- Email: [email protected]
Strictly potentiostatic current oscillations during bulk CO electro-oxidation on platinum in the presence of inhibiting anions
Electrochemistry Communications 7 (2005) 710–716 www.elsevier.com/locate/elecom
Strictly potentiostatic current oscillations during bulk CO electro-oxidation on platinum in the presence of inhibiting anions Souradip Malkhandi, Antoine Bonnefont, Katharina Krischer
*
Physik-Department E19, Technische Universita¨t Mu¨nchen, James-Franck-Str. 1, D - 85748 Garching, Germany Received 4 April 2005; received in revised form 19 April 2005; accepted 19 April 2005 Available online 26 May 2005
Abstract We investigate bulk CO electrooxidation on a rotating Pt disk electrode in concentrated acidic electrolytes containing small amounts of BF 4 anions with cyclic voltammetry and potentiostatic measurements. Under potentiodynamic conditions, the average reactivity decreases with increasing BF 4 -concentration. For fixed applied potential strictly potentiostatic current oscillations were observed. Both observations are reproduced with a mean field model for CO electrooxidation in which the anions are assumed to competitively adsorb on free electrode sites while the electrode potential is treated as a parameter. Bulk CO electrooxidation on Pt in the presence of inhibiting anions is thus an example of an electrochemical oscillator with purely chemical feedback loops. 2005 Elsevier B.V. All rights reserved. Keywords: Carbon monoxide oxidation; Oscillations; Tetrafluoroboric acid; Anion adsorption; Cyclic voltammetry
1. Introduction When adsorbed on an electrode, carbon monoxide is a poison for many electrocatalytic reactions, most notably hydrogen oxidation and the oxidation of small organic molecules, such as formic acid or methanol (see, e.g. [1,2] and references therein). Electrode poisoning by adsorbed CO is also a major problem concerning the anode reaction in low-temperature H2/O2 or direct methanol fuel cells. In the former type of fuel cells, the H2 feed gas might be contaminated with CO owing to the reformer process, which is the main source of H2. During methanol oxidation on Pt, CO is formed as an intermediate, leading thus to a self-poisoning of the electrode. Understanding the mechanism of CO electrooxidation is therefore essential to be able to minimize the poisoning effect of CO in electrocatalysis, and much *
Corresponding author. Tel.: +49 89 289 12535; fax: +49 89 289 12530. E-mail address: [email protected] (K. Krischer). 1388-2481/$ - see front matter 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.elecom.2005.04.022
work has been devoted to fundamental studies of CO electro-oxidation on noble metals under various conditions (for reviews, see e.g. [1–5]). An important result of these studies is that CO is oxidized through the Langmuir-Hinshelwood mechanism, i.e., the formation of CO2 is preceded by the adsorption of both reactants, CO and an oxygen-containing species which compete for the same reaction sites [6,7]. On Pt electrodes, the reactive oxygen species are formed through the oxidative chemisorption of water, which takes place only at potentials considerably more positive than the reversible oxidation potential of hydrogen or methanol. Hence, the extent of poisoning by CO should be reducible when providing an oxygen source at lower potentials. Most of the efforts along these lines aimed at ÔdecoratingÕ or alloying Pt catalysts with other noble metals, which form oxides at lower potentials than Pt [8]. Ru was found to be a promising metal in this respect. However, not only the anode material but also the nature of the supporting electrolyte influences the
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
overall reaction rate and thus has to be incorporated when one wishes to obtain a complete picture of the mechanism of CO electro-oxidation. Consequently, more recently also the influence of different anions on the oxidation rates of CO and of small organic molecules has been studied. In the course of such studies, an unusual anion effect was reported by Schell and coworkers [9–11]. These authors observed a remarkable enhancement in the oxidation rate of formic acid [9,10] and of methanol [11] when replacing a small amount of perchloric acid base electrolyte with HBF4. They offer an intriguing explanation for the enhancement effect, which could be described as the anionic counterpart to the metal-promoted oxide formation at low potentials: in aqueous solutions BF is in equilibrium with 4 BF3OH ions, which in turn can, directly or indirectly, adsorb on the electrode negative to the onset of oxidation of chemisorbed water. Schell et al. suggest that the OH-group of adsorbed BF3OH reacts with adsorbed CO, keeping thus the electrode reactive for the direct oxidation of formic acid or methanol at potentials at which it is poisoned by CO in a BF 4 -free electrolyte. In this paper, we study bulk CO electro-oxidation on a rotating Pt disk electrode in perchloric and sulfuric acid containing different amounts of HBF4 as well as in 0.5 M HBF4 solution to test the above described hypothesis.
2. Experimental The working electrode (WE) was a polycrystalline platinum (99.999%) disk electrode embedded in a Teflon block. The home-built electrode system was attached to a Pine instrument rotator. The electrode had a geometrical surface area of 0.1975 cm2 and a roughness factor of about 1.65. Note that below we normalize the current density to the geometrical surface area. Prior to each experiment the electrode was pretreated in the following way: three times in a row, it was first immersed in a 1:1 mixture of H2SO4 (96%) and H2O2 (30%) for 5 min, and then in Millipore water for 5 min. Both steps were done in an ultrasonic bath. Thereafter, the electrode was thoroughly rinsed with water and transferred to the electrochemical cell with a water droplet on the electrode surface. It was immersed in the electrolyte under potential control close to 0 V RHE. The counter electrode was a Pt wire bent to a ring and was placed symmetrically to the WE in the main compartment of the cell (for details see [12]). The reference electrode (RE) was a sat. Hg/ Hg2SO4 electrode and kept in a separate compartment. All potential values below are given with respect to this reference electrode. All water used in the experiments was obtained from a Millipore MilliQ system. The electrolytes were prepared from HClO4 (70%, Suprapure Merck) or H2SO4
711
(96%, Suprapure Merck), and HBF4 (50%, Aldrich, ÔPurumÕ). The glass cell as well as all other glass items used were cleaned in H2SO4 + H2O2 1:1 solution. The potentiostat employed was built by the Electronic Laboratory of the Fritz-Haber-Institut der MPG, Berlin, Germany and worked in parallel with a Sun Solaris based Real Time data acquisition system. Prior to the measurements with CO, the quality of the WE was checked by recording a cyclic voltammogram (CV) in an Ar (Linde, 99.999%) saturated electrolyte. Therefore, the electrolyte was purged with Ar for at least 30 min, during which the potential was scanned between 660 and 600 mV (vs. sat. Hg/Hg2SO4). After switching to CO (Air Liquide, 99.99%) bubbling, the electrode was cycled for approximately 25 min between 640 and 400 mV to guarantee saturation of the electrolyte with CO before CVs were recorded. Also in the further course of the experiments CO was continuously bubbled through the electrolyte. In all experiments shown the WE was rotated with 1200 rpm.
