The impact of spectator species on the interaction of H 2 O 2 with platinum – implications for the oxygen reduction reaction pathways

1. Concentration decay in chloride-containing solution Figure S1 shows the concentration of H2O2 vs. time (open circles) in an electrolysis experiment performed at +0.75 VRHE in a 0.1 M HClO4 + 1 10 M NaCl + 2 10 M H2O2 solution. The methodology followed in this experiment was described previously. The solid curve shows the expected concentration decay assuming a mass-transport-limited transfer of two electrons, following Levich and Faraday equations (see ref. 1 for more details). We previously showed that in a chloride-“free” 0.1 M HClO4 solution, the theoretical prediction and the experimental data coincide at any potential between +0.2 VRHE and +1.5 VRHE, which led to the conclusion that the total rate of H2O2 decomposition (reduction rate + oxidation rate) is limited by the transport of H2O2. The observed deviation between the experimental data and the expected decay upon addition of chloride ions in the electrolyte, as shown in Figure S1, highlights the inhibition of hydrogen peroxide reduction due to the chloride adsorption at this potential.


Introduction
9][10][11][12][13] In particular, the role of hydrogen peroxide in the reaction mechanism is still controversial.Apart from its fundamental importance, the understanding of the participation of H 2 O 2 in the ORR is also significant from a technological point of view, since H 2 O 2 is an undesirable product in low-temperature H 2 /air fuel cells having a detrimental impact on the fuel cell durability. 14he reduction of oxygen to water involves four electron transfer steps (coupled or decoupled with proton transfer) 15 and one O-O bond breaking step.Even though these elementary steps cannot occur simultaneously, 16 they may take place very fast one after the other, which makes the distinction between them rather difficult.When an oxygen molecule is adsorbed on a metal surface, the cleavage of the O-O bond may occur prior to the first, second or third electron transfer step.The position of the bond-breaking step in the reaction sequence is typically used as the criterion to distinguish between the three proposed mechanisms of the ORR (visualized in Scheme 1): [17][18][19] (i) the dissociative mechanism, in which the O-O bond breaking precedes all electron transfer steps (ii) the 1st associative mechanism, where one electron transfer step precedes the bond breaking and (iii) the 2nd associative (or ''peroxo''-) mechanism, in which two electron transfer steps occur forming hydrogen peroxide before the bond breaks.
0][21] In addition, the experimental exploration of the ORR mechanisms, for instance by in situ spectroscopic techniques, is not an easy task. 9he introduction of the rotating ring-disc electrode (RRDE) by Frumkin and Nekrasov 22 enabled the probing of the amount of H 2 O 2 that is formed during ORR and escapes from the vicinity of the working disc electrode, leading to a phenomenological distinction between two pathways (Scheme 2): [23][24][25] (i) the ''direct'' (or ''4-electron'') pathway, in which O 2 is reduced to water without the detection of H 2 O 2 on the ring of the RRDE, and (ii) the ''series'' (or ''2-electron'') pathway, in which H 2 O 2 is detected as an intermediate of the ORR.
Due to its convenience, this phenomenological distinction between pathways is widely used nowadays.However, it must be kept in mind that this offers only a macroscopic description of the overall reaction, thus obscuring details of the reaction mechanism at the microscopic level.In particular, the detection of H 2 O 2 is a proof that the 2nd associative mechanism is operative to a certain extent, as it is the only one that involves H 2 O 2 formation.However, a lack of detection of H 2 O 2 does not support or exclude any of the three mechanisms, as H 2 O 2 may be formed at the interface but dissociate before it is transported away from the diffusion layer.
6][37][38][39][40][41][42] These studies indicate that under these conditions, the ORR proceeds (at least in part) through the formation and desorption of H 2 O 2 , a fraction of the latter being re-adsorbed and eventually further reduced to water.Proving or disproving that the desorption-re-adsorptiondissociation scheme 36,38 is universal for the ORR would certainly be a major milestone for the understanding of the ORR.
In order to elucidate the participation of H 2 O 2 in the ORR, it is essential to understand the electrochemistry of H 2 O 2 under conditions relevant to the ORR.However, the current understanding of the interaction of H 2 O 2 with metal surfaces which is used to interpret observations during the ORR has evolved from old, poorly reproducible and eventually inconclusive data, as summarized in ref. 43.4][45][46] We previously showed that the kinetics of the hydrogen peroxide reduction and oxidation reactions (PROR) on Pt(poly) are very facile in electrolytes free from strongly adsorbing species. 43,44Depending on the electrode potential, either the hydrogen peroxide reduction reaction (PRR) to H 2 O or the hydrogen peroxide oxidation reaction (POR) to O 2 occur very fast upon interaction with the Pt(poly).In such electrolytes, the total rate of H 2 O 2 decomposition (given by the sum of the POR and PRR rates) is always controlled by mass transport of H 2 O 2 due to the fast kinetics of both reactions on Pt(poly).In the following, we demonstrate how strongly adsorbed anionic spectator species can change the picture by influencing the interaction of hydrogen peroxide with Pt(poly).Finally, we correlate the findings of this paper on the H 2 O 2 electrochemistry with previous RRDE measurements on the impact of such adsorbates on the ORR pathways.

