Spatial and velocity distributions of desorbing products in steady-state NO+CO and N 2 O+CO reactions on Pd(110) and Rh(110) ∗

The spatial and velocity distributions of desorbing products (N 2 , N 2 O and CO 2 ) were studied at various crystal azimuths in steady-state NO +CO and N 2 O+CO reactions on Pd(110) and Rh(110) by cross-correlation time-of-ﬂight techniques. The product N 2 desorption on both surfaces collimates far from the surface normal toward the [001] direction below 600 K. The N 2 distribution on Pd(110) was presented in a three-dimensional way. [DOI:


I. INTRODUCTION
The desorption dynamics (the spatial, velocity and internal energy distributions) of surface reaction products with hyper-thermal energy is sensitive to the structure of the sites on which the molecules are formed [1]. The correlation of the desorption dynamics to the site structure makes it possible to provide direct evidence of surface species participating in its reaction as well as the most direct site-identification method applicable in the course of a catalyzed reaction. Indeed, the inclined N 2 desorption in N 2 O decomposition on Pd(110) is useful to analyze the surface-nitrogen removal pathways in the de-NO x process because the concomitant associative desorption of N(a) emits N 2 sharply along the surface normal and the other nitrogen-containing products N 2 O and NH 3 show a cosine (broad) distribution [2]. N 2 O is not only an undesirable by-product in the catalytic NO reduction but also the key intermediate in controlling the selectivity to N 2 [3]. This paper is the first report of the angular and velocity distributions of desorbing N 2 at various crystal azimuths in steady-state NO+CO and N 2 O+CO reactions on Pd(110) and Rh(110). Similarities and remarkable differences in the angular distributions were found on these two reactions and surfaces.
Product desorption must be consistently characterized from the viewpoints of both chemical kinetics and reaction dynamics. Kinetics deals with the desorption frequency (reaction rate), while dynamics deals with energy partitioning. Surface chemical reactions show their own sensitivity toward surface structures to a certain extent. This sensitivity has been noticed as 'structure-sensitive' only when its rate is extremely dependent on the surface structure, i.e., a difference of a few orders of magnitude between different planes [4]. This classification is partly attributed to the lack of reliable methods to directly correlate the site structure to the reaction itself. Indeed, only the above-mentioned dynamics viewpoint has the potential to provide structural information through the reaction itself [1]. This structure-informative desorption dynam-ics has been primarily examined in the CO oxidation on noble metals, which has been classified as the structureinsensitive reaction because the reaction rate changes only a few times on different planes. Nevertheless, remarkable differences are found in the angular and velocity distributions as well as in the vibrational temperatures of desorbing CO 2 [1,5]. Desorbing product molecules hold structural information on the transition state, including the site and reactants, when they are desorbed before thermalization to the surface temperature [1].
The crystal azimuth dependence of the angular distribution of desorbing products has been extensively studied only for the reactive CO 2 desorption on Pd(110) and Pt(110) surfaces [1]. These data have provided the fundamental basis for the relationship between the collimated product desorption and reaction-site structures. Such a relationship will be examined in regard to the orientation of intermediate N 2 O molecules emitting products directly [6]. A detailed analysis of desorbing fragment N 2 distributions will provide information on the movement of parent N 2 O as well as its orientation. A similar anisotropy of desorbing species has been frequently reported in electron (or photon)-stimulated desorption ion angular distribution (ESDIAD) [7].
The angular distribution of desorbing products in the NO reduction has been analyzed with several relaxation methods, such as modulated molecular beams [8] and angle-resolved temperature-programmed desorption (AR-TPD) [9]. The steady-state conditions for the reaction, however, could not be established, and the dynamics analysis of product desorption was seriously limited. In the present work, AR-product desorption measurements were successfully performed at different crystal azimuths for the steady-state NO(or N 2 O)+CO reaction and the resultant spatial distributions were constructed on threedimensional coordinates.

