Imaging Interface and Particle Size Effects by In Situ Correlative Microscopy of a Catalytic Reaction

The catalytic behavior of Rh particles supported by three different materials (Rh, Au, and ZrO2) in H2 oxidation has been studied in situ by correlative photoemission electron microscopy (PEEM) and scanning photoemission electron microscopy (SPEM). Kinetic transitions between the inactive and active steady states were monitored, and self-sustaining oscillations on supported Rh particles were observed. Catalytic performance differed depending on the support and Rh particle size. Oscillations varied from particle size-independent (Rh/Rh) via size-dependent (Rh/ZrO2) to fully inhibited (Rh/Au). For Rh/Au, the formation of a surface alloy induced such effects, whereas for Rh/ZrO2, the formation of substoichiometric Zr oxides on the Rh surface, enhanced oxygen bonding, Rh-oxidation, and hydrogen spillover onto the ZrO2 support were held responsible. The experimental observations were complemented by micro-kinetic simulations, based on variations of hydrogen adsorption and oxygen binding. The results demonstrate how correlative in situ surface microscopy enables linking of the local structure, composition, and catalytic performance.


■ INTRODUCTION
Size and support effects of catalytically active metal particles have been among the most studied phenomena in heterogeneous catalysis for over 50 years. 1−4 A vast number of ex situ and in situ studies have been carried out by spectroscopic, diffractive, or imaging techniques, both on technological as well as on model catalytic systems. 5−7 Low-coordinated step/ edge sites and interfaces were often found most active. 8−14 In the present work, we study the effect of various metal/support interfaces on catalytic H 2 oxidation on Rh by in situ correlative microscopy, i.e., the same area of the same sample is imaged under identical reaction conditions by different microscopies, providing real-time complementary information. Accordingly, interface phenomena and particle size effects can be directly observed by spatially resolved evaluation of surface structure, surface composition, and catalytic performance.
The correlative microscopy concept stems from biological research, where the same cell/tissue structures were studied by light and electron microscopy already in the 1970s. 15,16 This has been extended to other imaging techniques such as atomic force microscopy, single molecule fluorescence, X-ray tomography, or scanning electron microscopy 17,18 and has made a huge impact in materials research. 19,20 In catalysis, correlative microscopy has also been applied, e.g., by combining transmission electron microscopy (TEM) with fluorescence microscopy, atom probe tomography (APT) with scanning transmission X-ray microscopy, scanning photoelectron microscopy (SPEM) with APT, or optical microscopy with confocal and X-ray fluorescence microscopy. 21 −25 On planar Rh samples, catalytic H 2 oxidation has been studied in a correlative way by various combinations of UV or X-ray photoemission electron microscopy [(UV or X)PEEM], SPEM, or low-energy electron microscopy. 26−28 Since the initial studies in the 1960s, these techniques have been developed into powerful surface science tools. 29−32 A wide range of surface processes were investigated, e.g., layer growth, 33 diffusion, 34 phase transitions, 35 adsorption, 36 oxidation, 37 or chemical reactions, 38 with Gerhard Ertls's work on the oscillating CO oxidation reaction on Pt 39 being the prime example. XPEEM alone can already provide correlative spectroscopic data and microscopy images, but its capabilities are limited in terms of spectroscopy of sub-micrometer areas and imaging areas larger than several micrometers. 28, 40 Thus, a combination with other techniques is often beneficial. Apart from the known steady states of high and low activity and bistability, 41−44 previous studies shed light on the spatiotemporal phenomena in catalytic H 2 oxidation on Rh, including mono-and multi-frequential self-sustaining oscillations, 12,45,46 coexisting multi-states, 47 or the emergence of chaos. 48 For supported Rh particles, such an approach has not yet been applied. Based on the thorough understanding of the mechanism of hydrogen oxidation, it can be used as a test reaction to probe catalytically inactive and active states. By assessing the parameter regions for these states, interface and size effects in different catalysts can be benchmarked.
