Structure Matters: Asymmetric CO Oxidation at Rh Steps with Different Atomic Packing.

Curved crystals are a simple but powerful approach to bridge the gap between single crystal surfaces and nanoparticle catalysts, by allowing a rational assessment of the role of active step sites in gas-surface reactions. Using a curved Rh(111) crystal, here, we investigate the effect of A-type (square geometry) and B-type (triangular geometry) atomic packing of steps on the catalytic CO oxidation on Rh at millibar pressures. Imaging the crystal during reaction ignition with laser-induced CO2 fluorescence demonstrates a two-step process, where B-steps ignite at lower temperature than A-steps. Such fundamental dissimilarity is explained in ambient pressure X-ray photoemission (AP-XPS) experiments, which reveal partial CO desorption and oxygen buildup only at B-steps. AP-XPS also proves that A-B step asymmetries extend to the active stage: at A-steps, low-active O-Rh-O trilayers buildup immediately after ignition, while highly active chemisorbed O is the dominant species on B-type steps. We conclude that B-steps are more efficient than A-steps for the CO oxidation.

Introduction

CO oxidation is a fundamental pillar for the understanding of surface catalytic reactions, and, accordingly, it is among the most studied reactions in surface science.[1−4] However, the majority of scientific investigations have been performed with either single crystals or under ultrahigh vacuum, far from the operando requirements of industry. Therefore, an extrapolation of their conclusions to real catalytic systems (atmospheric pressures and powder/nanoparticle catalysts) may not be correct.[5] On one side, metallic nanoparticles comprise several facets, and monitoring their specific activity and interplay during a chemical reaction is challenging.[6] On the other side, the information that can be obtained using conventional flat crystals is restricted to one plane alone, and hence it is not representative of the structure of a real catalyst.[7] One approach to partially close this structure gap is to use cylindrical sections of single crystals, since their curved surface provides a smooth variation of the crystal orientation that allows to systematically compare different facets under the very same reaction conditions.[8−11]

CO oxidation on Rh is characterized by two well-defined stages, depending on the catalyst temperature.[12] At low temperature, adsorbed CO molecules block the O2 dissociative adsorption. Co-adsorption of reactants is not possible, and hence there is very low CO2 turnover.[13] At higher temperatures, however, most of the CO molecules will desorb from the surface, leaving empty sites that are typically occupied by oxygen species. Co-adsorption of reactants is now possible, leading to a substantial CO2 production.[12] The transition between the poisoned (CO-covered) and active (O-covered) stages occurs at the so-called light-off (or ignition) temperature Ti. A detailed knowledge on the CO ignition at different crystal facets is mandatory to tailor new catalysts and improve the energy costs of industrially relevant processes.

The adsorption of both CO and O2 has been previously studied on various Rh surfaces with atomic-scale precision, which facilitates operando studies of the CO oxidation reaction with the same level of detail, such as the present one. At low temperature, well below the ignition, CO will adsorb in Rh(111) terraces in top and hollow positions,[14−16] while it will anchor to top and bridge sites at the steps of A- and B-type surfaces.[17−19] On the other hand, O2 adsorption on flat and stepped Rh surfaces will lead to dissociation into single oxygen atoms at face-centered cubic (fcc) hollow sites, eventually forming oxide stripes as the oxygen coverage increases.[20,21] At higher O coverages, several works at different Rh facets and nanoparticles report the formation of the Rh surface oxide consisting of O–Rh–O (RhO2) trilayers.[20−27] Such surface oxide is a less-efficient catalyst for the CO oxidation than the metallic surface covered with chemisorbed oxygen,[12,28] possibly due to the fact that CO does not stick on the surface oxide and the reaction is restricted to oxide-metal boundaries.[20]

Here, we explore the evolution of chemical species during the thermal activation of the CO oxidation at A- and B-type vicinal surfaces simultaneously, using a curved Rh(111) crystal. Planar laser-induced fluorescence (PLIF) reveals an asymmetric light-off, where B-steps ignite at a lower temperature than A-steps. As judged by near ambient-pressure X-ray photoemission spectroscopy (NAP-XPS), such asymmetry is caused by the partial CO-depletion and O-accumulation exclusively at B-steps. During the active stage of the reaction, after the ignition of the entire sample, we observe surface oxide formation at the (111) terraces and A-steps, while the oxygen at B-steps remains chemisorbed. Therefore, since B-steps do not oxidize further during reaction conditions, we conclude that they are more active toward the CO oxidation than A-type steps.