II. EXPERIMENTAL
The apparatus consists of three separately pumped chambers [6]. Briefly, the reaction chamber is equipped with low-energy electron diffraction (LEED) and an X-ray photoelectron spectroscopy (XPS) system, an ion gun, and a quadrupole mass spectrometer (QMS) for angleintegrated (AI) measurements. The chopper house has a large pumping rate of about 7 m 3 s −1 , yielding acceptable  1: (a) Definition of the crystal azimuth (φ) and desorption angle (θ) used in the experiments, (b) a new angle system (α, β) in the data analysis, and a top-view of Pd(110). The plane at a fixed α value is cross-hatched. (c) N2 angular distributions (β-dependence) in the inclined plane at α = 45 • for the N2O+CO reaction on Pd(110) and at α = 41 • for the NO+CO reaction. The spatial distribution of desorbing N2 on three-dimensional coordinates is inserted.
angle-resolved performance [10]. The chopper house has a narrow slit toward the reaction chamber and a crosscorrelation random chopper blade. Another QMS was set in the analyzer for AR-product desorption and time-offlight (TOF) analyses. The distance from the ionizer to the chopper blade was 377 mm and the time resolution was selected at 20 μs [11].
A palladium or rhodium crystal with a (110) plane (Surface Preparation Laboratory, Netherlands) in a diskshaped slice (with 1mm thickness and a 10-mm diameter) was rotated in front of the first slit to change the desorption angle (polar angle; θ) in the normally directed plane at various crystal azimuths between the [001] and [110] directions. The crystal azimuth (φ) is defined as the angle shift from the [001] direction, i.e., φ = 0 • at the [001] direction ( Fig. 1(a)). 15 N 2 O was introduced through a doser with a small orifice (diameter; 0.1 mm) while 13 CO and 15 NO were backfilled. The partial pressures of 13 C 16 O (P CO ) and 15 NO (P NO ) were kept constant by dosing the gases continuously. Hereafter, the isotopes 15 N and 13 C are simply described as N and C in the text. The product N 2 , CO 2 and N 2 O signals were monitored in both the AI and AR forms. The N 2 signals in both QMSs were corrected by the contribution due to the fragmentation of N 2 O in their electron-impact ionizers. The AR signal was obtained by the analyzer QMS as the difference between the signal at the desired angle and that when the crystal was away from the line-of-sight position. The flux of the incident N 2 O toward the surface decreased proportionally to cos θ when the angle shifted from the normal direction. This decrement was evaluated and the signal intensity was corrected [6]. The surface was cleaned by Ar + ion bombardments in the surface temperature (T S ) range between 800 and 900 K, heating in 5 × 10 −8 Torr oxygen at 850 K, and annealing to 1100 K for Pd(110) and to 1200 K for Rh(110).

Reaction rate
For the steady-state NO+CO reaction on Pd(110) under the condition of P NO /P CO = 1, the AR signals of N 2 , CO 2 , and N 2 O at θ = 0 • and the AR N 2 signals at θ = 41 • were measured ( Fig. 2(a)). The angles cited above are the collimation angles at which the desorption flux became maximum. The N 2 signal at θ = 41 • rapidly increased with increasing T S , reached a maximum rate at about 540 K and then decreased above it. This peculiar desorption is due to the intermediate N 2 O(a) decomposition [2].
The N 2 signal at θ = 0 • , on the other hand, increased slowly and peaked at about 600 K. This component comes from the associative desorption of nitrogen adatoms [1]. No hysteresis was found in their signals with decreasing T S , in contrast to the results on Rh(110) at P NO > P CO . In the latter, the surface was highly oxidized, showing significant hysteresis [6]. The AR CO 2 signal increased at around 500 K and decreased above 570 K with increasing T S ; however, it did so more slowly than the N 2 signal at θ = 41 • . The temperature dependence of the N 2 O signal was similar to that of N 2 at θ = 41 • . The N 2 signal at θ = 0 • was much less than that at θ = 41 • below 600 K. Above 650 K, the normally directed N 2 signal overcame the other. The reaction pathway of surface-nitrogen removal shifted to the associative process with increasing T S .
The maximum N 2 flux is located at θ = 43 − 46 • in the N 2 O+CO reaction. The signals at its collimation angle at 3.3 × 10 −6 Torr of N 2 O and 0.5 × 10 −6 Torr of CO are shown vs. T S in Fig. 2(b). The AR N 2 signal became noticeable above 450 K. The signal is maximized at 520 K and decreased quickly at higher temperatures. No AR N 2 signal was found in the normal direction. The other product CO 2 desorption collimated along the surface normal in a similar way to that in the NO+CO reaction. The higher signal intensity of N 2 over that of CO 2 is due to  their different angular distributions because these products should be balanced. Only the (1 × 1) LEED pattern is observed under a steady-state CO+NO reaction at the total pressure of 1 × 10 −7 Torr of the equi-molar mixture of NO and CO and in the range of T S = 400 − 800 K, indicating that CO adsorption is faster than NO dissociation. The LEED pattern was converted from a (1 × 1) structure into (2 × 3)-1D due to oxygen adsorption at the kinetic transition point for the steady-state N 2 O+CO reaction or above it. This point was observed when the N 2 O/CO pressure ratio was about 13. Both N 2 O and NO reductions are seriously retarded by the adsorbed oxygen [12][13][14], and the ratelimiting steps of the reactions (NO and N 2 O dissociation) are likely to proceed on clean parts that are free from oxygen, i.e., the inclined N 2 emission proceeds on the (1× 1) part.