Electrochemical measurements
The electrochemical measurements were conducted in a Teflon three-compartment electrochemical cell using a rotating disc electrode (RDE) setup with a Gamry Reference 600 potentiostat and rotation control (Radiometer Analytical).The volume of the electrolyte was 50 mL and the experiments were performed at room temperature.The potentiostat, the rotator and the gas flow were automatically controlled and programmed using an in-housebuilt LabVIEW software. 47The working electrode was a polycrystalline platinum disc of 5 mm diameter.The disc was embedded in a Teflon cylinder and polished with a silica suspension (final polishing suspension: 0.1 mm, Buehler; polishing cloth: Struers, MD Chem) prior to the measurements.A graphite rod and a saturated Ag/AgCl electrode (Metrohm), each separated by a This journal is c the Owner Societies 2013 Nafion membrane from the main cell compartment, served as the counter and the reference electrodes, respectively.All potentials in the paper are expressed with respect to the reversible hydrogen electrode (RHE) potential, which was determined versus the potential of the Ag/AgCl electrode in a hydrogen-saturated solution before each measurement.Positive feedback was used to compensate for the electrolyte resistance and the residual uncompensated resistance was less than 2 O in all experiments.The solutions were freshly prepared using ultrapure water (18 MO, TOC o 3 ppb, ELGA) and superpure chemicals (Merck, Suprapur s ).The gases were provided by Air Liquide (class: 5.0 N).

Computational details
For the quantum chemical investigations, a cluster ansatz was applied to model a Pt(111) surface using the bottom plane of a Pt 37 cluster of hemispherical shape. 48Former investigations showed that small clusters with at least three layers of Pt and a diameter of approx. 1 Å yield results sufficiently close to those of slab calculations. 19,49The distance between the platinum atoms was the bulk distance of 2.78 Å. 19 Furthermore, a ring of 12 water molecules was placed over the platinum atoms on the perimeter of the bottom plane to minimize the influence of edges as adsorption sites and to model the local environment of a water-covered surface.Seven Pt atoms are left free to adsorb H 2 O 2 or chlorine atoms.Different surface coverages with spectator species were simulated by adsorbing up to three chlorine atoms at the surface.
The calculations were carried out using the Orca program package 50 at the RI-BP86/Def2-TZVP level of theory, [51][52][53] including the Stuttgart pseudo potential (60 electrons) for platinum. 54,55ore details for the quantum chemical calculations are available in the ESI.†

Results
Influence of (bi)sulfate adsorption The corresponding voltammograms in 0.1 M HClO 4 are also depicted for comparison (dotted curves).7][58] On the basis of these results, the current responses recorded for Pt(poly) in different solutions (Fig. 1a) can be interpreted, which will be useful for the following analysis of the ORR and PROR responses.In particular, the background response in sulfuric acid shows sharper peaks in the hydrogen adsorption region as compared to perchloric acid.This is a result of the adsorption-desorption of (bi)sulfate ions and the concomitant desorption-adsorption of hydrogen. 29oreover, the area under the ''oxide'' reduction peak decreases and the onset potential for the adsorption of oxygenated species shifts more positively upon addition of sulfuric acid, both indicating the competition between the adsorption of oxygenated species and (bi)sulfate ions. 57,58t is well known that the adsorption of (bi)sulfate ions causes an inhibition of the ORR on all low-index Pt single-crystal surfaces. 29is can also be seen in Fig. 1b, where the onset potential for the ORR shifts towards more negative potentials as the concentration of H 2 SO 4 increases, while the potential region at which the current is controlled by both kinetics and mass transport (often called ''mixed'' region) is extended by approximately 0.2 V compared to 0.1 M HClO 4 .The kinetic current density for the ORR at +0.9 V RHE (SA 0.9V ) decreases in the order: 0.1 M HClO 4 (2.0 mA cm À2 ) > 0.05 M H 2 SO 4 (0.44 mA cm À2 ) > 0.1 M H 2 SO 4 (0.34 mA cm À2 ) > 0.5 M H 2 SO 4 (0.19 mA cm À2 ).Namely, the SA 0.9V decreases by a factor of 4 from 0.1 M HClO 4 to 0.05 M H 2 SO 4 , and by a factor of 2.5 when increasing the H 2 SO 4 concentration from 0.05 M to 0.5 M. Note that the lower diffusion-limited current density in the 0.5 M H 2 SO 4 solution compared to 0.1 M HClO 4 is due to the decreased solubility of O 2 in the former solution, while the difference in the viscosity (and thus the diffusion coefficient) is negligible.