Catalytic H 2 oxidation is also relevant in various areas of technology: On one hand, excess electric energy can be stored in the chemical bonds of H 2 and released on demand in a fuel cell, being one route toward renewable energy generation and storage. 49 On the other hand, for some applications, traces of H 2 must be safely removed. This includes, e.g., treatment of fuel cell off-gas, 50 nuclear power plants, 51 or feeds of catalytic reactors. 52 Furthermore, catalytic H 2 oxidation also occurs in hydrogen gas sensors. 53 Here, we present the first correlative microscopy study of catalytic H 2 oxidation on Rh powder aggregates (size of ∼5− 30 μm) supported by different materials: A single sample comprising Rh aggregates supported either by Rh, Au, or ZrO 2 was studied in situ by combining PEEM (providing local reaction kinetics) and SPEM (providing local chemical information). The Rh/Rh system represents a reference for unmodified (quasi-unsupported) Rh, with the particles nonetheless having the same curved/stepped surface morphology as in the Au and Zr-oxide supported systems. Rh/Au was initially prepared to provide an inert support but finally served as a model system for active particles modified by an inactive second metal, possibly by active site blocking 54 or surface alloying. This configuration is comparable to, e.g., Au-or Cumodified Ni-based steam reforming catalysts 55,56 or Aumodified Cu-based catalysts for water gas shift. 57 As an example for oxide supported metals, the Rh/ZrO 2 system was used, which is catalytically active in steam reforming, 58 CO and CO 2 hydrogenation, 59 or H 2 oxidation. 44,60 Late transition metals (including platinum group metals) supported on various oxides (including ZrO 2 ) are well-known for strong metal−support interaction (SMSI), 61−69 i.e., fully or partially oxide decorated metal surfaces. In the present study, we demonstrate that the mechanisms of surface modification are very different, affecting both steady state and oscillatory hydrogen oxidation, but the modification of steps was crucial in both cases.
The experiments are illustrated in Figure 1: H 2 oxidation occurring simultaneously on the three sample regions (Rh/Rh, Rh/Au, Rh/ZrO 2 ) was visualized in situ by PEEM. The sample is illuminated by UV-light and the emitted photoelectrons form an image on a fluorescent screen. The image brightness depends on the adsorbate coverage via the work function and is thus directly related to catalytic activity (kinetics by imaging). 70 Therefore, spatially resolved reaction kinetics, e.g., hysteresis curves and kinetic phase diagrams, can be extracted by analyzing PEEM video sequences.
To complement local kinetics with spatially resolved chemical information, the same sample regions were studied in situ by SPEM under identical reaction conditions, with the sample raster-scanned under a focused sub-micrometer synchrotron X-ray photon probe. The energy distribution of the emitted photoelectrons is measured, resulting in highresolution X-ray photoelectron spectra (XPS) and elemental/ chemical maps on a micrometer scale.
■ RESULTS AND DISCUSSION Kinetic PEEM Studies. PEEM studies were carried out in an ultrahigh vacuum (UHV) setup operated as a flow reactor (for details, see the Supporting Information). First, the kinetic transitions between the steady states of catalytic activity in H 2 oxidation were studied on Rh/Rh. The reaction follows the Langmuir−Hinshelwood mechanism, 41,71 i.e., both reactants adsorb on the catalyst before reacting. At low p(H 2 )/p(O 2 ), the system is in an inactive steady state (adsorbed oxygen blocks the hydrogen adsorption, ball model in Figure 1a). Upon increasing p(H 2 )/p(O 2 ), the system switches to a catalytically active steady state. This, and the reverse, occur via kinetic transitions, which can be visualized due to the dependence of the PEEM image intensity on surface coverage: The catalytically inactive oxygen-covered Rh surface has a higher work function than adsorbate-free Rh, resulting in darker PEEM contrast. In turn, the catalytically active surface with low adsorbate coverage appears much brighter. These image contrast variations allow extracting local kinetic information from in situ recorded PEEM videos (kinetics by imaging). 