Experimental Section

Curved Rh(111) Crystal

The c-Rh(111) sample (see bottom of Figure a) possesses the (111) plane at the apex of the crystal, and its cylinder axis is parallel to the [11̅0] direction. This leads to surfaces with a smooth increase of either A-type (100 microfacet) or B-type (111 microfacet) close-packed steps as one departs from the center of the crystal.[9,10,19,29] Vicinal angles (α), directly related to the step density,[19] up to α = ± 15°, can be probed using this sample. This allows to reach vicinal surfaces such as the A- and B-stepped (223) and (553) planes, with α = +11.4° and −12.3°, respectively. The curved surface was prepared by several cycles of Ar+ sputtering, O2 annealing and high-temperature flashes. As described in ref (19), this yields a contaminant-free surface with the expected variation of the step density across the sample.

Figure 1

(a) PLIF images acquired under a 24 mbar, 1:5 CO:O2 gas mixture. Argon was added to reach 100 mL/min and 150 mbar of total flow and pressures, respectively. The average heating slope was 12 K/min. Individual snapshots during the activation process, from the poisoned stage (485 K, no CO2 turnover) to the active stage (503 K), are shown. The sample design is illustrated in the bottom, together with the position of relevant surfaces across the curved crystal and the vicinal angle α scale (i.e., step density[38]). (b) CO2 signal in the QMS (mass 44), as a function of the temperature for individual experiments varying the CO:O2 pressure ratio. Vertical lines mark the two abrupt ignition steps that define the entire activation process. Movies of the ignition at CO:O2 pressure ratios of 1:5 (Figure a), 1:1, and 2:1 are provided in Supplementary Movies 1, 2, and 3.

Planar Laser-Induced Fluorescence

PLIF was employed to measure the CO2 production above the curved catalyst surface.[30,31] A broadband laser centered at a wavelength of λ = 2.7 μm is used to excite multiple vibrational transitions in the CO2 molecules. The most intense fluorescence transition occurs at λ = 4.3 μm, which is collected using a set of lenses and an IR camera. Since the laser is shaped into a sheet by a cylindrical lens, the fluorescence can be measured in 2D across the catalytically active sample.[10] Each image has a spatial resolution of 30 μm per pixel, and the acquisition rate was 10 frames per second. This yields a time resolution of 1 s after averaging 10 images. The experimental setup is described in more detail in ref (32). During measurements, the sample is placed in a flow reactor with optical access from four sides and heated using an inert boron nitride heater, where we measure the temperature according to ref (33).

Near-Ambient-Pressure X-ray Photoemission Spectroscopy

The NAP-XPS experiments were conducted at the In Situ and Operando Soft X-ray Spectroscopy (IOS, 23-ID-2) beamline of NSLS-II synchrotron, using normal emission geometry and 50° incidence angle, with respect to the (111) plane,[34] and Circe beamline of the ALBA synchrotron, using a 55° emission angle and a 15° incidence angle, with respect to the (111) plane.[35] The experiments at NSLS-II were performed using an inert boron nitride heater, and the chamber was pumped through the electron analyzer nozzle (<1 mL/min). For the experiments at ALBA, an encapsulated Pt filament was used as a heater, and the flux was ∼2.5 mL/min. Variations in the ignition temperature are expected due to the differences in the reactants ratio, heater, flow, and thermocouple position. Accordingly, the temperature for the 1:1 CO:O2 ratio NAP-XPS experiment shown in Figures d and 2e, as well as Figure S3b in the Supporting Information, is determined with ±15 K accuracy. No time evolution of the spectra was observed during the α-scans shown in Figures and 3 (∼2–3 h). In order to compare OAds during the early and late intermediate stages shown in Figure e, spectra at NSLS-II and ALBA were normalized to the height of TT-CO at the (111) plane. Signal-to-noise is lower in the ALBA experiments due to a faster acquisition mode, compared to the spectra measured at NSLS-II.