The current for the H 2 O 2 reduction is controlled by mass transport of H 2 O 2 only up to +0.6 V RHE in all concentrations of sulfuric acid studied; this is around 0.25 V more negative compared to 0.1 M HClO 4 (Fig. 1c).Above +0.6 V RHE the current in H 2 SO 4 deviates from that in HClO 4 leading to a broader ''transition region'' (as previously 43,44 defined to describe the potential region between the diffusion-limited PRR and POR regions).This deviation is more pronounced for the more concentrated H 2 SO 4 solutions.The sharp transition from negative to positive currents in the HClO 4 electrolyte above +0.85V RHE is due to the competition between the facile oxidation and reduction of H 2 O 2 . 43The observed deviation of the response in H 2 SO 4 compared to that in HClO 4 at potentials between +0.6 V RHE and +0.85 V RHE is related to the adsorption of (bi)sulfate ions and the consequent inhibition of the dissociative adsorption and reduction of H 2 O 2 at potentials below the ''oxide'' formation.As the potential increases, adsorbed (bi)sulfate ions are gradually displaced by adsorbed oxygenated species, 58 and the consequent decrease of the (bi)sulfate coverage allows the POR to occur eventually as fast as in HClO 4 .
If one takes into account the very high H 2 SO 4 concentrations that were used, the influence of (bi)sulfate adsorption on the H 2 O 2 reactions cannot be considered as very strong.Additionally, the increase of the concentration of H 2 SO 4 by one order of magnitude does not significantly alter the current response in Fig. 1c, which means that the inhibition of H 2 O 2 reactions due to (bi)sulfate adsorption cannot become severely stronger.Therefore, it can be concluded that even though (bi)sulfate adsorption causes some inhibition of the PRR in the potential region between +0.6 V RHE and +0.85 V RHE , the reaction can still proceed with a nearly diffusion-limited rate.It is worth mentioning that this is not the case for the ORR, where (bi)sulfate adsorption has a much more pronounced influence (see Fig. 1b).This difference in the strength of the effect of (bi)sulfate on the ORR and the PRR will be discussed below.2a shift towards more negative potentials and become sharper with increasing chloride concentrations.0][61] Increasing the potential above the potential of zero total charge during the positivegoing sweep increases the chloride coverage, 62 which in turn suppresses the adsorption of oxygenated species.Only at much more positive potentials Cl ad becomes replaced by oxygenated species. 33,61,63The suppression of the ''oxide'' formation due to chloride adsorption is also responsible for the decrease of the charge associated with the reduction of the irreversible ''oxide'', as the chloride concentration increases (Fig. 2a). 63hlorides are thus adsorbed on the Pt(poly) surface in the whole potential region between the hydrogen adsorption and the onset of adsorption of oxygenated species, and cause the inhibition of the ORR, as it is evidenced by the considerable shift of the onset potential for the ORR to higher overpotentials with the increase of the concentration of NaCl (Fig. 2b). 33,64In particular, the addition of 1 Â 10 À3 M Cl À in the 0.1 M HClO 4 solution leads to a decrease by two orders of magnitude of the SA 0.9V .Mass transport control of the reaction rate is achieved at much more negative potentials than in the chloride-''free'' electrolyte.For instance, in the solution containing 1 Â 10 À3 M Cl À the diffusion-limited current is reached only below +0.3 V RHE .

Influence of halide adsorption
Even though the presence of 1 Â 10 À5 M Cl À is already enough to influence the ORR, neither the PRR nor the POR is significantly affected by this concentration (Fig. 2c).However, higher chloride concentrations can severely inhibit the PRR: in a 1 Â 10 À4 M Cl À solution, the diffusion-limited current can be reached only in the hydrogen adsorption region, i.e. at potentials below +0.3 V RHE .Above +0.3 V RHE , the Cl ad coverage increases -as explained above -so the PRR is strongly suppressed and the negative current decreases, deviating from the diffusion-limited value, i.e. from the response in the chloridefree electrolyte.The deviation becomes more pronounced for the solutions containing higher chloride concentrations.A constant-potential electrolysis experiment (+0.75 V RHE ) in a 0.1 M HClO 4 + 1 Â 10 À3 M NaCl + 2 Â 10 À3 M H 2 O 2 electrolyte confirmed that the H 2 O 2 concentration decay deviates significantly from that expected for a diffusion-limited reaction (see Fig. S1, ESI †).Thus, in contrast to a chloride-''free'' solution where the experimentally observed concentration decay is limited by the mass transport of H 2 O 2 in the whole potential region studied, 43 the decrease of the negative current at potentials above +0.3V RHE is due to the inhibition of the PRR.
As the potential becomes more positive, the chloride coverage decreases due to the displacement of Cl ad by oxygenated species, therefore the inhibition of the H 2 O 2 interaction with the surface weakens, similarly to what was described above for (bi)sulfate.The positive current increases rapidly and becomes eventually controlled by H 2 O 2 diffusion at around +1.2 V RHE for the 1 Â 10 À3 M Cl À solution (Fig. 2c).A more positive potential is required to reach the diffusion-limited rate of the POR for higher concentrations of chloride in solution, as the latter leads to enhanced Cl ad coverage up to that potential, in agreement with the positively shifted adsorption of oxygenated species in Fig. 2a.