3. Results Fig. 1 shows asymptotic cyclic voltammograms of Pt in CO-saturated 0.5 M H2SO4 containing different amounts of BF 4 ions. The solid curve was obtained in pure supporting electrolyte, the dashed, dotted and dot-dashed curves in electrolytes containing 5, 10 and 20 mM HBF4, respectively. The CVs are all qualitatively similar and agree with those from the literature. During the positive scan, they exhibit a low current, COpoisoned state up to approximately 250 mV, at which the onset of the oxidation of the CO monolayer manifests itself in an ignition peak. For more positive
Fig. 1. Cyclic voltammograms of a rotating Pt disk electrode in CO saturated 0.5 M H2SO4 solution containing 0 (solid curve), 5 (dashed curve), 10 (dotted curve) and 20 (dash-dotted curve) mM HBF4; scan rate 50 mV/s; rotation rate 1200 rpm.
712
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
potentials the current density takes on a diffusion limited plateau. On the reverse scan, the system stays in this reactive, diffusion limited state also beyond the ignition peak. The CVs thus exhibit a hysteresis between a reactive and a poisoned state in a certain potential interval negative to the ignition peak. The quantitative potential value, at which the transition to the CO-covered lowcurrent state occurs depends on the BF 4 -concentration: the higher the BF 4 -concentration, the more positive is the transition from the reactive branch to the poisoned branch on the negative scan. With other words, the hysteresis becomes smaller and is shifted toward more positive potentials with increasing HBF4 concentration. The same observation was made when using 0.5 M HClO4 as base electrolyte: in the presence of HBF4 the transition from the reactive to the unreactive state occurred at more positive potentials compared to the one in pure base electrolyte. As an extreme case, we also recorded CVs in 0.5 M HBF4. Here, the ignition peak was shifted positive by about 100 mV, and on the negative scan the current decreased from the positive turning point (0.4 V) on such that the hysteresis changed its direction of circulation, the oxidation of CO being inhibited in a potential region which is considerably extended towards positive values. The cyclic voltammograms of Fig. 1 were obtained at a scan rate of 50 mV/s. To obtain a quasi-stationary current–voltage characteristic, we also investigated the current response at a scan rate of 1 mV/s for solutions containing 10 mM HBF4. Under these conditions, the width of the hysteresis was smaller than at a higher scan rate. We also note that it depended on the anodic turning point. The most remarkable difference to the CVs at higher scan rate was however, that the current exhibited oscillatory behavior in a small potential region around the ignition point. The oscillations sustained when fixing the applied voltage within a potential window of a few mV positive to the ignition potential. An example of the onset of oscillations in 0.5 M H2SO4 supporting electrolyte is shown in Fig. 2(a). Fig. 2(b) displays a portion of the current time series after transients decayed. The oscillations are fairly regular and periodic. They disappeared when the rotation of the electrode was stopped. Also when using 0.5 M HClO4 as supporting electrolyte, current oscillations developed in a small potential window just positive to the ignition potential. However, in this case, the oscillations were never regular. Instead, oscillations developed, which had a characteristic base period but apparently randomly varying amplitudes and a higher frequency, noise-like signal superimposed (Fig. 3). These irregularities occurred only in the oscillatory region, strongly suggesting that they are intrinsic to the dynamics rather than caused by instrumental or other types of ÔexternalÕ noise. When repeating the experiments shown in Figs. 2(b) and 3(b) in 0.5 M HBF4, sustained oscillations could
Fig. 2. Current oscillation during bulk CO oxidation on Pt in 0.5 M H2SO4 containing 10 mM HBF4. (a) Current–time (solid curve) and voltage time (dashed curve) plots showing the onset of oscillations after sweeping the voltage to 270 mV with 10 mV/s. (b) Current time series at U = 270 mV after the decay of transients. Rotation rate: 1200 rpm.
not be obtained. Instead, as can be seen in Fig. 4, a series of Ôignition peaksÕ with strongly decreasing amplitude was obtained.
4. Discussion The CVs displayed in Fig. 1 clearly show that the potential window in which CO is oxidized is shifted towards more positive potentials with increasing concentration. The simplest explanation of this observation seems to be that BF 4 ions, or species which result from the hydrolysis of BF 4 , adsorb at the electrode thereby occupying surface sites but not reacting with CO, as originally suggested in [9,10]. In other words, the results point to an inhibition of CO oxidation by BF 4 adsorption. To substantiate this conjecture, we
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
713
augmented the mean-field model for CO oxidation on Pt introduced in [13] by an equation describing the ad- and desorption of an unreactive anion, which competes with CO and OH for free Pt sites. The model is based on the following reaction steps COb ! COs
ðaÞ
COs þ ! COad H2 O þ $ OHad þ Hþ þ e
ðbÞ ðcÞ
COad þ OHad ! CO2 þ Hþ þ e þ 2
ðdÞ
X þ $ Xad
ðeÞ
describing the transport of CO from the bulk solution to the electrode surface (a), the adsorption of CO on a free site of the electrode, denoted by * (b), the oxidative chemisorption of water on the electrode (c), the reaction of adsorbed CO and adsorbed OH species (d) and the ad- and desorption of anions, X, on free electrode sites (e). This reaction mechanism can be cast into a set of four ordinary differential equations determining the temporal evolution of the CO concentration in front of the electrode, cs, and the coverages of the electrode with CO, hCO, OH, hOH and anions, hX, respectively
Fig. 3. Current oscillation during bulk CO oxidation on Pt in 0.5 M HClO4 containing 10 mM HBF4. (a) Current–time (solid curve) and voltage time (dashed curve) plots showing the onset of oscillations after sweeping the voltage to 261 mV with 10 mV/s. (b) Current time series at U = 261 mV after the decay of transients. Rotation rate: 1200 rpm.