Angular distribution and crystal azimuth
In the N 2 O+CO reaction, desorbing N 2 always splits in a two-directional way in the plane along the [001] direction and collimates at θ = 43 − 46 • off the surface normal in the temperature range studied, 400-800 K. No normally directed desorption was found even at 800 K. The angular distributions at different crystal azimuths at 520 K, P N2O = 3.3 × 10 −6 Torr, and P CO = 0.5 × 10 −6 Torr are shown in Fig. 3. This is under reducing conditions because the CO pressure is above the kinetic transition point (P N2O /P CO = 13 at 520 K), i.e., CO(a)>>O(a) [15]. The signal at φ = 0 • (along the [001] direction) was approximated as a cos 22 (θ + 45) + cos 22 (θ − 45) form. The distribution became broader with an increasing azimuth shift and the signal intensity decreased quickly and was suppressed above φ = 40 • .
The angular distribution of desorbing N 2 in the NO+CO reaction below 550 K shows very similar azimuth dependence to that in the N 2 O+CO reaction, indicating the same N 2 emission process. The distributions at 550 K and P NO = P CO = 5 × 10 −6 Torr are shown in Fig. 4. This is the optimum condition for N 2 formation at an equi-molar (NO+CO) mixture. The signal at φ = 0 • is approximated as a cos 28 (θ +41)+cos 28 (θ −41)+ 0.03 cos 5 (θ)+0.06 cos(θ) form. The latter two components are estimated from the velocity analysis described in the next section. The intensity of these components is independent of the φ position. The distribution again became broader with an increasing azimuth shift and the signal decreased quickly around φ = 40 • . The remaining signal at φ = 50 − 90 • was mostly due to the normally directed and cosine components.
At temperatures above 600 K and higher P CO /P NO ratios, the distribution changed significantly [16]. The angular distribution at 640 K involved three desorption components. The normally directed and cosine components were enhanced. The distribution at φ = 0 • was still in the two-directional form. Only the intensity of the inclined component decreased with increasing azimuth shift. The cosine component became major at crystal azimuths above φ = 25 • .
The CO 2 desorption collimated sharply along the surface normal in both N 2 O and NO reduction. No differences were found in the CO 2 distribution between them. The distribution in the NO+CO reaction was approximated as a {cos 13 (θ) + 0.2 cos(θ)} form at φ = 0 • and a {cos 4 (θ) + 0.2 cos(θ)} form at φ = 90 • .

Velocity distribution
Typical velocity distributions of desorbing N 2 at T S = 550 K and P NO = P CO = 5 × 10 −5 Torr are shown in cosine components are minor below 600 K; however, they are enhanced at higher temperatures. In this paper, the cosine component is considered to be formed in the associative process of nitrogen adatoms because it is not found in the N 2 O+CO reaction.
Each velocity distribution curve is too wide to fit to one modified Maxwellian form even after the subtraction of the thermalized component. In particular, the distribution at around θ = 40 • and φ < 18 • is wide, extending to 4 kms −1 (Fig. 5b,c). After the subtraction of the thermalized component, the distribution yields 1.1 at 550 K for the speed ratio (SR) defined as (< v 2 > / < v > 2 −1) 1/2 /(32/9π − 1) 1/2 , where v is the velocity of the molecule, < v > is the mean velocity, and < v 2 > is the mean square velocity. The SR value is usually below unity for a hyper-thermal component at around the collimation position [17]. Hence, the distribution curve was deconvoluted into two components of the modified Maxwellian distribution, where v 0 is the stream velocity and α is the width parameter. Here, we simply assumed a common α value for the deconvolution procedures [6]. The resultant deconvolutions are shown by broken curves. The faster component showed 5600-6100 K, and the slower one 2000-2400 K.

