of 50 mV/s, ω = 3000 rpm, (a) Idisk –Edisk , (b) Iring –Edisk (Ering = 1.4 V vs. RHE).

dance with the shift in the reduction peak potential of the disk electrode in Fig. 3(a). These results are similar to the interpretation of the experiments by Johnson et al. [25], which suggest that waves similar to those shown in Fig. 3(b) result from the electrochemical oxidation of Pt2+ at the ring electrode, as this species is generated at the disk electrode by the electrochemical reduction reaction described in Eq. (2). Based upon these experiments, oxidizable Pt species in solution appear to be generated at the disk only during the reduction of the oxide peak. We also note that experiments in which the ring was held at very low potentials did not reveal any significant current signatures as the disk potential was cycled; that is, no reducible forms of Pt were observed under these conditions. The results in Fig. 3(b) suggest that the cathodic dissolution of PtO2 is enhanced by increasing the upper potential limit of the disk electrode to 1.8 V vs. RHE, likely due to the increase in the concentration of PtO2 [14]. Fig. 4 shows the Pt mass loss based on the calculation of oxidation charge at the ring electrode (Er = 1.4 V vs. RHE) during cathodic scans in a N2 -saturated 0.5 M H2 SO4 solution. The ring oxidation tests were conducted at discrete cycles (cycle numbers 0, 50, 100, 200, 300, 400, 500) during the Pt dissolution test using the same reactor. The potentiostat controlling the ring potential was only active during those particular cycles. Therefore, the ring should have been electrochemically inert during other cycles. Based upon the efficiency of the RRDE used in the experiment (N = 0.37) and Faraday’s Law (Q = n × F × m/Mpt ), the amount of soluble Pt species at the disk electrode generated during cathodic scan can be calculated. For the purposes of this calculation, we assumed a mobile Pt2+ species, resulting in n = 2 for the oxidation of Pt to Pt4+ at the ring electrode. The shaded area below the line in Fig. 4 represents the total amount of soluble Pt species generated during the cathodic scan of the potential cycling test, resulting in an integrated

S. Kim, J.P. Meyers / Electrochimica Acta 56 (2011) 8387–8393

8391

experiment. In contrast, for the mass detected at the ring electrode, it is an estimate of the total mass lost. We note that even if our assumption of Pt2+ and n = 2 is incorrect in the prediction of mass loss from the disk, we cannot account for the total amount of mass lost from the disk by any reasonable combination of Pt dissolution and subsequent oxidation reactions. We believe that this discrepancy suggests either a highly irreversible electrochemical reaction or a non-electrochemical/physical degradation of the oxide films that results in a mass loss from the disk that is not detected at the ring electrode. Additional work is underway to corroborate this observation via scanning electrochemical microscopy (SECM) and to develop a mechanistic model to describe the mass loss based on these results. 3.3. Electrochemical Pt surface shielding using cation (Zn2+ ) UPD (underpotential deposition)

Fig. 4. Pt mass loss (ng) calculated by the integrated oxidation charge at the ring electrode during the cathodic scan (50 mV/s, bubbling N2 through solution). The collection efficiency of the RRDE is 0.37, and the assumption of the charge associated with mobile Pt species is n = 2.

value of 126.7 ng. There is a possibility that there were some cycles that released a much higher amount of electroactive Pt species into the solution that this test did not reveal, but we found no evidence of anything other than the fairly unremarkable trend in detectable Pt species shown in Fig. 4. The total amount of Pt detected at the ring corresponds to just 1.76% of total Pt mass loss during the potential cycling test of 0.5–1.6 V vs. RHE, using 500 cycles, at a rate of 500 mV/s in a N2 saturated 0.5 M H2 SO4 solution. Fig. 5 demonstrates the relative values of mass loss from the EQCN experiments, the total Pt placed into solution via ICP-MS (circle) after EQCN flow-cell testing, the ICP-MS (square) after RRDE testing, and the integrated Pt mass loss from Fig. 4. The change in mass from the EQCN experiments corresponds to a ∼12% mass loss over the course of 500 cycles and corresponds very well with the independent measurement of ICP-MS testing of the solution after those same 500 cycles. We note that there is a slightly greater mass loss, as measured by ICP-MS in the RRDE experiments over the course of 500 cycles, of roughly 20%, corresponding to 80% mass remaining. For comparison, the integrated mass detected at the ring electrode for RRDE is shown on an enlarged scale. Note that for the EQCN and ICP-MS results, the data shown are for the initial mass of Pt remaining as a function of cycle number or at the end of the

