Improved Performance of Ni- and Co-YSZ Anodes via Sulfidation to

Oct 3, 2007 - Chemically stable and coke resistant Sm1−xCexFeO3−δ perovskites for low temperature solid oxide fuel cell anode applications. Syed ...
0 downloads 5 Views 233KB Size
Download PDF
15446

J. Phys. Chem. C 2007, 111, 15446-15455

Improved Performance of Ni- and Co-YSZ Anodes via Sulfidation to NiS- and CoS-YSZ. Effects of Temperature on Electrokinetic Parameters Catherine M. Grgicak and Javier B. Giorgi* Centre for Catalysis Research and InnoVation, Department of Chemistry, UniVersity of Ottawa, Ottawa, Ontario, Canada. K1N-6N5 ReceiVed: May 8, 2007; In Final Form: August 10, 2007

Single-cell solid oxide fuel cell (SOFC) experiments using Ni yttria-stabilized zironia (YSZ) and Co-YSZ anodes were performed to examine the electrochemical oxidation of H2 and H2S/H2 mixtures. The introduction of 10%v/v H2S into the fuel stream resulted in anodes that initially showed significant signs of degradation. However, when all of the metal was changed to metal-sulfide, the performance was enhanced under most tested conditions, suggesting that metal-sulfides are viable anode materials for SOFC systems. Electrochemical experiments, mass spectrometry of the exhaust gas, and X-ray diffraction of the postrun anodes show that the main anodic reaction is hydrogen oxidation in both fuel streams. Direct current experiments at various temperatures were performed to determine the inverse Tafel slope (Lefat slope, b-1) and the resulting chargetransfer coefficients for Ni-, Co-, NiS- and CoS-YSZ anodes. For three of the four anodes tested, the Lefat slope decreased with increasing T but not with the required slope of RF/R. The relationship was better represented by RF/R + K. The charge-transfer coefficient was determined to be 1.5 for Ni-, Co- and NiSYSZ anodes, suggesting that the rate-determining step of hydrogen oxidation on these anodes is the electron transfer between Had and O2-. The CoS-YSZ exhibited traditional Tafel behavior such that R remained constant over a wide temperature range but yielded a charge-transfer coefficient much less than that of the other anodes (R ) 0.21 ( 0.04).

Introduction Solid oxide fuel cells (SOFC) are the most flexible fuel cells with respect to fuel selection. They have great potential for generating power from many sources including hydrogen, hydrocarbons, syngas, and biogas.1-3 Utilizing hydrocarbons as fuels has procured much attention recently because of their natural abundance, ease of transport, and accessibility, making SOFCs an ideal bridging technology toward H2 utilization. However, fuel impurities, especially sulfur compounds, which are found in most hydrocarbon feed stocks, cause degradation of the anodes, affecting the overall performance of the fuel cell. Therefore, it is of interest to examine the sulfur tolerance of SOFC anodes by understanding the sulfur poisoning mechanism and to develop sulfur tolerant anodes for SOFC systems. The Ni yttria-stabilized zironia (Ni-YSZ) based systems have shown little tolerance to fuels containing low (ppm) H2S concentrations, where sulfur tolerance decreases as a function of lower operating temperatures and higher H2S concentrations.4,5 However, up to 90% recovery of these systems occurred when H2S was removed from the fuel, signifying that poisoning is caused by blocking of the active Ni-YSZ sites by either adsorbed sulfur or H2S. The small irreversible component was suggested to be the result of Ni-S formation.5,6 Further studies of nickel-sulfide interactions by Raman spectroscopy at high temperatures showed that Ni-S is the prevalent species at SOFC operating conditions.7 Because of the incompatibility of Ni-YSZ based systems with sulfur, emphasis has recently been placed on improving the * To whom correspondence should be addressed. E-mail: jgiorgi@ uottawa.ca; fax: +1(613)562-5170; tel: +1(613)562-5800x6037.

tolerance of SOFC anodes by changing their composition. Nibased anodes where the electrolyte component was changed from YSZ to scandia-stabilized zirconia (SSC) showed increased tolerance at 100 ppm H2S concentrations at 800 °C.4 Additionally, Cu-ceria anodes exhibited high sulfur tolerance and were able to operate at levels up to 450 ppm at 800 °C without any appreciable loss in performance.8 An alternative approach was to utilize perovskite structures that can support the oxide-ion vacancies to give good oxide-ion conduction and that have a mixed-valent cation from the 4d or 5d block, which provides good electronic conduction, while exhibiting high sulfur tolerance. The double perovskites Sr2Mg1-xMnxMoO6-δ showed long-term stability in 50 ppm H2S/H2 at 800 °C, without the formation of sulfur species.9 Other perovskites also showed potential by exhibiting sulfur tolerance at 1000 °C for up to 8 h in a 1% H2S/H2 mixture.10 Although these anode materials show promise in H2S/H2 fuel mixtures at low concentrations, little evidence on their long term performance at high sulfur concentrations is available. Alternatively, the ability to electrochemically oxidize H2S in a fuel cell has also been established. This was first demonstrated by Pujare et al. by utilizing thiospinel CuFe2S4 as the anodes electrocatalytic site and pure H2S as the fuel.11 Additionally, anode catalysts comprising of composite metal sulfides derived from a mixture of sulfides of Mo and other transition metals (Fe, Co, and Ni) were stable and effective electrocatalysts for the conversion of H2S in SOFCs, with Co-Mo-S exhibiting superior activity and longevity.12,13 Although these data suggested that metal-sulfides are viable SOFC anodes for H2S containing fuels, very little is understood regarding the H2S

10.1021/jp073525n CCC: $37.00 © 2007 American Chemical Society Published on Web 10/03/2007

Improved Performance of Ni- and Co-YSZ Anodes

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15447

oxidation reaction that was emphasized by the larger-thanexpected open circuit voltage (OCV) values obtained in these cases. When utilizing H2S as fuel with an oxide-conducting electrolyte, the reaction that takes place at the cathode is the same as that in the conventional case when hydrogen is used and is represented by the following equation.

