Research Article pubs.acs.org/acscatalysis
A Comparison of CO Oxidation by Hydroxyl and Atomic Oxygen from Water on Low-Coordinated Au Atoms Matthijs A. van Spronsen,*,†,∥ Kees-Jan Weststrate,‡ and Ludo B. F. Juurlink§ †
Huygens-Kamerlingh Onnes Laboratory, Leiden University, P.O. Box 9504, 2300 RA Leiden, The Netherlands Syngaschem BV, c/o Dutch Institute for Fundamental Energy Research (DIFFER), De Zaale 20, 5612 AJ Eindhoven, The Netherlands § Leiden Institute of Chemistry, Leiden University, Einsteinweg 55, P.O. Box 9502, 2300 RA Leiden, The Netherlands ‡
S Supporting Information *
ABSTRACT: The catalytic oxidation of CO is studied at lowcoordinated Au atoms using a single-crystal approach. Electron irradiation activates an otherwise unreactive overlayer of undissociated D2O on Au(310). A low-coverage D2O/O mixture is subsequently allowed to react at surface temperatures from 105 K upward, with CO supplied from the gas phase. X-ray photoelectron spectroscopy shows the absence of Au oxides and quantifies various O-containing species during the reaction. The dependency of the reaction rate on the surface temperature yields an activation energy for the Langmuir−Hinshelwood reaction of O(ads) and CO(ads) between 26 ± 4 and 42 ± 5 kJ/mol. The presented results provide evidence that O(ads) and not OH(ads) is the active reactant on small Au nanoparticles. In addition, the observations suggest that water has a negative effect on the reactivity of O(ads). KEYWORDS: fragmentation of water, CO oxidation, electron irradiation, hydroxide, Au model catalyst, low-coordinated atoms
1. INTRODUCTION The recent history of Au catalysis started with the pioneering work of Haruta et al.1 In this study and following studies, Au nanoparticles (NPs) were found to be active for CO oxidation, water−gas shift reaction, selective oxidation, and hydrogenation/isomerization.2−4 In addition to being of fundamental, chemical interest, Au catalysis can hold the answer to some challenging problems in catalysis. One of these is the low-temperature and selective oxidation of CO. Traditionally used catalysts, such as Pt, operate only at high temperatures. Au does not have this requirement and has the potential to increase the efficiency of the automotive catalysts before the catalyst is at the operating temperature. Additionally, Au can selectively oxidize adverse CO in hydrogen fuel cells.5 Advances in the understanding of Au catalysis have recognized the importance of three factors: highly dispersed NPs,6 an active oxide support,7 and the presence of H2O vapor to enhance reactivity.8 These have inspired various explanations for the high reactivity of Au. Among these hypotheses are the modified electronic structures of small clusters,9 enhanced activity of low-coordinated atoms,6,10,11 the existence of active sites on the NP−support perimeter,12−14 and spillover, i.e., diffusion of reaction intermediates from the support to NPs or the NP−support perimeter.7,15 Proving or disproving these theories with technical catalysts is very difficult. This is because precise characterization of the © XXXX American Chemical Society
active sites is needed, both chemically and structurally. This is where model catalysts come into play. They can represent one aspect of a real catalyst while greatly reducing the inherent complexity of the technical catalyst. Ideally, the full structural and chemical characterization is achieved under reaction conditions. Although recently developed surface science tools can provide some of the desired structural16−18 or chemical information,19 they have limitations. These tools are very challenging and usually operate at the signal-to-noise limit. As a consequence, these so-called in situ or operando techniques still depend on vacuum-based, surface science experiments to interpret the experimental results. In this and our previous work,20,21 we undertook a surface science study using X-ray photoelectron spectroscopy (XPS) and temperature-programmed desorption (TPD) to answer elementary questions about the interaction of H2O and Au, the oxidation of CO, and the role of low-coordinated atoms. In our previous work,21 the interaction was probed between H2O and a stepped model catalyst, the Au(310) single-crystal surface. This surface is visualized in Figure 1. The surface consists of narrow (two or three-atom wide) terraces with a (100) structure separated by monatomic (110) steps. The high step Received: June 17, 2016 Revised: August 28, 2016
7051
DOI: 10.1021/acscatal.6b01720 ACS Catal. 2016, 6, 7051−7058
Research Article
ACS Catalysis
stable on the Au surface. This mixture was tested for reactivity to CO at several temperatures to study the kinetic details of this reaction.
