Competition of CO and H 2 for Active Oxygen Species during the Preferential CO Oxidation ( PROX ) on Au / TiO 2 Catalysts

Aiming at an improved mechanistic understanding of the preferential oxidation of CO on supported Au catalysts, we have investigated the competition between CO and H2 for stable, active oxygen (Oact) species on a Au/TiO2 catalyst during the simultaneous exposure to CO and H2 with various CO/H2 ratios at 80  ̋C and 400  ̋C by quantitative temporal analysis of products (TAP) reactor measurements. It is demonstrated that, at both higher and lower temperature, the maximum amount of active oxygen removal is (i) independent of the CO/H2 ratio and (ii) identical to the amount of active oxygen removal by CO or H2 alone. Hence, under preferential CO oxidation (PROX) reaction conditions, in the simultaneous presence of CO and H2, CO and H2 compete for the same active oxygen species. In addition, also the dependency of the selectivity towards CO oxidation on the CO/H2 ratio was evaluated from these measurements. Consequences of these findings on the mechanistic understanding of the PROX reaction on Au/TiO2 will be discussed.


Introduction
Gold nanoparticles supported on a variety of metal oxides are well known for their exceptional high activity and selectivity for various oxidation and reduction reactions already at very low temperatures [1][2][3][4].One of the most prominent examples is the CO oxidation reaction on supported Au catalysts, e.g., on Au/TiO 2 , since it also serves as a test reaction for gas phase oxidation catalysis in heterogeneous Au catalysis in general [5].Practical applications of monometallic supported Au catalysts are rare, so far, but they are claimed to be promising candidates for the selective removal of CO from H 2 -rich reformates to values below 10 ppm [6][7][8] for use in low temperature Polymer Electrolyte Membrane (PEM) fuel cells [6,9].The ultra-purification of H 2 -rich fuels via the selective CO oxidation (or preferential CO oxidation-PROX) reaction requires catalysts that are highly active for the CO oxidation (Equation (1)).At the same time, the catalysts should have a very low activity for the continuous oxidation of H 2 to water (Equation (2)), although H 2 reaches concentrations up to 75% (depending on the process of H 2 generation) and is, hence, present in large excess compared to CO in the reaction gas [6,[9][10][11].One should note, however, that an oxidation of H 2 to adsorbed hydroperoxy-like species, which have previously been proposed to represent reaction intermediates in the preferential CO oxidation in the presence of H 2 , may even be beneficial for the PROX activity of supported Au catalysts [10,12].
to consider that under PROX reaction conditions the formation of hydroxyl groups and adsorbed water on the catalysts surface upon interaction with H 2 may significantly influence its CO oxidation activity [14,[37][38][39][40].
In the following we report results of TAP reactor measurements, where the absolute amount of active oxygen removal by CO and H 2 and its replenishment by O 2 was evaluated upon exposure of a Au/TiO 2 catalyst to alternate sequences of CO/H 2 /Ar and O 2 /Ar pulses, using different CO/H 2 mixtures.From these data we will derive whether CO and H 2 compete for the same active oxygen species also in the simultaneous presence of CO and H 2 .Moreover, the selectivity for CO oxidation during mixed CO/H 2 pulses can be determined.Finally we will discuss the consequences of the present finding for the mechanism of the PROX reaction on Au/TiO 2 catalysts.

Active Oxygen Removal by CO and H 2
Before focusing on the reaction with mixed CO/H 2 /Ar pulses, we determined the removal of stable, active oxygen from the Au/TiO 2 catalyst surface by CO/Ar and H 2 /Ar pulses only for calibration.The absolute amounts of active oxygen determined in these measurements will then be used as references for the active oxygen removal in the simultaneous presence of CO and H 2 .Similar to our previous study [35], the catalyst was alternately exposed to several sequences of (i) CO/Ar and O 2 /Ar or (ii) H 2 /Ar and O 2 /Ar, respectively, in order to reactively remove and replenish the available active oxygen from the catalyst surface.These measurements were performed after in situ calcination (O400) of the Au/TiO 2 catalyst and subsequent saturation of its surface by adsorbed water by exposure to 5000 H 2 /Ar pulses at 80 ˝C.Additional experiments including the exposure of the catalyst to various amounts of H 2 /Ar pulses and subsequent temperature programmed desorption (TPD) measurements revealed that (i) there is formation of stable adsorbed water upon exposure to H 2 /Ar pulses at 80 ˝C and (ii) treatment with 5000 H 2 /Ar pulses results in a surface saturation and, hence, equal surface concentrations of adsorbed water during all subsequent experiments.Note that, at 400 ˝C, in contrast, water formed during the H 2 oxidation readily desorbs, and does not need to be considered in the following measurements.