43,45,70 The observed kinetic transitions are illustrated in Figure 2: Experiments always started from the catalytically inactive state (shown for Rh particles on Rh foil in Figure 2a). Upon increasing p(H 2 ) at constant p(O 2 ) and T, a kinetic transition to the catalytically active state takes place at a certain p(H 2 ) From the PEEM video sequences, hysteresis curves and kinetic phase diagrams can be obtained for selected regions (ROIs). (b) In SPEM, the sample is raster-scanned under a sub-μm synchrotron X-ray beam, and energy analysis of the emitted photoelectrons provides local X-ray photoelectron spectra (XPS) and elemental/chemical maps on various length scales. value τ A . On Rh/Rh this is accompanied by H ads fronts nucleating at the perimeter of the particles 44 and spreading over the whole field of view ( Figure 2b). After the kinetic transition, the system stays in the active steady state ( Figure  2c). Upon decreasing p(H 2 ), a reverse kinetic transition to the inactive state will take place at a certain p(H 2 ) value τ B ( Figure  2d), which differs from τ A , exhibiting a hysteresis and indicating bistability. 41−44 From recorded video sequences, the local PEEM intensity can be read out for any region of interest (ROI). For Rh/Rh, a ROI was placed, e.g., on the "big" particle in the middle (marked in Figure 2a) and the local PEEM intensity vs p(H 2 ) upon cyclewise variation of p(H 2 ) is shown as black trace in Figure 2d. The curve exhibits the expected pronounced hysteresis between the kinetic transition points τ A and τ B . To examine the effect of different supports, identical experiments were performed for Rh/Au and Rh/ ZrO 2 . Similarly sized Rh particles (diameter ∼20 μm) were selected for analysis, and the results for Rh/Au (red) and Rh/ ZrO 2 (blue) are shown in Figure 2d. A hysteresis was also observed for both but with a much wider and shifted loop.
By conducting experiments at different temperatures and plotting the measured τ A and τ B values vs the inverse temperature, kinetic phase diagrams can be constructed. For Rh/Rh (Figure 2e, black), the unexpected sharp bend in the τ A line can be attributed to oxygen-induced surface Rh restructuring and corresponding roughness changes at a certain temperature, reported for several single crystals, 72−75 Rh nanoparticles, 62,76 or needle-shaped specimens. 77 Due to the rounded shape of the Rh particles/aggregates, exposing plenty of differently oriented stepped surfaces ( Figure S5), similar behavior can be expected in the present case.
The kinetic phase diagram of Rh/Au (red) is similar to the one of Rh/Rh, but τ A and τ B are both shifted to higher p(H 2 ). Since (bulk) Au is catalytically inactive in H 2 oxidation due to the lack of dissociative adsorption of the reactants, 78,79 and no interaction between the Rh particles and the Au support is expected, this seems unusual and will be explained below.
For Rh/ZrO 2 (blue diagram), the τ A and τ B values are also shifted to higher p(H 2 ) in comparison to Rh/Rh, but the bistability area does not exhibit a sharp bend in τ A . The shift toward higher p(H 2 ) has been discussed in our previous work. 44 Due to the electron density jump across the metal/ oxide interface, the binding energy for oxygen is modified in close vicinity of the interface. Since the energetics govern the adsorption kinetics, the local oxygen/hydrogen adsorption equilibrium is shifted toward oxygen, necessitating compensation by higher p(H 2 ) to induce kinetic transitions. Furthermore, it appears that the presence of the metal/oxide interface inhibits the oxygen-induced restructuring observed on Rh/Rh and Rh/Au.
Besides the "big" Rh particles, other differently sized particles are present in the field of view (Figure 2a−c) and the in situ recorded video sequences contain kinetic information for all particles at inherently identical conditions, allowing the addressing of size effects. Several particles of various sizes were analyzed, and representative examples were selected. Figure 2f,g shows the kinetic phase diagrams analogous to Figure 2e, but additionally for a set of "small" particles (∼5 μm). The corresponding PEEM images are shown in Figure S1. Comparison of diagrams for differently sized particles reveals interesting features: For Rh/Rh, the kinetic phase diagrams are identical within experimental accuracy, which is expected as there are no interface effects between the Rh particles and Rh support. The same seems true for Rh/Au, but the shift in comparison to Rh/Rh remains to be explored below. In contrast, for Rh/ZrO 2 , there is a clear particle size effect: The τ A values of the "small" particle are shifted toward higher p(H 2 ), while the τ B values remain basically identical. This demonstrates the contribution of the particle/support interface to the catalytic activity. For smaller particles, the ratio of perimeter/interface length to surface area is higher than that for bigger particles: Smaller particles are more strongly influenced by the interface than bigger ones.