Fig. 4. Current–time and voltage time plots of CO oxidation on a rotating Pt electrode in 0.5 M HBF4. Rotation rate 1200 rpm.
dcs 2S tot 2D ¼ mCO;ads þ 2 ðcb cs Þ; dt d d dhCO ¼ mCO;ads mreac ; dt dhOH ¼ mOH;ads mreac mOH;des ; dt dhX ¼ mX;ads mX;des ; dt
ð1Þ ð2Þ ð3Þ ð4Þ
where Stot denotes the number of surface sites per unit electrode area, d is the thickness of NernstÕs diffusion layer, and D the diffusion constant of CO in aqueous solution. The various rates of adsorption, mi,ads, desorption, mi,des, and reaction, mreac, are given by mCO;ads ¼ k CO;ads cs ð0.99 hCO hOH hX Þ; aFU mOH;ads ¼ k OH;ads exp RT
ð5Þ
ðhm OH ð1 hCO hX Þ hOH Þ; aFU mreac ¼ k reac exp hOH hCO ; RT ð1 aÞFU mOH;des ¼ k OH;des exp hOH ; RT nX aFU mX;ads ¼ cX k X;ads exp RT
ð6Þ
mX;des
ð0.99 hCO hX hOH Þ; nX ð1 aÞFU ¼ k X;des exp hX . RT
ð7Þ ð8Þ
ð9Þ ð10Þ
714
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
Here, the kis denote the rate constants of adsorption, desorption and reaction, respectively, U is the applied voltage (which is for the high electrolyte concentrations of the experiment equal to the electrode potential), and R, F, a and T have their usual meanings, cX is the anion bulk concentration, and nX the fractional number of electrons transferred upon ad- and desorption of anions. As in [13], a maximum coverage of OHad, hm OH , was assumed to avoid a complete blocking of CO adsorption by OHad at high potential, and the CO and anion coverages were limited to 0.99 to allow for some formation of OHad on the CO-poisoned state at low potentials such that the reaction can ignite spontaneously. For all reaction steps involving charge transfer, a Butler–Volmertype expression was used and the total Faradaic current density flowing is iF ¼ FS tot ½mOH;ads þ mreac þ nX mX;ads mOH;des nX mX;des : ð11Þ The values of the parameters used in the calculations are given in Table 1. Calculated cyclic voltammograms for four different concentrations of anions are presented in Fig. 5. They exhibit a hysteresis between a CO-poisoned and a reactive state, as typical for bulk CO electro-oxidation. Moreover, the transition from the reactive to the poisoned state during a negative scan occurs the more positive the higher the anion concentration is, while the onset potential of CO oxidation during the positive scan is independent of the anion concentration. This is exactly what was also observed in the experiments. Thus, the experimental observations at high scan rate can be explained assuming that BF 4 anions competitively adsorb on the electrode thereby blocking free surface sites for OHad formation and CO adsorption. Also the experiments at fixed applied voltage are reproduced by the model. For an outer voltage close to the transition from the poisoned to the reactive state,
Fig. 5. Calculated cyclic voltammograms for different anion concentrations cX. cX = 0 mM (solid line), 5 mM (dashed line), 10 mM (dotted line), 20 mM (dot-dashed line). Scan rate: 50 mV/s (other parameters see Table 1).
the calculated current density exhibits sustained oscillations (Fig. 6). This result gives further evidence that BF 4 anions or species deriving from their hydrolysis, such as BF3OH, competitively adsorb at the electrode and block surface sites for CO oxidation rather than supplying a source for a reactive Ôlow potentialÕ oxygen species as originally conjectured in [9,10]. Hence, the present study suggests that the reported enhancement in the formic acid oxidation rate by BF 4 , is caused by a more intricate and so far unidentified interaction between reactants, reaction intermediates and anions. There are further implications of the proposed reaction mechanism. Reaction step (e) does not take into account the nature of the adsorbing anion. Hence, any anion which competes with oxidative chemisorption of
Table 1 Values of the parameters in Eqs. (1)–(11) used in the calculations Parameters cb D d Stot kx,ads kx,des nX kOH,ads kOH,des kCO,ads kr hm OH a T
Value 3
10 5 · 105 0.0019 2.2 · 109 500 5 0.2 104 105 108 105 0.333 0.5 300
Unit mol l1 cm2 s1 cm mol cm2 cm3 mol1 s1 s1 s1 s1 cm3 mol1 s1 s1
K
Fig. 6. Calculated current timeseries for U = 0.858 V and cX = 20 mM (other parameters see Table 1).
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
water for free surface sites should induce oscillations in some concentration window. In fact, when replacing BF 4 by Cl , we obtained current oscillations in a potential window just positive to the ignition potential, giving evidence that there is a wide spectrum of physically relevant situations in which sustained oscillations during CO electro-oxidation exist. The fact that the reaction mechanism (a)–(e) can lead to oscillatory behavior has also an interesting aspect concerning nonlinear phenomena in electrochemical systems in general. The reaction mechanism takes into account only chemical steps, the electrode potential being reduced to a parameter. In all other models of electrochemical oscillations, we are aware of, the electrode potential is an essential variable, i.e., when it is not free to vary, such as under potential control in concentrated base electrolytes, the systems do not exhibit oscillations. Let us discuss how the oscillations can then be understood. The Langmuir-Hinshelwood mechanism of CO electro-oxidation (Reactions (b)–(d) coupled to slow mass transport of CO (a) provides a positive (autocatalytic) feedback loop, giving rise to an S-shaped current–potential curve [13,14]. It thus falls into the category of electrochemical systems possessing a negative differential resistance (NDR) of the S-type, in short it is classified as an S-NDR system [15,16]. In all S-NDR systems, the autocatalysis is of chemical nature, while the electrode potential tends to suppress or inhibit the autocatalytic growth. In the language of nonlinear dynamics the electrode potential is an inhibitor, and an S-NDR system can be classified as an activator– inhibitor system. Activator–inhibitor systems exhibit oscillations whenever the characteristic time scale of the activator is faster than the one of the inhibitor. In S-NDR systems, however, characteristic changes of the electrode potential are much faster than of the chemical species comprising the autocatalytic loop. For this reason, S-NDR systems are very unlikely to exhibit (homogeneous) oscillations [15,16]. However, they were found to undergo spatial instabilities leading to Turing structures or spatial domains [17–20]. In addition, the instabilities will only be found for nonnegligible IR-drop or under galvanostatic conditions. In our experiments, the oscillations were found in 0.5 M acidic solutions under potential control, i.e., under conditions, under which the ohmic potential drop through the electrolyte is minimal. Together with the fact that in the simulations the oscillatory instability was obtained with the electrode potential being a parameter, we can exclude that the oscillations are of the S-NDR-type. An analysis of the reaction mechanism (a)–(e), on the other hand, reveals that the anion coverage acts like an inhibitor, and the oscillations arise owing to the interplay of the positive feedback loop stemming from the Langmuir-Hinshelwood mechanism coupled to the mass transfer limited step and the nega-