/2O2 + 2e- f O2-

1

(1)

After migration through the electrolyte the oxide ions can react with hydrogen sulfide in two ways (eqs 2 and 3),

H2S + O2- f H2O + 1/2S2 + 2e-

(2)

H2S + 3O2- f H2O + SO2 +6e-

(3)

where the reversible potentials are 0.726 and 0.750 V at 800 °C, respectively. The process may further be complicated by the thermal decomposition of hydrogen sulfide to sulfur and hydrogen via the following equation;

H2S f H2 + 1/2S2

(4)

thereby, introducing hydrogen, which also may be oxidized. Therefore, the higher than expected OCV values (∼1 V) observed in these systems may be explained by the reaction of H2. Furthermore, elemental sulfur produced through eqs 2 and 4 may be oxidized according to eq 5. 1

/2S2 + 2O2- f SO2 + 4e-

(5)

By studying the product distribution of the H2S, Pt|(CeO2)0.8)(SmO1.5)|Pt, air system, Peterson et al. suggested that the complete oxidation of hydrogen sulfide to sulfur dioxide, as opposed to sulfur, was the preferred product.14 However, it was impossible to determine whether complete oxidation of hydrogen sulfide occurred via eqs 2 and 5 or via eq 3, thereby avoiding the S2 intermediate. Reactions between excess H2S and SO2 may also take place via the second step of the Claus process (eq 6),

2 H2S + SO2 a 2H2O + 3/nSn

n ) 2-8

(6)

where n is the average molecular species of the sulfur product. Because reaction 6 is an equilibrium reaction favored at low temperatures and high SO2 partial pressures, significant H2O and Sn production from this reaction path is not expected at high H2S concentrations and SOFC operating temperatures. Because SOFC operation in the presence of H2S/H2 fuel mixtures represents a complex electrochemical process where the oxidation of hydrogen has been shown to be strongly influenced by the presence of H2S, a detailed study to determine the electrochemical reactions involved is of interest. This paper represents the first detailed study of Ni-YSZ and Co-YSZ anodes subjected to 10% v/v H2S/H2 fuel over an extended period of time. Product analysis via mass spectrometry, as well as a postrun anode analysis, elucidated the nature of the active anode, whereas electrochemical impedance spectroscopy and linear sweep voltammetry at various temperatures was used to determine fundamental kinetic parameters, such as the chargetransfer coefficient (R) and electrochemical activation energies. Experimental The fuel cell was prepared by mixing Triton X-100 (Research Chemical Ltd)/water and 1 g of anode powder. The YSZ

supported metal (M-YSZ) (55 wt %) anode materials were prepared via coprecipitation by dissolving the appropriate amounts of ZrCl4, Y2O3 (dissolved in 10 mL of 37 wt % HCl) and NiCl2‚6H2O or CoCl2‚6H2O (Aldrich) in 165 mL of distilled water. To ensure YSZ phase purity and stability, the electrolyte component of the anode consisted of 21 mol % YSZ as opposed to 8 mol %.15 The precipitation was achieved through introduction of 2 M NaOH (Aldrich) to a final pH of 13. Following precipitation, the product was filtered, washed, dried at 120 °C, and calcined at 750 °C for 2h. The resulting powder composition and characterization has been described previously.16,17 A 0.15 g portion of the slurry was coated onto an YSZ greendisk (Tosoh) and heated to 1400 °C for 4 h in air, resulting in an anode >50µm thick. A 10 wt % Rh/Pt mesh current collector was embedded into the anode by adding an additional coat of anode slurry and sintering to 1400 °C. This process was used to minimize any losses due to contact resistance.18 Although the active thickness of fine Ni-YSZ cermet anodes has been demonstrated to be in the range of 10µm,19 while the rest of the anode acts as current carrier, issues related to enhanced activity due to the mesh were, nevertheless, tested and found to have an insignificant impact on performance. An anode consisting of YSZ and 10 wt % Rh/Pt mesh exhibited a current density of 4 mA/cm2 (η ) 1 V) at 850 °C, whereas typical Ni-YSZ anodes and embedded 10 wt % Rh/Pt mesh consistently exhibited current densities 2 orders of magnitude larger. The low current densities observed in the ‘blank’ showed that contributions from the embedded mesh are minimal and, therefore, were not taken into consideration during electrochemical analysis. A 70 wt % Lanthanum Strontium Maganite (LSM)/30 wt %YSZ mixture (LSM, Nextech Materials; YSZ, Tosoh) was used as the cathode and reference electrode material. The cathode was prepared by coating 0.15 g of the LSM/YSZ powder symmetrically opposing the sintered anode, and the reference electrode was approximately 4 mm to the side of the cathode.20 The 10 wt % Pt/Rh mesh was embedded as described above. After sintering, the YSZ electrolyte was 0.5 mm thick, and the areas of the electrodes were 0.4 cm2 for the anode and cathode and 0.2 cm2 for the reference. A schematic of the fuel cell and holder is shown in Figure 1. The cell geometry was chosen based on both numerical and experimental findings that suggest that an ideal SOFC threeelectrode geometry would include a symmetric cell where the counter and working electrodes are symmetrically opposed to one another with the reference electrode as far away from them as possible. Also, by increasing the electrolyte thickness, errors associated with electrode misalignment would be minimized. However, a compromise regarding electrolyte thickness is needed, given that it must be thin enough to allow reasonable current densities and thick enough to keep electrode alignment errors minimal. Numerical evaluations suggest there is very little room for error when it comes to electrode alignment. Misalignment of the working and counter electrodes by as little as 0.5 electrolyte thicknesses resulted in a 20% error in polarization resistances.21 Therefore, for an electrolyte thickness of 500 µm, only a 50 µm misalignment is endurable. Other numerical studies found similar results; a misalignment of working and counter of 1 mm on a 500 µm electrolyte led to errors as high as 70%.22 However, experimental findings show a significant amount of misalignment can be tolerated as long as the electrolyte is thicker than 200 µm. Jiang et al. found that a misalignment of 0.5 mm on a 200 µm electrolyte resulted in separation of anodic