Figure 2

O 1s (hν = 650 eV) and C 1s (hν = 400 eV) spectra at (a) 0.1 mbar CO at 300 K (far from light-off) and (b) 0.7 mbar in a 1:4 CO:O2 gas mixture at 460 K, right after the first ignition step. Top, middle, and bottom rows correspond to the data acquired at the (553), (111), and (223) planes, respectively. TT, TH, ST and SB refer to CO molecules anchored in Terrace-Top, Terrace-Hollow, Step-Top and Step-Bridge sites. OAds and “C” stand for atomic oxygen at fcc sites and amorphous carbon, respectively. Insets illustrate the surface composition of the facets in each case. (c, d) O 1s photoemission intensity scans along the curved Rh surface (α-scan) for (c) the 1:4 mixture of panel (b) (0.7 mbar, 460 K), and (d) for a 1:1 CO:O2 gas mixture (1.0 mbar, ∼530 K). They respectively reflect an early and a late intermediate ignition stage. (e) OAds peak area across the c-Rh(111) sample, as determined from line fit analysis of the spectra in (c, d) α-scans (see also Figure S2 in the Supporting Information). NAP-XPS data in panels (b) and(c) have been acquired at NSLS-II with a flux smaller than 1 mL/min, and in panel (d) at ALBA with 2.5 mL/min flux.

Peak Fitting of the Photoemission Spectra

Peak fitting was performed with Python’s lmfit package.[36] Surface components were fitted to Doniac-Šunjić lineshapes[37] convoluted with a Gaussian profile, while Voigt profiles were considered for gas-phase peaks. Using the spectra at the (111) plane at 0.1 mbar CO at 300 K as a reference, chemisorbed CO peaks were constrained to have a similar width and asymmetry, while subtle changes in position (<50 meV) were allowed at the different temperatures to improve the fit. During the active stage, the asymmetry of the O-species was alike, although the high binding energy component of the RhO2 doublet was wider than the other contributions. After pumping the CO from the chamber, the RhO2 doublet become significantly more asymmetric, compared to reaction conditions.

Results

Asymmetric Light-off at A- and B-Steps

We first investigate the CO ignition across the curved Rh(111) sample [c-Rh(111)] by PLIF, which allows a spatial and temporal imaging of the CO2 production above the curved surface.[10,11] The c-Rh(111) crystal is identical to that described in ref (19), and appears sketched at the bottom of Figure a. It provides a smooth variation of the density of steps as one leaves the Rh(111) plane located at the center. At one side, one finds A-type close-packed steps (square, 100 microfacet), while B-Steps (triangular, 111 microfacet) are encountered at the other side. The step density is directly related to the vicinal angle α.[38] With this particular sample design, one can probe up to α = ±15°, allowing to reach the densely stepped (223) and (553) planes at each of the sample edges. Both (223) and (553) facets feature 5-atom-wide terraces (corner atoms included) separated by either A- or B-type monatomic steps, respectively. For convenience, we use α > 0 for A-Steps and α < 0 for B-Steps.

In PLIF experiments the clean c-Rh(111) sample was exposed to 24 mbar of a 1:5 CO:O2 gas mixture and subsequently heated to trigger the reaction. Argon was added as a carrier gas in order to reach a total flow of 0.1 l/min and a total pressure of 150 mbar. In addition, a quadrupole mass spectrometer (QMS) was employed to monitor the gases in the outlet of the cell. CO2 PLIF snapshots were continuously taken at different temperatures along the curved surface to track the CO ignition (see Figure a). At 485 K, no CO2 production is observed due to the CO poisoning of the entire surface. The CO2 cloud arises at 498 K at α ≈ – 5° in the B-side of the crystal (tick mark). At 501 K, the CO2 cloud extends toward the (111) plane located at the center, and at higher temperatures (503 K) it is detected above the entire c-Rh(111) sample. The PLIF images reveal an earlier ignition of B-type Rh(111) vicinals, followed by a progressive extension of the reaction toward the (111) center and the A-side. The activation of the entire c-Rh(111) sample marks the transition to the complete active stage. A somewhat different asymmetric ignition is observed in Pd,[10] while A-type and B-type surfaces ignite at the same temperature in Pt.[11]