2. EXPERIMENTAL SECTION The experimental details have been thoroughly discussed previously,21 and only the most important aspects will be highlighted. All experiments were performed at the SuperESCA beamline, Elettra Sincrotrone Trieste, Italy. An ultrahigh vacuum (UHV) system was used, designed to study surfaces with high-resolution or time-resolved XPS. This system was equipped with a liquid nitrogen cryostat, a quadrupole mass spectrometer (QMS), and equipment to measure low-energy electron diffraction (LEED). A single crystal of Au, polished to the (310) plane (accuracy of 2.3°) was cleaned with multiple cycles of Ar+ sputtering (energy of 1 keV, for a few minutes) and annealing (in UHV, 860 K). Surface crystallinity was checked with LEED, and cleanliness was confirmed via XPS. High-purity (D2O, 99.95 atom % D, Aldrich) D2O and He (He, 6.6 N purity) were co-dosed after repetitive freeze− pump−thaw cycles to remove dissolved air. D2O was used, because of the low background levels of both D2O and D2 in the residual gas of the vacuum chamber. Reproducible dosing was achieved by monitoring the O 1s signal from the surface while admitting D2O/He to the vacuum chamber. After adsorption at a surface temperature of ∼100 K, the D2O(ads) was exposed to electron irradiation for 5−720 s. These electrons were generated using the electron gun of the LEED system at an energy of 100 eV. For this purpose, the electron gun was completely defocused, with a resulting beam width of ∼5 mm and a sample current of 7−8 μA. The electron dose is reported in monolayers [with 1 monolayer (ML) equal to the number of Au atoms on the (310) surface, 1.1× 1019 m−2] and was based on the integrated sample current and the estimated beam size. Electron irradiation led to a sample temperature increase of 4−5 K. XP spectra were recorded at three different photon energies: 170 eV for Au 4f7/2, 650 eV for O 1s, and 400 eV for C 1s. Spectra were recorded with normal emission and an incident angle of 70°. Furthermore, they were corrected by measuring the Fermi level. After this calibration, every spectrum was scaled by the average background level, obtained over a 0.5 eV interval on the low-binding energy side, to correct for changes in beam intensity. After scaling, a linear fit to the background was subtracted. Fitting of the resulting spectra was achieved using a Doniach−Šunjić function27 convoluted with a Gaussian line shape. The D2O coverage was estimated by comparing the O 1s integral with that of a saturated CO(ads) overlayer at 105 K, which equals a coverage of 50% of the step atoms or 0.167 ML.20,28 These integrals were measured at 1205 eV to weaken the effect of energy-dependent fluctuations of the photoemission cross section, caused by photoelecton diffraction. This effect is more pronounced closer to the adsorption edge. The calibration of the coverage was confirmed by analyzing the relative intensities of the O 1s and Au 4f7/2 peaks (see ref 21). During the experiments, some increase in the amount of C was detected (see the Supporting Information). It was identified as CO(ads), maybe some amorphous or graphitic C, and possibly carbidic AuCx. This C buildup was unavoidable because of the lengthy nature of these experiments, typically several tens of minutes. These species were unrelated to X-ray beam exposure and showed only a slow increase over time and,
Figure 1. (310) surface of a face-centered cubic metal, such as Au, as seen from (a) the top view, (b) a side view, and (c) an angled view. Coordination numbers are indicated, and the surface unit cell is depicted (transparent blue rectangle), measuring 0.645 nm × 0.408 nm.