The PEEM results unambiguously show interface and particle size effects but lack detailed chemical information. In particular, the behavior of the Rh/Au system cannot be explained by PEEM alone, calling for a chemically sensitive technique.
Chemically Resolved SPEM Studies. Therefore, SPEM experiments, with the SPEM chamber also operating as a flow reactor, were performed using a photon energy of 720 eV at the ESCA Microscopy beamline (Elettra synchrotron), as described in detail in the Supporting Information along with information on the spectra deconvolution procedures.
In Figure 3, the same "big" Rh particle on Au as in PEEM was probed in the catalytically active state (the Rh/Rh system provides no chemical contrast). Figure 3a shows highresolution in situ Au 4f 7/2 spectra of the two spots marked in Figure 3d: Spot A on the support and spot B on the Rh particle. While spot A shows expected bulk Au, significant Au amounts are detected on the Rh particle as well. There are three possible reasons: First, the cleaning procedure includes Ar + -ion sputtering, possibly leading to deposition of support material onto the Rh particles. Such redeposition has been observed on several substrates and under various sputter conditions. 80,81 Second, Au adatoms can diffuse on platinum group metals at temperatures as low as 300 K. 82,83 Third, kinetic transitions in a surface reaction may cause a large-scale redistribution of adsorbed Au. 84 Furthermore, the spectra show two distinct Au species, bulk Au (support material, red) and another one only on the Rh particle (green). We attribute this second component to the formation of a RhAu surface alloy, in analogy to studies of thin Au films on Al, Rh, or Ru. 85−87 The corresponding Rh 3d 5/2 spectra (Figure 3b) indicate no presence of Rh on the support material. We conclude that the presence of Au on the Rh particles is caused by a combination of the effects mentioned above: Au atoms are deposited on the Rh particle in close vicinity to the interface due to sputtering, from where they diffuse and migrate, finally covering the whole particle. The spectrum on the particle, deconvoluted following a previously established procedure, 47 consists of a bulk Rh component (Rh bulk , orange) and a component related to oxygen-bound Rh (Rh O1/4 , red), which unambiguously indicates the catalytically active state. 47 The O 1s spectra (Figure 3c) highlight that the Au support does not take part in the reaction (no oxygen is present in spot A), while the spectrum for the Rh particle (spot B) agrees with our previous observations. 47 Assuming a simple homogeneous layer model for analysis of the Au 4f 7/2 versus Rh 3d 5/2 absolute intensities yields an average Au thickness of below one monolayer. Au is, however, known to form islands and multi-layered clusters 84,87 on Rh, resulting in a significantly more complex morphology.
Using the Au 4f 7/2 and Rh 3d 5/2 signals after correction for the different inelastic mean free paths and X-ray cross-sections, a chemical map of the atomic concentrations of Au (blue) and Rh (orange) was constructed (Figure 3d). The map reveals that there is hardly any gradient in the Au concentration on the Rh particle with the entire particle being covered. A Au 4f 7/2 chemical map detailing the lateral distribution of the two different Au species is displayed in Figure 3e, whereby the relative contributions of the individual species to the total Au 4f 7/2 signal are shown. The map completes the picture and supports our component assignment of RhAu on the particle. The individual contributions to the maps displayed in Figure  3d,e are depicted in Figure S2.
Based on SPEM, the unexpected behavior of Rh/Au in PEEM can be explained: Au atoms migrating onto the Rh particle are very likely located at the Rh step edges, in line with previous observations on Pt, Rh, or Ru. 54,83,87 Since the Rh step sites are crucial for dissociative hydrogen adsorption 88 and thus for switching to the active state, their lower activity due to  Table  S2.
RhAu alloy formation must be compensated by a higher p(H 2 ), shifting the kinetic phase diagram as observed.