The A-B asymmetry noted in PLIF is mirrored in the simultaneously acquired QMS signal, which, in turn, reveals a two-step ignition process. In Figure b we display the CO2 QMS intensity during separate ignition cycles with different CO:O2 gas ratios. Heating ramp, total pressure and total flow were kept constant. Looking to all curves, we immediately note the increase of the ignition temperature with the CO content, which is a well-known phenomenon.[1] Considering the 1:5 CO:O2 mixture of Figure a, we observe that the CO2 signal steadily increases as the sample is heated, but it steeply boosts at ∼488 and 501 K. PLIF images in Figure a allow one to correlate these steps with the two abrupt ignition events occurring at the B- (T, 488 K) and A-sides (T, 501 K) of the crystal, marking a ΔTAB of 13K. The minimum ΔTAB gap occurs at the 1:1 CO:O2 pressure ratio, and then becomes larger with both the amount of CO or O2 in the gas mixture. As we shall discuss below, such processes are the successive activation of B- and A-type atomic steps, separated by the ignition of the (111) terraces in the T-T temperature range, which we call the intermediate stage. The CO:O2 gas ratio strongly influences both the ignition temperature and the relative CO coverage at terraces and steps, and hence the ΔTAB gap is heavily dependent on it.

Evolution of Chemical Species during Light-off

NAP-XPS experiments were performed to explore the chemical species involved in the two-step ignition of the c-Rh(111) sample. For a quantitative characterization of the CO poisoning layer, it is useful to examine the CO adsorption alone, and compare with reference chemisorption experiments performed at low pressures.[19] To this aim, we exposed the c-Rh(111) crystal to 0.1 mbar CO at 300 K. Spectra acquired at three relevant Rh facets, namely, the (111), A-type (223) and B-type (553) surfaces, are shown in Figure a.

At the (111) plane, two well-resolved peaks are observed in both the O 1s and C 1s core level regions. We assign them to CO adsorbed in top (TT, 532.4 and 286.1 eV) and hollow (TH, 530.8 and 285.5 eV) terrace sites, respectively.[14,15,39] The intensity ratio TH/TT is close to 2, pointing toward the arrangement of the CO molecules in the (2 × 2)-3CO superstructure, with 0.75 ML (ML = monolayer, adsorbed molecules per substrate atom).[14−16] In the stepped (553) and (223) (top and bottom rows of Figure a, respectively), two more features are resolved in the C 1s region. They correspond to CO molecules adsorbed in top (ST, 285.8 eV) and bridge (SB 285.4 eV) positions at steps, and as previously reported they equally cover the surface.[18] Taking into account the 0.75 ML CO coverage at the (111) plane, one can calculate the ST- and SB-CO coverages, resulting in 0.07 ML each. These values agree well with the saturation model proposed in an earlier publication (1 CO molecule in ST and SB sites per 3 Rh atoms[19]). The CO saturation structure of the aforementioned surfaces are shown in the insets of Figure a.

The intensity of the TH-CO peak decreases from the (111) surface to the (223) and (553) planes, while that of TT-CO does the reverse. This is explained by considering a transition of some CO molecules from TH to TT sites, as TH-CO is known to become less favorable as the step density increases and the terraces narrow.[40] Step-related CO contributions cannot be resolved in the O 1s region,[39] hence the peaks are renamed as (SB + TH) and (ST + TT). The CO(g) emission peaks at 536.5 eV. Residual amorphous carbon (“C”, at 284 eV) at the stepped surfaces may either arise from the CO dissociation or adsorption of residual hydrocarbons.

Next, we proceed to activate the catalytic CO oxidation by adding O2 to the mixture and then increasing the temperature. The mixture is fixed to a CO:O2 ratio of 1:4 with a total pressure of 0.7 mbar. The temperature is increased until a significant increase of the CO2 production is detected at 460 K. At this temperature, the CO2 turnover is roughly half of that achieved upon full sample activation, which indicates that we lie between the first and the second ignition steps, i.e., in the intermediate stage. Spectra are shown in Figure b. Measurements at temperatures below the first ignition step are discussed in Figure S1 in the Supporting Information. Note that absolute temperature values may differ from those in PLIF experiments, which were performed under very different conditions, particularly flux.[11] Because of the high turnover, CO(g) coexists now with CO2(g) in the spectra (shown in Figure S2 in the Supporting Information).