715
tive feedback loop generated by anion ad- and desorption. Thus, the system can be characterized as an ÔS-NDR system with chemical inhibitionÕ or as a strictly potentiostatic oscillator, in which the electrolyte resistance does not play any role for the oscillations. This class of electrochemical oscillators had been speculated to exist, however, so far an unambiguous experimental identification was missing. There are two other reports of oscillations during CO electro-oxidation on Pt. Azevedo et al. [21] observed current oscillations in 0.1 M HClO4 on Pt that occurred in a small potential window positive to the ignition peak, which coincides with the voltage range in which the above described oscillations occurred. The authors speculate that the oscillations arise owing to a hidden negative differential resistance, without however presenting convincing arguments for this interpretation. On the other hand, in our experiments we never observed oscillatory behavior when using suprapure HClO4 as supporting electrolyte without addition of adsorbing anions. Thus, the simplest interpretation of the origin of the oscillations reported in [21] seems to be that there were traces of foreign anions in the perchloric acid electrolyte, such as chloride ions, and the oscillations are also caused by the competitive ad- and desorption of these anions. We emphasize, however, that as long as the specification of the acid used is unknown this interpretation remains very speculative. Koper et al. [13] observed potential oscillations under galvanostatic conditions in perchloric acid and in sulfuric acid on Pt single crystal electrodes, again without addition of inhibiting anions. The above described homogeneous model does not possess oscillatory solutions under galvanostatic conditions, such that the origin of these oscillations appears to be different from the present ones, and it remains unclear whether there is any relation between the oscillations observed here and in [13]. However, it is striking that in our work as well as in their work the oscillations are irregular in HClO4 base electrolyte while regular oscillations are reported for Pt(1 1 1) electrodes in H2SO4 solutions [13] and were found in our studies when using H2SO4 as base electrolyte. Irregular oscillations might be induced by spatial instabilities. We are currently investigating whether this explanation applies here.
5. Conclusions Using a simple reaction mechanism, we demonstrated that oscillations during CO electrooxidation on rotating Pt electrodes in concentrated base electrolytes, which were observed in our experiments in the presence of tetrafluoroborate anions, may solely be induced by surface electrochemical steps, i.e., the electrode kinetics. An interaction between electrode processes and the IR drop
716
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
through the electrolyte which generates the positive feedback loop in all other electrochemical oscillators we are aware of is not involved in the instability. Hence, this is a long sought-after experimental example of a strictly potentiostatic electrochemical oscillator. This finding expands not only the experimentally verified categories of electrochemical oscillators but makes it also likely that spatio-temporal patterns of the reaction-diffusion type, as opposed to reaction-migration patterns that form in Ôtraditional electrochemical oscillating systemsÕ, may establish on electrode surfaces. Concerning fundamental aspects of CO electrocatalysis, the observed dynamic instabilities underline that it is necessary to incorporate the possibility for cooperative phenomena when analyzing experimental data, and especially when trying to bridge the gap between microscopic measurements and mean field behavior.
Acknowledgment This work was supported by the Deutsche Forschungsgemeinschaft (DFG) through a Research Grant.
References [1] T. Iwasita, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells, vol. 2, Wiley, Chichester, 2003, Chapter 41.
[2] N.M. Markovic, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells, vol. 2, Wiley, Chichester, 2003, Chapter 26. [3] J.M. Weaver, S. Zhou, in: R.J.H. Clark, R.E. Hester (Eds.), Infrared and Raman Spectroscopy, vol. 26, Wiley, 1998, Chapter 5. [4] M.T.M. Koper, N.P. Lebedeva, C.G.M. Hermse, Faraday Discuss. 121 (2002) 301. [5] B. Beden, C. Lamy, N.R. DeTacconi, A.J. Arvia, Electrochim. Acta 35 (1990) 691. [6] N.M. Markovic, C.A. Lucas, B.N. Grgur, P.N. Ross, J. Phys. Chem. B 103 (1999) 9616. [7] N.M. Markovic, B.N. Grgur, C.A. Lucas, P.N. Ross, J. Phys. Chem. B 103 (1999) 8568. [8] M.T.M. Koper, Surf. Sci. 548 (2004) 1. [9] S.L. Chen, D. Lee, M. Schell, Electrochim. Acta 46 (2001) 3481. [10] S.L. Chen, D. Lee, M. Schell, Electrochem. Commun. 3 (2001) 81. [11] B.E.K. Swamy, C. Vannoy, J. Maye, M. Schell, Electrochem. Commun. 6 (2004) 1032. [12] P. Grauel, H. Varela, K. Krischer, Faraday Discuss. 120 (2002) 165. [13] M.T.M. Koper, T.J. Schmidt, N.M. Markovic, P.N. Ross, J. Phys. Chem. B 105 (2001) 8381. [14] P. Strasser, M. Eiswirth, G. Ertl, J. Chem. Phys. 107 (1997) 991. [15] K. Krischer, J. Electroanal. Chem. 501 (2001) 1. [16] K. Krischer, in: D.M. Kolb, R.C. Alkire (Eds.), Advances in Electrochemical Sciences and Engineering, vol. 8, Wiley–VCH, 2003, p. 89. [17] N. Mazouz, K. Krischer, J. Phys. Chem. B 104 (2000) 6081. [18] Y.J. Li, J. Oslonovitch, N. Mazouz, F. Plenge, K. Krischer, G. Ertl, Science 291 (2001) 2395. [19] A. Bonnefont, H. Varela, K. Krischer, Chem. Phys. Chem. 4 (2003) 1260. [20] A. Bonnefont, H. Varela, K. Krischer, J. Phys. Chem. B 109 (2005) 3408. [21] D.C. Azevedo, A.L.N. Pinheiro, E.R. Gonzales, Electrochem. Solid-State Lett. 5 (2002) A51.
