Patterns in the NO + H2 Reaction on Rh(110) Modified by Coadsorbed

Feb 18, 2009 - Adding potassium as coadsorbate to the pattern forming system Rh(110)/NO + H2 generates new types of chemical wave patterns that have ...
0 downloads 0 Views 3MB Size
4174

J. Phys. Chem. C 2009, 113, 4174–4183

Patterns in the NO + H2 Reaction on Rh(110) Modified by Coadsorbed Potassium Liu Hong, Martin Hesse, and R. Imbihl* Institut fu¨r Physikalische Chemie and Elektrochemie, Leibniz-UniVersitu¨a¨t HannoVer, Callinstrasse 3-3a, D-30167 HannoVer, Germany ReceiVed: December 1, 2008; ReVised Manuscript ReceiVed: January 19, 2009

Adding potassium as coadsorbate to the pattern forming system Rh(110)/NO + H2 generates new types of chemical wave patterns that have been investigated in the 10-6 and 10-5 mbar range using photoelectron emission microscopy as spatially resolving method. The modified patterns were studied for varying K coverages with θK ) 0.03, 0.05, 0.07, and 0.12. As in the unpromoted system, one finds elliptically and rectangular shaped target patterns, but in addition one observes stationary Turing-like concentration patterns, fingering instabilities, and the mass transport of potassium by propagating pulses. Large scale (a few tenths of a millimeter) redistribution of potassium by the chemical waves is seen. The propagation direction of traveling wave fragments is rotated by 90° in the presence of coadsorbed potassium. At high K, coverages the existence range for patterns is shifted drastically toward higher pH2. 1. Introduction Pattern formation on catalytic surfaces has been intensely investigated in the past two decades focusing on catalytic CO oxidation and catalytic NO reduction on metals of the platinum group.1-4 Theoretically, these pattern forming systems are modeled by reaction-diffusion (RD) systems assuming Fickian diffusion for the mobile adsorbates. Adding an alkali metal to such a system introduces a new aspect because the alkali metal is very mobile and very reactive and it modifies the catalytic properties of a metal in its surroundings.5,6 The latter effect is the basis for the role of alkali metals as so-called electronic promoters in a number of industrially important reactions such as ammonia synthesis via Haber-Bosch over Fe-based catalysts. Because of the strong energetic interaction of alkali metals with coadsorbates diffusion is no longer Fickian diffusion and new types of patterns may develop that do not exist in RD systems with pure Fickian diffusion. This was demonstrated with the O2 + H2 reaction on a potassium promoted Rh(110) surface where reactive phase separation occurs leading to a stationary concentration pattern of K + O coadsorbate islands of macroscopic size.7-10 The unpromoted system Rh(110)/NO + H2 exhibits a rich variety of different chemical wave patterns owing to the large structural variability of this surface which makes the diffusional anisotropy state-dependent; rectangular and elliptical target patterns, wave fragments traveling along certain crystallographic directions, and transitions between varying front geometries have all been found in this system.11-17 Adding potassium this excitable system retains its excitability but one observes new types of patterns caused by the interaction of the reacting species with the reactive alkali metal.18,19 The most spectacular of the new effects is the mass transport of potassium by propagating pulses as reported before.19 In this paper, we systematically investigate the influence of different K coverages on chemical wave patterns in the K-promoted system.20 As will be shown, with potassium one observes stationary Turing-like concentration patterns, fingering instabilities, and the formation of complex * To whom correspondence should be addressed. E-mail: imbihl@ pci.uni-hannover.de.

patterns associated with a large scale redistribution of potassium on the catalytic surface. 2. Experimental Section The studies were performed in a conventional UHV system equipped with LEED, a scanning Auger electron spectrometer (PHI), a Kelvin probe, and a photoelectron emission microscope (PEEM). In PEEM, the sample is illuminated with photons from a D2 discharge lamp (5.5-6 eV emission maximum) generating photoelectrons that are then focused with a three-lens system onto a phosphorus screen. The local work function can thus be imaged with a spatial resolution of 0.1-1 µm and with the temporal resolution of conventional video images (20 ms). The Rh(110) sample was prepared in a standard way using cycles of Ar ion sputtering at 800 K/ heating in oxygen (T ) 1000 K, pO2 ) 2 × 10-6 mbar)/annealing to 1400 K. Gases of high purity (5.3 N for H2 and 2.7 N for NO) were introduced into the reaction chamber via a feedback-stabilized gas inlet system. The UHV chamber (≈100 L) was continuously pumped by a turbomolecular pump (≈100 L/s). To a good approximation, the UHV system can be considered as a gradient-free flow reactor. Potassium was deposited using a SAES getter source. For calibrating the K coverage, we take advantage of the fact that the phase diagram for Rh(110)/K has been determined and use the LEED patterns to estimate the K coverage.21 Potassium induces a series of (1 × n) reconstruction phases (n ) 2, 3, 4) that are all based on the missing-row principle, that is, every nth of the close-packed [11j0]-oriented rows is missing in these phases. In the phase diagram, symmetrically around the (1 × 2) structure, (1 × n) reconstruction phases (n ) 3, 4) exist, both, at high K coverage (HC) and at low K coverage (LC). In the direction of increasing K coverage, one observes a (1 × 4)-LC with θK ) 0.04, a (1 × 3)-LC with θK ) 0.07, and finally a (1 × 2) with θK ) 0.12; at higher K coverage, deconstruction sets in leading over a (1 × 3)-HC with θK ) 0.16 to a (1 × 4)-HC with θK ) 0.21. 3. Results a. Existence Range for Wave Patterns. A typical PEEM image showing the effect of coadsorbed potassium is displayed