O 1s and C 1s spectra have significantly changed in the (111) plane, compared to 300 K. CO anchored at TH sites has reduced its intensity by more than 60% at 460 K (middle row of Figure b). The reason for this is that the CO adsorption energy is lower in TT sites, compared to TH sites.[14,15] An extra contribution close to 529.5 eV is also detected, which is ascribed to chemisorbed atomic oxygen (OAds) at hollow fcc sites.[20,21,41] The desorption of TH-CO creates empty surface sites, which allow the O2 dissociative adsorption to occur and explains the presence of OAds. Furthermore, the decrease in TH-CO also accounts for the slight increase of TT-CO, since a portion of the remaining adsorbed CO molecules may rearrange into R30°-CO domains. These feature a larger TT-CO coverage than the (2 × 2)-3CO superstructure, hence explaining its growth just before the ignition.

The C 1s spectrum at the (223) plane at 460 K (top row of Figure b) is quite similar to that acquired at 300 K in the CO atmosphere alone: the step peaks (ST- and SB-CO) are identical, while both terrace contributions decrease, but particularly TH-CO. In contrast, the (553) plane (bottom row of Figure b) has evolved substantially: in addition to the large decrease of TH-CO (∼70%), the feature related to SB-CO has vanished from the C 1s region. Therefore, the photoemission spectra reveal a significant asymmetry after the first ignition step: A-type steps remain close to CO saturation, while CO has significantly desorbed from B-steps. Meanwhile, a significant amount of CO has also desorbed from TH sites everywhere. Since desorption is the opposite to adsorption, these results agree with the fact that sticking at TH and SB sites occurs at a less efficient rate, compared to TT and ST positions at Rh vicinals,[17−19] while CO adsorption at TH and SB sites is faster at A-type Rh steps.[19]

CO-related peaks show the same effect in the O 1s spectra, i.e., nearly the saturation intensity at A-steps and a strong decay at B-steps. Here, we focus on the OAds peak. Its intensity is roughly half of the TH-CO feature at the (111) plane, and it is very small at the (223) surface. This points toward adsorption of oxygen at terrace fcc terrace sites at (111) and (223) facets. In contrast, OAds is almost double that of the (SB + TH) contribution in the (553) plane, and it is also larger than OAds at the (111) surface. We conclude that, together with a minor adsorption at terraces, oxygen adsorbs majorly at fcc sites of B-stepped edges. In fact, the C 1s analysis reveals partially CO-depleted B-steps, which allow oxygen to accumulate and start the CO oxidation earlier than at the CO-poisoned A-type steps, as observed in PLIF. The chemical composition of each surface is illustrated in the insets in Figures a and 2b.

The smooth variation of the surface orientation of curved crystals allows us to finely investigate the surface species as a function of the vicinal angle α. By scanning the curved surface with the small synchrotron ligth beam, one can create photoemission intensity maps that image the distribution of adsorbates across the different vicinal planes.[9,11] In Figure c, we show the O 1s α-scan for the 1:4 CO:O2 mixture of Figure b. The (SB + TH) feature steadily grows from the center to the top of the image (A-side, α > 0), while it decreases from the center to the bottom (B-side, α < 0). This confirms that CO adsorbs to SB sites solely at A-steps, and not at B-steps. OAds follows the reverse behavior, hence oxygen sticks to the CO-depleted B-steps and not to the CO-poisoned A-steps.

The dashed white line in Figure c marks the crossover from a (SB + TH)-covered surface to a OAds-covered one at α ≈ −5°. This α angle is coincident with the center of the emerging CO2 cloud in the 498 K PLIF image of Figure b, which may simply reveal that the α-scan in Figure c characterizes an “early” intermediate stage, immediately after the first ignition step. In fact, altering the reaction conditions is possible to reach a “late” intermediate stage, closer to the second ignition step, where the intensity and distribution of (SB + TH) and OAds species across the c-Rh(111) surface are markedly different. This is shown in the α-scan of Figure d, which has been obtained for a 1:1 CO:O2 mixture at ∼530 K. We immediately notice that the (SB + TH)-OAds crossover point has shifted toward the A-side the crystal, as it happens with the CO2 cloud in PLIF as the ignition progresses (Figure b).