Strictly potentiostatic current oscillations during bulk CO electro-oxidation on platinum in the presence of inhibiting anions Souradip Malkhandi, Antoine Bonnefont, Katharina Krischer
*
Physik-Department E19, Technische Universita¨t Mu¨nchen, James-Franck-Str. 1, D - 85748 Garching, Germany Received 4 April 2005; received in revised form 19 April 2005; accepted 19 April 2005 Available online 26 May 2005
Abstract We investigate bulk CO electrooxidation on a rotating Pt disk electrode in concentrated acidic electrolytes containing small amounts of BF 4 anions with cyclic voltammetry and potentiostatic measurements. Under potentiodynamic conditions, the average reactivity decreases with increasing BF 4 -concentration. For fixed applied potential strictly potentiostatic current oscillations were observed. Both observations are reproduced with a mean field model for CO electrooxidation in which the anions are assumed to competitively adsorb on free electrode sites while the electrode potential is treated as a parameter. Bulk CO electrooxidation on Pt in the presence of inhibiting anions is thus an example of an electrochemical oscillator with purely chemical feedback loops. 2005 Elsevier B.V. All rights reserved. Keywords: Carbon monoxide oxidation; Oscillations; Tetrafluoroboric acid; Anion adsorption; Cyclic voltammetry
1. Introduction When adsorbed on an electrode, carbon monoxide is a poison for many electrocatalytic reactions, most notably hydrogen oxidation and the oxidation of small organic molecules, such as formic acid or methanol (see, e.g. [1,2] and references therein). Electrode poisoning by adsorbed CO is also a major problem concerning the anode reaction in low-temperature H2/O2 or direct methanol fuel cells. In the former type of fuel cells, the H2 feed gas might be contaminated with CO owing to the reformer process, which is the main source of H2. During methanol oxidation on Pt, CO is formed as an intermediate, leading thus to a self-poisoning of the electrode. Understanding the mechanism of CO electrooxidation is therefore essential to be able to minimize the poisoning effect of CO in electrocatalysis, and much *
Corresponding author. Tel.: +49 89 289 12535; fax: +49 89 289 12530. E-mail address: [email protected] (K. Krischer). 1388-2481/$ - see front matter 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.elecom.2005.04.022
work has been devoted to fundamental studies of CO electro-oxidation on noble metals under various conditions (for reviews, see e.g. [1–5]). An important result of these studies is that CO is oxidized through the Langmuir-Hinshelwood mechanism, i.e., the formation of CO2 is preceded by the adsorption of both reactants, CO and an oxygen-containing species which compete for the same reaction sites [6,7]. On Pt electrodes, the reactive oxygen species are formed through the oxidative chemisorption of water, which takes place only at potentials considerably more positive than the reversible oxidation potential of hydrogen or methanol. Hence, the extent of poisoning by CO should be reducible when providing an oxygen source at lower potentials. Most of the efforts along these lines aimed at ÔdecoratingÕ or alloying Pt catalysts with other noble metals, which form oxides at lower potentials than Pt [8]. Ru was found to be a promising metal in this respect. However, not only the anode material but also the nature of the supporting electrolyte influences the
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
overall reaction rate and thus has to be incorporated when one wishes to obtain a complete picture of the mechanism of CO electro-oxidation. Consequently, more recently also the influence of different anions on the oxidation rates of CO and of small organic molecules has been studied. In the course of such studies, an unusual anion effect was reported by Schell and coworkers [9–11]. These authors observed a remarkable enhancement in the oxidation rate of formic acid [9,10] and of methanol [11] when replacing a small amount of perchloric acid base electrolyte with HBF4. They offer an intriguing explanation for the enhancement effect, which could be described as the anionic counterpart to the metal-promoted oxide formation at low potentials: in aqueous solutions BF is in equilibrium with 4 BF3OH ions, which in turn can, directly or indirectly, adsorb on the electrode negative to the onset of oxidation of chemisorbed water. Schell et al. suggest that the OH-group of adsorbed BF3OH reacts with adsorbed CO, keeping thus the electrode reactive for the direct oxidation of formic acid or methanol at potentials at which it is poisoned by CO in a BF 4 -free electrolyte. In this paper, we study bulk CO electro-oxidation on a rotating Pt disk electrode in perchloric and sulfuric acid containing different amounts of HBF4 as well as in 0.5 M HBF4 solution to test the above described hypothesis.
2. Experimental The working electrode (WE) was a polycrystalline platinum (99.999%) disk electrode embedded in a Teflon block. The home-built electrode system was attached to a Pine instrument rotator. The electrode had a geometrical surface area of 0.1975 cm2 and a roughness factor of about 1.65. Note that below we normalize the current density to the geometrical surface area. Prior to each experiment the electrode was pretreated in the following way: three times in a row, it was first immersed in a 1:1 mixture of H2SO4 (96%) and H2O2 (30%) for 5 min, and then in Millipore water for 5 min. Both steps were done in an ultrasonic bath. Thereafter, the electrode was thoroughly rinsed with water and transferred to the electrochemical cell with a water droplet on the electrode surface. It was immersed in the electrolyte under potential control close to 0 V RHE. The counter electrode was a Pt wire bent to a ring and was placed symmetrically to the WE in the main compartment of the cell (for details see [12]). The reference electrode (RE) was a sat. Hg/ Hg2SO4 electrode and kept in a separate compartment. All potential values below are given with respect to this reference electrode. All water used in the experiments was obtained from a Millipore MilliQ system. The electrolytes were prepared from HClO4 (70%, Suprapure Merck) or H2SO4
711
(96%, Suprapure Merck), and HBF4 (50%, Aldrich, ÔPurumÕ). The glass cell as well as all other glass items used were cleaned in H2SO4 + H2O2 1:1 solution. The potentiostat employed was built by the Electronic Laboratory of the Fritz-Haber-Institut der MPG, Berlin, Germany and worked in parallel with a Sun Solaris based Real Time data acquisition system. Prior to the measurements with CO, the quality of the WE was checked by recording a cyclic voltammogram (CV) in an Ar (Linde, 99.999%) saturated electrolyte. Therefore, the electrolyte was purged with Ar for at least 30 min, during which the potential was scanned between 660 and 600 mV (vs. sat. Hg/Hg2SO4). After switching to CO (Air Liquide, 99.99%) bubbling, the electrode was cycled for approximately 25 min between 640 and 400 mV to guarantee saturation of the electrolyte with CO before CVs were recorded. Also in the further course of the experiments CO was continuously bubbled through the electrolyte. In all experiments shown the WE was rotated with 1200 rpm.
