J. Phys. Chem. B 2001, 105, 7771-7774
7771
Evidence for the Formation of Nitrogen Islands on Rhodium Surfaces Francisco Zaera,* Stefan Wehner, and Chinnakonda S. Gopinath Department of Chemistry, UniVersity of California, RiVerside, California 92521
Jose´ Luis Sales, Vero´ nica Gargiulo, and Giorgio Zgrablich Laboratorio de Ciencias de Superficies y Medios Porosos, UniVersidad Nacional de San Luis, Chacabuco 917, 5700 San Luis, Argentina ReceiVed: April 17, 2001
Kinetic evidence is presented for the formation of nitrogen-atom islands on Rh(111) surfaces. The deposition of atomic nitrogen on that surface was carried out via the steady-state conversion of NO with CO by using effusive collimated beams. Isotopic labeling experiments were performed where 14N-dosed surfaces were subsequently exposed to 15NO + CO gas mixtures for varying times. Subsequent temperature-programmed desorption indicated a nonstatistical distribution of isotopes in the resulting molecular nitrogen, the yield of the mixed 14N15N isotopomer being significantly lower than that expected on statistical grounds. Monte Carlo simulations were used to explain the observed isotopic distributions in terms of the formation of islands with the nitrogen isotopes distributed in a layered structure, the 14N atoms in a core surrounded by a 15N outer shell.
Introduction Interactions among adsorbates on solids often play a determining role in the final arrangement of those molecules on the surface.1,2 In particular, strong adsorbate-adsorbate attractions commonly lead to the formation of surface islands.3,4 A typical example of this is provided by the (2 × 2) ordered twodimensional clusters formed by surface oxygen on the (111) planes of many late transition metals.5-7 On Pt(111), both lowenergy electron diffraction (LEED)8 and scanning tunneling microscopy (STM)9,10 have indicated that (2 × 2) oxygen islands form at coverages as low as 0.04 ML, and that they are stable up to at least 450-500 K. The formation of these islands affects the kinetics of subsequent surface reactions, sometimes in dramatic ways.11,12 For instance, two kinetically distinct types of oxygen atoms have been identified during the isothermal oxidation of CO on Pt(111) even though they all sit in identical sites at the start of the reaction.13,14 In addition, a Pt(111) surface with O atoms clustered in ordered islands displays CO oxidation rates as much as four times faster than when the same number of oxygen adsorbates are randomly distributed on the same metal surface.13 Despite the similarities between oxygen and nitrogen atoms, much less is known about the adsorption of the latter on metals.15 Disordered structures have been often proposed for N atoms on some hexagonal basal planes, but (x3 × x3)R30° and (2 × 2) ordered phases are known on Ru(0001).16,17 On Rh(111), a diffuse (1 × 1) LEED pattern has been interpreted as a sign of disordered adsorption.18,19 On the other hand, a (2 × 1) structure has been reported for nitrogen on Rh(111) when in the presence of some coadsorbed oxygen.20 Unfortunately, the indirect STM evidence supporting this conclusion is less than conclusive.20 In our recent kinetic studies on the reduction of NO by CO on Rh(111),21-27 it was determined * Corresponding author. Fax: 1 (909) 787-3962. E-mail: francisco.zaera@ ucr.edu.