34000

EQCN TW ICP TW (EQCM)

Pt mass (ng)

32000 88% 30000

ICP TW (RRDE) 80% 28000

150 100 50 0

Integrated Pt mass at Ring

0

100

200

300

400

500

Cycle number Fig. 5. Total Pt mass change vs. cycle for in situ EQCN, ICP-MS (EQCN) after EQCN test, ICP-MS (RRDE) after RRDE test and integrated Pt mass loss (ng) from Fig. 4. EQCN and ICP-MS experiments reveal original electrode mass remaining; integrated Pt mass loss reveals mass lost to the solution.

As mentioned in Section 3.1, cycling at a lower potential limit in the range of the hydrogen underpotential region appears to lower the cyclic dissolution rate significantly below that which results when the lower limit is maintained in the oxide- and HUPD -free region. In the absence of any information about the specific interaction between platinum and the electrolyte under these conditions, we surmised that the hydrogen underpotential deposition process somehow influences the nature of the platinum surface and the manner in which it subsequently behaves. We noted in the introduction that the Zn UPD changes the nature of the HUPD region by surface reordering and blocking [21,22]; while we do not anticipate that cations such as Zn2+ will likely be present in significant quantities during fuel cell operation, we considered that the opportunity to interfere with hydrogen underpotential deposition, while still sweeping to the same potentials, might help illuminate the role this process plays on electrode degradation. In these experiments, we used underpotential deposition (UPD) of Zn2+ on a Pt catalyst in a 0.5 M H2 SO4 solution to determine the influence of the hydrogen underpotential region on the overall dissolution rate. Underpotential deposition produces a surface ad-layer of less than a monolayer of a metal on a foreign metal surface and has been studied for electrochemical process, such as electroplating, electrocatalysis, and electrocrystallization [26–28]. Although the standard reduction potential of Zn2+ /Zn is −0.76 V vs. RHE, the UPD of Zn2+ ions on Pt surface in an acidic solution, such as 0.5 M H2 SO4 (pH = 1), is approximately +0.34 V vs. RHE. This 1.1 V potential shift of Zn UPD from standard reduction potential is the work function difference between Pt and Zn [29]. For the 5 mM ZnSO4 concentration used in our experiments, the Nernstian potential for the UPD of Zn2+ ions at room temperature is approximately +0.27 V vs. RHE. These UPD properties are, of course, strongly influenced by the surface structure of the electrochemically oriented Pt electrode [21]. Fig. 6 shows a cyclic voltammogram and the resulting EQCN mass change of the electrode versus the electrode potential in a 0.5 M H2 SO4 solution, using a scan rate of 100 mV/s, both with and without the addition of 5 mM ZnSO4 to the electrolyte. At potentials lower than 0.4 V vs. RHE without Zn ion additives, the mass of the electrode decreases when cycling into the HUPD region, relative to the mass in the capacitive region of the CV. In the presence of Zn2+ , the electrode mass increases relative to the baseline of the capacitive region of the CV. It is worth noting that the addition of Zn ion to the 0.5 M H2 SO4 acid solution does not appear to affect the processes in the double-layer region or the oxide formation/reduction on the surface of Pt. Fig. 7 shows the Pt dissolution rate of triangular-wave and square-wave cycling both with and without Zn ion additives. Compared to Fig. 1, in the presence of Zn2+ , the dissolution rate at EL = 0.05 V vs. RHE increases to roughly the same rate as the

8392

S. Kim, J.P. Meyers / Electrochimica Acta 56 (2011) 8387–8393 -3

C

0.0

-5.0x10

-4

-1.0x10

-3

20 -1.5x10

-3

50

40

40 30 30 20 20 10

10

0

0

(b)

No Addiv

Zn

15

2

Delta mass(ng/Hzcm )

Na

K

Cd

Fig. 8. Pt dissolution rate (bar) and ECA loss (dot) at EL = 0.05 V and EH = 1.6 V vs. RHE with various cation additives. The mass loss is shown as loss per cycle; the ECA loss is cumulative over the course of 500 cycles.