In analogy to Rh/Au, in situ SPEM of the "big" Rh particle on ZrO 2 in the active state is shown in Figure 4. In Figure 4a− c, high-resolution Zr 3d, Rh 3d 5/2 , and O 1s XPS spectra are displayed for spots A to E placed along a line from the oxide support to the Rh particle center. The Zr 3d spectra ( Figure  4a) display three different Zr x+ species (Zr 4+ bulk-oxide, red; Zr 3+ sub-oxide, green; and Zr 2+ sub-oxide, blue), identified based on previous XPS studies of thin ZrO x islands/films on Rh(111) 61 or Pt(111) 68 and of the initial oxidation of Zr. 89, 90 Similarly, detailed analysis of the Rh 3d 5/2 spectra (Figure 4b) reveals the presence of bulk Rh (Rh bulk , orange), two components associated with oxygen-bound Rh (Rh O1/4 , red, and Rh O2/3 , green) as well as a component related to Rh surface oxide (Rh Ox , blue), in accordance with our previous work. 27,47 The O 1s spectra (Figure 4c) complete the picture: One component is associated with bulk ZrO 2 (red), one component is related to the Zr sub-oxides (green), and one component represents oxygen adsorbed on Rh (blue).
Combining all of the information yields the following: At spot A (i.e., on the support, several μm away from the particle), there is bulk-like ZrO 2 . At spot B, still on the support, but near the metal/oxide interface, Zr sub-oxides are present in addition, but there are no significant amounts of Rh. This can be understood by partial reduction of ZrO 2 , possibly by hydrogen spillover, as, e.g., observed for Rh or Pt particles on ZrO 2 . 61,91,92 Computational studies have shown that reduction of nano-sized ZrO 2 is facilitated in comparison to bulk ZrO 2 , making partial reduction feasible under the present conditions. 93 On the Rh particle near the metal/oxide interface (spot C), significant amounts of zirconia are present mostly as Zr sub-oxides, while the RhO 1/4 component again indicates the active state. Further toward the particle center (spot D), the amount of Zr sub-oxides decreases and the Rh 3d 5/2 spectrum reveals an unexpected shoulder due to a distinct Rh surface oxide. It seems that small amounts of Zr sub-oxides locally enhance the oxidation of Rh, probably as a result of the stronger oxygen bonding close to the metal/oxide interface. 44,94 Even more toward the particle center (spot E), there is no Zr present and the Rh surface is in the active state.
Based on reports that the growth of ZrO x 2D-islands preferentially started at step and defect sites of a Pt substrate, 95,96 it can be suggested that the ZrO x layers on Rh near the interface may originate from a limited mobility of Zr− O species at temperatures around 450 K. However, the  Table S2. observed formation of small ZrO x islands closer to the Rh center rather takes place by sputter-induced support deposition on the Rh particles, where it forms small, hardly mobile oxide islands.
A chemical map (Figure 4d) of the atomic concentrations of O (red), Rh (orange), and Zr (blue) was constructed from the O 1s, Rh 3d 5/2 , and Zr 3d signals. The map shows a significant coverage of ZrO x on the Rh particle. The island density of ZrO x is, however, not as uniform as for Au on Rh, but decreases toward the particle center, which remains free of ZrO x . In addition, the Zr/O atomic ratio differs up to 50% between the support (appearing purple) and near the interface (appearing blue). This is also reflected in the Zr 3d chemical map (Figure 4f), constructed using the same three components as in the spectra (Figure 4a), which shows the absolute background-normalized contributions of the individual species to the total Zr 3d signal. The Rh 3d 5/2 chemical map ( Figure  4f), constructed using the absolute, background-normalized contributions of the individual species to the total Rh 3d 5/2 signal reveals mainly bulk metallic Rh (orange) being present on the "big" particle, while small patches of Rh surface oxide (blueish dots) are also formed. Smaller particles (e.g., in the top right corner) appear to be strongly oxidized. The corresponding O 1s chemical map (Figure 4g), again displaying the absolute background-normalized contributions of the individual species to the total O 1s signal, shows the expected gradient between support (appearing mainly red), interface region (appearing green), and the center region (appearing in a slight bluish tint). The individual contributions to the maps displayed in Figure 4d−g are depicted in Figure  S3.