density provides many atoms with low coordination numbers, going as low as 6. The lowest-coordinated atoms are found in the (110) step at the edge of the terrace. The high concentration of low-coordinated Au atoms makes it a good model for small NPs. However, because it is a single crystal, the electronic structure remains that of bulk Au. Furthermore, the lack of an oxide support makes it possible to completely separate the effect of different aspects of the technical catalyst. The main results of the previous studies are summarized as follows: low-coordinated atoms have an enhanced interaction with H2O(ads); however, they cannot dissociate H2O(ads).20,21 As the high concentrations of low-coordinated atoms cannot facilitate direct reaction between CO(ads) and H2O(ads), the focus, here, shifted to the reaction of CO with possible intermediates derived from H2O, namely, OH(ads) and O(ads). Both these species were proposed as intermediates in reaction pathways that start with the activation of O2 by H2O, which can occur on the facets of both unsupported22 and supported Au NPs,23 and on the NP−support interface.24 Activation of O2 on the support or the particle−support perimeter could be followed by diffusion of the reaction intermediates to the facets of the NP, where they potentially react with CO.7,15 Alternatively, CO can diffuse to the perimeter via the formation of transient Au−CO complexes.25,26 Because the currently used model catalyst, the Au(310) surface, has no oxide support, the reaction intermediates were produced by exposing adsorbed D2O to low-energy electron irradiation. This treatment led to D2O/O mixtures on the surface, based on the XPS measurements. No significant amount of OD(ads) was detected, suggesting that it was not 7052
DOI: 10.1021/acscatal.6b01720 ACS Catal. 2016, 6, 7051−7058
Research Article
ACS Catalysis
Figure 3 presents three fitted spectra. Figure 3a depicts a spectrum prior to electron irradiation, showing intact D2O-
for CO(ads), a decrease upon electron irradiation. The total amount was rather low, typically 2−3% with a maximum of 4.7%, which mainly consisted of CO(ads). The coverage estimation was based on the comparison with the saturated CO(ads) overlayer (105 K, 400 eV).20,28 Furthermore, the C species were considered to be spectators, because the results showed no correlation with their coverage.
3. RESULTS 3.1. Formation of Reaction Intermediates. H2O does not adsorb dissociatively on Au(310) under UHV conditions, and hence, the co-adsorption of H2O and CO did not lead to CO oxidation.20,21 To study the reaction of H2O with CO on this Au model catalyst without oxide support, electron irradiation was used to activate D2O(ads). Figure 2 shows the XP O 1s spectra after increasing amounts of 100 eV electron irradiation. The experiment started after
Figure 3. Fits of several O 1s spectra recorded at ∼100 K showing the effects of electron irradiation. Peaks represent the following: D2O, weakly bonded (black); D2O, Au-bonded (red); new peak at ∼531.6 eV (blue); and O(ads) (green). (a) No electron irradiation, (b) 36 ML electron dose, and (c) 304 ML electron dose. The bottom parts of the panels give the residual spectra.
Figure 2. XP O 1s spectra showing the effect of an increasing electron irradiation dose on the adsorbed D2O layer. Several changes are notable: a decrease in the intensity of the main peak, a shift to lower binding energies, and the growth of the peak around 530 eV. Spectra were recorded at ∼100 K.