5 0 -5 2+

w/o Zn 2+ w/ Zn

-10

-20 0.0

0.2

0.4

0.6

0.8

1.0

1.2

Potential(V vs. RHE) Fig. 6. Cyclic voltammogram and EQCN mass change versus electrode potential in a 0.5 M H2 SO4 solution with and without 5 mM ZnSO4 . Scan rate is 100 mV/s.

dissolution rate observed when the lower potential limit is maintained above the HUPD range for both the triangular-wave and square-wave experiments. EQCN measurements from cycling experiments (EL = 0.05 V with EH = 1.6 V vs. RHE) are qualitatively consistent with the results of ICP analysis (28.96 ng cycle−1 cm−2 without Zn additive, versus 62.43 ng cycle−1 cm−2 with Zn additive). Thus, the nature of what happens in the UPD region (region A in Fig. 6) is presumed to have a strong influence on Pt dissolution in cycling experiments for various lower potential limits when the upper limit is held fixed at 1.6 V vs. RHE, even though the amount of dissolution observed when the electrode is held in this potential

window without cycling has been shown by others to be very low [3]. In addition, we conducted a series of experiments by changing the type of cation used in the electrolyte in the acidic solution, while maintaining all other conditions constant, and the results are shown in Fig. 8. The addition of alkaline metal ions such as Li2 SO4 , Na2 SO4 , K2 SO4 that have more negative UPD potentials than the HUPD have no noticeable effects on the Pt dissolution rate or the ECA loss compared to the sample with no additives. This finding means that these ions present in the outer Helmholtz plane do not appear to influence the Pt surface adsorption/desorption. On the other hand, the addition of CdSO4 that has an UPD at approximately 0.71 V vs. RHE (within the H2 O adsorption/desorption region) results in an increased Pt dissolution rate and ECA loss, although not as much as is observed with ZnSO4 . We further investigated the influence of Zn additives in a series of experiments shown in Fig. 9. This figure shows the Pt catalyst dissolution resulting from potential cycling for several different sets of conditions. For curves indicated by unfilled markers, the solution is free of Zn additives, and the lower potential limit is maintained at 0.05 V vs. RHE, well within both the HUPD and ZnUPD potential range; for the set of curves with filled triangles oriented

35

70

EQCM TW w/o Zn2+ EQCM TW w/ Zn2+ EQCM SW w/o Zn2+ EQCM SW w/ Zn2+

60

30

Mass loss (%)

-1

-2

40 30

EL=0.05V,

0rpm

EL= 0.5V,

0rpm

2+

Zn , EL=0.05V 1600 rpm

EL=0.05V, 1600rpm

25

50

(ngcm cycle )

Li

Cation additives (5mM)

10

-15

Catalyst dissolution rate

ECA loss (%)

-4

B

-1

5.0x10

A

w/o Zn 2+ w/ Zn

-2

-3

Current(A)

1.0x10

50

60 2+

(a)

Dissolution rate (ngcm cycle )

1.5x10

EL= 0.5V, 1600rpm

20 15 2+

Zn , E =0.05V 0 rpm

10

20

L

2+

Zn , E L=0.05V

5

10

0

0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

EL (V vs. RHE) Fig. 7. Pt dissolution rate as a function of EL with EH = 1.6 V vs. RHE at 298 K in a 0.5 M H2 SO4 solution for 500 mV/s triangular wave (TW) and 1 Hz square-wave (SW) over 500 cycles, with and without Zn ion additives.

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

EH (V vs. RHE) Fig. 9. ICP results of the RRDE test at 298 K in a N2 saturated 0.5 M H2 SO4 solution for 500 cycles of 500 mV/s scan rate in terms of potential limits (EL , EH ) and rotation speed.