The SPEM results demonstrate an unexpected complexity of the seemingly simple metal/support interfaces: Due to the mobility of Au atoms, the Rh particles were decorated by significant amounts of Au, strongly modifying the catalytic properties by RhAu alloy formation. In the Rh/ZrO 2 system, due to the formation of ZrO x islands, a complex interface region spanning several micrometers has been formed, resembling strong metal−support interaction (SMSI) states observed on several platinum group metal nanoparticles on various oxides. 61−64,66,68,69,97 Furthermore, the previously identified modification of binding energies at the metal/oxide interface 44,94 appears to be complemented by the formation of sub-stoichiometric Zr oxides, hydrogen spillover onto the ZrO x support near the particle, and a narrow region where Rh surface oxide is present, altogether resulting in the observed support-induced modifications of the catalytic properties ( Figure 2).
Oscillating Reaction Mode. Catalytic H 2 oxidation may exhibit self-sustaining oscillations between the active and inactive state at particular constant reaction parameters, as observed on polycrystalline Rh foils 45−47 and hemispherical apexes of nanometer-and micrometer-sized Rh specimens. 12,13,98 The oscillation frequency and the parameter window for oscillations are very sensitive to surface structure and composition, 12,45−47 turning the oscillating H 2 oxidation on Rh into a sensitive probe. The well-established mechanism of oscillations ( Figure S4) is based on the formation and depletion of subsurface oxygen. 45−47 However, as for supported Rh particles, self-sustained oscillations in H 2 oxidation have so far not been observed, their existence and parameter space should be explored.
At constant external parameters, oscillations were visualized by PEEM for Rh/Rh and Rh/ZrO 2 (but did not occur for Rh/ Au). For Rh/Rh, selected PEEM image cutouts for the "big" and "small" particles already studied in the steady states and the corresponding local PEEM intensity time series are shown in Figure 5a. Both particles exhibit the same oscillation frequency within the experimental accuracy. It is important to note that while the Rh substrate acts as excitable medium and therefore also exhibits oscillations, the nucleation of the reaction fronts transmitting the oscillations always takes place on the Rh particle perimeters acting as pacemakers. 13,94 For Rh/Au, PEEM images and local intensity curves are displayed in Figure 5b. No oscillations occur, with the particles remaining in the inactive state. This can be explained in light of SPEM experiments and the mechanism of oscillations: Au atoms migrated onto the Rh particle are located preferentially at/in steps, hindering the formation of subsurface oxygen and thus the switch to the active state.
In Figure 5c, the corresponding results for Rh/ZrO 2 are shown. In addition to the "big" and "small" particles, even "very small" ones were studied, for which no oscillations were observed (exemplary curve in Figure 5c), apparently due to their oxidized state resulting from ZrO x -promoted Rhoxidation. Oscillations observed for "small" and "big" Rh particles exhibited significantly lower frequency than those observed for Rh/Rh, which depended on particle size. We attribute this to the Rh steps being partly blocked by ZrO x and the resulting stronger oxygen bonding. Accordingly, the oscillation cycle slows down. Differently sized Rh particles are differently affected and thus exhibit different oscillation frequencies, due to their different ratios of ZrO x -modified (interfacial) vs -unmodified Rh surface areas.
In order to assess the curved/stepped particle shape and structure, post-reaction TEM analysis ( Figure S5) was performed on cross-sections of some of the studied particles. The composition of particle surfaces and metal/support interface regions was characterized by combined scanning-TEM/energy dispersive X-ray fluorescence (EDX), confirming the in situ results.

Micro-Kinetic Model Simulations.
To better understand the experiments, micro-kinetic model simulations of H 2 oxidation were performed, based on the Langmuir−Hinshelwood mechanism and by adapting a model of McEwen et al. 98,99 For representing Rh/Rh, parameters similar to those used in our previous work were used. 12,45−47 Reflecting the PEEM and SPEM results to model Rh/Au, the hydrogen sticking coefficient was adjusted to simulate hindered dissociative hydrogen adsorption and the parameter corresponding to surface "roughness" was decreased, both resulting from RhAu surface alloy formation. To model Rh/ZrO 2 , parameters were modified to simulate the stronger oxygen binding/adsorption 44 and the changed energetics of subsurface oxygen formation in the metal/oxide interface regions (for details, see the Supporting Information). Figure 6 presents the results of the micro-kinetic model simulations: In Figure 6a, traces of the reaction rate (turnover frequency; TOF) upon cyclewise variation of p(H 2 ) are displayed for the three different model configurations, i.e., Rh/ Rh (black), Rh/Au (red), and Rh/ZrO 2 (blue). Simulations were performed for the external parameters of the PEEM experiments, allowing a direct comparison (cf. Figure 2d). By repeating simulations for different temperatures, a kinetic phase diagram can be constructed (Figure 6b), similar to the experimental diagram in Figure 2e. Simulations were performed only for 433−473 K, because oxygen-induced surface restructuring was not included in the model. Simulations of the oscillating mode were also performed for all three configurations. The simulated TOF time series are shown in Figure 6c, corresponding to the big Rh particles in Figure 5.