15448 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Grgicak and Giorgi

Figure 1. Schematic of electrochemical cell.

and cathodic overpotentials with minimal error, and only cells with YSZ electrolytes 200 mV. Therefore, simple gas diffusion limitations based on Fick’s Law were not used to describe this element. Other possible sources of the low-frequency element may include, but are not limited to, hydrogen and/or oxygen ‘diffusion’ through the metal and/or electrolyte.29 Further analysis on the nature of the third low-frequency response was not performed, and subsequent discussions regarding equivalent circuit modeling refer to the first two arcs of the spectra. Bias-dependent measurements were used to identify the origins of each arc. Figure 2b shows the dependence of each resistance as it related to overpotential for the Ni-YSZ anodes at 850 °C in H2. The bias independence of Rs and R1 strongly indicated that R2 is the only arc/process associated with ‘true’

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15449

Figure 3. (a) Typical impedance spectra obtained for the Ni-YSZ system at 850 °C and 50 sccm flow of 10% v/v H2S/H2 at OCV; (0) impedance data, and (-) fitted data. The apex of each arc is labeled with the frequency in Hz. (b) Parameters of equivalent circuit as a function of electrode overpotentials for the Ni-YSZ system at 850 °C in 10% v/v H2S/H2; (9)Rs, ([)R1, and (2)R2.

electrode processes (i.e., hydrogen oxidation). The bias independence of Rs is not surprising because it is associated with the series resistance and represents contact resistances between cell components and conductivity of the electrolyte and electrodes. The bias independence of R1 was consistent with previous findings and suggested that this process was also related to bulk transport.30,31 Figure 3a shows impedance of a typical Ni-YSZ anode at 850 °C in H2S/H2 after significant exposure (t ) 15 h). It also shows two overlapping depressed arcs at high- and midfrequency ranges. Figure 3b illustrates the dependence of each resistance as it relates to overpotential for the Ni-YSZ anode at 850 °C in H2S/H2 after significant exposure time (15 h). It also exhibits the same overpotential trends as those found with the H2, Ni-YSZ system, suggesting the interpretation of impedance spectra remains similar to that found in hydrogen, where Rs may be related to serial resistance, R1 to bulk transport, and R2 to the charge-transfer process. The progression of impedance behavior is shown in Figure 4 for Ni-YSZ and Co-YSZ anodes with time after H2S addition. The first measurement was taken with hydrogen as the fuel. The results indicated that the Ni-YSZ and Co-YSZ cermet anodes were initially degraded by the sulfide impurity. All resistances for both anodes increased when H2S was added to the fuel stream (t ) 1 min). However, after t ) 1.5 h the resistances noticeably decreased. After t ) 3 h, when the metal has undergone a transition to a metal sulfide species, the resistances remain relatively constant and are significantly smaller than those obtained when H2 was the fuel. The Rs values also significantly changed after H2S addition. Convergence of Rs for the Co-YSZ anode to a value that was approximately half of the original indicated improved contact between fuel cell components or decreased electrolyte resistance. Increased electrical conductivity of the anode was excluded as a source of decreased Rs because most metal sulfides are small band gap

15450 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Grgicak and Giorgi

Figure 4. Impedance plots of (a) Ni-YSZ and (b) Co-YSZ anodes in H2S/H2 at OCV and T ) 850 °C at various times. (×) t ) 0 (H2 only), (9) t ) 1 min, (0) t ) 1.5 h, ([) t ) 3 h, (4) t ) 5 h, (O) t ) 10 h, (+) t ) 15 h. Panels c and d are zoomed in images of panels a and b, respectively.

semiconductors and are expected to have a lower electrical conductivity than the metals.32 Cross-sectional SEM/energy dispersive X-ray (EDX) analysis of the postrun SOFC showed the absence of sulfur in the YSZ electrolyte, demonstrating that conversion from metal to metal-sulfide influenced the serial resistance by improving contact between anode and electrolyte. However, the same trend was not present in the Ni-YSZ system, where the Rs increased from 1.6 to 2.2 Ω cm2 after H2S addition and was linked to the instability of the anode at high temperatures because of the low melting temperatures of nickel sulfides (Tm(Ni3S2) ) 787 °C and Tm(NiS) ) 976 °C). In comparison, the melting temperature of CoS (1182 °C) would account for the higher stability seen in the cobalt anodes. The R2 resistances also change when H2S is added to the fuel stream. The R2 for Ni-YSZ decreased from 2.0 Ω cm2 in hydrogen to 1.1 Ω cm2 after 15 h of 10% v/v H2S/H2 exposure. The Co-YSZ anode exhibited a drastic increase in exchange current densities where R2 decreased from 4.5 to 1.4 Ω cm2 in hydrogen and H2S/H2, respectively. The XRD analysis of postrun anodes indicated that the addition of high levels of H2S into the fuel stream results in anodes that consist of Ni-S and Co-S species. More specifically, the Ni-YSZ anode consisted of mostly Ni3S2 (PDF No. 44-1418) with traces of Ni1-xS (PDF No. 02-1273), whereas the Co-YSZ anode was Co1-xS(PDF No. 25-1081) with traces of Co9S8 (PDF No. 02-1459). The presence of metal-sulfide species demonstrated that the metals were converted to metal sulfides during operation when H2S was introduced. However, the phases present at room temperature are not assumed to be present during fuel cell operation at high temperatures. In situ analysis is required to determine the exact species of metalsulfide present during operation. Sulfidation studies of cobalt and nickel suggest the active species at fuel cell operating temperatures are Co1-xS and Ni3+xS2 + Ni1-xS respectively.33,34 Because the exact metal sulfide species at high temperatures is not known, the sulfide anodes synthesized in situ by H2S addition are hereby referred to as NiS- and CoS-YSZ. The increase in polarization of the anode during the first minutes of H2S introduction is consistent with results previously obtained at low concentrations of H2S/H2 and may be the result