10.1021/jp810540g CCC: $40.75  2009 American Chemical Society Published on Web 02/18/2009

Patterns in the NO + H2 Reaction on Rh(110)

J. Phys. Chem. C, Vol. 113, No. 10, 2009 4175

Figure 1. PEEM image showing K enrichment in the collision area of target patterns in the Rh(110)/K/ NO+H2 system. Experimental conditions: θK ∼ 0.05 ML, T ) 580 K, PNO )1.5 × 10-6 mbar, pH2 ) 5.9 × 10-6 mbar (a) t ) 80 s (b) t ) 170 s with t ) 0 denoting the start of the experiment.

in Figure 1. The target patterns known from the unpromoted system still exist with potassium, but one notes that an enhanced brightness develops in the area where pulses coming from two different trigger centers collide. The bright area was shown to be due to a higher K concentration in this area caused by pulses that are transporting coadsorbed potassium with them.19,22 As demonstrated in Figure 2 in measurements with a Kelvin probe, potassium lowers the work function (WF) of the Rh(110) surface by up to 4 eV. Even in the presence of coadsorbed oxygen the WF is still below the level of the clean Rh(110) surface for K coverages above θK ) 0.08. For the unpromoted system, the interpretation of the gray levels in PEEM was straightforward. The dark area is assigned to oxygen covered surface which increases the WF by 1.1 eV at maximum.14 The traveling bright bands, that is, the pulses represent nitrogen covered Rh(110) surface with the (2 × 1)-N at θN ) 0.5 exhibiting a WF increase of 280 meV relative to the clean Rh(110) surface.14 With a small amount of coadsorbed potassium as in the experiment displayed in Figure 1 this assignment remains valid. In Figure 1, one can clearly distinguish between a medium gray, which is attributed to the traveling nitrogen pulses, and a very bright gray level, which is indicative of K enrichment. As shown in Figure 1b, the K enrichment is observed mostly with the [001]-oriented parts of the pulses. The existence range of the different chemical wave patterns of the unpromoted system Rh(110) has been mapped out quite in detail.16 This diagram also serves as the basis for studying the effect of coadsorbed potassium. In our study, we stepwise increased the potassium coverage in the sequence θK ) 0.03, 0.05, 0.07, up to 0.12 as highest K coverage. This coverage is still much below the θK ) 0.22 that has been determined as saturation coverage for one monolayer but beyond θK ) 0.12 desorption starts to become significant in the temperature range up to 600 K used in our experiments.21 At the lowest K coverage, LEED displayed a weak 1 × 4 pattern; assuming that we are close to the low coverage boundary of this phase we assign a coverage θK ) 0.03 ( 0.01 to the surface. For the next higher coverage with LEED showing a mixture of a (1 × 4) and (1 × 3) pattern a coverage of θK ) 0.05 ( 0.01 was determined from the phase diagram. Similarly, with very intense spots of the (1 × 3) phase a coverage θK ) 0.07 ( 0.01 was deduced and with very intense (1 × 2) beams a coverage of θK ) 0.12 ( 0.01 was determined.

Figure 2. Variation of the work function (WF) with potassium coverage for a clean and for an oxygen covered Rh(110) surface. The WF was measured with a Kelvin probe. (a) K adsorption on Rh(110). The observed LEED patterns used for determination of the K coverage are indicated. (b) The coadsorption system Rh(110)/K + O. The experiment was carried out such that first potassium was dosed onto the clean Rh(110) surface followed by exposure to 135 L O2. The latter dose corresponds to nearly saturation coverage. The LEED structures refer to the surface with only potassium being present.

For mapping out the bifurcation diagram of the unpromoted a certain procedure had been applied.16 After adjusting pNO and the temperature chemical wave patterns were initiated by first increasing pH2 up to 1 × 10-5 mbar so that the whole surface become uniformly reduced followed by a decrease of pH2 to the desired value. This procedure cannot be applied to a system

4176 J. Phys. Chem. C, Vol. 113, No. 10, 2009

Hong et al.