3. Results Fig. 1 shows asymptotic cyclic voltammograms of Pt in CO-saturated 0.5 M H2SO4 containing different amounts of BF 4 ions. The solid curve was obtained in pure supporting electrolyte, the dashed, dotted and dot-dashed curves in electrolytes containing 5, 10 and 20 mM HBF4, respectively. The CVs are all qualitatively similar and agree with those from the literature. During the positive scan, they exhibit a low current, COpoisoned state up to approximately 250 mV, at which the onset of the oxidation of the CO monolayer manifests itself in an ignition peak. For more positive
Fig. 1. Cyclic voltammograms of a rotating Pt disk electrode in CO saturated 0.5 M H2SO4 solution containing 0 (solid curve), 5 (dashed curve), 10 (dotted curve) and 20 (dash-dotted curve) mM HBF4; scan rate 50 mV/s; rotation rate 1200 rpm.
712
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
potentials the current density takes on a diffusion limited plateau. On the reverse scan, the system stays in this reactive, diffusion limited state also beyond the ignition peak. The CVs thus exhibit a hysteresis between a reactive and a poisoned state in a certain potential interval negative to the ignition peak. The quantitative potential value, at which the transition to the CO-covered lowcurrent state occurs depends on the BF 4 -concentration: the higher the BF 4 -concentration, the more positive is the transition from the reactive branch to the poisoned branch on the negative scan. With other words, the hysteresis becomes smaller and is shifted toward more positive potentials with increasing HBF4 concentration. The same observation was made when using 0.5 M HClO4 as base electrolyte: in the presence of HBF4 the transition from the reactive to the unreactive state occurred at more positive potentials compared to the one in pure base electrolyte. As an extreme case, we also recorded CVs in 0.5 M HBF4. Here, the ignition peak was shifted positive by about 100 mV, and on the negative scan the current decreased from the positive turning point (0.4 V) on such that the hysteresis changed its direction of circulation, the oxidation of CO being inhibited in a potential region which is considerably extended towards positive values. The cyclic voltammograms of Fig. 1 were obtained at a scan rate of 50 mV/s. To obtain a quasi-stationary current–voltage characteristic, we also investigated the current response at a scan rate of 1 mV/s for solutions containing 10 mM HBF4. Under these conditions, the width of the hysteresis was smaller than at a higher scan rate. We also note that it depended on the anodic turning point. The most remarkable difference to the CVs at higher scan rate was however, that the current exhibited oscillatory behavior in a small potential region around the ignition point. The oscillations sustained when fixing the applied voltage within a potential window of a few mV positive to the ignition potential. An example of the onset of oscillations in 0.5 M H2SO4 supporting electrolyte is shown in Fig. 2(a). Fig. 2(b) displays a portion of the current time series after transients decayed. The oscillations are fairly regular and periodic. They disappeared when the rotation of the electrode was stopped. Also when using 0.5 M HClO4 as supporting electrolyte, current oscillations developed in a small potential window just positive to the ignition potential. However, in this case, the oscillations were never regular. Instead, oscillations developed, which had a characteristic base period but apparently randomly varying amplitudes and a higher frequency, noise-like signal superimposed (Fig. 3). These irregularities occurred only in the oscillatory region, strongly suggesting that they are intrinsic to the dynamics rather than caused by instrumental or other types of ÔexternalÕ noise. When repeating the experiments shown in Figs. 2(b) and 3(b) in 0.5 M HBF4, sustained oscillations could
Fig. 2. Current oscillation during bulk CO oxidation on Pt in 0.5 M H2SO4 containing 10 mM HBF4. (a) Current–time (solid curve) and voltage time (dashed curve) plots showing the onset of oscillations after sweeping the voltage to 270 mV with 10 mV/s. (b) Current time series at U = 270 mV after the decay of transients. Rotation rate: 1200 rpm.
not be obtained. Instead, as can be seen in Fig. 4, a series of Ôignition peaksÕ with strongly decreasing amplitude was obtained.
4. Discussion The CVs displayed in Fig. 1 clearly show that the potential window in which CO is oxidized is shifted towards more positive potentials with increasing concentration. The simplest explanation of this observation seems to be that BF 4 ions, or species which result from the hydrolysis of BF 4 , adsorb at the electrode thereby occupying surface sites but not reacting with CO, as originally suggested in [9,10]. In other words, the results point to an inhibition of CO oxidation by BF 4 adsorption. To substantiate this conjecture, we
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
713
augmented the mean-field model for CO oxidation on Pt introduced in [13] by an equation describing the ad- and desorption of an unreactive anion, which competes with CO and OH for free Pt sites. The model is based on the following reaction steps COb ! COs
ðaÞ
COs þ ! COad H2 O þ $ OHad þ Hþ þ e
ðbÞ ðcÞ
COad þ OHad ! CO2 þ Hþ þ e þ 2
ðdÞ
X þ $ Xad
ðeÞ
describing the transport of CO from the bulk solution to the electrode surface (a), the adsorption of CO on a free site of the electrode, denoted by * (b), the oxidative chemisorption of water on the electrode (c), the reaction of adsorbed CO and adsorbed OH species (d) and the ad- and desorption of anions, X, on free electrode sites (e). This reaction mechanism can be cast into a set of four ordinary differential equations determining the temporal evolution of the CO concentration in front of the electrode, cs, and the coverages of the electrode with CO, hCO, OH, hOH and anions, hX, respectively
Fig. 3. Current oscillation during bulk CO oxidation on Pt in 0.5 M HClO4 containing 10 mM HBF4. (a) Current–time (solid curve) and voltage time (dashed curve) plots showing the onset of oscillations after sweeping the voltage to 261 mV with 10 mV/s. (b) Current time series at U = 261 mV after the decay of transients. Rotation rate: 1200 rpm.