The multi-pulse experiments following this treatment (O400 and H 2 pulses) revealed an oxygen consumption of the fully-reduced catalyst during O 2 /Ar pulses at 80 ˝C of 3.3 ˆ10 18 O atoms¨g cat ´1.This is almost identical to the CO consumption/CO 2 formation in the subsequent sequence of CO/Ar pulses (3.4 ˆ10 18 molecules¨g cat ´1), as well as to the oxygen consumption during subsequent O 2 /Ar pulses (after 250 CO/Ar pulses).Hence, this (partial) reduction and re-oxidation of the catalyst surface is fully reversible under present conditions.Furthermore, in agreement with previous findings, the amount of CO 2 formation in each CO/Ar sequence always equals the amount of CO consumed, indicating that there is no significant build-up of stable adsorbed, carbon containing surface species (surface carbonates, etc.) under present reaction conditions [25,41].Finally, all CO 2 is formed during CO pulses (and not during O 2 pulses), as expected for CO weakly adsorbed on the Au NPs at 80 ˝C, and completely desorbed before O 2 is introduced [31].
The oxygen storage capacity (OSC) at 80 ˝C is slightly lower in the present study (3.3 ˆ10 18 O atoms¨g cat ´1) than determined previously for a comparable Au/TiO 2 catalyst, which was also saturated with adsorbed water by H 2 pulses prior to the pulse experiments (4.6 ˆ10 18 O atoms¨g cat ´1) [35], but of a similar order of magnitude.More important for the present study, however, is the comparison between the absolute amounts of active oxygen on (fully oxidized) Au/TiO 2 for the CO oxidation and the H 2 oxidation.From the oxygen uptake during O 2 /Ar pulses after 1500 H 2 /Ar pulses on the Au/TiO 2 catalyst at 80 ˝C we obtained a value of 3.3 ˆ10 18 O atoms¨g cat ´1 for the total amount of stable oxygen which is active for the H 2 oxidation (see Table 1).Note that compared to CO/Ar a significantly higher number of H 2 /Ar pulses is necessary to completely remove the active surface oxygen.While for a complete removal of all availabe O act species via the CO oxidation about 200 CO/Ar pulses were needed, almost 1500 H 2 /Ar pulses are necessary for its removal via the H 2 oxidation.This reflects the much lower efficiency of hydrogen for active oxygen removal compared to CO, as was already demonstrated previously [31].The total amounts of active oxygen for CO and H 2 oxidation, however, are almost identical also for the present Au/TiO 2 catalyst [31].Additionally, for reaction at 400 ˝C, the total amounts of active oxygen for CO oxidation and for H 2 oxidation are also almost identical (12.7 ˆ10 18 O atoms¨g cat ´1 and 12.6 ˆ10 18 O atoms¨g cat ´1, respectively; see Table 2).The higher amount of O act species available and, accordingly, a higher amount of TiO 2 surface lattice oxygen removal at 400 ˝C is in agreement with findings in previous studies, where this was explained by the migration of TiO 2 surface lattice oxygen to Au-TiO 2 perimeter sites at elevated temperatures [27].This enables the removal also of O act species which were originally located further away from the Au-TiO 2 interface perimeter.At 80 ˝C, in contrast, the mobility of surface lattice oxygen is negligible, and only surface lattice oxygen species located directly at Au-TiO 2 interface perimeter sites can be removed by reaction with CO ad or H ad [25,35].Considering the Au loading and the Au particle size distribution (from TEM imaging), and assuming hemispherical Au nanoparticles, the total amount of these species can be estimated to be 6.2 ˆ10 18 O atoms¨g cat ´1.
Hence, the data are in full agreement with our previous proposal that CO and hydrogen are both oxidized by the same active oxygen species at both 80 ˝C and 400 ˝C, by TiO 2 surface lattice oxygen close to Au-TiO 2 perimeter sites [27].For more details see Ref. [35].During the first 10 CO/H2/Ar pulses on the fully oxidized Au/TiO2 catalyst there is a significant consumption of CO, indicated by the much lower CO intensity compared to that after about 50 or more pulses, and in parallel the formation of CO2 (Figure 1).Hence, also under these conditions CO oxidation proceeds by reaction with stable surface oxygen of the catalyst.With increasing number of pulses the CO consumption/CO2 formation during each pulse decreases, until there is no visible discrepancy to the steady-state situation after about 10 pulses.The absolute amount of CO consumed/CO2 formed during the 250 CO/H2/Ar pulses in this sequence at 80 °C was 2.2 × 10 18 molecules•gcat −1 .The significantly lower amount of CO2 formation during the exposure of the fully oxidized Au/TiO2 catalyst to CO/H2 pulses as compared to CO pulsing (3.3 × 10 18 molecules•gcat −1 , see above) indicates already that stable adsorbed active oxygen is not only consumed by reaction with CO, but also by reaction with hydrogen.From the H2 signal shown in Figure 1, however, no (or only very little) H2 consumption can be detected, in contrast to the CO signal.A somewhat lower H2 signal is observed only during the first 1-2 pulses, with the lower intensity measured during the first pulse originating from a generally lower overall pulse size for the first pulse in every sequence (an artifact of our TAP reactor system).Hence, from the raw data there is no indication for reduction of Au/TiO2  During the first 10 CO/H 2 /Ar pulses on the fully oxidized Au/TiO 2 catalyst there is a significant consumption of CO, indicated by the much lower CO intensity compared to that after about 50 or more pulses, and in