Fig. 3 shows the hydrodynamic voltammograms recorded in 0.1 M HClO 4 solutions containing 1 Â 10 À4 M of NaX, where X: F À , Cl À or Br À .The influence of the halides on the background responses (Fig. 3a) becomes stronger following the order F À o Cl À o Br À (Fig. 3a), reflecting the increasing halide coverage on platinum for identical concentrations. 65In fact, a 10 À4 M concentration of fluoride in the HClO 4 solution is low enough to have practically no effect on the background responses.This is because the adsorption strength of fluoride and perchlorate ions, and thus their influence on the adsorption of hydrogen and of oxygenated species, is about the same. 60,65In contrast, the same concentration of chloride and bromide significantly suppresses both processes: (i) the adsorption of oxygenated species, as can be seen by the positive shift of the onset potential in the positive-going sweep, and (ii) the hydrogen adsorption, as can be seen by the negative shift of the onset potential for H upd adsorption in the negative-going sweep.Both shifts increase with the size of the halide ions (see Fig. 3a). 61hile the oxygen reduction reaction is not influenced by the adsorption of fluoride (Fig. 3b), a considerable shift of the onset potential for the ORR is observed in the presence of chloride and bromide, resulting in a stronger inhibition of oxygen reduction.In the bromide-containing solution, the ORR commences at around +0.6 V RHE , which is approximately 0.35 V more negative than the onset potential in the presence of the same concentration of fluoride.It should be noted that the inflection point in the curve corresponding to the ORR in the bromide-containing solution, which signals the switch of the ORR from a 2-electron to a 4-electron reduction, coincides with the onset of the PRR.
The same trend with respect to the different anions is qualitatively followed for the hydrogen peroxide reactions (Fig. 3c).Again, the interaction of H 2 O 2 with platinum in the presence of fluoride is practically unaffected and comparable to that in pure HClO 4 .The significantly stronger inhibition of H 2 O 2 reactions by Cl ad has been described above, while Br ad has an even stronger effect.Indeed, the adsorption of bromide is so strong that it suppresses completely the H 2 O 2 reactions in the potential region between +0.4 V RHE and +1.1 V RHE .This implies that neither the PRR nor the POR are possible in this potential region.The POR commences only above +1.1 V RHE due to the displacement of Br ad by oxygenated species, and the current increases gradually, reaching the diffusion-limited value at +1.3 V RHE .The good agreement in all halide-containing solutions between the potentials at which the POR starts (Fig. 3c) and at which the displacement of halides by oxygenated species takes place (Fig. 3a) is an additional indication that the inhibition of POR is caused by the adsorbed halides.On the other hand, the PRR in the bromide-containing electrolyte starts only below +0.4 V RHE , which is 0.2 V more negative than the onset potential for ORR in the same solution.
It is noteworthy that, while Cl ad hinders the ORR more strongly than the PRR, Br ad influences the two-electron reduction of O 2 to a lesser extent than the further reduction of H 2 O 2 to H 2 O.A similar effect was observed in an additional experiment in the presence of iodide (0.1 M HClO 4 + 1 Â 10 À4 M NaI; see Fig. S2, ESI †): the PRR was completely suppressed in the potential region studied, while the two-electron reduction of oxygen could take place at potentials below +0.4 V RHE .This behaviour will be discussed in the Discussion section.

Electronic structure calculations
Electronic structure calculations were carried out in order to obtain trends of the atomistic details that give rise to the major influences of co-adsorbed species on the reactions discussed above.For this purpose, three model surfaces were used to resemble the local environment of a given reacting species on the surface.While these systems are far from a detailed model that includes all relevant parameters that influence reactivity, they still allow assessing whether adsorbates influence the adsorption and/or bond breaking due to local electronic or steric effects.
In a previous paper, 43 we showed that H 2 O 2 dissociates into two OH ad with a barrier of 0.22 eV, when adsorbed on a bare Pt(111) surface (an adsorption energy of À0.34 eV), modeled by a cluster.The dissociation energy is À1.55 eV (see Fig. 4a).Experimental evidence for the dissociative adsorption of H 2 O 2 on Pt has been given previously using surface-enhanced Raman spectroscopy. 66If two of the seven platinum atoms are covered by chlorine atoms, the adsorption energy of H 2 O 2 out of the gas phase remains exothermic (À0.37 eV), as shown in Fig. 4b.Taking the solvation energy of H 2 O 2 into account (literature values range from À0.3 eV to À0.6 eV as indicated by a grey bar in Fig. 4), the tendency for H 2 O 2 to adsorb is small, both on the clean Pt(111) surface and on a Pt(111) surface with two chlorine atoms adsorbed.It should be noted, however, that interactions of the adsorbed H 2 O 2 with the solvent were neglected, so this is just a lower limit for the adsorption energy.Taking solvation effects of the adsorbed H 2 O 2 into account, a further stabilization and thus a further increase of the adsorption energy must be expected.The barrier for H 2 O 2 dissociation for this relatively low coverage with chloride is 0.14 eV, which is similar to that for the chlorine-free surface (the slight difference between the two barriers is due to Cl-H hydrogen bonding, see ESI †).The dissociation is still highly exothermic, with a dissociation energy of À1.40 eV.Hence, the results of the simulations indicate that the presence of up to two chlorine atoms does not introduce major differences in the local reactivity of H 2 O 2 on the surface.