Figure 3. Bifurcation diagram of the unpromoted Rh(110)/NO + H2 system illustrating the influence of the procedure of parameter adjustment on the boundaries in the diagram. Shown is the excitable range (EX) with the boundaries toward the stable oxygen-covered range (OX), and the bistable range of the reaction (BS). The full lines represent the boundaries as determined from the procedure applied in this study, the dashed lines mark the results from Schaak et al. who applied a different experimental procedure.16 A fixed pNO ) 1.5 × 10-6 mbar was used.

with coadsorbed potassium because already the initial pH2 increase leads to an inhomogeneous K distribution on the surface which does not homogenize again in the second step of pH2 reduction. Therefore in the modified procedure, pH2 is raised directly from zero to the desired value without going via a completely reduced surface. This modified procedure results in a shift of the boundaries of the pattern forming parameter range as shown in Figure 3 which represents a kind of benchmark study for subsequent measurements with the K-promoted system. In particular, the low pH2 boundary between a stable oxygen covered state of the surface and the excitable range shifts to a pH2 value that is higher by nearly a factor of 3 as compared to the original procedure.16 With the new procedure the excitable range of the system was mapped out for the different K coverages in pH2,T-parameter space with pNO being kept fixed at 1.5 × 10-6 mbar. The overall effect of coadsorbed potassium is that the boundaries of the existence range for chemical wave patterns shift to larger pH2 values as compared to the unpromoted system. This effect is demonstrated in Figure 4a for a relatively large K coverage of θK ) 0.12. The excitable range is strongly broadened which is due to the drastic shift of the high pH2 boundary to larger values. Qualitatively, this shift can be explained in a straightforward way. Potassium reduces the reactivity of codsorbed oxygen so that a larger pH2 is required in order to reactively remove oxygen from the surface.23 The boundaries do not shift linearly with the amount of coadsorbed potassium as demonstrated in Figure 4b,c. At low K coverage up to θK ) 0.07, the boundaries are hardly affected by coadsorbed potassium or the boundaries even move to lower pH2 with coadsorbed potassium. A drastic effect is seen only at high K coverage, as θK goes up to 0.12. b. Traveling Wave Fragments. In the unpromoted system, wave fragments traveling along the [001]-direction were found close to the low pH2 boundary of the pattern forming parameter range.13 Their formation has been explained with a statedependent anisotropy of surface diffusion caused by adsorbateinduced reconstructions of different types. With coadsorbed potassium, wave fragments still exist at the low pH2 boundary but their orientation and their direction of propagation is rotated

by 90°. At very low K coverage, at θK ) 0.03, wave fragments traveling in the [001]-direction and [11j0]-direction coexist. A typical example showing for an intermediate K coverage, θK ) 0.07, how a target pattern with progressive time decomposes into individual wave fragments is displayed in Figure 5. The whole process from the start of the experiment to the fragments in Figure 5d extends over a relatively long period of ≈20 min. The wave pattern continues to transform becoming more and more disordered. After 2600 s, mainly cloudlike features remain with few surviving wave fragments as illustrated by Figure 5e. The PEEM images in Figure 5d,e exhibit clearly three distinct gray levels: bright fragments, a medium gray, and a dark gray background. At high K coverage, at θK ) 0.12, no wave fragments are found. c. Wave Trains and Front Instabilities. Further inside the excitable range propagation failure in one direction no longer occurs. Target patterns with closely packed wave trains will develop but one observes new phenomena. As was shown before, the collision of pulse trains may lead to an enrichment of potassium in the collision area.19 Figure 6a displays the initial situation when wave trains from two target patterns are going to collide in the central area. We can use the varying distance between the individual pulses to deduce a kind of dispersion relation for pulses propagating on a K-covered surface. Since each pulse carries a certain fraction of the K coverage with it, the K coverage varies along the profile of a pulse train. The dispersion relation obtained in this way is therefore only a crude approximation. The amount of transported potassium varies with the reaction conditions, and it is therefore difficult to estimate the K concentration profile. Taking the wave trains from the top and the bottom part we obtain the two dispersion curves displayed in Figure 6b. The pulse velocity varies in a range from 0.7 µm/s at large separation to 0.45 µm/s at the shortest wavelength. Compared to the dispersion relation of the unpromoted system where the velocity in the [11j0]-direction range extends from 2.5 to 1.5 µm/s, the pulse velocity is reduced by the addition of potassium by roughly a factor of 3.15 The reason why coadsorbed potassium slows down the pulse velocity is easy to understand. Coadsorbed

Patterns in the NO + H2 Reaction on Rh(110)

J. Phys. Chem. C, Vol. 113, No. 10, 2009 4177

Figure 5. PEEM images showing how the decompositions of target patterns leads to the formation of wave fragments traveling in the [11j0]direction. Experimental conditions: θK ) 0.07 ML, T ) 520 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 5.9 × 10-6 mbar. (a) t ) 455 s, (b) t ) 725 s, (c) t ) 910 s, (d) t ) 1305 s, and (e) t ) 2640 s. The time t ) 0 refers to the start of the experiment.