that below 500 K the surface of the rhodium catalyst is partially covered by atomic nitrogen during the steady-state reaction. Moreover, isotope-labeling experiments indicated that the rate of replacement of 14N by 15N on the surface upon switching from 14NO + CO to 15NO + CO reaction mixtures follow a complex kinetic behavior.23 Those kinetics have been explained by the preferential removal of atoms from the periphery of surface N islands.23 Finally, the nitrogen molecules made during the steady-state conversion of the NO + CO mixtures were determined to always contain at least one 15N atom, even immediately after the isotopic switchover, an observation that was interpreted as the result of reactions between newly adsorbed 15NO molecules and nitrogen atoms from the edges of the surface N islands, and of the formation of a N-NO intermediate.25 Here we provide more direct evidence for the formation of surface nitrogen islands on Rh(111) surfaces. Experimental Section The apparatus used in the experiments reported here has been described in detail elsewhere.28 Briefly, it consists of a stainless steel ultrahigh vacuum (UHV) vessel pumped to a base pressure below 2 × 10-10 Torr and equipped with a computercontrolled mass quadrupole for gas identification, kinetic determinations, and temperature-programmed desorption (TPD) experiments. A collimated multichannel microcapillary array doser 1.2 cm in diameter is used for directional exposure of the sample to the reactants, a sputtering ion gun for sample cleaning, and a crystal holder for three-dimensional translation and onaxis rotation as well as for resistive heating and liquid nitrogen cooling. The Rh(111) single-crystal, a 1.10 × 0.56 cm2 rectangle, was cleaned in situ, initially by Ar+ sputtering and before each experiment by cycles of oxygen exposures (1 × 10-7 Torr at 900 K for up to 20 min) and annealing to 1200 K until the NO TPD spectra reported in the literature21 could be reproduced. The surface temperature was monitored continuously with a chromel-alumel thermocouple spot-welded to the
10.1021/jp011427a CCC: $20.00 © 2001 American Chemical Society Published on Web 07/19/2001
7772 J. Phys. Chem. B, Vol. 105, No. 32, 2001
Zaera et al.
Figure 1. Temperature-programmed desorption (TPD) traces from Rh(111) surfaces dosed with atomic nitrogen by following a procedure described in detail in the text. All five panels display traces for the desorption of 14N14N, 14N15N, and 15N15N from 14N + 15N mixed layers. The different panels correspond to surfaces prepared by exchanging an initial pure 14N layer with some 15N via reaction with a 15NO + CO beam at 480 K for different times t (from left to right, t ) 0, 2.5, 10, 50, and 200 s). The fraction of 14N exchanged in each case, X(15N), calculated by integration of the TPD traces, is reported as a percentage in the corresponding panel. The total nitrogen atom coverage on the surface was 0.17 monolayers in all cases.
back of the crystal, kept constant at 480 K during all of the exposures to the NO + CO beams, and linearly ramped at a rate of 10 K/s for the TPD experiments with a homemade precision temperature controller. Isotopically labeled 15NO (CIL, 98% 15N purity) and regular 14NO (Matheson, 99.9% purity) and CO (Matheson, 99.9% purity) were used as supplied. Nitrogen atoms were deposited on the Rh(111) surface via isothermal kinetic experiments using a variation of the molecular beam method originally developed by King and Wells.14,29,30 The Rh(111) single-crystal surface was initially exposed to a 1:1 14NO + CO mixed collimated effusive beam for a period of time sufficiently long (210 s) to reach the appropriate steadystate catalytic condition. It has previously been determined that a steady-state coverage of strongly bonded atomic nitrogen (ΘN ≈ 0.17 ML) is present at that point on the surface, and that no significant amount of oxygen remains adsorbed.23,24 The original 14NO + CO beam was then rapidly replaced by an identical 15NO + CO beam, after which the reaction was run for variable periods of time t before shutting it off completely. After that, the system was pumped to its base pressure, and nitrogen TPD spectra for 14N2 -28 amu-, 14N15N -29 amu-, and 15N2 -30 amu- were taken. It is important to note that, since the second nitrogen isotope-exchange part of the experiment was carried out under the same steady-state reaction conditions used to deposit the initial 14N atoms, the surface coverage obtained was the same in all cases; only the isotopic composition of the nitrogen overlayer changed. Results and Discussion Typical molecular nitrogen TPD data from these experiments are shown in Figure 1 for five different times t. As expected, some of the 14N initially deposited by the conversion of the 14NO + CO mixture are replaced by 15N upon exposure to the second 15NO + CO beam. This is manifested in the TPD data by a clear reduction in the 14N14N peak at the expense of the growth of the 15N15N feature with increasing t. The yield for 14N15N, meanwhile, goes through a maximum after about 50 s of reaction with the 15N-labeled nitrogen monoxide before