Kinetics
The product formation rates in the NO+CO reaction on Rh(110) are about one order lower than those on Pd(110). Both N 2 and CO 2 products are desorbed sharply along the surface normal. The N 2 desorption became sharper with increasing T S at the total pressure of 7.5 × 10 −5 Torr with the (1:1) mixture, from cos 2 (θ) at 470 K to cos 7 (θ) at 750 K. This suggests the contribution from the inclined N 2 desorption because the sharp angular distribution usually becomes broader at higher temperatures [6]. Nevertheless, we cannot clearly identify the inclined N 2 desorption from the N 2 O intermediate pathway on Rh(110) in the pressure range of 10 −7 − 10 −4 Torr of NO and T S = 300 − 800 K. This is to be expected because the NO decomposition is very fast on this surface, i.e., there is no NO(a) [18,19] and then the N 2 O formation becomes significant at P NO above 0.1 Torr [20]. On the other hand, the N 2 O+CO reaction on Rh(110) is faster than that on Pd(110) because of the larger sticking probability of N 2 O [18]. The reaction kinetics was reproducible when the ratio of P CO /P N2O was larger than 0.8, i.e., under reducing conditions. Under these conditions, the AR N 2 signal was maximized at around 57 • in the plane along the [001] direction. The temperature dependence of the N 2 signal at this collimation angle was very close to that at 45 • on Pd(110), i.e., the AR N 2 signal became significant at around 450 K, increased steeply with increasing T S , reached a maximum value at around 550 K and decreased slowly above it.

Angular and velocity distributions
The angular distribution of desorbing N 2 at 500 K is shown in Fig. 6 (P CO = 2.8 × 10 −7 Torr and P N2O = 3.0 × 10 −7 Torr). It is approximated as cos 10 (θ + 57) + cos 10  range of 500-650 K and at P CO > P N2O . This is very close to the results in the AR-TPD and AR-pressure jump below T S = 150 K [19]. The AR N 2 signal was noticeable in the normal direction. The velocity distributions at different desorption angles are shown in Fig. 7. The average translational temperature is inserted in < >. The value is maximized to 3100-3200 K at around 55 • with SR values of 1.05 − 1.15. It is very close to that on Pd(110) and less sensitive to the desorption angle, consistent with a broader angular distribution. The translational temperature after subtraction of the thermalized component reached about 4000 K and was insensitive to the desorption angle. The velocity curve was further deconvoluted into two components by assuming a common α value and the results are shown by the broken curves. The highvelocity component reached about 5900 − 7500 K with an SR value of around 0.6 and the slower one reached about 2400 − 2600 K with an SR value of around 0.92. It is obvious that the velocity is less sensitive to the desorption angle on Rh(110) although the maximum value is very close to that on Pd(110).