Figure 4. Influence of the potassium coverage on the bifurcation diagram of Rh(110)/K/NO + H2 system. The NO partial pressure was kept fixed at pNO ) 1.5 × 10-6 mbar. (a) Bifurcation diagram for θK ) 0.12 ML. Only the boundaries of the excitable range are shown: EX ) excitable range, OX ) stable oxygen covered surface, BS ) bistable range. The dashed lines represent the boundaries for the unpromoted system. The inset in the top section indicates the range where different fingering instabilities (F1, F2) were observed (see text). (b) Dependence of the boundary stable oxygen covered/excitable on the potassium coverage for varying temperatures. (c) Dependence of the boundary excitable/bistable on the potassium coverage for varying temperatures.

potassium reduces the reactivity of oxygen toward hydrogen thus making it more difficult for the primary front to propagate inside the oxygen surface area. One also notes that the two dispersion curves taken from top and bottom of Figure 5a differ significantly. The simplest explanation is that the mass transport of potassium during the 5-10 pulses preceding the start of the measurements has already led to an inhomogeneous potassium distribution in the imaged area. The faster pulses in the bottom part of Figure 6a could indicate that a depletion of potassium has taken place relative to the situation in the top part. The

stronger K enrichment seen in the bottom part of the central collision area supports such an explanation. In addition to the mass transport of potassium with the pulses, one observes the formation of conelike structures within a target pattern. An example of such a conelike structure found at θK ) 0.07 is displayed in Figure 7a. One notes that the pulses in the cone oriented along the [11j0]-direction all exhibit a strongly enhanced brightness, which is a property that we can tentatively assign to potassium transported with the pulses. The opening angle of the cone is roughly 30°. A closer look at the pulses reveals that the pulses are no longer straight lines of uniform brightness but they exhibit an intensity modulation as well as a modulation of the front line. The modulation becomes stronger toward the edges of the cone. Bright lines running perpendicular to the pulses are connecting the spots of enhanced brightness in the different pulses giving the whole structure the appearance of a spider web. This structure is also visible in Figure 7b, which was recorded from a different spot on the sample where the pulses are less closely packed. The tip of the cone is initially located at the trigger center of the target pattern. As the cone is fully developed, the whole cone starts to move in the [11j0]-direction with a velocity of ≈ 0.2 µm/s. Another example of cone structures observed at lower K coverage, at θK ) 0.05 is displayed in Figure 8. On the left half of Figure 8a one still observes the target pattern similar to the unpromoted system, while on the right half the pulse propagation only occurs along certain crystallographic directions

4178 J. Phys. Chem. C, Vol. 113, No. 10, 2009

Figure 6. Approximate dispersion relation for pulses propagating on a potassium covered Rh(110) surface with θK ) 0.07 ML in the [11j0]direction. Experimental conditions: T ) 510 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 1.8 × 10-5 mbar. (a) PEEM image showing the formation of a bright zone in the collision area of two pulse trains from target patterns centered outside the imaged area. The enhanced brightness in the central area is attributed to K enrichment. The lines along which the wave velocities have been measured are indicated. (b) Dispersion relation for pulses propagating in the top section of panel a and along the bottom section of panel a.

that differ by roughly 30°. Judging from the brightness difference the potassium concentration is significantly larger on the right half than on the left half. As observed before in the collision areas of pulse trains, potassium accumulates causing these areas to grow into large areas of enhanced brightness. With progressive time, these areas grow considerably in extension as shown in Figure 8a-c. Inside the bright regions of K accumulation spatiotemporal dynamics still persists but the patterns are quite irregular. Evidently the mass transport of potassium with pulses leads to a large scale redistribution of potassium on the reacting surface over distances of at least a few tenths of a millimeter. This large scale redistribution is apparently connected with the formation of the cone structures as suggested by the development visible in Figures 7 and 8. At relatively high K coverage, at θK ≈ 0.07-0.12 one observes pulse trains running in opposite directions parallel to the [11j0]-direction as reported before.18 An example recorded at θK ) 0.12 is displayed in Figure 9. These patterns are quite stable and persist over several hours. d. Stationary and Nearly Stationary Patterns. i. T-Range 500-550 K. Stationary patterns were found over the whole range of potassium coverages investigated here from θK ) 0.05