decreasing again at longer times. The fraction of the total nitrogen replaced, as calculated by integration of the TPD peaks, also increases monotonically with time, but in a fashion difficult to reconcile with any simple kinetic model.23 It is this temporal behavior of the isotopic exchange of the adsorbed atomic nitrogen during the catalytic conversion of NO to N2 that first led us to propose the formation of surface N islands. The molecular nitrogen isotopomer desorption yields from the different TPD experiments, Y(xNyN), were quantified via integration of desorption traces such as those illustrated in Figure 1, and plotted against the fraction of 15N on the surface,
X(15N) )
(1/2) × Y(14N15N) + Y(15N15N) Y(14N14N) + Y(15N15N) + Y(15N15N)
in Figure 2 (filled symbols). Error bars, calculated by several repetitions of the experiments under identical conditions, were added to the data for 14N15N; they correspond to a margin of error of approximately (15% (using Student’s t for 95% confidence). Also plotted in Figure 2 (solid lines) are the yield fractions for all three N2 isotopomers estimated by assuming a random distribution of atomic isomers (Y(xNxN) ) X2(xN) and Y(xNyN) ) 2‚X(xN)‚X(yN)). One thing becomes immediately apparent from simple inspection of this figure: the distribution of nitrogen isotopes in the desorbing molecular nitrogen does not follow the statistical behavior expected from homogeneous mixtures. Specifically, the yield of the mixed-isotope (14N15N) nitrogen molecules is noticeably smaller than that expected on statistical grounds. These results imply a nonrandom distribution of nitrogen atoms on the surface and a limited diffusivity of those atoms under reaction conditions. It is argued here that such nonstatistical distribution supports the initial proposal of nitrogen island formation. Simulations To better justify the isotopomeric distributions of the desorbing molecular nitrogen observed in these N/Rh(111) studies,
Formation of Nitrogen Islands on Rh(111) Surfaces
Figure 2. 14N14N, 14N15N, and 15N15N TPD yield fractions as a function of the fraction of 15N within the total surface atomic nitrogen. The filled symbols correspond to the data obtained experimentally by integration of TPD traces such as those shown in Figure 1. The open symbols correspond to the results from a Monte Carlo simulation starting with hexagonal islands containing 61 atoms (5 layers) each and by assuming surface diffusion (see text for details). The lines correspond to the yields expected on statistical grounds.
J. Phys. Chem. B, Vol. 105, No. 32, 2001 7773 a number of Monte Carlo simulations were carried out for the desorption process. The starting point for such simulations was our initial model of nitrogen surface islands with preferential reactivity at the periphery.23 For simplicity, perfect hexagonal nitrogen-atom islands were assumed, the initial sizes and shapes of which were kept the same for all isotope-switching experiments. Only their isotopic composition was varied as a function of exposure time to the 15NO + CO beam, that is, as a function of X(15N). Because of the proposed preferential exchange at the edges of the islands, they were initially set with a core of 14N atoms surrounded by a 15N shell in a layered fashion. Two sets of simulations were carried out for each island size. In the first, no atom mobility was allowed. Pairs of nearest neighbor atoms were successively chosen at random and desorbed. Once all those possible pairs were consumed, the process continued through the desorption of randomly chosen next nearest-neighbor pairs, then of second-next nearestneighbors, and so on until all atoms were removed from the surface. As this simulated desorption process was carried out, record was kept of each desorbing pair in terms of its isotopomeric composition, as 14N14N, 14N15N, or 15N15N molecules. In the second set of simulations, surface N atoms were allowed to diffuse to fill in the holes left in the islands by the desorbing atom pairs. That process can be better visualized as a random diffusion of holes toward the periphery of the island. The second simulation method was similar to the first, except that each time a randomly chosen nearest-neighbor pair was desorbed, a diffusion process for the pair of holes was then carried out by moving each hole to a randomly chosen occupied nearest-neighbor site. Such a diffusion process was continued until both holes reached the island periphery. This hole diffusion process eliminated the necessity of having to desorb pairs farther apart than nearest neighbor, so that only nearest-neighbor desorption needed to be considered in this case. For each system,
Figure 3. Results from Monte Carlo simulations on the desorption of molecular nitrogen from atomic-nitrogen surface islands by using the model and algorithm described in the text. The simulations started with perfect hexagonal islands of varying sizes, ranging from 2 to 5 layers (from 7 to 61 atoms per island, from left to right). Two sets of simulations were performed for each island size, with (bottom) and without (top) surface atom diffusion. Plotted are molecular nitrogen yield fractions for all three possible isotopomers as a function of 15N fraction, the same as in Figure 2.