parallel the formation of CO 2 (Figure 1).Hence, also under these conditions CO oxidation proceeds by reaction with stable surface oxygen of the catalyst.With increasing number of pulses the CO consumption/CO 2 formation during each pulse decreases, until there is no visible discrepancy to the steady-state situation after about 10 pulses.The absolute amount of CO consumed/CO 2 formed during the 250 CO/H 2 /Ar pulses in this sequence at 80 ˝C was 2.2 ˆ10 18 molecules¨g cat ´1.The significantly lower amount of CO 2 formation during the exposure of the fully oxidized Au/TiO 2 catalyst to CO/H 2 pulses as compared to CO pulsing (3.3 ˆ10 18 molecules¨g cat ´1, see above) indicates already that stable adsorbed active oxygen is not only consumed by reaction with CO, but also by reaction with hydrogen.From the H 2 signal shown in Figure 1, however, no (or only very little) H 2 consumption can be detected, in contrast to the CO signal.A somewhat lower H 2 signal is observed only during the first 1-2 pulses, with the lower intensity measured during the first pulse originating from a generally lower overall pulse size for the first pulse in every sequence (an artifact of our TAP reactor system).Hence, from the raw data there is no indication for reduction of Au/TiO 2 by H 2 during these pulses, and also the quantitative evaluation of all CO/H 2 /Ar pulses could not resolve significant H 2 consumption.From previous TAP reactor studies on Au/TiO 2 as well as on Au/CeO 2 it is already known that the efficiency of H 2 for active oxygen removal is considerably lower compared to CO [35,42].The H 2 consumption during a single pulse is, accordingly, too low to be resolved in these measurements, and the same is true for the formation of water during exposure to H 2 or CO/H 2 pulses at 80 ˝C [35].As described in Section 3.2 (see below), the consumption of active oxygen by H 2 can, nevertheless, be calculated from the total oxygen uptake during subsequent O 2 /Ar pulses, from the more than stoichiometric consumption of O 2 compared to the preceding CO 2 formation during two consecutive runs at 80 ˝C.
The O 2 signal measured during re-oxidation of Au/TiO 2 by O 2 /Ar pulses, after reduction of the catalyst by CO and H 2 , is also shown in Figure 1  in both cases), but significantly higher than the CO 2 formation during the preceding CO/H 2 /Ar pulses (2.2 ˆ10 18 molecules¨g cat ´1).Accordingly, the amount of active oxygen removed by reaction with hydrogen amounts to 0.9 ˆ10 18 O atoms¨g cat ´1.These results clearly demonstrate that under present reaction conditions, with CO and hydrogen being simultaneously present on the surface, these species compete for the same active oxygen species.The higher amount of active oxygen removal by CO, in spite of the considerable excess of H 2 during the CO/H 2 /Ar pulses (CO/H 2 ratio of 1/2) illustrates the high selectivity of Au/TiO 2 for the CO oxidation (71% in this case) under present reaction conditions.
The results of additional measurements with different CO/H 2 ratios during the CO/H 2 /Ar pulses, where the amount of CO molecules per pulse was kept constant at 3 ˆ10 15 molecules CO per pulse, while the amount of H 2 was varied between 1.5 ˆ10 15 and 24 ˆ10 15 molecules H 2 per pulse, are plotted in Figure 2a and listed  by H2 during these pulses, and also the quantitative evaluation of all CO/H2/Ar pulses could not resolve significant H2 consumption.From previous TAP reactor studies on Au/TiO2 as well as on Au/CeO2 it is already known that the efficiency of H2 for active oxygen removal is considerably lower compared to CO [35,42].The H2 consumption during a single pulse is, accordingly, too low to be resolved in these measurements, and the same is true for the formation of water during exposure to H2 or CO/H2 pulses at 80 °C [35].As described in Section 3.2 (see below), the consumption of active oxygen by H2 can, nevertheless, be calculated from the total oxygen uptake during subsequent O2/Ar pulses, from the more than stoichiometric consumption of O2 compared to the preceding CO2 formation during two consecutive runs at 80 °C.The O2 signal measured during re-oxidation of Au/TiO2 by O2/Ar pulses, after reduction of the catalyst by CO and H2, is also shown in Figure 1 (lower panel).Similar to CO consumption during CO/H2/Ar pulses, O2 consumption during the first 10-15 O2/Ar pulses is obvious already from the raw data by the lower intensity compared to that after saturation.Quantitative evaluation of the O2 consumption during the whole sequence (150 O2/Ar pulses) revealed that 3.1 × 10 18 O atoms•gcat −1 are needed to fully re-oxidize the catalyst at 80 °C (see also Table 1).This value is almost identical to the one determined after reduction by CO or H2 only (3.3 × 10 18 O atoms•gcat −1 in both cases), but significantly higher than the CO2 formation during the preceding CO/H2/Ar pulses (2.2 × 10 18 molecules•gcat −1 ).Accordingly, the amount of active oxygen removed by reaction with hydrogen amounts to 0.9 × 10 18 O atoms•gcat −1 .These results clearly demonstrate that under present reaction conditions, with CO and hydrogen being simultaneously present on the surface, these species compete for the same active oxygen species.The higher amount of active oxygen removal by CO, in spite of the considerable excess of H2 during the CO/H2/Ar pulses (CO/H2 ratio of 1/2) illustrates the high selectivity of Au/TiO2 for the CO oxidation (71% in this case) under present reaction conditions.