of initial blocking of the reactive sites by sulfur species or by the agglomeration of metal particles accelerated by the presence of sulfur, thereby decreasing the triple phase boundary (TPB) area. However, with increased exposure time, all of the metal was converted to metal-sulfide, and a decrease in polarization was observed. This suggests the deactivation observed at low concentrations over short periods of time is the result of the transition between metal and metal-sulfide phases. Once all the metal is converted, the active metal-sulfide catalyst is a viable SOFC anode material for H2S-containing fuels. However, it should be noted that reduction of metal-sulfide back to metal is possible. Therefore, to maintain the structural and compositional purity of metal-sulfide anodes, H2S would continuously be required in the fuel stream. Very little is known regarding the actual electrochemical oxidation of H2S to SO2, particularly if the reaction is complicated by the presence of another fuel such as H2. The analysis of exhaust gas by mass spectrometry indicated that there was not a significant amount of SO2 production and that it was below the detection limit of the instrument (PSO2/Ptot ) 0.001), suggesting that the main fuel was actually H2 and not H2S. Additionally, there was no evidence of solid sulfur formation once the fuel cell was examined after a 15 h period, except for the metal-sulfide anodes themselves. Therefore, the main increase in performance over time was because of the ability of the metal-sulfide electrodes to oxidize hydrogen at an appreciable and continuous rate. The OCVs of the Ni-YSZ and Co-YSZ anodes in H2S/H2 at 850 °C were 1.1 and 1.2 V, respectively. The slight discrepancy between the OVCs of each anode was attributed to reproducibility of the seal between fuel cells. These values are similar to those obtained when H2 was used as fuel, whereas the OCVs (1.2 V for both anodes) are very different from those calculated from thermodynamic data for the oxidation of H2S, which is approximately 0.75 V at 850 °C, but are similar to the equilibrium OCV for a 10% mixture (1.1 V). Although small contributions from the oxidation of H2S cannot fully be excluded, on the basis of the aforementioned observations, all corresponding kinetic analyses assume the oxidation of hydrogen is the main electrochemical reaction for all four anodes.

Improved Performance of Ni- and Co-YSZ Anodes

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15451

Hydrogen Oxidation on Ni-, NiS-, Co- and CoS-YSZ. A detailed understanding of the rate-determining steps of SOFC cermet anodes is necessary to appropriately describe the fundamental factors related to their kinetics, hence, allowing for optimization of the fuel cell. Fundamental factors that influence the charge-transfer reactions are generally accepted to occur at the TPB, where the metal, electrolyte, and open pores meet as paths for the electrons, oxide ions, and gas to carry out the hydrogen oxidation reaction of the SOFC anode (eq 8).

H2(g) + O2-(electrolyte) T H2O(g) + 2e-(electrode)

(8)

Several reaction schemes for hydrogen oxidation on electrodes have been described. Mogensen et al. proposed a mechanism based on impedance data where the high-frequency semicircle was assumed to arise from the transfer of ions across the TPB and electrode particles, and the mid frequency semicircle was associated with the formation of water.31 Reaction 11 was assumed to be rate-limiting.

H2 T 2 Had

(9)

2(Had T H+ ad + e )

(10)

Figure 5. LSV’s for (-) Ni-YSZ and (- -) Co-YSZ in (a) H2 and (b) 10% v/v H2S/H2.

22(H+ T OH-) ad + O

(11)

2OH- T H2O +O2-

(12)

where i is the current density, io is the exchange current density, R is the transfer coefficient, η is the overvoltage, and F, R, and T are the Faraday constant, ideal gas constant, and temperature, respectively. This relation takes into account both the forward and backward reaction directions of the process and the exponential potential dependence of the current components through the transfer coefficients Ra and Rc. However, these factors only apply to one-electron processes, and more complex relations must be established for multistep processes. It is only in the case of one-electron transfers, where the first step is the rate-determining one, that R t β, the symmetry factor. Therefore, a link between multistep electrochemical reaction mechanisms and the Tafel slopes is considered. From eq 16, for appreciable positive polarizations, the second term becomes negligible, and the expression simplifies to that shown in eq 17.

Another possible reaction mechanism is represented by the following equations.