In order to estimate whether a situation exists in which steric hindrance might prevent dissociation of H 2 O 2 by O-O bond cleavage, a model for an extreme situation was chosen.In this model, three out of the seven Pt atoms of the surface were covered by chlorine atoms.Under these conditions, the distance between two chlorine atoms is 3.8 Å, which is very close to the sum of their van der Walls radii (3.5 Å), 67 thus leaving only minimal space for further adsorbates (see Fig. S3, ESI †).As a consequence, the adsorption of H 2 O 2 at a platinum atom of this model is not favorable energetically.For a structure including H 2 O 2 adsorption, the distance between the chlorine atoms increases to 5.1 Å (see Fig. S5, ESI †), while the sum of the van der Waals radii of the two chlorine atoms and the diameter of the oxygen atom (that is located between two chlorine atoms) is 6.3 Å.However, the barrier to dissociate H 2 O 2 under these circumstances is only 0.05 eV and the dissociation energy is À1.25 eV.As these values are of comparable magnitude to the results obtained on the other models, one can assume that irrespective of local spatial restrictions due to high chlorine coverage, the kinetics of H 2 O 2 dissociation on the surface should be more or less the same.While detailed and accurate data for adsorption/solvation equilibria would be required for a more complete description, the limited results from this study allow at least presuming that the inhibition of the PRR at high chlorine coverages might be due to an inhibited adsorption of H 2 O 2 , rather than due to an impeded dissociation.
Turning to the behavior of oxygen, it first needs to be noted that adsorption of an oxygen molecule is always more favorable than that of a H 2 O 2 molecule, due to the poorer stabilization of O 2 in aqueous solution.The dissociation of O 2 at a bare Pt(111) surface model (see Fig. 5a) proceeds with a barrier of +0.72 eV and a dissociation energy of À0.53 eV, in agreement with values obtained by Jacob and Goddard. 19When two of the seven Pt atoms are covered with chlorine atoms, breaking the O-O bond of oxygen becomes highly endothermic (+1.01 eV).Therefore, the dissociation of O 2 in this model system is inhibited by the local environment, in contrast to H 2 O 2 dissociation that remains unaffected in the same model, as described above.This contrast is caused by the different binding modes preferred by the products, O ad in the case of O 2 and OH ad in the case of H 2 O 2 : while OH ad preferentially binds on top of a platinum atom (like Cl ad ) and hence just one platinum atom is required, atomic oxygen preferentially adsorbs on an fcc-site, which requires three platinum atoms. 68,69For O 2 dissociation on the model system in Fig. 5b, the nearby chlorine atoms restrict the two oxygen atoms to adsorb on top sites, accompanied by a drastic increase of relative energy of this product structure (compare relative energies for the products of O 2 dissociation in Fig. 5a and b).In this context, it is interesting to note that Eikerling and co-workers calculated binding energies of atomic oxygen for different binding sites on Pt nanoparticles and reported values of approximately À0.85 eV for fcc, À0.50 eV for hcp, À0.38 eV for bridge and 0.41 eV for top sites. 48This strong decrease of oxygen binding strength from fcc to top sites indicates that the unfavorable O 2 dissociation in the presence of chlorine atoms mostly stems from the spatial This journal is c the Owner Societies 2013 restrictions imposed by the co-adsorbates, preventing the product O ad to adsorb in the preferred fcc hollow sites.

Discussion
Correlation between the PRR and the ORR In a previous paper, we showed that the kinetics of both H 2 O 2 reactions, the PRR at low potentials and the POR at high potentials, are very fast in the absence of inhibiting spectators in a highly pure, weakly adsorbing electrolyte (0.1 M HClO 4 ). 43he coverage with inhibiting spectator species is negligible in this case in the potential region between +0.2 V RHE and +1.5 V RHE .‡ As a consequence, the H 2 O 2 decomposition occurs with a total rate that is controlled only by the mass transport of H 2 O 2 in the whole potential window between +0.2 V RHE and +1.5 V RHE .This translates into an infinitesimally low concentration of H 2 O 2 at the electrode-electrolyte interface.In particular, due to the fast kinetics of the reactions of H 2 O 2 , the local concentration of hydrogen peroxide must be close to a limit imposed by thermodynamics, which equals 8.67 Â 10 À19 M. 43 The results presented in this paper show that the PRR and to a lesser extent the POR are slowed down when the coverage with inhibiting species increases.The extent of the inhibiting effect is related to the surface coverage with such spectators.The electronic structure calculations indicate that the observed inhibition of the PRR at high coverage is related to an inhibition of the adsorption of H 2 O 2 .Depending on the coverage with inhibiting spectator species, the total rate of H 2 O 2 decomposition will eventually be limited by the slow kinetics of the PRR and the POR (rather than by the mass transport of H 2 O 2 , as for the weakly adsorbing electrolytes).As a consequence, considerable local H 2 O 2 concentrations are expected at the vicinity of the electrode surface.