A quantitative comparison of the OAds intensity between Figures c and 2d surface allows deeper insights into the evolving chemical composition of the intermediate stage. This analysis is shown in Figure e, where data points are determined from the peak fit of individual spectra in the respective α-scans (selected spectra are shown in Figure S2). Generally, there is a larger OAds signal in the 1:1 case, as expected for a surface that is more depleted of CO at a more advanced ignition stage at a higher temperature. The α-dependent trends are also different in both cases, revealing a different OAds filling of terrace and step sites in each case. To understand the intensity variation in these curves one must note that, because of the lateral extension of the steps, terrace and step signals in a curved surface are expected to respectively decrease and increase as a function of α.[38] In the 1:1 case, the OAds intensity is virtually constant at the B-side, indicating that both steps and terraces are equally occupied by oxygen. In contrast, at the A-side, where steps remain CO-poisoned, the intensity decreases linearly as a function of α, reflecting the decaying contribution of OAds-covered terraces to the total peak emission. In the 1:4 case, the linear decrease in the A-side also reveals exclusive, but minor OAds occupation of terraces. In contrast, the B-side exhibits a α-dependent increase of the OAds intensity, which reflects the oxygen adsorption of B-step sites.

In summary, the surface chemistry analysis of Figure allows one to explain the two-step ignition of the c-Rh(111) surface discussed in Figure . Abrupt changes in the CO2 turnover are due to the sequential activation of B-steps at T, due to CO desorption from SB sites and OAds adsorption, and afterward A-steps at T. Activation of (111) terraces appears to be progressive and occurs at intermediate temperatures, when CO desorption allows OAds occupation of TH sites. Although CO desorption from TT sites from both sides of the crystal may be most significant at T, a sizable amount remains at higher temperatures in nonoxidized facets, as we discuss next.

Chemical Species in the Active Stage

Further heating of the 1:4 CO:O2 mixture to 470 K triggers the reaction. Most of the CO leaves the surface, CO(g) vanishes from the spectra and the maximum CO2 production is reached, marking the activation of the entire sample. Figure a shows O 1s α-scans acquired at different temperatures above the ignition point (470, 510, and 570 K), while Figures b–d display fitted spectra for selected positions across the curved crystal. They unveil, in the most direct way, a striking step-related trend in surface oxidation during the active stage: the progressive formation of O–Rh–O (RhO2) trilayers from the A-side toward the B-side of the crystal as the temperature is increased. At 470 K, only the spectra at α > 6° (A-side of the sample) feature a pair of peaks at 529.5 and 528.5 eV, while a single contribution is observed in the rest of the crystal. Such doublet represents the superficial and interstitial O-layers of the RhO2 surface oxide trilayer.[20,21] This means that, at this temperature, a O–Rh–O trilayer forms at densely stepped surfaces with A-steps (two well-defined features at 529.5 and 528.5 eV), while oxygen remains in its more active chemisorbed form at (111) terraces and B-steps (single peak at ∼529 eV). At the B-side, OAds exhibits an α-dependent binding energy shift, as well as a peak broadening toward the (553) edge of the sample (Figure b), which likely reflects the presence of terrace and step species that cannot be resolved.[20,21] The scenario is slightly different after reaching 510 K: the RhO2 doublet arises at the (111) plane, indicating that terraces start to form O–Rh–O trilayers as the temperature of the sample increases. The oxidation of the terraces is further enhanced after reaching 570 K, and the RhO2 signal extends to the B-side up to α = −6°. RhO2 steadily decreases in the range of −6° < α < −10°, while beyond α = −10°, only OAds is detected. At 570 K (Figure c), the oxygen remains chemisorbed at B-steps under reaction conditions, while the other facets of the sample have fully developed RhO2 trilayers.

Figure 3

(a) O 1s α-scans acquired at 470, 510, and 570 K under a 1:4 CO:O2, 0.7 mbar gas mixture. Spectra were taken at hν = 650 eV at NSLS-II, after those shown in Figures b and 2c. (b, c) Individual fitted spectra of the (223), (111), and (553) surfaces at 470 and 570 K under reaction conditions, as well as at (d) 570 K after pumping the CO (light-beige panel). OAds, RhO2, and Oxide stand for chemisorbed atomic O, surface oxide trilayers, and an uncharacterized Rh oxide, while CO molecules anchored to Terrace-Top sites and carbonate/carboxyl species are denoted as TT and COX, respectively. Vertical dashed lines are included in panels (b) and (c) to illustrate the shift of Oxide, OAds, and CO2(g) with α. The latter is caused by a varying work-function across the curved crystal,[9] which, in turn, reflects the strong differences in the local surface composition and structure.