The results of additional measurements with different CO/H2 ratios during the CO/H2/Ar pulses, where the amount of CO molecules per pulse was kept constant at 3 × 10 15 molecules CO per pulse, while the amount of H2 was varied between 1.5 × 10 15 and 24 × 10 15 molecules H2 per pulse, are plotted in Figure 2a and listed in Table 1, together with the absolute amounts of active oxygen consumption/replenishment for CO only (0% H2) and H2 only (100% H2).These data clearly demonstrate that the total amount of oxygen that can be removed by CO/H2 is almost identical for all mixtures and, moreover, also for "CO only" and "H2 only" experiments.Obviously, under present reaction conditions, in the simultaneous presence of CO and H2, these reactants are oxidized by the same active oxygen species as we have already concluded for the individual reactants before (see Section 2.1 and [35]).The existence of other active oxygen species, These data clearly demonstrate that the total amount of oxygen that can be removed by CO/H 2 is almost identical for all mixtures and, moreover, also for "CO only" and "H 2 only" experiments.Obviously, under present reaction conditions, in the simultaneous presence of CO and H 2 , these reactants are oxidized by the same active oxygen species as we have already concluded for the individual reactants before (see Section 2.1 and [35]).The existence of other active oxygen species, which may be active only for H 2 oxidation when CO and H 2 are present simultaneously, can be ruled out, since this would result in a higher amount of active oxygen species for the simultaneous CO and H 2 oxidation as compared to the CO oxidation only.
On a first glance these findings seem to contradict recent conclusions from Saavedra et al., who studied, in detail, the role of water in the catalytic CO oxidation on Au/TiO 2 [40].Based on experimental results and DFT calculations they concluded (i) that weakly adsorbed water on TiO 2 determines the catalysts activity at room temperature by changing the effective number of active sites; and (ii) that O 2 activation in the dominant reaction pathway under these conditions requires the presence of both water and CO [40].Considering that our experiments were performed at 80 ˝C and under vacuum conditions (not during pulses, but between individual pulses and pulse sequences), the presence of weakly-adsorbed water on the catalyst surface can be excluded.Temperature programmed desorption (TPD) measurements performed after saturation of the catalyst with water by 1500 H 2 /Ar pulses had shown that stable adsorbed water is formed upon H 2 pulsing, which desorbs with a maximum in the TPD spectra at about 230 ˝C.This is the only water species present during O 2 pulsing at 80 ˝C.Hence, our results clearly demonstrate that as long as only stable adsorbed water is present on the catalyst surface, coexisting adsorbed CO and hydrogen both are oxidized by TiO 2 surface lattice oxygen and compete for this O act species.Weakly-adsorbed water species, whose presence during continuous PROX reaction cannot be ruled out from our data, may also enable another oxygen activation mechanism as described in the recent study by Saavedra et al. [40].In that case, however, the dominant reaction mechanisms of CO and H 2 oxidation under PROX reaction conditions are expected to sensitively depend on the reaction temperature, due to the facile desorption of these species at higher temperatures.Similar kinds of temperature effects have been proposed earlier for the CO oxidation reaction under dry reaction conditions [31].Another possibility that has to be considered under continuous PROX conditions, in the simultaneous presence of CO, O 2 , and H 2 in the catalyst bed, is the H-assisted O 2 activation.Rousset and coworkers previously demonstrated that the presence of H 2 promotes the CO oxidation activity over various Au catalysts supported on different (reducible and non-reducible) metal oxides [10,12].According to them, this beneficial effect of H 2 results most probably from the reaction of adsorbed hydrogen atoms with molecular oxygen to active hydroperoxy-like species that represent additional reaction intermediates in the CO oxidation.In the present approach there is an inherent difficulty in studying this effect, since CO/H 2 and O 2 are never present simultaneously in the catalyst bed during the alternate pulse sequences.Here, further work is planned to more closely investigate the influence of H 2 on the O 2 activation in pulse experiments, in particular by exposing the catalyst to mixed CO/H 2 /O 2 /Ar pulses.In that approach it will also be highly interesting to compare the oxygen activation in the presence of H 2 on different Au catalysts supported on reducible and non-reducible metal oxides (Au/TiO 2 , Au/ZnO, Au/Al 2 O 3 , and Au/SiO 2 , for example).Numerous previous studies have demonstrated that the activity for CO oxidation largely depends on the reducibility of the metal oxide support, with Au catalysts supported on reducible oxides being much more active [43,44].Applying TAP reactor pulse experiments we could even demonstrate a quantitative relation between the catalysts reducibility (measured by its oxygen storage capacity) and its activity for CO oxidation [41].In the presence of H 2 , however, the activity of Au catalysts supported on reducible and non-reducible metal oxides becomes very similar [10,12], reflecting a more pronounced promotional effect of H 2 on the CO oxidation activity (see above) for non-reducible than for reducible metal oxide supports.