Based on this difference between weakly and strongly adsorbing electrolytes, we will summarize and discuss below our observations for the PRR in relation to previous RRDE investigations for the selectivity of the ORR.
Hydrogen adsorption.7][28][29] Thus, it can be safely argued that the observed transition from an effective ''4-electron'' to a ''2-electron'' reduction of oxygen under these conditions originates from the inhibition of the PRR due to adsorbed hydrogen.7][28][29] In that case, the topmost atoms remain available, 70,71 and can still carry out the PRR, so that the transition from the ''4-electron'' to a ''2-electron'' reduction of oxygen is absent.
(Bi)sulfate adsorption.(Bi)sulfate adsorbs specifically on Pt, but the surface coverage is not very high (the maximum sulfate coverage on Pt(111) is ca.0.22 ML, 72,73 and is probably not much higher on other facets).As can be observed in Fig. 1a, the negative-going sweeps of the background voltammograms in the presence of different concentrations of sulfuric acid coincide above ca.+0.95 V, suggesting that (bi)sulfate can be adsorbed on the Pt surface only below this potential.Interestingly, above ca.+0.95 V all the curves in Fig. 1c also coincide, and differences in the PROR in the presence of different concentrations of H 2 SO 4 can only be observed below that potential.This is an indication that the latter differences are in turn due to the inhibition of the PRR by adsorbed (bi)sulfate.Nevertheless, due to the low (bi)sulfate coverage, this inhibition is not significant and thus the total reaction rate remains almost diffusion-limited, resulting in low hydrogen peroxide concentrations at the interface.This also explains why typically no H 2 O 2 can be detected during ORR in sulfuric acid solutions, as has been reported previously, despite the inhibition caused to the ORR by (bi)sulfate adsorption. 29The stronger inhibition of the ORR caused by bisulfate, as compared with that of the PRR, should rather be attributed to the retardation of one of the initial steps of the ORR.
Halide adsorption.The results presented in this paper show that the addition of a small amount of chloride, bromide or iodide to the HClO 4 solution causes a much stronger inhibition of the PRR than the addition of much higher concentrations of (bi)sulfate, due to their stronger specific adsorption and, consequently, to the higher coverages reached.The inhibiting effect increases with the increase of the potential-dependent coverage.][33] Two important conclusions can be drawn from the above discussion.First, there is a clear correlation between the inhibition of the PRR and the selectivity of the ORR towards the complete reduction to H 2 O.In particular, H 2 O 2 formation during the ORR has been observed only under conditions in which the PRR is retarded, while the complete reduction of O 2 to H 2 O on Pt is only observed under conditions in which H 2 O 2 reduction to H 2 O is very fast.This consistency between previous RRDE investigations of the ORR and our observations made for the PRR suggests that the facile reduction of H 2 O 2 to H 2 O is a prerequisite for the complete reduction of O 2 to H 2 O, under the same conditions.Even though this can be so far stated only for platinum, literature data on other noble and non-noble materials, [74][75][76][77][78] as well as preliminary results that we have obtained on Au(poly) and Rh(poly), seem to confirm this concept.This strongly implies an important role of H 2 O 2 in the oxygen reduction reaction, as it will be also discussed below.
The second important conclusion is that the nature of the inhibiting effect that the spectators cause on the PRR (and on the ORR selectivity) is the same, regardless of the nature of the spectators.The effect becomes stronger for higher surface coverage, which can be reached either by increasing the strength of the adsorption (increasing from fluoride, where an effect can hardly be seen, to (bi)sulfate, chloride, bromide, and finally to iodide), or by increasing the anion concentration.The relevance of the surface coverage and the apparent insensitivity to the chemical nature of the adsorbed anions suggest that the latter act mainly as blocking species, and that the inhibition of the PRR is essentially due to an atomic-ensemble effect, 79 rather than to an alteration of the electronic structure of the Pt atoms that remain exposed to the solution (electronic effect).As shown by the theoretical calculations above, the spatial restrictions imposed by spectators on the adsorption of reactants, intermediates and products affect their corresponding adsorption energies.

Mechanistic insight into the ORR -desorption versus dissociation
The mechanistic details of the role of H 2 O 2 in the oxygen reduction, and of the observed changes in the selectivity during the ORR, are not yet fully elucidated.However, the findings of this work regarding the impact of spectators on the interaction of H 2 O 2 with platinum and the discussion made above converge to a plausible scheme for the ORR that can describe all the macroscopic observations that have been made so far for the ORR selectivity, as follows.