Hong et al. up to θK ) 0.12.22 Starting from a homogeneous surface, the stationary patterns usually develop via reaction fronts. Fronts nucleate at different spots on the surface. The fronts expand until they collide thus transforming the surface into a domain structure. This pattern then evolves forming structures with finer and finer length scale until a stationary pattern with a characteristic length scale of a few micrometers remains. This development is illustrated in Figure 10 for θK ) 0.07. After ignition of several reduction fronts (Figure 10a), the area behind the primary reaction fronts begins to darken similar to the development observed in the range of “double metastability” of the unpromoted system. In the image taken at t ) 111 s in Figure 10d bright spots are seen to nucleate at many different places. Finally, a stable stationary pattern develops which does not exhibit any detectable changes over a period of several hours. The pattern at t ) 154.5 s in Figure 10e is not well ordered and exhibits structures with a length scale of 10-20 µm. The PEEM image in Figure 11a taken from a different spot of the surface shows better ordering. One can identify three distinct gray levels: bright, medium gray, and dark gray. The thin bright stripes are elongated in the [11j0]-direction. A stationary pattern obtained at lower K coverage, at θK ) 0.05 is reproduced in Figure 11b. The structural similarity to Figure 11a is evident. ii. T-Range around 600 K. At 600 K and at a high K coverage of θK ) 0.12, a fingering instability of reduction fronts and the formation of nearly stationary patterns is observed. The development of such a structure is displayed in Figure 12. Because of the strong lowering of the WF of the oxygen covered surface by coadsorbed potassium the reduced area appears dark whereas the surface covered by the K + O coadsorbate is imaged as bright area. The reduction fronts in Figure 12a initially exhibit a nearly rectangular shape. One notes bright zones elongated along the [11j0]-direction which run ahead of the front. In analogy with the system Rh(110)/K/O2+H2 where similar observations were made, this zone can be assigned to an enhanced K concentration building up in the region ahead of the reduction front.24 Soon after the reduction fronts starts to expand, the initially rectangular shape of the front becomes unstable and fingerlike structures start to grow. As shown by Figure 12b-f, the fingers develop in the bright halo surrounding the reduction front. After about 4 min, the surface is covered with black elongated islands. Within the duration of this experiment of about 1500 s the patterns do not become stationary but keep on slowly changing their shape. Turning off the hydrogen gas does not lead to a fast homogenization of the surface but just the contrast weakens and the fine structures become blurred as demonstrated by Figure 13. As is known from the O2 + H2 reaction on the K-promoted Rh(110) surface, the mobility of potassium is strongly reduced in the presence of coadsorbed oxygen.7 The observation that the gray levels in PEEM are not reversed by turning off hydrogen confirms the previous assignment: the bright area in Figures 12 and 13 represent the K + O coadsorbate while the dark areas have to be attributed to K-deficient surface, that is, a nitrogen-covered surface under reaction conditions. This area is probably also covered by some oxygen after turning off H2. At higher pH2 the tendency for fingering is strongly reduced as demonstrated by Figure 14. Similarly to Figure 12, reduction fronts with a bright anisotropically propagating zone ahead of the fronts are seen. The collision of two fronts in the [11j0]direction generates a bright area as shown in Figure 14b but, remarkably, collision of the fronts in the [001]-direction produces a dark area. These dark areas divide the large bright island in the middle of Figure 14b into several parts separated

Patterns in the NO + H2 Reaction on Rh(110)

J. Phys. Chem. C, Vol. 113, No. 10, 2009 4179

Figure 7. Development of cone structures in pulse trains on a K-covered surface. Note that bright lines are running perpendicular to the pulses (“spider web”). Experimental conditions: θK ) 0.07 ML, T ) 530 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 1.2 × 10-5 mbar. The PEEM images in (a) and (b) were recorded from different spots on the sample surface. The movement of the cones is indicated by dashed black arrows.

from each other. The bright island exhibits a ragged shape reminiscent of the fingering structures discussed above. As in the previous example, the islands are not really stationary but are slowly changing their shape. As shown by Figure 14c, turning off hydrogen causes the contrast to weaken and the structures become blurred similar to the experiment displayed in Figure 13. 4. Discussion The unpromoted system Rh(110)/NO + H2 was shown to exhibit an enormous variety of different wave patterns: elliptically and rectangularly shaped target patterns and spiral waves, wave fragments traveling in the [001]-direction, and the simultaneous presence of different front geometries in the socalled range of dynamic bistability.11,13,16,17 With coadsorbed potassium, the basic excitation cycle of pulses remains intact but the patterns of the unpromoted system are modified and one observes the following new types of phenomena: (i) mass transport of potassium by pulses; (ii) formation of stationary Turing-like structures; (iii) rotation of the propagation direction of traveling wave fragments by 90°; (iv) fingering instability in reduction fronts; (v) formation of large scale structures of wave trains, that is, conelike structures with two-dimensional (2D)-net pattern (spider web). A good level of understanding has been reached with the mass transport of potassium by pulses. Spectromicroscopic studies with LEEM/MEM, µ-XPS, and µ-LEED provided detailed structural and chemical information about the relevant surface phases so that a concentration profile of a pulse was obtained.19,25 On the basis of two existing models for the system Rh(110)/ NO + H2 and Rh(110)/O2 + H2, a realistic mathematical model for the NO + H2 reaction on the K-promoted Rh(110) surface was constructed which reproduced in 1D-simulations the experimentally observed mass transport of potassium with pulses.10,17,19,26 Pulses are generated by the excitation cycle of the unpromoted system, which remains intact, but what had to be added were the energetic interactions of potassium with coadsorbates. The attraction between K and oxygen atoms is well known, but what turned out to be essential was the effective repulsion between nitrogen and coadsorbed potassium.19 Nitrogen accordingly acts as a diffusion barrier for potassium that

Figure 8. Formation of cone structures in pulse trains associated with a large scale redistribution of potassium on the imaged area. The movement of the cones is indicated by dashed white arrows. The full line white arrows mark the movement of the pulses. Experimental conditions: θK ) 0.05 ML, T ) 540 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 8.4 × 10-6 mbar. (a) t ) 540 s (b) t ) 740 s (c) t ) 940 s. The time t)0 refers to the start of the experiment.