(ads), fitted with two separate peaks. After a small amount of electron irradiation [36 ML (Figure 3b)], the feature corresponding to O(ads) developed into a significant contribution to the spectrum and a peak at ∼531.6 eV appeared, while the peaks assigned to intact D2O(ads) strongly decayed. The total O 1s intensity decreased considerably, from 0.5 to 0.3 ML, which was attributed to electron-stimulated desorption. Thermal desorption could be excluded, because the temperature increase during electron irradiation was too small. This was confirmed by TPD experiments that showed no desorption of H2O at 100 ± 10 K.20,21 With an increasing irradiation dose (304 ML), the total O coverage decreased to 0.1 ML. The surface was covered by roughly equal amounts of O(ads) and the unknown species, which was reduced by a factor 3, between 36 and 304 ML of electron irradiation. The fitted peaks were integrated, normalized to obtain the O coverages, and plotted as a function of electron dose as shown in Figure 4. The figure shows that the D2O coverage decayed exponentially upon electron irradiation. Simultaneously, the intensity of the peak at ∼531.6 eV was rapidly increasing for an electron dose up to 8 ML, after which it showed a modest decrease. The intensity of the O(ads) peak increased initially, after which it was saturated at ∼0.05 ML, followed by a slow, gradual decrease. The relative intensity of the peak at ∼531.6 eV rapidly increased to 8 times that of the O(ads) peak within
dosing D2O up to a coverage of 0.5 ML, termed pristine D2O(ads). Electron irradiation induced a large decrease in the intensity of the major peak. Furthermore, the peak shifted to lower binding energies (532.3 to 531.4 eV) with an increasing electron dose. In addition, the peak around 530 eV grew significantly compared to that of the pristine D2O(ads) layer and shifted from 529.9 to 529.7 eV. A quantitative view of the effect of electron irradiation was obtained by fitting the O 1s spectra with four peaks. Two were used to account for adsorption of pristine D2O (black, 532.8 eV; red, 532.2 eV). The origin of these molecular peaks is discussed in ref 21. In short, the peak at 532.8 eV was assigned to both multilayer D2O(ads) and D2O(ads) bonded to 9-foldcoordinated Au atoms. The other peak, at 532.2 eV, was attributed to D2O(ads) bonded to 6- and 8-fold-coordinated Au atoms. To account for the changes induced by electron irradiation, one new peak (blue, ∼531.6 eV) and one peak for O(ads) (green) were needed. To obtain the best fit, the binding energy for the former shifted from 531.9 to 530.9 eV, with an average of 531.6 eV. The appearance of the new species on the surface indicated the formation of hydroxides or the O stabilization of D2O(ads). The actual assignment will be given in the Discussion. 7053
DOI: 10.1021/acscatal.6b01720 ACS Catal. 2016, 6, 7051−7058
Research Article
ACS Catalysis
3.2. CO Oxidation by Activated D2O. For the next set of experiments, a reproducible amount of D2O (0.5 ± 0.1 ML) was adsorbed while the O 1s signal was being monitored. After adsorption, the surface was irradiated with a fixed electron dose (36 ML) to obtain a mixture close to that indicated by the gray bar in Figure 4. This led to a mixture of D2O(ads), O(ads), and either OD(ads) or O-stabilized D2O(ads). The total coverage of this mixture was ∼0.3 ML. In this mixture, the O(ads) content was 21 ± 1%. This procedure resulted in mixtures with very reproducible compositions, with only modest variations in absolute coverage. The prepared mixture was exposed to CO, at the highest attainable pressure (∼1 × 10−7 mbar) that allowed us to measure the O 1s spectra simultaneously. This corresponded to a CO flux of 0.03 ML/s. Even at temperatures as low as 105 K, reactivity was observed. Figure 6 presents the areas of the fitted peaks as a Figure 4. Coverage of different O-containing species plotted vs electron irradiation dose. A large decrease in D2O coverage (red and black) caused by desorption and fragmentation was observed for small electron doses. In parallel, the intensity of the peak at ∼531.6 eV increased, reaching a maximum after an electron dose of 8 ML, after which it decreased. The O(ads) peak showed similar behavior but was saturated at 0.05 ML. The gray bar indicates an electron dose of 36 ML for which the XPS fits are shown in Figure 3b and which was used for the CO oxidation experiments (Figure 6).
the first 8 ML of electron irradiation. After this initial increase, the relative intensity started to decrease, starting from 30 ML. Figure 5 shows four Au 4f7/2 spectra. None of the Au 4f7/2 spectra showed any sign of Au oxides. These oxides would be
Figure 5. Au 4f7/2 spectra obtained during various stages of the experiment: clean Au(310) surface (orange), after adsorption of 0.5 ML D2O prior to electron irradiation (red), after an electron dose of 8 ML (blue), and after an electron dose of 973 ML (yellow). Possible Au oxides would be expected in the yellow region between 85.2 and 86.0 eV.