S. Kim, J.P. Meyers / Electrochimica Acta 56 (2011) 8387–8393

upwards, the solution is free of Zn additives and the lower potential limit is maintained at 0.5 V, well above the UPD regions. All experiments were conducted at 500 mV/s, with two different rotation speeds: stagnant conditions (no rotation = circular markers) and a rotation speed of 1600 rpm (triangular markers). The data for solutions with Zn added to the acidic solution are shown for a handful of cases by points marked with arrows on the graph and indicated by filled, downward-pointing triangular markers. We consider the influence of Zn additives in three distinct conditions: at the EH = 1.2 V, EH = 1.6 V under stagnant conditions, and EH = 1.6 V at a rotation speed of 1600 rpm. We observe that the Pt catalyst mass loss increases with increasing rotating speed when the upper potential limit is set above 1.2 V vs. RHE, while we observe less separation as a function of rotation speed when the upper potential limit is set below 1.2 V. Moreover, the influence of Zn additives on Pt dissolution is smaller at less positive upper potential limits, as can be seen by the data points plotted at EH = 1.2 V. Furthermore, in cases with high rotation speeds, the addition of Zn to the acidic solution does not enhance the dissolution rate for EL = 0.05 V nearly as much as it does under stagnant conditions, or with the comparatively low flow rates of the EQCN flow cell, as shown in Fig. 7. While hardly conclusive, these data are consistent with a scenario in which the hydrogen underpotential region prepares the platinum surface to be more conducive to redeposition or recapture of the mobile species when mobile species are generated during cycling. The fact that increasing the rotation speed increases the overall rate of dissolution at higher potentials suggests that convection can carry mobile species away from the surface before they can redeposit. If the effect of Zn was invariant with rotation speed, we might infer that the Zn interference with hydrogen underpotential deposition changed the magnitude or the nature of dissolution; the fact that the system is more sensitive to Zn additives under stagnant conditions than under conditions of higher rotation speeds suggests changes in solution-phase saturation or redeposition rates relative to convection. We note that the relatively small changes in the rate of mass loss with Zn additives at potentials where PtO2 has yet to be formed makes it more likely that the UPD region has a greater effect on the reaction in Eq. (2) than the passivation or protection in Eq. (1), although both mechanisms do appear to be at play in the system. From the discussion above, it is clear that the influence of Zn UPD and the hydrogen underpotential deposition processes are not sufficiently explored in the literature, especially the relationship between the UPD and Pt dissolution. Moreover, there are little data about the surface structure of the monolayer with regard to the number of active sites occupied by each ad-atom, as well as to the maximum coverage of the surface. To the best of our knowledge, this use of an incorporated Zn UPD in explaining the Pt dissolution rate is novel and has not been reported previously. While we do not anticipate that addition of cations will be used to tailor stability in actual fuel cell operation, we believe that the results suggest future work to clarify how modifications to the surface in the HUPD region are sustained even as the metal is cycled to higher potentials. 4. Conclusion The dissolution behavior of Pt catalyst was investigated by EQCN flow cell, RRDE and ICP-MS techniques. We found that when the upper potential limit was maintained at 1.6 V vs. RHE, the dissolution rate from both the triangular-wave and square-wave potential cycling revealed a broad maximum as a function of the lower potential cycling limit; cyclic tests in which the lower potential limit

8393

remained either entirely within a region of oxide stability, or in a region where HUPD occurs, resulted in a significantly lower dissolution rate. The results from the EQCN experiment are quite consistent with the results obtained by ICP-MS. We also found a significant effect of the metal UPD on the dissolution rate from cycling under conditions under which the lower potential limit was less than 0.2 V. By adding 5 mM of Zn2+ ion, the Pt dissolution rate under conditions when EL was less than 0.2 V increased up to the same relatively largely rates when EL was set to 0.5–0.6 V. These results suggest that Zn UPD changed the properties of the Pt surface that influence dissolution or hinder soluble Pt species from re-depositing on Pt surface. However, the nature of what occurs on the surface at lower potentials and its effect on oxide formation and stripping are not yet fully understood. From the RRDE test, it was shown that the rotation speed plays a key role in accelerating the Pt dissolution over EH = 1.2 V vs. RHE. The soluble Pt species can be detected at the ring electrode (Er = 1.4 V vs. RHE) during cathodic scans of disk electrode at potentials more positive than EH = 1.3 V vs. RHE. The amount of Pt(II) detected at the ring electrode during the cathodic scan is estimated to be approximately 1.7% of total amount of dissolved Pt during cycling, suggesting the formation of other dissolved or detached species that are not detectable electrochemically. Acknowledgement Financial support from the Welch Foundation (Grant F-1690) is gratefully acknowledged. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