In addition to the absolute amount of active oxygen on Au/TiO 2 we also calculated the selectivity for active oxygen removal by CO at 80 ˝C for different CO/H 2 ratios, where the selectivity was given by the ratio between the absolute amount of CO 2 formation and the overall oxygen removal during CO/H 2 /Ar pulses (see Section 3.2). Figure 2a illustrates already that the amount of CO consumption/CO 2 formation continuously decreases with increasing H 2 content in the CO/H 2 pulses at 80 ˝C, despite the almost constant amount of O act removal (for combined CO and hydrogen oxidation).This results in a decrease of the selectivity for CO 2 formation from 92% in the presence of 33% H 2 to slightly below 50% in the presence of 89% H 2 (see Figure 2b and Table 1).A similar trend of a decreasing selectivity with decreasing CO/H 2 ratio was also observed in kinetic measurements of the PROX reaction on a Au/TiO 2 catalyst under continuous flow, but otherwise similar reaction conditions (80 ˝C) [14,45].The high selectivities fully agree with the much lower efficiency for O act removal by H 2 (including H 2 adsorption, dissociation, and reaction with O act ) compared to reaction with CO.This was already demonstrated by the much higher number of H 2 pulses needed for the complete removal of the available active oxygen species compared to reaction with CO in multi-pulse TAP reactor experiments (see Section 2.1 and Refs.[35,42]).Considering the selectivity of about 50% in the presence of 89% H 2 (CO/H 2 ratio of 1/8), this difference in the efficiency for O act removal can be estimated to be about 8.This value is also in close agreement to the approximately eight times higher amount of H 2 molecules compared to CO molecules that are needed for a complete removal of all available O act species (see Section 2.1).One has to keep in mind, however, that the latter averages over the entire pulse sequence, and does not necessarily reflect the difference in efficiency of a single CO or H 2 molecule.The selectivity largely originates from the competition between adsorbed CO and hydrogen for the same active oxygen species when simultaneously present on the catalyst surface.Additional effects arising, e.g., from the competition of the two adsorbing species (CO, H 2 ) for adsorption sites and from (de-)stabilizing interaction between the coadsorbed species, which had been suggested previously [13,14], can neither be excluded nor supported from our TAP reactor experiments measurements.It should be noted, however, that the steady-state coverages under typical reaction conditions (ď0.5% CO, 60%-70% H 2 , 80 ˝C, atmospheric pressure) are well below saturation.Accordingly, such effects are expected to be small, and similar conclusions were put forward also in the above studies [13,14].
Finally, we performed similar multi-pulse experiments with alternate sequences of O 2 /Ar pulses and CO/H 2 /Ar pulses at 400 ˝C, using the same reaction gas mixtures for O act removal from Au/TiO 2 as described above (CO, H 2 , and CO/H 2 = 1/0.5 ´CO/H 2 = 1/8).The main characteristics during reduction and re-oxidation were identical to those obtained in the experiments at 80 ˝C: For all measurements, CO 2 and H 2 O are only formed during CO and/or H 2 pulses, respectively, and the processes of catalyst reduction and re-oxidation are fully reversible, as evidenced by at least three cycles of reduction and re-oxidation with identical absolute amounts of O act removal/replenishment for each gas mixture.The absolute amounts of active oxygen removal upon exposure to CO only, H 2 only, and CO/H 2 mixtures are presented in Figure 3 and listed in Table 2.
Similar to the results at 80 ˝C also in this case the absolute amount of active oxygen (12 ˆ10 18 O atoms¨g cat ´1) is essentially independent of the composition of the reaction gas pulses used for reduction.Hence, also at 400 ˝C CO and H 2 seem to compete for TiO 2 surface lattice oxygen close to the Au nanoparticles when present simultaneously.The only difference to the measurements at lower reaction temperature is that at 400 ˝C the increasing thermal mobility of surface lattice oxygen and surface oxygen vacancies enables the migration and, therefore, the removal and replenishment of oxygen and oxygen vacancies which were originally located further away from the perimeter sites.At 80 ˝C, in contrast, only oxygen species directly adjacent to the Au nanoparticles participate in the oxidation reactions [27].Nevertheless, also at 400 ˝C removal of O act species upon reaction with CO or H 2 is expected to occur only at Au-TiO 2 interface perimeter sites.
reduction and re-oxidation were identical to those obtained in the experiments at 80 °C: For all measurements, CO2 and H2O are only formed during CO and/or H2 pulses, respectively, and the processes of catalyst reduction and re-oxidation are fully reversible, as evidenced by at least three cycles of reduction and re-oxidation with identical absolute amounts of Oact removal/replenishment for each gas mixture.The absolute amounts of active oxygen removal upon exposure to CO only, H2 only, and CO/H2 mixtures are presented in Figure 3 and listed in Table 2.