The electronic structure calculations presented in this paper, as well as the findings in ref. 43, show that platinum is intrinsically a good catalyst for the dissociation of H 2 O 2 .In particular, the barrier for breaking the O-O bond of adsorbed H 2 O 2 is independent of the local environment and is consistently very low at all spectator coverages.In contrast, the calculations indicate that spectators affect strongly the adsorption of H 2 O 2 , which is slightly exothermic at low to medium coverage, but becomes highly endothermic at high coverage.Once adsorbed H 2 O 2 is formed during the ORR, a competition between (spectator coverage-dependent) desorption and (spectator coverageindependent) dissociation will occur.The desorbed H 2 O 2 may experience an additional competition between re-adsorption and diffusion away from the electrode surface.
At low spectator coverage, H 2 O 2 desorption is not very favorable and it is very likely that it does not supersede the very facile H 2 O 2 dissociation.Even if some H 2 O 2 desorbs (note that this process is slightly endothermic at low spectator coverage) re-adsorption is still very probable, unless the mass transport of the produced H 2 O 2 is very fast.Therefore, at low spectator coverage, H 2 O 2 cannot accumulate on the surface and exceed the infinitesimal equilibrium concentration imposed by thermodynamics (8.67 Â 10 À19 M).
At high spectator coverage, the H 2 O 2 desorption-adsorption equilibrium changes and now desorption of H 2 O 2 is very favorable, competing strongly (or even dominating) over dissociation, while re-adsorption becomes endothermic.Thus, there is a high probability that the formed H 2 O 2 will desorb rather than dissociate and H 2 O 2 will be detected in the electrolyte as long as it diffuses away before re-adsorbing.Partial re-adsorption of the desorbed H 2 O 2 will be possible only at spectator coverages at which H 2 O 2 adsorption is only slightly endothermic.
Interestingly, the effect described above for high coverage of spectators was reported even in weakly adsorbing electrolytes in previous papers mainly by the groups of Kucernak 7 and Behm. 36The H 2 O 2 produced upon increasing mass transport rates (e.g. by the use of ultramicroelectrodes) was attributed to a higher probability for H 2 O 2 to escape prior to re-adsorption, which cannot be observed under the slow (compared to ultramicroelectrodes) mass transport conditions of an RRDE.It needs to be noted, however, that only under equilibrium conditions, the spectator coverage (at a given surface) depends on the nature and the concentration of the electrolyte ions as well as on the electrode potential.In contrast, under nonequilibrium conditions (e.g. during potentiodynamic experiments and low anionic concentration, like in the current and the previous studies cited above) the spectator coverage depends additionally on the mass transport of the inhibiting ions from the solution to the surface at a given time, and thus does not necessarily equal the equilibrium spectator coverage for the corresponding potential.Therefore, it is not only the mass transport rate of the produced H 2 O 2 away from the surface that is enhanced, e.g. for the case of ultramicroelectrodes, but also the mass transport rate of trace impurities towards the surface, as also suggested previously. 80In principle, the increase of the mass transport rates of both H 2 O 2 and chloride cannot be decoupled in these experiments, and will eventually lead to a higher coverage of chloride than the one expected for the slow mass transport conditions of an RRDE, even for the highest purity weakly adsorbing electrolyte.Under these conditions, and in agreement with the case of high spectator coverages described above, the H 2 O 2 desorption and release will therefore be stimulated at high mass transport rates of both H 2 O 2 and chloride.Obviously, the re-adsorption of H 2 O 2 prior to diffusion may still be possible as described in the desorption/re-adsorption/reaction model, 7,36 but only in case that the H 2 O 2 adsorption is still only slightly endothermic at the given spectator coverage.
Therefore, only one H 2 O 2 -mediated ORR pathway (equivalent to the ''series'' pathway) is enough to explain and unify all the observations that have been made so far for the ORR selectivity.In this scheme (Scheme 3), the decisive step on whether H 2 O 2 or H 2 O will be formed is the competition between dissociation and desorption of the produced H 2 O 2 intermediate, which is related to the surface and/or interface structure.For surfaces that are excellent catalysts for the dissociation of H 2 O 2 , as for example platinum, this competition depends strongly on the presence of adsorbates that can stimulate H 2 O 2 desorption and make it capable of competing dissociation only at high spectator coverages.Hence, the ''direct'' pathway can be eliminated as it is only a limiting case in which the dissociation is much faster compared to desorption.It has to be emphasized, however, that this is still not sufficient proof that the ORR always and solely proceeds through the 2nd associative mechanism, and hence to exclude the other two mechanisms.

Conclusions
We will summarize below the main findings of this work: Hydrodynamic voltammetry studies of the PRR on Pt(poly) in the presence of adsorbing anions show an increasingly inhibiting effect along the series ClO 4 À E F À o HSO 4 À o Cl À o Br À o I À and with the increase of the concentration, introducing kinetic limitations on the PRR rate.At concentrations of 10 À4 M, chloride hinders the ORR more than the PRR, while bromide and iodide block the PRR for a wider range of potentials than the ORR.Electronic structure calculations show that once hydrogen peroxide is adsorbed on Pt(111), it can dissociate with a low barrier at all chloride coverages.On the other hand, the adsorption of H 2 O 2 becomes strongly endothermic (and desorption highly exothermic) at high chloride coverage.