A. Reaction pathway and intermediate N2O
Three branching of the N 2 O path to the N(a) recombination path was estimated by integrating the AR signals over the angle around the collimation axis [23] to get the ratio of the sum of the inclined N 2 and the cosine N 2 O desorption to that of the normally directed N 2 and the cosine N 2 desorption (Fig. 2(c)). Process (i) overcomes the other above about 650 K. The reaction proceeds mostly through the N 2 O pathway on Pd(110). On the other hand, process (i) is always predominant on Rh(110) on which process (ii) can be noticed under limited conditions [6]. Nitrous oxide (N 2 O) is the key intermediate in controlling the selectivity to N 2 in catalytic de-NO x treatments on both surfaces.
A clean Pd(110) plane shows a stable (1 × 1) form, whereas it is reconstructed into missing-row forms when it is covered by oxygen. The resultant surface consists of three-or four-atom-wide terraces of a (111) structure with the zigzag oxide chain. The terrace declines alternatively about +30 • or −30 • in the [001] direction [24]. However, no inclined CO 2 desorption is found on Pd(110) even when the c(2 × 4)-O lattice due to the missing-row structure is observed. The lack of inclined CO 2 desorption on this surface is well known in AR-TPD work below 400 K [1,25]. It is due to the instability of the (1×2) reconstruction without oxygen toward the (1 × 1) form above 355 K and/or to the stable c(2 × 4)-O lattice highly covered by oxygen. In the former, normally directed desorption is expected when the reaction takes place only on the (1×1) facets. In the latter, the desorption of bulky CO 2 from the inclined terrace may be affected by the zigzag oxide chain on the nearest terrace [1]. The oxygen coverage is 0.5 on Pd(110)-c(2 × 4)-O or on Rh(110)-(2 × 2)p2mg-O when the zigzag chain fully covers the surface. In fact, the CO 2 desorption is well split on Pt(110)(1 × 2) when the oxygen coverage is below about 0.35 monolayer and commonly shifts to the surface normal on reconstructed Pd(110), Rh(110), and Pt(110) surfaces when the (1 × 2) surface is highly covered by oxygen [1,26]. Neither NO nor N 2 O dissociation proceeds on the oxygen-covered surface [12][13][14]. Thus, the N 2 inclined emission proceeds on oxygen-free (1 × 1) parts.
The reaction pathway of NO reduction through the N 2 O decomposition pathway is operative at low temperatures, where both N(a) and NO(a) are significant. Nitrous oxide (N 2 O) is decomposed even below 150 K on both Pd(110) and Rh(110) surfaces, emitting N 2 in the inclined way [14,19]. This peculiar desorption is induced in the decomposition of N 2 O oriented along the [001] direction. This [001]-oriented N 2 O has been confirmed on Pd(110) and Rh(110) by density-functional theory with generalized gradient approximations (DFT-GGA) [27,28], scanning tunneling microscope (STM) [29], and near-edge Xray-absorption fine structure (NEXAFS) [30] work below 60 K. These studies commonly support the presence of two adsorption forms, i.e., one lying form oriented along the [001] direction and a tilted form with the terminal nitrogen atom interacting with the metal. This is also consistent with the results from vibrational spectroscopy at around 100 K [31]. No N 2 O adsorption was detected on Pd in the steady-state NO reduction above 400 K by spectroscopic methods [32,33].

B. Three-dimensional presentation
The above collimation angle of desorbing N 2 shows noticeable differences on Pd(110) between the NO and N 2 O reduction, i.e., 41 ± 2 • and 45 ± 2 • , respectively. It is unlikely that the intermediate N 2 O in the NO reduction is in different adsorption states from that of N 2 O(a) supplied from gaseous N 2 O [21] because part of the intermediate desorbs without dissociation after thermalization as observed in its cosine distribution. Rather, the interme-diate N 2 O may be affected by co-adsorbed species, such as N(a) and/or NO(a). In fact, the collimation angle of desorbing N 2 from the N 2 O decomposition was reported to be affected by the co-adsorbed species and the kind of metal. In AR-TPD work of adsorbed N 2 O on Pd(110), the collimation angle of desorbing N 2 shifted from 50 • at low N 2 O density to 44 ± 2 • at higher N 2 O(a) and/or O(a) density in the range of 100 − 160 K.
For the anisotropy analysis of the inclined N 2 desorption, the following transformation from polar coordinates to the other polar angle system is necessary (see Fig. 1). According to the rotation defining the Eulerian angles, the relation between angles (α, β) and (θ, φ) is given by cos θ = cos α cos β and tan φ = − tan β/ sin α [34]. α is the longitude measured from the normal [110] direction about the [110] axis when the polar axis is taken to be parallel to the crystal azimuth [110], whereas β becomes the longitude shifted from the plane along the [001] direction when the polar axis is parallel to the [001] axis.
The experimental AR signals at definite (θ, φ) values were converted into the signal intensity at new coordinates (α, β) after smoothing the data points against a varying θ value at fixed φ values. The signals estimated at α = 45 • for the N 2 O+CO reaction are shown as a function of β in Fig. 1(c). The resultant distribution at α = 45 • was approximated in a cos n β form with n = 17 ± 3. Very similar β dependences were obtained at α = 41 • from the inclined N 2 desorption in the NO+CO reaction at 550 K.
The spatial distribution of desorbing N 2 is shown on three-dimensional polar coordinates in the inset in Fig.  1(c). This was drawn by assuming a distribution with two-fold symmetry around the collimation axis at θ = 45 • and φ = 90 • for the inclined N 2 desorption component. In a three-dimensional way, the N 2 distribution for the N 2 O+CO reaction was approximated as cos 22 (α ± 45) cos 17 (β). The N 2 distribution for the NO+CO reaction was in a form of cos 28 (α ± 41) cos 17 (β) at 550 K.
Similarly, the fast CO 2 component was approximated as a cos 13 α cos 4 β form, commonly for the N 2 O+CO and NO+CO reactions. This anisotropy is very close to that in the CO+O 2 reaction on Pd(110) [35].