H2 T 2Had

(13)

Had + O2- f OH- + e-

(14)

Had + OH- f H2O + e-

(15)

Where the first step is assumed to be the dissociative adsorption of H2 followed by surface diffusion of Had to the TPB. The oxidation of Had by oxygen ions from the electrolyte would constitute the first of the electron transfers. If it is assumed that each electron transfer occurs individually, then water formation would proceed via hydroxide intermediates. On the basis of the DC techniques, Holtappels et al. suggested the electron transfer involving adsorbed hydrogen (Had, eq 15) is the rate-determining step. However, these findings were conclusive only for the low-temperature range below 845 °C.35 The anodic charge-transfer coefficient (Ra) did not vary at low temperatures, but it decreased at higher temperatures, which led the authors to conclude that a mechanistic change occurs at high temperatures and that another reaction step becomes important for the overall reaction rate. However, it was previously suggested that changes in chargetransfer coefficients with respect to temperature or overpotential may not indicate complex changes in rate-determining steps but an expected dependence of R on temperature. Determination of the charge-transfer coefficient is obtained by using the generalized kinetic equation for a one-electron-transfer process proposed by Erdey-Gruz and Volmer and Butler,36-39

{ [ ]

i ) io exp

[

]}

RcηF RaηF - exp RT RT

(16)

ln i ) ln io + b-1η

b-1 )

R aF RT

(17)

Therefore, by plotting ln i vs η, one may obtain both the exchange current density for a particular reaction and the inverse of the Tafel slope (also referred to as the Lefat slope). The charge-transfer coefficient may be determined if the temperature is known. If equilibrium measurements were obtained at various temperatures, one would expect a plot of b-1 versus 1/T to result in a slope of RaF/R and a y-intercept of 0. However, the expected linear proportionality of b-1 to 1/T, where R is a temperatureindependent constant, has been shown to be the exception rather than the rule. Previous work has indicated that b-1 is not dependent on temperature, according to eq 17, but varies with temperature in a different manner, because R itself may be a function of temperature. Various cases of R-dependence on T are discussed below. The charge-transfer coefficient has been shown to vary with temperature in three distinct ways. The classical expression for b-1 with classical T-1 dependence, that is, when R is independent of T, has been observed for simple ionic redox reactions that have minimal chemical coupling with the electrode surface.40 Investigations of the hydrogen evolution reaction with

15452 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Grgicak and Giorgi

Figure 7. Charge-transfer coefficients as a function of temperature for (9) H2, Ni-YSZ; ([) H2, Co-YSZ; (gray-colored square) H2S/H2, NiS-YSZ; and (gray-colored diamond) H2S/H2, CoS-YSZ SOFC systems.

Figure 6. (a) Enlarged LSV of Ni-YSZ in 10% v/v H2S/H2 atmospheres at 850 °C. (b) Full range CV of the anode in H2S-contaminated fuel at 850 °C.

various metals under different conditions have indicated that the traditional dependence of b-1 with respect to temperature is not followed and that it is actually fitted by the following equation (eq 18)40-43,

b-1 )

R′F +K RT

(18)

which also implies that R is, apparently, temperature dependent because of eq 19

b-1 )

RF R′F t +K RT RT

(19)

or eq 20.

R ) R′ +

KRT F

(20)

A third possible type of b-1-dependence on T -1 has been observed and was expressed in the anodic Br2 evolution from Br- by Conway et al.41 In this case, b-1 was found to be independent of T over an 80 K temperature range. This phenomenon was also observed for the N2 and O2 evolution reactions and O2 reduction.40 In this case, we obtain eq 21,

b-1 )

R(T)F RT

(21)

where R varied with T, making b-1 seemingly independent of T -1. Figure 5 shows an example of iRs-corrected DC measurements for Ni-YSZ and Co-YSZ samples in H2 and H2S/H2 and the subsequent Tafel behavior of each. Linear voltammograms for the H2S/H2 system were obtained after 15 h of exposure to ensure all the metal had converted to metal-sulfide. The Rs term was obtained as the high-frequency intercept of the impedance spectrum for each respective temperature. Qualitatively, it is observed that near OCV, before reaching Tafel behavior, an

anodic peak is observed in the H2S/H2 atmosphere that is not present when H2 is used as fuel. An enlarged partial cyclic voltametry (CV) of this region demonstrates that it consisted of at least two overlapping peaks (Figure 6a). Previous CV experiments on similar samples performed in this laboratory show that the process is irreversible in H2S atmospheres (Figure 6b). Cyclic voltametry studies at various scan rates, temperatures, and partial pressures are required to fully determine the nature of these peaks; however, CV studies on copper sulfide (Cu2S) suggest they may be related to serial oxidation of the metal-sulfides through defective sulfides.44 The overpotential range was not restricted because there was no indication of passivation. Qualitatively, it is observed that the cathodic and anodic branches of the current-overpotential curves are not symmetrical. This indicates different reaction kinetics for the hydrogen oxidation reaction (anodic branch) and hydrogen evolution reaction (cathodic branch). Therefore, the assumption that R t β ) 0.5 is not valid and will be examined in more detail. Additionally, even after iR compensation it was observed that the ln i versus η plots are curved at high overpotentials. Therefore, to choose an appropriate Tafel region, three criteria were required: 1) the exchange current densities between DC and alternating current (AC) techniques were comparable, 2) the Tafel region began at high overpotentials (i.e., >180 mV according to η g RT/βF), and 3) the correlation coefficient was >0.99. As an exception, the linear region for Co-YSZ at 800 °C started at an overpotential less than 180 mV, as seen in Figure 5a. Generally, the linear segment ranged approximately 150 mV and 0.5 decades. Exchange current densities obtained via AC and DC techniques were compared and showed good agreement (Table 1) between the two techniques. Figure 7 shows the relationship between the charge-transfer coefficient and temperature for both anodes in H2 and H2S/H2 atmospheres. The traditional relationship between R and T was not observed and was more readily described by eq 20 for all but one of the anodes. The CoS-YSZ anode exhibited a relatively constant R at all temperatures (0.21 ( 0.04) as per eq 17, where the traditional relationship between b-1 and T-1 was observed. For the three anodes exhibiting a change in R with T, one may determine whether R(T) or R changes because it apparently varies with T by utilizing a simple mathematical comparison between the slope and intercept of eqs 18 and 20. That is, in the case of Type II dependence, the y-intercept of Figure 6 should be equal to the R′ calculated from db-1/dT-1)R′F/R. Table 2 includes the resulting R′ values and its temperature dependence for Ni-, NiS- and Co-YSZ, where the R′ errors are the standard deviation of the y-intercept obtained from the linear regression of eq 20.