Important details inside the α-scan images are better observed in the single spectra shown in Figures b and 3c. Similarly to Pd,[9] Ir,[42] and Ru,[43] and in clear contrast to Pt,[42,44] a sizable amount of CO can still be detected at both the (111) and (553) Rh facets after the ignition. The significantly smaller CO peak at the (223) plane reflects the fact that RhO2 quenches the CO adsorption.[20] An additional feature arises at 533.3 eV at the (223) surface. As judged by its binding energy, we assign this peak to carbonates/carboxyls (COX)[45] anchored exclusively at the A-steps. Small extra contributions [531.0 eV at the (111) plane, 530.5 eV at the (223) plane] are observed in parallel to the surface oxide doublet, which we attribute to a different Rh oxide. This feature vanishes once the CO is removed from the gas mixture, suggesting a reaction-stabilized oxide[46] or reaction intermediate.[44]

Under the oxidative reaction conditions of Figure c, OAds can only be removed from the B-edge of the sample by closing the CO valve, which leads to a progressive pumping of CO from the chamber, and to a pure O2 atmosphere within minutes. The resulting spectra are shown in Figure d. The surface oxide covers the entire sample, but peaks become more asymmetric. This points to the buildup of additional, unresolved oxidic species, which would contribute to the high binding energy tail of the RhO2 doublet. Similar spectra are obtained in the reference oxidation shown in Figure S3 in the Supporting Information. The fact that removing the CO from the gas feed is required to further oxidize B-stepped surfaces is very meaningful. It reveals a larger CO oxidation rate at B-Steps, compared to A-Steps, since the formation of the surface oxide would be hindered if the CO oxidation kinetics are faster than those of the metal oxidation.[47,48] Surface reduction kinetics are further accelerated, and no surface oxide is formed if the CO:O2 ratio is closer to stoichiometry (1:1 pressure ratio, see Figure S3). Although faster CO oxidation kinetics are expected for B-steps rather than A-steps, we could not estimate this effect with the data available. After the ignition, the reaction reaches the mass-transfer limit (MLT), where the CO2 cloud blocks the diffusion of reactants toward the Rh surface and limits the reaction rate.[49] As discussed in Figure S4 in the Supporting Information, the CO2(g) peak exhibits the same random intensity variation as the O2(g) feature across the c-Rh(111) surface. This indicates that, because of insufficient pumping, the CO2 cloud equally covers the sample, and hence differences in turnover frequencies cannot be estimated.

Klikovits and co-workers studied the oxidation of A- and B-type Rh(111) steps, observing closed and open oxide structures for A- and B-type steps, respectively.[50] The open oxide would be tentatively easier to reduce than the closed one, simply explaining the larger activity of B-steps over A-steps for the CO oxidation that we postulate. In any case, RhO2 is known to be less active toward CO oxidation than OAds.[12,28] On the other hand, Gustafson et al. concluded that specific Rh crystal planes exposed during catalysis will not directly influence the activity.[27] However, we have shown that there is a clear A-B asymmetry in both the ignition temperature and composition of the active stage of the CO oxidation under the very same experimental conditions. Finally, we must not discard faceting of any of the aforementioned surfaces, since Rh vicinals are known to undergo faceting under oxidative[21,22] and CO oxidation conditions.[23] This calls for a combined effort for the simultaneous probing of the chemical species and surface structure during the CO oxidation reaction.[51]

Conclusions

In summary, we have investigated the role of A- and B-type steps during the CO oxidation at millibar pressures using a curved Rh(111) crystal. PLIF images reveal an asymmetric, two-step light-off, where B-type vicinals ignite at lower temperature than A-type surfaces. NAP-XPS points toward a significant CO-depletion and O-accumulation exclusively at B-steps just before the full sample ignition, while A-steps remain CO-poisoned. After the light-off is completed, A-steps readily develop the less-active RhO2 trilayers, while oxygen at B-steps remains chemisorbed and no RhO2 is formed under reaction conditions. Therefore, we conclude that B-steps are more active toward the CO oxidation than A-steps. Our experiments using curved surfaces emphasize the need of operando studies on the influence of steps and their interplay with other surface sites in chemical reactions.