The calculations/simulations reproduce the experimental behavior well: Realistic variations of hydrogen adsorption and oxygen binding, mimicking the effects of the various Rh/ support interfaces, quantitively reproduce the shifts in the kinetic phase diagrams as well as the alteration of the oscillatory behavior (oscillation frequency and inhibition of oscillations).

■ CONCLUSIONS
In summary, the catalytic behavior of stepped Rh particles supported by three different materials (Rh, Au, and ZrO 2 ) has been visualized in situ by PEEM for H 2 oxidation, a probe reaction that is also relevant in many areas of technology. Kinetic transitions between the catalytically inactive and active steady states were studied and self-sustaining oscillations on supported Rh particles were observed for this reaction for the first time. Supported Rh particles show different catalytic properties in H 2 oxidation, depending on the support material and particle size. The oscillating reaction mode varies from particle-size-independent (Rh/Rh) via size-dependent (Rh/ ZrO 2 ) to fully inhibited oscillations (Rh/Au).
To interpret the PEEM results, in situ SPEM studies were performed for the same sample under the same reaction conditions, i.e., in a correlative way. The SPEM results identified the presence of support material on the Rh particles during the reaction to cause their different behavior. For Rh/ Au, the formation of a RhAu surface alloy was observed, and for Rh/ZrO 2 , the formation of substoichiometric Zr oxides, hydrogen spillover onto the ZrO 2 support, and a narrow region of enhanced Rh-oxidation were detected. Combining the realtime PEEM data and chemical information from SPEM, the catalytic properties of Rh particles on different supports can be explained, underpinning the effectiveness of the correlative approach. The experimental observations were complemented by micro-kinetic model simulations of kinetic phase diagrams and oscillating time series for catalytic H 2 oxidation on various Rh surfaces. Through realistic variations of hydrogen adsorption and oxygen binding, resulting from the observed modification of stepped Rh surfaces, the experimental findings could be confirmed. A small particle size and support materials library was created by combining various Rh particle sizes and several support materials in a single sample, which was studied by a combination of microscopy techniques able to provide local reaction kinetics and local chemical information. As the mechanism of catalytic H 2 oxidation on Rh is well understood, this probe reaction can be used to benchmark interface and particle size effects in catalysis, an approach that could be extended to other important catalytic reactions. ■ ASSOCIATED CONTENT
Experimental and computational procedures, PEEM images of the supported Rh particles, SPEM maps of the individual contributions to the chemical maps, details on the oscillation mechanism, TEM/EDX characterization of some of the Rh particles, and energy windows used in constructing the chemical maps (PDF) P.W. and L.M.R. performed the PEEM measurements and evaluated the data. P.W., M.R., J.Z., M.A., R.P., L.G., and G.R. performed the SPEM measurements, and P.W. evaluated the data. P.W. and M.S.-P. performed TEM and data evaluation. M.R. performed the micro-kinetic model simulations. Y.S. and G.R. supervised the work. P.W., Y.S., and G.R. prepared the manuscript. All authors contributed to the discussion and approved the manuscript.

Funding
Open Access is funded by the Austrian Science Fund (FWF).

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
This work was supported by the Austrian Science Fund (FWF) (P 32772 N and F81−P08). We acknowledge Elettra Sincrotrone Trieste for providing access to its synchrotron radiation facilities and for financial support under the IUS internal project. The authors are grateful to Prof. Henrik Grönbeck for helpful discussions on micro-kinetic modeling.