C.K. Dyer, J. Power Sources 106 (2002) 31. P. Bindra, S.J. Clouser, E. Yeager, J. Electrochem. Soc. 126 (1979) 1631. X.P. Wang, R. Kumar, D.J. Myers, Electrochem. Solid-State Lett. 9 (2006) A225. M.S. Wilson, F.H. Garzon, K.E. Sickafus, S. Gottesfeld, J. Electrochem. Soc. 140 (1993) 2872. Y. Shao-Horn, W.C. Sheng, S. Chen, P.J. Ferreira, E.F. Holby, D. Morgan, Top. Catal. 46 (2007) 285. D.A. Stevens, M.T. Hicks, G.M. Haugen, J.R. Dahn, J. Electrochem. Soc. 152 (2005) A2309. B.E. Conway, Prog. Surf. Sci. 49 (1995) 331. B.E. Conway, G. Jerkiewicz, J. Electroanal. Chem. 339 (1992) 123. D.A. Harrington, J. Electroanal. Chem. 420 (1997) 101. G. Jerkiewicz, G. Vatankhah, J. Lessard, M.P. Soriaga, Y.S. Park, Electrochim. Acta 49 (2004) 1451. S. Mitsushima, S. Kawahara, K.I. Ota, N. Kamiya, J. Electrochem. Soc. 154 (2007) B153. S. Mitsushima, Y. Koizumi, S. Uzuka, K.I. Ota, Electrochim. Acta 54 (2008) 455. E. Guilminot, A. Corcella, M. Chatenet, F. Maillard, F. Charlot, G. Berthome, C. Iojoiu, J.Y. Sanchez, E. Rossinot, E. Claude, J. Electrochem. Soc. 154 (2007) B1106. A. Sun, J. Franc, D.D. Macdonald, J. Electrochem. Soc. 153 (2006) B260. D.o. Energy, DOE Cell Component Accelerated Stress Test Protocols for PEM Fuel Cells, 2007. G. Benke, W. Gnot, Hydrometallurgy 64 (2002) 205. D.C. Johnson, D.T. Napp, S. Bruckenstein, Electrochim. Acta 15 (1970) 1493. K. Kinoshita, J.T. Lundquist, P. Stonehart, J. Electroanal. Chem. Interfacial Electrochem. 48 (1973) 157. D.A.J. Rand, R. Woods, J. Electroanal. Chem. Interfacial Electrochem. 35 (1972) 209. R.M. Darling, J.P. Meyers, J. Electrochem. Soc. 150 (2003) A1523. A.A. El-Shafei, J. Electroanal. Chem. 447 (1998) 81. S. Taguchi, A. Aramata, J. Electroanal. Chem. 396 (1995) 131. G. Vatankhah, J. Lessard, G. Jerkiewicz, A. Zolfaghari, B.E. Conway, Electrochim. Acta 48 (2003) 1613. A.P. Yadav, A. Nishikata, T. Tsuru, Electrochim. Acta 52 (2007) 7444. D.C. Johnson, D.T. Napp, S. Buckenst, Electrochim. Acta 15 (1970), 1493. T. Hachiya, H. Honbo, K. Itaya, J. Electroanal. Chem. 315 (1991) 275. K. Sashikata, N. Furuya, K. Itaya, J. Electroanal. Chem. 316 (1991) 361. A. Aramata, M.A. Quaiyyum, W.A. Balais, T. Atoguchi, M. Enyo, J. Electroanal. Chem. 338 (1992) 367. S. Taguchi, A. Aramata, M.A. Quaiyyum, M. Enyo, J. Electroanal. Chem. 374 (1994) 275.