7774 J. Phys. Chem. B, Vol. 105, No. 32, 2001 the desorbing yield results were averaged over 105 independent runs to reduce statistical fluctuations (which amounted to less than 1%). Typical results from our Monte Carlo simulations are shown in Figure 3 for the two sets mentioned above, namely, with (bottom) and without (top) atom diffusion. In these, the island size was varied from 2 (7 atoms total) to 5 (61 atoms total) layers. A number of observations are worth highlighting from these data: (1) the isotopomeric distributions obtained in these simulations are nonstatistical, qualitatively deviating in the same way as the experimental results (that is, by yielding more 14N14N and 15N15N at the expense of 14N15N production); (2) the nonstatistical deviations become more pronounced as the size of the islands is increased; (3) the behavior of the yield for 14N15N desorption shows an asymmetry with respect to the fraction of 15N in the island (recall that these islands contain a 14N core surrounded by a 15N shell); (4) both the deviations from statistical behavior and the asymmetry mentioned above are more pronounced in the absence of atom mobility. It was found that the Monte Carlo simulations reproduced the experimental data quite adequately. The best fit was obtained with the five-layered islands where atomic diffusion was allowed. The results from that simulation were included in Figure 2 (open symbols) for a better direct comparison between simulations and experiments. It can be seen by simple inspection that the agreement is excellent, especially in the high 15Nfraction region. To quantify the goodness of the fit, a couple of statistical tests were performed. First, reduced χ2 parameters (χν2) were calculated to compare the experimental data against both a statistical scrambling of isotopes within the desorbing nitrogen molecules and the results from the Monte Carlo simulations chosen for Figure 2. Values for χν2 of 2.80 and 0.89 where obtained for those cases, corresponding to probabilities of less than 0.1 and 65%, respectively. This means that the first model, the complete statistical scrambling of the isotopes, is highly improbable. It also implies that, statistically, there is only about a 35% chance of finding a better model than our Monte Carlo results to describe the TPD experimental data. However, these conclusions should be taken with some degree of skepticism, because χ2 measures not only the discrepancy between the estimated models and the data but also deviations within the data themselves. An F test was therefore applied to contrast the relative validity of our two models. The ratio of the χ2 parameters, approximately 3.1 (24 degrees of freedom), leads to the conclusion that the statistical isotope scrambling model can be rejected in favor of our Monte Carlo simulations with better than 99% confidence. The agreement between our data and the Monte Carlo simulations is quite encouraging, since the model used here was quite simple, containing no adjustable parameters. Also, atomic migration to fill in the island holes is easily justified by the presumed N-N attractive interactions that also account for the formation of the islands and by the expected surface atom mobility at the high temperatures reached above the onset of N2 desorption. It should be emphasized, however, that all we can say at this point is that statistical isotope scrambling can be safely ruled out and that the islanding model used in our Monte Carlo simulations is justifiable on statistical grounds. This does not mean that a better model cannot be found. In fact, our calculations are not meant to imply that the nitrogen atoms only form perfect five-layered hexagonal islands on the surface. Most likely, there is a distribution of island sizes and shapes around that average. A sense of how island size affects the isotope-labeling TPD results can be obtained from inspection