Improved Performance of Ni- and Co-YSZ Anodes

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15453

TABLE 1: Comparison of Exchange Current Densities Obtained for Ni-, Co-, NiS- and CoS-YSZ Anodes via DC and AC Techniques Ni-YSZ

Co-YSZ

NiS-YSZ

CoS-YSZ

temperature (°C)

io (EIS) (mA/cm2)

io (DC) (mA/cm2)

io (EIS) (mA/cm2)

io (DC) (mA/cm2)

io (EIS) (mA/cm2)

io (DC) (mA/cm2)

io (EIS) (mA/cm2)

io (DC) (mA/cm2)

800 850 900 950 1000

15 27 39 64 104

19 32 35 68 88

1 3 6 8 24

2 3 6 7 16

20 40 62 70 63

19 40 64 71 81

3 18 71 54 76

10 49 83 75 84

TABLE 2: Summary of Electrochemical Data for Ni-, NiS-, Co-, CoS-YSZ SOFC Anodes temp (°C)

R′

b-1 (mV-1)

R f(T)

Ea (kJ/mol)

R ) R′ + KRT/F

90 (10)

0.0046 0.0032 0.0028 0.0022 0.0019

1.5 (0.2)

120 (10)

0.0028 0.0019 0.0017 0.0022 0.0019

0.21 (0.04)

b-1 (mV-1)

Ni-YSZ 800 850 900 950 1000

0.0049 0.0040 0.0038 0.0025 0.0024

1.5 (0.2)

800 850 900 950 1000

0.0072 0.0068 0.0059 0.0059 0.0046

1.5 (0.3)

Ea (kJ/mol)

R ) R′ + KRT/F

130 (10)

CoS-YSZ

R ) R′ + KRT/F

The ‘true’ transfer coefficient (R′ ) 1.5) had a value appreciably different from 0.5, which is normally assumed. Furthermore, the true transfer coefficient was substantially different from the apparent transfer coefficients obtained at various temperatures, where the measured Rs were in the range 0.26-0.46, 0.50-0.66, and 0.20-0.42 for Ni-, Co-, and NiSYSZ, respectively, suggesting mechanistic differences in hydrogen oxidation between Ni-, NiS-, and Co-YSZ. Furthermore, if the traditional form of the Tafel equation was assumed, changes in R with respect to T would lead to the conclusion that there are different rate-determining steps at high and low temperatures. By identifying the real form of the Tafel equation and the true relationship between R and T, it was determined that the true charge-transfer coefficient did not vary between anodes and only the temperature-independent K-term did. Nontrivial explanations of the significance of the K-term have been offered in the literature. Generally, it has been suggested that if the entropy of activation (∆Sq) is affected through the term RηF, then the Lefat slope will be expected to contain a temperature-independent component.41

(

)

q 1 d ∆H - [T(∆S ( RηF)] + R′ηF )b dη RT

(22)

1 F )(R′ ( RT) b RT

(23)

This result can account for cases where the apparent transfer coefficient linearly increases or decreases with temperature. If the entropy of activation is not affected through overpotential, then the traditional form of the Tafel equation would ensue. The b-1 dependence on inverse temperature thereby influences the conventional representations of the activation processes usually described by the Arrhenius type equation and must be considered. Traditionally, the apparent energy of activation is dependent on overpotential and is described by the following equation,

[(

R f(T)

NiS-YSZ

Co-YSZ

in ) k0 exp -

R′

)]

∆G°q - nRηF RT

(24)

R ) R′

220 (60)

where k0 is the rate constant, in is the current density at a particular overpotential, n is the number of electrons, and ∆G°q represents the apparent activation energy. However, the relationship between in and T-1 will not be linear unless R is independent of T. If R is dependent on T, then ∆G°q can be calculated only at OCV when η is 0. In the case where R′ linearly varies with temperature by virtue of the potential dependence of entropy, the traditional Arrhenius relation would not be observed at higher overpotentials. Figure 8 shows the log in versus T-1 relationships of the Ni-YSZ sample tested in H2 over five overpotentials ranging from 0 to 400 mV in 100 mV intervals, using AC data. Linearity between the two variables is lost at high overpotentials and temperatures. Nonlinearity is most noticeable at η ) 400 mV, again suggesting a nontraditional R dependence on T. In the case of a reaction involving only one electron transfer, the derived Lefat slopes may be directly related to the characteristics of the transition state determining the reaction rate of the process by evaluating the symmetry factor β. However, when a reaction involves more than one electron, where the transfer steps usually occur in discrete steps, the transfer coefficients may also be used to give information on reaction mechanisms. On the basis of the quasiequilibrium approximation, Bockris and Reddy developed transfer coefficients in terms of mechanistic parameters (eq 25),45

Ra )

γf + zrds(1 - β) ν

(25)

where the important quantities are the numbers of electrons transferred during (zrds) and following (γf) a rate-determining step and the number of times the rate-determining step occurs (ν). Therefore, if β ) 0.5 and there is only one electron-transfer per rate-determining step, then a transfer coefficient of 1.5 would suggest a successive two-electron transfer, where the first electron transfer is the rate-determining step and the second electron transfer follows it. This work supports the mechanism in which the rate-determining step was proposed to involve adsorbed hydrogen35 and not an electron transfer between H+ and OH-.31 However, we do not suggest that changes in transfer

15454 J. Phys. Chem. C, Vol. 111, No. 42, 2007

Grgicak and Giorgi Conclusion

Figure 8. Arrhenius plots of anodic current density on Ni-YSZ cermet anodes at various overpotentials and their effect on linearity.