Similar to the results at 80 °C also in this case the absolute amount of active oxygen (12 × 10 18 O atoms•gcat −1 ) is essentially independent of the composition of the reaction gas pulses used for reduction.Hence, also at 400 °C CO and H2 seem to compete for TiO2 surface lattice oxygen close to the Au nanoparticles when present simultaneously.The only difference to the measurements at lower reaction temperature is that at 400 °C the increasing thermal mobility of surface lattice oxygen and surface oxygen vacancies enables the migration and, therefore, the removal and replenishment of  As described below (Section 3.3) it is not possible to calculate the selectivity for reaction at 400 ˝C, since even in the absence of H 2 there is no stoichiometric consumption of O 2 and CO (with respect to the CO oxidation).Hence, the differences between O 2 consumption and CO consumption in consecutive sequences of O 2 /Ar and CO/H 2 /Ar pulses cannot be assigned to active oxygen removal via the H 2 oxidation only.Nevertheless, from the continuously-decreasing amount of CO 2 formation with increasing H 2 ratio in CO/H 2 /Ar pulses, from 9.3 ˆ10 18 CO 2 molecules¨g cat ´1 to 6.1 ˆ10 18 CO 2 molecules¨g cat ´1 at CO/H 2 = 1/0.5 and CO/H 2 = 1/8 (see Table 2), it is evident that also at 400 ˝C the reaction is dominated by the same trend of decreasing selectivity towards CO 2 formation with increasing H 2 content.Moreover, considering that even for the highest amount of H 2 during CO/H 2 pulses (CO/H 2 = 1/8) and, hence, in an eight-fold excess of H 2 compared to CO, still 6.1 ˆ10 18 CO 2 molecules¨g cat ´1 are formed, which is about 60% of the amount of CO 2 formed upon reduction by CO only (see Table 2), it can be concluded that also at 400 ˝C the catalyst still selectively catalyses the CO oxidation under present reaction conditions.
In total, these results, obtained at different temperatures as well as for reaction in different atmospheres, point to a competition of CO and H 2 for the same active oxygen species when present simultaneously and, hence, under PROX reaction conditions.This competition will mainly determine the catalyst selectivity towards CO oxidation, and strategies for even more selective Au catalysts under these conditions (80 ˝C and higher) should accordingly focus on the efficiency of H 2 and CO for active oxygen removal, which may be achieved by a systematic modification of the support material and/or the Au-support interface perimeter sites, which both significantly influence the activity for CO and H 2 oxidation via a Au-assisted Mars-van Krevelen mechanism [31,41].

Preparation and Characterization of Au/TiO 2
For all measurements we used a home-made Au/TiO 2 catalyst with a Au loading of 2.2 wt.%, which was prepared via a deposition-precipitation method using commercial, non-porous TiO 2 as support material (P5, surface area 56 m 2 ¨g´1 ).Prior to all measurements the catalyst was pre-treated in situ by calcination at 400 ˝C in a continuous flow of 10% O 2 /N 2 (20 Nml¨min ´1) at atmospheric pressure for 30 min (hereafter denoted as O400), resulting in a surface-volume mean diameter of 2.4 ˘0.5 nm.The mean Au particle size was determined by transmission electron microscopy (TEM), evaluating the diameter of more than 500 Au nanoparticles after pre-treatment (O400).

TAP Reactor Measurements
Pulse experiments were performed in a TAP reactor system, which was built in our laboratory and which is described in detail in Ref. [46].This system consists of a gas mixing unit, two piezo-electrically driven pulse valves, and a quartz glass micro-reactor (length: 90 mm, inner diameter: 4 mm), which is connected to an analysis chamber (under ultrahigh vacuum conditions), which houses the mass spectrometer for gas phase analysis.For in situ catalyst pre-treatment at atmospheric pressure, the micro-reactor with the catalyst inside can be separated from the analysis chamber via a differentially-pumped gate valve.This latter gate valve enables fast transitions from vacuum to ambient conditions in the micro-reactor and vice versa [46].
Gas pulses containing between 8 ˆ10 15 and 3 ˆ10 16 molecules per pulse were generated by the piezo-electrically driven pulse valves and directed to the micro-reactor.At the centre of this reactor the catalyst bed was fixed by stainless steel sieves.For all measurements the catalyst bed consisted of two outer layers of SiO 2 powder (inert under present reaction conditions) and a central catalyst zone with 10 mg of the Au/TiO 2 catalyst diluted with 10 mg SiO 2 (dilution 1:1) in between the SiO 2 layers.