Zaera et al. of Figure 3. On the basis of those data, we speculate that the distribution of nitrogen island sizes may be highly peaked around 5 ( 2 layers. Better fits could be obtained by weight-averaging the results from simulations with different island sizes, but that would require the addition of adjustable parameters to describe the island size distribution. At this point, the quality of the experimental data does not justify such extension. Conclusions In summary, isotope-labeling data were presented here to support the nonstatistical nature of atomic nitrogen recombination on Rh(111). The desorption behavior observed was justified by the formation of N islands on the surface, a model validated by Monte Carlo simulations. The information reported here adds to our previous evidence, which comprises isotope-exchange kinetics for the replacement of labeled nitrogen atoms on the surface23 as well as for the production of nitrogen molecules during the steady-state reduction of NO by CO.25 Future tests of the islanding model will encompass simulations of N2 TPD22 and isothermal desorption kinetic traces.31 Acknowledgment. Funding for this research was provided by a Grant from the National Science Foundation (CTS9812760). Additional funding came from a Los AlamosUniversity of California joint program. The Consejo Nacional de Investigaciones Cientı´ficas y Te´cnicas (CONICET) of Argentina is also acknowledged for financial support of the Argentinean researchers. References and Notes (1) Goymour, C. G.; King, D. A. J. Chem. Soc., Faraday Trans. 1 1973, 69, 749. (2) Adams, D. L. Surf. Sci. 1974, 42, 12. (3) Lagally, M. G.; Wang, G.-C.; Lu, T.-M. CRC Crit. ReV. Solid State Mater. Sci. 1978, 7, 233. (4) Silverberg, M.; Ben-Shaul, A. J. Chem. Phys. 1987, 87, 3178. (5) Thiel, P. A.; Yates, J. T., Jr.; Weinberg, W. H. Surf. Sci. 1979, 82, 22. (6) Brundle, C. R. In The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis; King, D. A., Woodruff, D. P., Eds.; Elsevier: Amsterdam, 1990; Vol. 3A (Chemisorption Systems); pp 132-388. (7) Besenbacher, F.; Nørskov, J. K. Prog. Surf. Sci. 1993, 44, 5. (8) Gland, J. L.; Sexton, B. A.; Fisher, G. B. Surf. Sci. 1980, 95, 587. (9) Stipe, B. C.; Rezaei, M. A.; Ho, W. J. Chem. Phys. 1997, 107, 6443. (10) Zambelli, T.; Barth, J. V.; Wintterlin, J.; Ertl, G. Nature (London) 1997, 390, 495. (11) Akhter, S.; White, J. M. Surf. Sci. 1986, 171, 527. (12) Zgrablich, G.; Sales, J. L.; Unac, R.; Zhdanov, V. P. Surf. Sci. 1993, 290, 163. (13) Xu, M.; Liu, J.; Zaera, F. J. Chem. Phys. 1996, 104, 8825. (14) Zaera, F.; Liu, J.; Xu, M. J. Chem. Phys. 1997, 106, 4204. (15) Comelli, G.; Dhanak, V. R.; Kiskinova, M.; Prince, K. C.; Rosei, R. Surf. Sci. Rep. 1998, 32, 165. (16) Trost, J.; Zambelli, T.; Wintterlin, J.; Ertl, G. Phys. ReV. B 1996, 54, 17850. (17) Dietrich, H.; Jacobi, K.; Ertl, G. J. Chem. Phys. 1996, 105, 8944. (18) Berko, A.; Solymosi, F. Appl. Surf. Sci. 1992, 55, 193. (19) Belton, D. N.; DiMaggio, C. L.; Ng, K. Y. S. J. Catal. 1993, 144, 273. (20) Xu, H.; Ng, K. Y. S. Surf. Sci. 1996, 365, 779. (21) Aryafar, M.; Zaera, F. J. Catal. 1998, 175, 316. (22) Gopinath, C. S.; Zaera, F. J. Catal. 1999, 186, 387. (23) Zaera, F.; Gopinath, C. S. J. Chem. Phys. 1999, 111, 8088. (24) Gopinath, C. S.; Zaera, F. J. Phys. Chem. B 2000, 104, 3194. (25) Zaera, F.; Gopinath, C. S. Chem. Phys. Lett. 2000, 332, 209. (26) Zaera, F.; Gopinath, C. S. In Studies in Surface Science and Catalysis Series; Elsevier: Amsterdam, 2000; Vol. 130 (Proc. 12th Int. Congr. Catal., Granada, Spain, July 9-14, 2000); pp 1295-1300. (27) Zaera, F.; Gopinath, C. S. J. Mol. Catal. A 2001, 167, 23. (28) Liu, J.; Xu, M.; Nordmeyer, T.; Zaera, F. J. Phys. Chem. 1995, 99, 6167. (29) King, D. A.; Wells, M. G. Surf. Sci. 1972, 29, 454. (30) O ¨ fner, H.; Zaera, F. J. Phys. Chem. 1997, 101, 396. (31) Zaera, F.; Gopinath, C. S. 2000. In preparation.