Under conditions in which the PRR is severely kinetically limited, the oxygen reduction forms hydrogen peroxide.However, when the PRR rate is limited by mass transport only, ORR proceeds with a complete 4-electron reduction.This coincidence strongly suggests that the feasibility of H 2 O 2 reduction is a decisive prerequisite for the complete reduction of oxygen to water.
The oxygen molecules that are reduced to hydrogen peroxide may either desorb or dissociate and further be reduced to water.The competition between the dissociation of H 2 O 2 and the spectator coverage-dependent desorption of H 2 O 2 is decisive for the selectivity of the ORR towards peroxide or water: at low coverage of inhibiting species, the desorptionadsorption equilibrium favors the adsorbed state while at high anion coverage, desorption of H 2 O 2 becomes exothermic and basically irreversible.
Overall, the considerations made in this work have very important implications for the selectivity of the ORR and for the change in the reaction pathways for Pt-based materials often mentioned in the literature.In fact, only a H 2 O 2 -mediated pathway that includes a competition between the dissociation and the spectator coverage-dependent desorption of the H 2 O 2 intermediate is enough to explain and unify all the observations that have been made so far on the selectivity of the ORR.The ''direct'' pathway is indeed a limiting case of the ''series'' pathway which arises when the rate of desorption of the formed H 2 O 2 is sufficiently low compared to that of the H 2 O 2 dissociation and further reduction.Previous macroscopic observations on the ORR selectivity using the RRDE, that have been interpreted as a suppression of the ORR without a change in the reaction pathway (e.g. for (bi)sulfate) or with a change in the reaction pathway (e.g. for H upd , chloride or bromide), are only phenomenological and do not originate from an intrinsic property of platinum that determines the selectivity of the reaction.Instead, such macroscopically observed changes in the reaction selectivity are related to the impact of the adsorbed spectators on the competition between H 2 O 2 desorption and dissociation.

Fig. 1
Fig. 1 shows the hydrodynamic voltammograms recorded in solutions of different concentrations of H 2 SO 4 , saturated with (a) argon (b) oxygen and (c) argon additionally containing 1 Â 10 À3 M H 2 O 2 .The corresponding voltammograms in 0.1 M HClO 4 are also depicted for comparison (dotted curves).The voltammogram features of Pt single-crystal surfaces have already been described in detail in the literature.29,[56][57][58]On the basis of these results, the current responses recorded for Pt(poly) in different solutions (Fig.1a) can be interpreted, which will be useful for the following analysis of the ORR and PROR responses.In particular, the background response in sulfuric acid shows sharper peaks in the hydrogen adsorption region as compared to perchloric acid.This is a result of the adsorption-desorption of (bi)sulfate ions and the concomitant desorption-adsorption of hydrogen.29Moreover, the area under the ''oxide'' reduction peak decreases and the onset potential for the adsorption of oxygenated species shifts more positively upon addition of sulfuric acid, both indicating the competition between the adsorption of oxygenated species and (bi)sulfate ions.57,58It is well known that the adsorption of (bi)sulfate ions causes an inhibition of the ORR on all low-index Pt single-crystal surfaces.29

Fig. 2
Fig. 2 shows the hydrodynamic voltammograms recorded in 0.1 M HClO 4 solutions with an increasing concentration of NaCl and saturated in (a) argon, (b) oxygen and (c) argon additionally containing 1 Â 10 À3 M H 2 O 2 .The background current peaks in the hydrogen adsorption-desorption region in Fig.2ashift towards more negative potentials and become sharper with increasing chloride concentrations.These changes indicate that chloride adsorption-desorption and hydrogen desorption-adsorption occur concomitantly in the potential region below +0.3 V RHE .[59][60][61]Increasing the potential above the potential of zero total charge during the positivegoing sweep increases the chloride coverage,62 which in turn suppresses the adsorption of oxygenated species.Only at much more positive potentials Cl ad becomes replaced by oxygenated species.33,61,63The suppression of the ''oxide'' formation due to chloride adsorption is also responsible for the decrease of the charge associated with the reduction of the irreversible ''oxide'', as the chloride concentration increases (Fig.2a).63Chlorides are thus adsorbed on the Pt(poly) surface in the whole potential region between the hydrogen adsorption and the onset of adsorption of oxygenated species, and cause the inhibition of the ORR, as it is evidenced by the considerable shift of the onset potential for the ORR to higher overpotentials with the increase of the concentration of NaCl (Fig.2b).33,64In particular, the addition of 1 Â 10 À3 M Cl À in the 0.1 M HClO 4

Fig. 4
Fig. 4 Energy diagram for the dissociation of H 2 O 2 on the Pt(111) surface model with (a) no, (b) two and (c) three adsorbed chlorine atoms.The energies are given relative to H 2 O 2 in the gas phase (dashed line).The grey bar corresponds to H 2 O 2 in solution (see the text).

Fig. 5
Fig. 5 Energy diagram for the dissociation of O 2 on the Pt(111) surface model with (a) no and (b) two adsorbed chlorine atoms.The energies are given relative to O 2 in the gas phase (dashed line).