Fig. 4. Current–time and voltage time plots of CO oxidation on a rotating Pt electrode in 0.5 M HBF4. Rotation rate 1200 rpm.
dcs 2S tot 2D ¼ mCO;ads þ 2 ðcb cs Þ; dt d d dhCO ¼ mCO;ads mreac ; dt dhOH ¼ mOH;ads mreac mOH;des ; dt dhX ¼ mX;ads mX;des ; dt
ð1Þ ð2Þ ð3Þ ð4Þ
where Stot denotes the number of surface sites per unit electrode area, d is the thickness of NernstÕs diffusion layer, and D the diffusion constant of CO in aqueous solution. The various rates of adsorption, mi,ads, desorption, mi,des, and reaction, mreac, are given by mCO;ads ¼ k CO;ads cs ð0.99 hCO hOH hX Þ; aFU mOH;ads ¼ k OH;ads exp RT
ð5Þ
ðhm OH ð1 hCO hX Þ hOH Þ; aFU mreac ¼ k reac exp hOH hCO ; RT ð1 aÞFU mOH;des ¼ k OH;des exp hOH ; RT nX aFU mX;ads ¼ cX k X;ads exp RT
ð6Þ
mX;des
ð0.99 hCO hX hOH Þ; nX ð1 aÞFU ¼ k X;des exp hX . RT
ð7Þ ð8Þ
ð9Þ ð10Þ
714
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
Here, the kis denote the rate constants of adsorption, desorption and reaction, respectively, U is the applied voltage (which is for the high electrolyte concentrations of the experiment equal to the electrode potential), and R, F, a and T have their usual meanings, cX is the anion bulk concentration, and nX the fractional number of electrons transferred upon ad- and desorption of anions. As in [13], a maximum coverage of OHad, hm OH , was assumed to avoid a complete blocking of CO adsorption by OHad at high potential, and the CO and anion coverages were limited to 0.99 to allow for some formation of OHad on the CO-poisoned state at low potentials such that the reaction can ignite spontaneously. For all reaction steps involving charge transfer, a Butler–Volmertype expression was used and the total Faradaic current density flowing is iF ¼ FS tot ½mOH;ads þ mreac þ nX mX;ads mOH;des nX mX;des : ð11Þ The values of the parameters used in the calculations are given in Table 1. Calculated cyclic voltammograms for four different concentrations of anions are presented in Fig. 5. They exhibit a hysteresis between a CO-poisoned and a reactive state, as typical for bulk CO electro-oxidation. Moreover, the transition from the reactive to the poisoned state during a negative scan occurs the more positive the higher the anion concentration is, while the onset potential of CO oxidation during the positive scan is independent of the anion concentration. This is exactly what was also observed in the experiments. Thus, the experimental observations at high scan rate can be explained assuming that BF 4 anions competitively adsorb on the electrode thereby blocking free surface sites for OHad formation and CO adsorption. Also the experiments at fixed applied voltage are reproduced by the model. For an outer voltage close to the transition from the poisoned to the reactive state,
Fig. 5. Calculated cyclic voltammograms for different anion concentrations cX. cX = 0 mM (solid line), 5 mM (dashed line), 10 mM (dotted line), 20 mM (dot-dashed line). Scan rate: 50 mV/s (other parameters see Table 1).
the calculated current density exhibits sustained oscillations (Fig. 6). This result gives further evidence that BF 4 anions or species deriving from their hydrolysis, such as BF3OH, competitively adsorb at the electrode and block surface sites for CO oxidation rather than supplying a source for a reactive Ôlow potentialÕ oxygen species as originally conjectured in [9,10]. Hence, the present study suggests that the reported enhancement in the formic acid oxidation rate by BF 4 , is caused by a more intricate and so far unidentified interaction between reactants, reaction intermediates and anions. There are further implications of the proposed reaction mechanism. Reaction step (e) does not take into account the nature of the adsorbing anion. Hence, any anion which competes with oxidative chemisorption of
Table 1 Values of the parameters in Eqs. (1)–(11) used in the calculations Parameters cb D d Stot kx,ads kx,des nX kOH,ads kOH,des kCO,ads kr hm OH a T
Value 3
10 5 · 105 0.0019 2.2 · 109 500 5 0.2 104 105 108 105 0.333 0.5 300
Unit mol l1 cm2 s1 cm mol cm2 cm3 mol1 s1 s1 s1 s1 cm3 mol1 s1 s1
K
Fig. 6. Calculated current timeseries for U = 0.858 V and cX = 20 mM (other parameters see Table 1).
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
water for free surface sites should induce oscillations in some concentration window. In fact, when replacing BF 4 by Cl , we obtained current oscillations in a potential window just positive to the ignition potential, giving evidence that there is a wide spectrum of physically relevant situations in which sustained oscillations during CO electro-oxidation exist. The fact that the reaction mechanism (a)–(e) can lead to oscillatory behavior has also an interesting aspect concerning nonlinear phenomena in electrochemical systems in general. The reaction mechanism takes into account only chemical steps, the electrode potential being reduced to a parameter. In all other models of electrochemical oscillations, we are aware of, the electrode potential is an essential variable, i.e., when it is not free to vary, such as under potential control in concentrated base electrolytes, the systems do not exhibit oscillations. Let us discuss how the oscillations can then be understood. The Langmuir-Hinshelwood mechanism of CO electro-oxidation (Reactions (b)–(d) coupled to slow mass transport of CO (a) provides a positive (autocatalytic) feedback loop, giving rise to an S-shaped current–potential curve [13,14]. It thus falls into the category of electrochemical systems possessing a negative differential resistance (NDR) of the S-type, in short it is classified as an S-NDR system [15,16]. In all S-NDR systems, the autocatalysis is of chemical nature, while the electrode potential tends to suppress or inhibit the autocatalytic growth. In the language of nonlinear dynamics the electrode potential is an inhibitor, and an S-NDR system can be classified as an activator– inhibitor system. Activator–inhibitor systems exhibit oscillations whenever the characteristic time scale of the activator is faster than the one of the inhibitor. In S-NDR systems, however, characteristic changes of the electrode potential are much faster than of the chemical species comprising the autocatalytic loop. For this reason, S-NDR systems are very unlikely to exhibit (homogeneous) oscillations [15,16]. However, they were found to undergo spatial instabilities leading to Turing structures or spatial domains [17–20]. In addition, the instabilities will only be found for nonnegligible IR-drop or under galvanostatic conditions. In our experiments, the oscillations were found in 0.5 M acidic solutions under potential control, i.e., under conditions, under which the ohmic potential drop through the electrolyte is minimal. Together with the fact that in the simulations the oscillatory instability was obtained with the electrode potential being a parameter, we can exclude that the oscillations are of the S-NDR-type. An analysis of the reaction mechanism (a)–(e), on the other hand, reveals that the anion coverage acts like an inhibitor, and the oscillations arise owing to the interplay of the positive feedback loop stemming from the Langmuir-Hinshelwood mechanism coupled to the mass transfer limited step and the nega-