C. Desorption anisotropy
A remarkable anisotropy in the product desorption was first found in the CO 2 desorption in the CO(a)+O(a) reaction on Pd(110) [36]. The desorption was approximated in a cos 3 (θ) form at φ = 90 • and a cos 10 (θ) form at φ = 0 • . This anisotropy was well reproduced for the CO 2 in the NO (or N 2 O)+CO reaction. Similar anisotropic CO 2 distributions in the CO oxidation have been found on Pt(110)(1×2), Ir(110)(1×2) and stepped platinum surfaces [1]. The distribution is commonly broad along the trough or narrow terraces and becomes sharp perpendicular to it. Thus, the anisotropy is due to the anisotropic structures of the product formation site. Facile movements of the transition state (TS) will yield a broader angular distribution. This model originates in the theoretical treatment of the reactive desorption of hydrogen molecules [37,38] and is available for associative desorption [1,35].
On the other hand, the anisotropy in dissociative desorption has been frequently reported in ESDIAD. This is due to the anisotropic motion of the parent molecule, i.e., the angular distribution becomes broader in the plane where the molecule is less restricted to vibrate [7]. The N 2 desorption from adsorbed N 2 O is sharper along the [001] direction than that perpendicular to it, suggesting that the intermediate N 2 O movement is more restricted along the [001] direction than the [110] direction. The movement of N 2 O with the terminal oxygen bonding to metal atoms has not been examined in DFT work. This form is not stabilized on clean Pd(110) and Rh(110) within DFT work [27,28]. Of course, the situation may be different for the transition state or in the presence of adsorbed oxygen.
The collimation angle of the product N 2 depends on the kind of metal, i.e., below 150 K, 70 • ± 5 • on Rh(110), 65 • ±5 • on Ir(110) and 43 • −50 • on Pd(110) [39]. This sequence is consistent with the hot atom-assisted desorption model [40]. In this model, a nascent oxygen atom (hotatom) provides a surface-parallel momentum to desorbing N 2 when N 2 O(a) molecules oriented along the [001] direction are decomposed. During stabilization of the nascent product O(a) in N 2 O dissociation, a large amount of the energy due to the O-metal bonding must be dissipated [41]. The available energy delivered to the product comes primarily from the O-metal bond formation because the heat of adsorption of N 2 O is close to the activation energy of dissociation [14]. This hot atom-assisted model predicts larger collimation angles and higher kinetic energy on rhodium and iridium than on palladium because the larger amounts of energy are released in the O-metal bond formation [42]. However, we have not confirmed this high energy component on Rh(110), i.e., the kinetic energy is close on both surfaces. The sharpness of the angular distribution is sensitive to the surface condition. Indeed, the Rh(110) surface is more difficult to flatten than Pd(110).

V. CONCLUSIONS
Product N 2 desorption in the steady-state NO+CO and N 2 O+CO reactions on Pd(110) and Rh(110) was studied through analysis of its angle and velocity distributions. The following results were obtained.
(1) N 2 desorption in both reactions on Pd (110)  (2) Above 600 K, the normally directed N 2 desorption is enhanced in the NO reduction on Pd(110). This comes from the associative desorption of N(a).
(3) The translational temperature of desorbing fast N 2 is maximized at the collimation angle to about 3400 K on Pd(110) and ca. 4000 K on Rh(110).