Figure 9. Arrhenius plots of the anodic exchange current density on (9) H2, Ni-YSZ; ([) H2, Co-YSZ; (gray-colored square) H2S/H2, NiSYSZ; and (gray-colored diamond) H2S/H2, CoS-YSZ SOFC systems.

coefficient values at high temperatures point to changes in the reaction mechanism but that they are an expected deviation from the assumed constancy of R. For the CoS-YSZ case, where the charge-transfer coefficient did not change with respect to temperature, an average value of 0.21 ( 0.04 for R was the result. This suggested a different rate-determining step for hydrogen oxidation involving the CoS anode, which is not easily determinable by eq 24 and possibly involves factors associated with adsorption/desorption kinetics, where the rate-determining step may include the dissociation of reactants or recombination of products on the surface.45 All processes were activated by temperature. Activation energies were calculated using the temperature dependence of ln io (i.e., when η ) 0). Figure 9 examines the relationship between ln io and temperature for all anodes. Linearity existed within the whole temperature range for H2 oxidation on metals. The activation energy for hydrogen oxidation was 90 ( 10 kJ/ mol for the Ni-YSZ anode and was slightly lower than previous findings, which range from 110-135 kJ/mol.35,46,47 The apparent activation energy associated with the Co-YSZ anode was 120 ( 10 kJ/mol. For the metal-sulfide systems, Arrhenius behavior was present in the temperature range of 800-900 °C. At temperatures above 900 °C the exchange current density did not change. Activation energies for the NiS-YSZ and CoS-YSZ system were calculated as 130 ( 10 and 220 ( 60 kJ/mol, respectively. The activation energy associated with NiS-YSZ was slightly larger than the one for Ni-YSZ, but it is in the range previously reported for hydrogen oxidation. The activation energy associated with CoS-YSZ is significantly larger than those energies previously reported, again suggesting the H2 oxidation on CoS-YSZ anodes occurs via a different path.

Metal-sulfides, specifically NiS- and CoS-YSZ anodes have been shown to be active over long periods of time for the hydrogen oxidation reaction. Impedance analysis showed an initial decrease in SOFC performance when H2S was added to the fuel stream, but after approximately 3 h, when all of the metal was converted to metal-sulfide, the performance was rectified, suggesting metal-sulfides are viable anode materials for SOFC systems. CoS-YSZ anodes are preferential to NiSYSZ because of their higher melting temperature, stability, and activity. Mass spectrometry and electrochemical results showed the main fuel was H2, as opposed to H2S, and subsequent kinetic analyses on metal and metal-sulfide systems for the oxidation of hydrogen at various temperatures followed. Proper determination of the charge-transfer coefficients to elucidate mechanistic differences for Ni-, Co-, NiS- and CoSYSZ SOFC anodes was performed. Impedance spectra for all anodes resulted in two main components, where the two processes were associated with charge transport and faradaic processes, respectively. The temperature dependence of Tafel slopes were examined and compared to three types of temperature dependence. The Ni, Co and NiS anodes resulted in a Type II dependence, where the Lefat slope is linearly proportional to the inverse temperature but with an additional temperature-independent term. The CoS anode exhibited traditional dependence, where R was constant. The presence of the temperature-independent term caused significant differences with the traditionally calculated R values at each temperature, which would lead to erroneous mechanistic conclusions. The true R calculated over a range of temperatures is 1.5 for Ni-, Co-, and NiS-YSZ, signifying the electrochemical oxidation of hydrogen is a two-step reaction where the first electron transfer is rate determining. The CoS-YSZ anode exhibited a constant R (0.21 ( 0.04) at all temperatures, suggesting a complex ratedetermining step for hydrogen oxidation on CoS anodes. Acknowledgment. The authors thank the Natural Sciences and Engineering Research Council of Canada (NSERC), the Premier’s Research Excellence Award, and the Centre for Catalysis Research and Innovation (CCRI) at the University of Ottawa for financial support. References and Notes (1) Sukeshini, A. M.; Habibzadeh, B.; Becker, B. P.; Stoltz, C. A.; Eichhorn, B. W.; Jackson, G. S. J. Electrochem. Soc. 2006, 153, A705. (2) Zhu, W.; Deevi, S. Mater. Sci. Eng. 2003, A362, 228. (3) Gorte, R. J.; Park, S.; Vohs, J. M.; Wang, C. AdV. Mater. 2000, 12, 1465. (4) Sasaki, K.; Susuki, K.; Iyoshi, A.; Uchimura, M.; Imamura, N.; Kusaba, H.; Teraoka, Y.; Fuchino, H.; Tsujimoto, K.; Uchida, Y.; Jingo, N. Sulfur Tolerance of Solid Oxide Fuel Cells; International Symposium of Solid Oxide Fuel Cells IX (SOFC-IX); Quebec City, Quebec, 2005. (5) Matsuzaki, Y.; Yasuda, I. Solid State Ionics 2000, 132, 261. (6) Xia, S.; Birss, V. DeactiVation and RecoVery of Ni-YSZ anode in H2 fuel containing H2S; International Proceedings of Solid Oxide Fuel Cells IX (SOFC-IX); Quebec City, Quebec, 2005. (7) Dong, J.; Zha, S.; Liu, M. Study of Sulfur-Nickel Interactions Using Raman Spectroscopy; International Symposium on Solid Oxide Fuel CellsIX (SOFC-IX); Quebec City, Quebec, 2005. (8) He, H.; Gorte, R. J.; Vohs, J. M. Electrochem. Solid State Lett. 2005, 8, A279. (9) Huang, Y. H.; Dass, R. I.; Xing, Z.-L.; Goodenough, J. B. Science 2006, 312, 254. (10) Mukundan, R.; Brosha, E. L.; Garzon, R. H. Electrochem. Solid State Lett. 2004, 7, A5. (11) Pujare, N. U.; Semkow, K. W.; Sammells, a. F. J. Electrochem. Soc. 1987, 134, 2639. (12) Liu, M.; Wei, G.; Luo, J.; Sanger, A. R.; Chuang, K. T. J. Electrochem. Soc. 2003, 150, A1025.