Multi-pulse experiments were performed with a separation time of 5 s between two consecutive pulses.During these measurements the catalyst was alternately exposed to sequences of (i) CO/Ar; (ii) H 2 /Ar; or (iii) CO/H 2 /Ar and O 2 /Ar pulses for removal and replenishment of stable, active oxygen species.The number of pulses in each sequence was chosen such that there is no measurable reactive removal or oxygen uptake at the end of each sequence, i.e., at the end of CO/Ar, H 2 /Ar, and CO/H 2 /Ar pulse sequences or after the O 2 /Ar pulse sequences.The exact numbers of pulses in each sequence are stated together with the results obtained (see Sections 2.1 and 2.2).Moreover, these cycles with alternate reduction and re-oxidation of the catalyst surface were always repeated at least three times in order to demonstrate the reversibility of the observed processes and, hence, to exclude permanent changes of the catalysts state during pulse experiments.The absolute amount of reversible oxygen removal and replenishment is defined as the catalysts oxygen storage capacity (OSC).Note that in each pulse Ar was included in order to enable the quantitative evaluation of the amount of CO, H 2, or O 2 consumed during every single pulse on an absolute scale.This is based on the comparison of the ratio of the intensities of the respective species and Ar during reaction (uptake/consumption of CO, H 2 and O 2 ) to that after saturation, where no more consumption of the corresponding species was detected.CO/Ar and O 2 /Ar pulses always consisted of 50% CO or 50% O 2 and 50% Ar at a pulse size of 8 ˆ10 15 molecules per pulse.Hence, the catalyst was exposed to about 4 ˆ10 15 molecules CO or O 2 during each of the pulses.For pulses with a CO/H 2 /Ar mixture, in contrast, the overall pulse size was between 1.2 ˆ10 16 and 3.0 ˆ10 16 molecules per pulse.Here we kept the amount of CO constant at 3 ˆ10 15 molecules CO per pulse, while varying the amount of H 2 between 1.5 ˆ10 15 and 24 ˆ10 15 molecules H 2 per pulse.This way, CO/H 2 ratios between 1/0.5 and 1/8 were realized for the CO/H 2 /Ar pulses.Note that in the discussions the H 2 ratio during CO/H 2 /Ar pulses is always stated with respect to CO and H 2 only (not considering Ar), which are accordingly between 33% H 2 (CO/H 2 = 1/0.5)and 89% H 2 (CO/H 2 = 1/8).
In order to avoid complications in the direct comparison of active oxygen removal by CO and H 2 at 80 ˝C due to adsorption of water, which is always formed on the catalyst surface during H 2 oxidation and has a distinct impact on the CO oxidation activity (see above), we saturated the Au/TiO 2 catalyst surface with adsorbed water by exposing it to 5000 H 2 /Ar pulses before starting the measurements.Note that this treatment additionally resulted in a complete removal of stable reactive oxygen (O act ) species for H 2 oxidation.The starting point of subsequent measurements is, hence, a fully reduced Au/TiO 2 catalyst.

Calculation of the Selectivity
Usually the selectivity for CO oxidation in CO/H 2 mixtures is defined by the ratio between the CO 2 formation rate and the overall oxygen consumption rate (for CO oxidation and H 2 oxidation, see also Equation ( 3)), which are determined in kinetic measurements.This is not possible in the present study, where we focused on the absolute amounts of oxygen removal and replenishment during alternate sequences of CO/H 2 /Ar and O 2 /Ar pulses rather than on their kinetics.In this case we calculated the selectivity for CO oxidation at 80 ˝C by the ratio between the absolute amount of CO consumption/CO 2 formation (N CO during a sequence of CO/H 2 /Ar pulses) and the absolute amount of atomic oxygen consumption (2¨N O2 ) during a subsequent sequence of O 2 /Ar pulses (Equation ( 4)).Note that it was not possible to determine the absolute amount of H 2 O formation during CO/H 2 /Ar pulses at 80 ˝C, since water desorption is too slow to be resolved.

S "
N CO 2 ¨NO 2 (4) For reaction at 400 ˝C it was not possible to calculate the selectivity in that way.From previous studies it is known that at this temperature, even in the absence of H 2 , the O 2 uptake is always slightly higher than the oxygen consumption during subsequent exposure to CO during alternate sequences of O 2 /Ar and CO/Ar pulses [27,35].Therefore, the difference between CO consumption/CO 2 formation during exposure to CO/H 2 /Ar pulses to that of O 2 consumption during subsequent exposure to O 2 /Ar pulses is not solely due to active oxygen consumption by H 2 oxidation/H 2 O formation, and can not be used for determination of the selectivity.For this reason we can only provide values for O act removal and CO 2 formation at 400 ˝C.

Conclusions
From the comparison of the active oxygen removal from a Au/TiO 2 catalyst by CO, H 2 , and CO/H 2 pulses with varying CO/H 2 ratios in quantitative temporal analysis of products (TAP) reactor measurements at 80 ˝C and 400 ˝C, we arrived at the following conclusions on the PROX reaction mechanism: (1) CO and hydrogen are oxidized by the same stable active oxygen species under present reaction conditions, also in the simultaneous presence of CO and H 2 in the reaction atmosphere, as evidenced by the similar amounts of active oxygen removal from a O 2 /Ar pulse oxidized Au/TiO 2 catalyst upon exposure to multi-pulse sequences of CO, H 2 , and CO/H 2 mixtures.This is independent of the CO/H 2 ratio.Hence, also under PROX reaction conditions CO and H 2 compete for TiO 2 surface lattice oxygen close to the Au nanoparticles/at the perimeter of the Au-TiO 2 interface as active oxygen species.(2) The selectivity of Au/TiO 2 catalysts for CO oxidation in CO/H 2 containing gas mixtures observed in kinetic measurements is proposed to mainly result from the much higher efficiency of CO for active oxygen removal compared to hydrogen under present reaction conditions, in the simultaneous presence of H 2 and CO in the reaction atmosphere.The latter is illustrated also by the much higher amount of H 2 pulses required for complete removal of active oxygen from the catalyst compared to CO pulsing.(3) Since we can rule out the presence of weakly-bound adsorbed water under present reaction conditions, in multi-pulse sequences, the much higher efficiency of CO for reaction with active oxygen species or the preference for this reaction pathway cannot result from effects caused by weakly-bound water species.On the other hand, stable adsorbed (strongly bound) water species, which are known to be present on the surface upon H 2 pulsing, leave the dominant reaction pathway for CO oxidation under present reaction conditions, namely reaction of CO ad with surface lattice oxygen at the perimeter of the Au NP-support interface, unchanged, although the reaction rate may change.