4180 J. Phys. Chem. C, Vol. 113, No. 10, 2009

Figure 9. PEEM image showing wave trains running in opposite directions parallel to [11j0]. Experimental conditions: θK ) 0.12 ML, T ) 510 K, pNO )1.5 × 10-6 mbar, pH2 ) 4.5 × 10-5 mbar.

has the consequence that the nitrogen adlayer in a pulse is pushing the K atoms which are initially homogeneously distributed in the oxygen adlayer ahead of the pulse. A substantial K concentration therefore piles up in front of a pulse. The observation of stationary Turing-like patterns was not too much of a surprise because already in the simpler system Rh(110)/K/O2 + H2, which represents a subsystem of the NO + H2 reaction system, such structures were found consisting of K + O coadsorbate islands of macroscopic size.7 Their formation was driven by the strong chemical attraction between oxygen and potassium. Since in contrast to a Turing structure, not differences in the diffusivities but energetic interactions were shown to be the driving force for the formation of the stationary patterns, the term reactive phase separation was introduced to describe this type of structure.8,10,27 For the stationary patterns in the system Rh(110)/K/NO + H2, it was shown using µ-XPS that also there a phase separation into K + O and into nitrogen containing phase occurs.22,25 An important difference to many other pattern forming systems is that many of the chemical wave patterns are just transient patterns existing for a relatively short period of time. For example, often a target pattern develops which becomes unstable as time progresses and transforms into a different structure. The stable wave trains depicted in Figure 9 were formed in this way. Other examples are the traveling wave fragments in Figure 5 and the pulses transporting potassium in Figures 6 and Figures 8. These transient patterns can be plausibly explained with the mass transport of potassium which modifies the local reaction conditions. Alternatively, also a modification of the substrate either by oxidation or by reaction-induced substrate roughening has to be taken into account (see below). A general problem in investigating the different types of chemical wave patterns on K-promoted surfaces is always that only the initially coverage of potassium is known with certainty. Subsequent redistribution of the potassium by chemical waves patterns will lead to locally varying reaction conditions. The formation of the cone structures and of the spider web appears to be linked to such a redistribution of potassium as evidenced by Figures 7 and 8. Since potassium is known to facilitate the oxidation of noble metal catalysts, the question arises whether oxide formation plays a role in the patterns observed here.28 If a high K concentration would in fact promote oxide formation, then such

Hong et al. a memory effect could explain potentially the formation of complex structures like the spider web in the pulse trains (Figure 7). If a high K concentration causes a local oxide formation that in turn slows down the propagation of the pulse then such a mechanism would presumably destabilize a planar front. The front shape would be modulated with a high K concentration being present in those parts of the front which are lacking behind. The bright stripes that are connecting the regions of enhanced brightness and that are running perpendicular to the front would then reflect a local oxidation of the surface. In the spectromicroscopic studies performed so far, no indication was found that oxide formation occurs under pattern forming conditions. Alternatively, one could think of reaction-induced roughening modifying the local transport properties of the Rh(110) surface thus giving rise to a similar instability as that caused by local oxidation. Since the different densities of the various N, O, and K-induced reconstructions require the mass transport of Rh atoms,29 the propagation of a pulse should induce some roughening on the surface. Although such a mechanism is quite plausible, experimentally firm evidence that substantial reactioninduced roughening occurs is still missing. The absence of a strong roughening can be rationalized by a high mobility of Rh atoms at elevated temperatures which allow a for very efficient thermal reordering of rough surfaces. What so far has not been included in the mechanistic discussion are 2D effects like diffusional anisotropy and the curvature dependence of front propagation. In the 1D-simulations carried out so far 2D effects are naturally absent. Since a state-dependent anisotropy has been made responsible for all of the unusual patterns found in the unpromoted system like the rectangular shaped patterns and the traveling wave fragments any changes in the diffusional anisotropy caused by potassium are expected to have a very strong influence on pattern formation.17,30 A drastic influence of coadsorbed potassium on diffusional anisotropy is in fact seen as demonstrated by the rotation of the direction in which traveling wave fragments move by 90° (see Figure 5). In other phenomena like in the conelike structure of wave trains or in the fingering instability, one can suspect that also K-induced changes in the diffusional anisotropy play a major role. A fingering instability arises at the interface of two immiscible fluids if the invading fluid has a smaller velocity than the fluid that is being invaded.31 The origin of the front instability here has yet to be identified but one can suspect that it is most likely the change in diffusional anisotropy caused by coadsorbed potassium that is responsible for the phenomenon. Theoretically, the adsorption and diffusion of potassium on a clean Rh(110) surface and in coadsorption with oxygen has been investigated in density function theory (DFT) calculations.32 Experimentally only few data are available about the influence of coadsorbates on K diffusion. For K diffusion on the oxygen-covered Rh(110) surface, an activation energy of 85 kJ/mol was determined that is larger by 47 kJ/mol than the 38 kJ/mol estimated for K diffusion on the bare Rh(110) surface.10 Clearly oxygen slows down K diffusion considerably. In the DFT calculations, the activation barrier for potassium diffusion on an oxygen-covered Rh(110) surface was 11.6 kJ/ mol, considerably below the experimental value of 85 kJ/mol.32 The reason for this discrepancy could be that K diffusion on the oxygen-covered Rh(110) surface is associated with restructuring of the substrate and with the mass transport of Rh atoms. From concentration profiles of pulses in the present system, it was concluded that nitrogen acts as a diffusion barrier for K diffusion. No experimental data exist on how the diffusional