715
tive feedback loop generated by anion ad- and desorption. Thus, the system can be characterized as an ÔS-NDR system with chemical inhibitionÕ or as a strictly potentiostatic oscillator, in which the electrolyte resistance does not play any role for the oscillations. This class of electrochemical oscillators had been speculated to exist, however, so far an unambiguous experimental identification was missing. There are two other reports of oscillations during CO electro-oxidation on Pt. Azevedo et al. [21] observed current oscillations in 0.1 M HClO4 on Pt that occurred in a small potential window positive to the ignition peak, which coincides with the voltage range in which the above described oscillations occurred. The authors speculate that the oscillations arise owing to a hidden negative differential resistance, without however presenting convincing arguments for this interpretation. On the other hand, in our experiments we never observed oscillatory behavior when using suprapure HClO4 as supporting electrolyte without addition of adsorbing anions. Thus, the simplest interpretation of the origin of the oscillations reported in [21] seems to be that there were traces of foreign anions in the perchloric acid electrolyte, such as chloride ions, and the oscillations are also caused by the competitive ad- and desorption of these anions. We emphasize, however, that as long as the specification of the acid used is unknown this interpretation remains very speculative. Koper et al. [13] observed potential oscillations under galvanostatic conditions in perchloric acid and in sulfuric acid on Pt single crystal electrodes, again without addition of inhibiting anions. The above described homogeneous model does not possess oscillatory solutions under galvanostatic conditions, such that the origin of these oscillations appears to be different from the present ones, and it remains unclear whether there is any relation between the oscillations observed here and in [13]. However, it is striking that in our work as well as in their work the oscillations are irregular in HClO4 base electrolyte while regular oscillations are reported for Pt(1 1 1) electrodes in H2SO4 solutions [13] and were found in our studies when using H2SO4 as base electrolyte. Irregular oscillations might be induced by spatial instabilities. We are currently investigating whether this explanation applies here.
5. Conclusions Using a simple reaction mechanism, we demonstrated that oscillations during CO electrooxidation on rotating Pt electrodes in concentrated base electrolytes, which were observed in our experiments in the presence of tetrafluoroborate anions, may solely be induced by surface electrochemical steps, i.e., the electrode kinetics. An interaction between electrode processes and the IR drop
716
S. Malkhandi et al. / Electrochemistry Communications 7 (2005) 710–716
through the electrolyte which generates the positive feedback loop in all other electrochemical oscillators we are aware of is not involved in the instability. Hence, this is a long sought-after experimental example of a strictly potentiostatic electrochemical oscillator. This finding expands not only the experimentally verified categories of electrochemical oscillators but makes it also likely that spatio-temporal patterns of the reaction-diffusion type, as opposed to reaction-migration patterns that form in Ôtraditional electrochemical oscillating systemsÕ, may establish on electrode surfaces. Concerning fundamental aspects of CO electrocatalysis, the observed dynamic instabilities underline that it is necessary to incorporate the possibility for cooperative phenomena when analyzing experimental data, and especially when trying to bridge the gap between microscopic measurements and mean field behavior.
Acknowledgment This work was supported by the Deutsche Forschungsgemeinschaft (DFG) through a Research Grant.
References [1] T. Iwasita, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells, vol. 2, Wiley, Chichester, 2003, Chapter 41.
[2] N.M. Markovic, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handbook of Fuel Cells, vol. 2, Wiley, Chichester, 2003, Chapter 26. [3] J.M. Weaver, S. Zhou, in: R.J.H. Clark, R.E. Hester (Eds.), Infrared and Raman Spectroscopy, vol. 26, Wiley, 1998, Chapter 5. [4] M.T.M. Koper, N.P. Lebedeva, C.G.M. Hermse, Faraday Discuss. 121 (2002) 301. [5] B. Beden, C. Lamy, N.R. DeTacconi, A.J. Arvia, Electrochim. Acta 35 (1990) 691. [6] N.M. Markovic, C.A. Lucas, B.N. Grgur, P.N. Ross, J. Phys. Chem. B 103 (1999) 9616. [7] N.M. Markovic, B.N. Grgur, C.A. Lucas, P.N. Ross, J. Phys. Chem. B 103 (1999) 8568. [8] M.T.M. Koper, Surf. Sci. 548 (2004) 1. [9] S.L. Chen, D. Lee, M. Schell, Electrochim. Acta 46 (2001) 3481. [10] S.L. Chen, D. Lee, M. Schell, Electrochem. Commun. 3 (2001) 81. [11] B.E.K. Swamy, C. Vannoy, J. Maye, M. Schell, Electrochem. Commun. 6 (2004) 1032. [12] P. Grauel, H. Varela, K. Krischer, Faraday Discuss. 120 (2002) 165. [13] M.T.M. Koper, T.J. Schmidt, N.M. Markovic, P.N. Ross, J. Phys. Chem. B 105 (2001) 8381. [14] P. Strasser, M. Eiswirth, G. Ertl, J. Chem. Phys. 107 (1997) 991. [15] K. Krischer, J. Electroanal. Chem. 501 (2001) 1. [16] K. Krischer, in: D.M. Kolb, R.C. Alkire (Eds.), Advances in Electrochemical Sciences and Engineering, vol. 8, Wiley–VCH, 2003, p. 89. [17] N. Mazouz, K. Krischer, J. Phys. Chem. B 104 (2000) 6081. [18] Y.J. Li, J. Oslonovitch, N. Mazouz, F. Plenge, K. Krischer, G. Ertl, Science 291 (2001) 2395. [19] A. Bonnefont, H. Varela, K. Krischer, Chem. Phys. Chem. 4 (2003) 1260. [20] A. Bonnefont, H. Varela, K. Krischer, J. Phys. Chem. B 109 (2005) 3408. [21] D.C. Azevedo, A.L.N. Pinheiro, E.R. Gonzales, Electrochem. Solid-State Lett. 5 (2002) A51.