Since the selectivity of pure Au/TiO 2 catalysts originates mainly from the intrinsic difference in the reactivity of TiO 2 surface lattice oxygen located at Au-TiO 2 perimeter sites for CO oxidation and H 2 oxidation under present reaction conditions, further improvements in the selectivity of TiO 2 -based Au catalysts should include a modification of these perimeter sites, for example by doping of the TiO 2 support.
We cannot rule out from the present data, however, that other reaction pathways contribute as well at lower reaction temperature and/or during reaction under continuous flow conditions at atmospheric pressure.Under those conditions, weakly-adsorbed water, as well as adsorbed H ad , are present and may also be involved in the reaction, as recently proposed by Saavedra et al. [40] and Rousset et al. [12].
Overall, the present findings confirm previous ideas on the CO oxidation mechanism and the physical origin of the high selectivity for CO oxidation in the PROX reaction on supported Au catalysts, specifically on Au/TiO 2 catalysts, which were based on plausibility arguments extending mechanistic insights from CO oxidation and H 2 oxidation to the present case of reaction in CO/H 2 mixtures.

2 Figure 1 .
Figure 1.Mass spectrometer signals of CO, CO 2 , H 2 , and H 2 O measured during a sequence of CO/H 2 /Ar pulses (CO/H 2 ratio = 1/2) on Au/TiO 2 at 80 ˝C after oxidation by O 2 /Ar pulses, as well as the O 2 signal measured during subsequent O 2 /Ar pulses at 80 ˝C.

Table 1 .
Absolute amounts of active stable oxygen (O act ) removal and CO 2 formation during CO, H 2 , and CO/H 2 pulses with different CO/H 2 ratios on the fully oxidized Au/TiO 2 catalyst at 80 ˝C, and corresponding selectivities for CO oxidation.
* Error estimations are based on the repeated reproduction of individual pulse sequences.

Table 2 .
Absolute amounts of active stable oxygen (O act ) removal and CO 2 formation during CO, H 2 , and CO/H 2 pulses with different CO/H 2 ratios on the fully oxidized Au/TiO 2 catalyst at 400 ˝C.

Ratio CO/H 2 H 2 /(CO+H 2 )/% O act Removal */ 10 18 O atoms¨g cat ´1 CO 2 Formation */ 10 18 molecules¨g cat
Ar pulses with various CO/H 2 ratios.At both temperatures (80 ˝C and 400 ˝C) five different The trends in the pulse signals of CO, CO 2 , H 2 , H 2 O, and O 2 during reduction and re-oxidation with CO/H 2 /Ar pulses (CO/H 2 = 1/2) and subsequent O 2 /Ar pulses, respectively, are illustrated in Figure1for reaction at 80 ˝C on Au/TiO 2 .Only the first 50 pulses of each sequence are shown, where differences in the measured intensities due to consumption of educts (CO, H 2 , and O 2 )/formation of products (CO 2 , H 2 O) are visible in the raw data, while the actual number of pulses per sequence are significantly higher (see Section 3.2).For CO/H 2 /Ar and O 2 /Ar at 80 ˝C the number of pulses in each sequence was 250 and 150, respectively.
(lower panel).Similar to CO consumption during CO/H 2 /Ar pulses, O 2 consumption during the first 10-15 O 2 /Ar pulses is obvious already from the raw data by the lower intensity compared to that after saturation.Quantitative evaluation of the O 2 consumption during the whole sequence (150 O 2 /Ar pulses) revealed that 3.1 ˆ10 18 O atoms¨g cat ´1 are needed to fully re-oxidize the catalyst at 80 ˝C (see also Table 1).This value is almost identical to the one determined after reduction by CO or H 2 only (3.3 ˆ10 18 O atoms¨g cat ´1 in Table1, together with the absolute amounts of active oxygen consumption/replenishment for CO only (0% H 2 ) and H 2 only (100% H 2 ).
Absolute amounts of active oxygen (O act ) removal from a fully oxidized Au/TiO 2 catalyst (after O 2 /Ar pulses) by CO and/or H 2 during CO, H 2 , and CO/H 2 pulses with varying ratios of H 2 (33%-89% H 2 ) at 400 ˝C.