Patterns in the NO + H2 Reaction on Rh(110)

J. Phys. Chem. C, Vol. 113, No. 10, 2009 4181

Figure 10. Development of a stationary “Turing-like” concentration pattern starting from a homogeneous K-covered surface. Experimental conditions: θK ) 0.07 ML, T ) 510 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 2.85 × 10-5 mbar. (a) t ) 28 s (b) t ) 43 s (c) t ) 97 s (d) t ) 111 s (e) t ) 154.5 s. The time t ) 0 refers to the start of the experiment.

Figure 11. PEEM images showing Turing-like stationary patterns with some ordering being visible. Experimental conditions: (a) θK ) 0.07 ML, T ) 510 K, pNO )1.5 × 10-6 mbar, pH2 ) 2.5 × 10-5 mbar; (b) θK ) 0.05 ML, T ) 510 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 2.85 × 10-5 mbar.

anisotropy of potassium on Rh(110) is influenced by other adsorbates and on how potassium changes the diffusional anisotropy of other adsorbates. The latter may occur via K-induced reconstructions or via energetic interactions of K with other adsorbates. The bright halo in Figures 12 and 14 that precedes the reduction fronts and that we attribute to potassium diffusing on the oxygen covered Rh(110) surface indicates that K diffusion is fast along the [11j0]-oriented troughs of Rh(110). This assignment is consistent with the idea that mainly the geometric corrugation determines the diffusional anisotropy. All reconstructions of Rh(110)/O, Rh(110)/K, and of the coadsorbate system Rh(110)/K + O are of the missing row type and would therefore give rise to diffusion being fast in the [11j0]-direction and slow perpendicular to this direction. Only the nitrogeninduced reconstructions with Rh-N-Rh rows in the [001]direction should give rise to a different anisotropy.29 However, the idea that it is only the geometric corrugation that determines the diffusional anisotropy is apparently too simplistic. Energetic

interactions and site blocking effects by coadsorbates will also have a very strong effect. 5. Conclusions It was demonstrated that the addition of potassium to the system Rh(110/NO + H2 (i) modifies the patterns of the unpromoted system and (ii) creates new types of chemical wave patterns. Stationary concentration patterns, fingering instabilities, and the mass transport of potassium with chemical waves are new types of patterns that have not been observed in the unpromoted system. Most of the effects can be understood as being a consequence of the strong energetic interactions of potassium with coadsorbates. The shift of the pattern forming parameter range toward higher pH2 and the slowing down of front velocities in the presence of potassium can be attributed to potassium making coadsorbed oxygen less reactive toward H2. In analogy to the related reaction system Rh(110)/K/O2 + H2, the formation of stationary concentration patterns can be

4182 J. Phys. Chem. C, Vol. 113, No. 10, 2009

Hong et al.

Figure 12. Development of a fingering instability in the reduction fronts leading to a nearly stationary pattern. The corresponding parameter range is denoted as F1 in Figure 4a. Experimental conditions: (a) θK ) 0.12 ML, T ) 600 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 1.2 × 10-5 mbar. (a) t ) 10 s, (b) t ) 30 s, (c) t ) 50 s, (d) t ) 100 s, (e) t ) 170 s, and (f) t ) 260 s. The time t ) 0 refers to the start of the experiment.

Figure 13. PEEM images showing the change in the nearly stationary pattern when H2 is turned off. The pattern is obtained in the experiment depicted in Figure 12. (a) Pattern under reaction conditions at t ) 1450 s. (b) Pattern at t ) 1800 s after turning off H2.

Figure 14. PEEM images showing the development of a nearly stationary pattern via reduction fronts. The corresponding parameter range is denoted as F2 in Figure 4a. The propagation directions of some reduction fronts are marked by black arrows. Experimental conditions: θK ) 0.12 ML, T ) 600 K, pNO ) 1.5 × 10-6 mbar, pH2 ) 1.5 × 10-5 mbar. (a) t ) 50 s, (b) t ) 410 s, and (c) t ) 420 s with H2 turned off. The time t ) 0 refers to the start of the experiment.

assigned to the strong chemical attraction between oxygen and potassium. The change in the propagation direction of traveling wave fragments by 90° indicates an influence of potassium on the diffusional anisotropy. The mass transport of potassium with

propagating pulses can be explained by adsorbed nitrogen acting as a diffusion barrier for the K atoms. What so far has not been satisfactorily explained are the complex patterns like the cone structures that are associated with a large scale redistribution