Figure 6. Reaction of the fragmented D2O mixture with CO at two different temperatures: (a) below the H2O desorption temperature, 137 K, and (b) at the desorption temperature, 153 K. Peak areas are plotted vs time for the combined D2O(ads) peaks of 532.2 and 532.8 eV (dashed, red line with circles), the peak at ∼531.6 eV (blue, solid line with empty squares), O(ads) (dotted, green line with filled squares), and CO(ads) (purple line with pluses). Note the shorter time scale on the axis of panel b.
expected around 85.2−86.0 eV29−42 (indicated by the yellow region in Figure 5). The largest difference can be observed between the spectrum of the clean Au(310) surface and that obtained with 0.5 ML D2O(ads). Electron irradiation induced only small changes in the spectra. The total intensity increased with an increasing electron dose, because of desorption of D2O. Other changes were an increase in the magnitude of the shoulder around 84.6 eV and the decrease in the magnitude of the shoulder at 83.7 eV. All changes were very modest, and the latter two changes were reversed by larger doses of electron irradiation.
function of time, measured while exposing the surface to CO. Two distinct cases are shown here. One (Figure 6a) is that of a reaction at 137 K, which is below the onset of H2O desorption. The other (Figure 6b) shows a data set recorded at 153 K, i.e., at the onset of H2O desorption.20,21 Both situations clearly showed a strong decrease in the level of O(ads) and in the intensity of the ∼531.6 eV peak and were directly correlated with CO exposure. 7054
DOI: 10.1021/acscatal.6b01720 ACS Catal. 2016, 6, 7051−7058
Research Article
ACS Catalysis
nearest-neighbor distance of the CO molecules, adsorbed on the Au step edges.20 This resulted in activation energies for the reaction of O(ads) with CO(ads) of 26 ± 4 and 42 ± 5 kJ/mol or 0.27 and 0.44 eV, respectively, depending on the CO adsorption site. These barriers agreed well with the barriers found for CO oxidation with O(ads) on Au(211)43 and are between the barriers for Au(111)44 and Au(110)45,46 (see Table 1).
In all measurements, the XP spectra showed an increase in the intensities of the peaks associated with pristine D2O(ads). In the lower-temperature experiment (Figure 6a), this is most prominent, because the surface temperature was not sufficiently high to facilitate desorption. At 153 K, D2O can desorb and only a transient peak in D2O coverage was observed, which quickly decayed. Interestingly, D2O(ads) showed a higher thermal stability before the reaction as it remained adsorbed on the surface. This suggests that the heat of reaction assisted in D2O desorption. 3.2.1. Activation Energies. To determine the apparent activation energies, CO oxidation experiments were performed under a constant pCO of (8 ± 1.8) × 10−8 mbar. The experiments were repeated for three temperatures: 125 ± 2, 137 ± 0, and 153 ± 2 K. The decrease in the level of O(ads) and in the intensity of the ∼531.6 eV peak was fitted with an exponential decay:
Table 1. CO Oxidation Apparent and Reaction Activation Energies (in kilojoules per mole) for Single-Crystal Surfacesa surface
T range
Ea,app
Ea,r
ref
Au(111) Au(110)
250−375 275−440 200−400 200−400 125−153
−10 ± 3 8±2 −1.8 ± 0.9 −7 5−7 ± 2
10 ± 3