Improved Performance of Ni- and Co-YSZ Anodes (13) Wei, G.; Luo, J.; Sanger, A.; Chuang, K. J New Mater. Electochem. Syst. 2005, 8, 59. (14) Peterson, D. R.; Winnick, J. J. Electrochem. Soc. 1998, 145, 1449. (15) Grgicak, C. M.; Giorgi, J. B. YSZ Phase Purity and Stability in Chemically Prepared SOFC anodes; Proceedings of the Ninth International Symposium on Solid Oxide Fuel Cells; Quebec City, Quebec, 2005. (16) Grgicak, C. M.; Green, R. G.; Du, W.-F.; Giorgi, J. B. J. Am. Ceram. Soc. 2005, 88, 3081. (17) Grgicak, C. M.; Green, R. G.; Giorgi, J. B. J. Mater. Chem. 2006, 16, 885. (18) Jiang, S.; Love, J.; Apateanu, L. Solid State Ionics 2003, 160, 15. (19) Brown, M.; Primdahl, S.; Mogensen, M. J. Electrochem. Soc. 2000, 2, 475. (20) Krishnan, V. V.; McIntosh, S.; Gorte, R. J.; Vohs, J. M. Solid State Ionics 2004, 166, 191. (21) Winkler, J.; Hendriksen, P.; Bonanos, N.; Mogensen, M. J. Electrochem. Soc. 1998, 145, 1184. (22) Chan, S.; Chen, X.; Khor, K. J. App. Electrochem. 2001, 31, 1163. (23) Jiang, S. J. Appl. Electochem. 2004, 32, 1045. (24) Boer, B. d. SOFC Anode: Hydrogen oxidation at porous nickel and nickel/yttria stabilised zirconia cermet electrodes. Ph.D. Thesis, University of Twente, 1998. (25) Fleig, J. J. Electroceram. 2004, 13, 637. (26) Fleig, J.; Maier, J. J. Electrochem. Soc. 1997, 144, L302. (27) Fleig, J.; Maier, J. J. Electroceram. 1997, 1, 73. (28) Fleig, J.; Pham, P.; Sztulzaft, P.; Maier, J. Solid State Ionics 1998, 113-115, 739. (29) Holtappels, P.; Vinke, I.; Haart, L. d.; Stimming, U. J. Electrochem. Soc. 1999, 146, 2976. (30) Kek, D.; Mogensen, M.; Pejovnik, S. J. Electrochem. Soc. 2001, 148, A878.

J. Phys. Chem. C, Vol. 111, No. 42, 2007 15455 (31) Primdahl, S.; Mogensen, M. J. Electrochem. Soc. 1997, 144, 3409. (32) Xu, Y.; Schoonen, M. A. Amer. Mineralogist 2000, 85, 543. (33) Mrowec, S.; Danielewski, M.; Wojtowicz, A. J. Mater. Sci. 1998, 33, 2617. (34) Seim, H.; Fjellvag, H.; Gronvold, F.; Stolen, S. J. Solid State Chem. 1996, 121, 400. (35) Holtappels, P.; Haart, L. d.; Stimming, U. J. Electrochem. Soc. 1999, 146, 1620. (36) Butler, J. Trans. Faraday Soc. 1923, 19, 729. (37) Butler, J. Trans. Faraday Soc. 1932, 28, 379. (38) Erdey-Gruz, T.; Volmer, M. Z. Phys. Chem. A 1930, 150, 203. (39) Erdey-Gruz, T.; Wick, H. Z. Phys. Chem. A 1932, 162, 53. (40) Conway, B. E. The Temperature and Potential Dependence of Electrochemical Reaction Rates and the Real Form of the Tafel Equation. In Modern Aspects of Electrochemistry; Bockris, J. O. M., Ed.; Plenum Press: New York, 1985; Vol. 16, pp 103. (41) Conway, B.; Tilak, D. M. B. Trans. Faraday Soc. 1969, 66, 1203. (42) Conway, B.; Salomon, M. J. Chem. Phys. 1964, 41, 3169. (43) Post, B.; Hiskey, C. J. Am. Chem. Soc. 1950, 72, 4203. (44) Nava, D.; Gonzalez, I. Electrochem. Acta 2006, 51, 5295. (45) Lefebvre, M. C. Establishing the Link Between Multistep Electrochemical Reactio Mechanisms and Experimental Tafel Slopes. In Modern Aspects of Electrochemistry; White, R. E., Ed.; Plenum Publishers: New York, 1999; Vol. 32, pp 249. (46) Holtappels, P.; Vinke, I.; Haart, L. d.; Stimming, U. J. Electrochem. Soc. 1999, 146, 2976. (47) Sukenshini, A. M.; Habibzadeh, B.; Becker, B. P.; Stoltz, C. A.; Eichhorn, B. W.; Jackson, G. S. J. Electrochem. Soc. 2006, 153, A705.