Patterns in the NO + H2 Reaction on Rh(110) of potassium. The spider web structure can be tentatively assigned to a memory effect of the chemical waves but this needs to be substantiated in future experiments. References and Notes (1) Eiswirth, M.; Ertl, G. In Chemical WaVes and Patterns; Kapral, R., Showalter, R., Kapral, K. S., Eds.; Kluwer: Dordrecht, 1994. (2) Imbihl, R.; Ertl, G. Chem. ReV. 1995, 95, 697. (3) Imbihl, R. Nonlinear dynamics on catalytic surfaces. In Handbook of Surface Science; Hasselbrink, E., Lundquist, B., Eds.; Elsevier: Amsterdam, 2008; Vol. 3. (4) See the focus issues: J. Phys. Chem. 100/49 (1996), Catal. Today 70/4 (2001), Chaos 12/1 (2002), J. Phys. 5 ( 2003). (5) Bonzel, H. P. E.; Bradshaw, A. M. E.; Ertl, G. E. Physics and Chemistry of Alkali Metal Adsorption; Elsevier: Amsterdam, 1989. (6) Kiskinova, M. Poisoning and Promotion in Catalysis Based on Surface Science Concepts; Elsevier: New York, 1992; Vol. 70. (7) Marbach, H.; Gunther, S.; Luerssen, B.; Gregoratti, L.; Kiskinova, M.; Imbihl, R. Catal. Lett. 2002, 83, 161. (8) De Decker, Y.; Marbach, H.; Hinz, M.; Gu¨nther, S.; Kiskinova, M.; Mikhailov, S.; Imbihl, R. Phys. ReV. Lett. 2004, 92, 198305. (9) De Decker, Y.; Mikhailov, S. J. Phys. Chem. B 2004, 108, 14759. (10) Hinz, M.; Guenther, S.; Marbach, H.; Imbihl, R. J. Phys. Chem. B 2004, 108, 14620. (11) Mertens, F.; Imbihl, R. Nature 1994, 370, 124. (12) Gottschalk, N.; Mertens, F.; Baer, M.; Eiswirth, M.; Imbihl, R. Phys. ReV. Lett. 1994, 73, 3483. (13) Mertens, F.; Gottschalk, N.; Baer, M.; Eiswirth, M.; Mikhailov, A.; Imbihl, R. Phys. ReV. E 1995, 51, R5193.

J. Phys. Chem. C, Vol. 113, No. 10, 2009 4183 (14) Mertens, F.; Schwegmann, S.; Imbihl, R. J. Chem. Phys. 1997, 106, 4319. (15) Mertens, F.; Imbihl, R. J. Chem. Phys. 1996, 105, 4317. (16) Schaak, A.; Imbihl, R. J. Chem. Phys. 1997, 107, 4741. (17) Makeev, A.; Hinz, M.; Imbihl, R. J. Chem. Phys. 2001, 114, 9083. (18) Marbach, H.; Gu¨nther, S.; Neubrand, T.; Imbihl, R. Chem. Phys. Lett. 2004, 395, 64. (19) Hong, L.; Uecker, H.; Hinz, M.; Qiao, L.; Kevrekidis, G.; Gu¨nther, S.; Mentes, O.; Locatelli, A.; Imbihl, R. Phys. ReV. E 2008, 78, 055203. (20) For simplification, the term promoter is used in this study in a more general sense of a nonreacting coadsorbate increasing or decreasing the catalytic activity of the surface. This includes the original definition of promoter as well as catalytic poisons. (21) Gu¨nther, S.; Hoyer, R.; Marbach, H.; Imbihl, R.; Esch, F.; Africh, C. J. Chem. Phys. 2006, 124, 014706. (22) Hong, L. Ph. D. Thesis, University of Hannover, 2008. (23) Gu¨nther, S.; Marbach, H.; Imbihl, R.; Baraldi, A.; Lizzit, S.; Kiskinova, M. J. Chem. Phys. 2003, 119, 12503. (24) Marbach, H.; Hinz, M.; Gu¨nther, S.; Gregoratti, L.; Kishkinova, M.; Imbihl, R. Chem. Phys. Lett. 2002, 364, 207. (25) Gu¨nther, S. Unpublished work. (26) Qiao, L. Unpublished work. (27) Locatelli, A.; Mentes, O.; Aballe, L.; Mikhailov, A.; Kiskinova, M. Phys. Chem. B 2006, 110, 10108. (28) Pirug, G.; Dziembaj, R.; Bonzel, P. Surf. Sci. 1989, 221, 553. (29) Kiskinova, M. Chem. ReV. 1996, 96, 1431. (30) Mikhailov, A. Phys. ReV. E 1994, 49, 5875. (31) Mathiesen, J.; Procaccia, I.; Swinney, H. L.; Thrasher, M. Europhys. Lett. 2006, 76, 257. (32) Xu, Y.; Marbach, H.; Imbihl, R.; Kevrekidis, G.; Mavrikakis, M. J. Phys. Chem. C 2007, 111, 7446.

JP810540G