The geometry of ordered benzene layers on Ru(001)

The geometry of ordered benzene layers on Ru(001)

surface science ELSEVIER Surface Science 325 (1995) L379-L384 Surface Science Letters The geometry of ordered benzene layers on Ru (001 ) C. Stellw...

556KB Sizes 1 Downloads 11 Views

surface science ELSEVIER

Surface Science 325 (1995) L379-L384

Surface Science Letters

The geometry of ordered benzene layers on Ru (001 ) C. Stellwag, G. Held t, D. Menzel * Physik-Department E20, Technische Universit~itMiinchen, 85747 Garching, Germany Received 4 November 1994; accepted for publication 6 December 1994

Abstract

We have measured LEED-/V-curves for all known commensurate benzene structures on Ru(001), carefully excluding contamination and radiation damage effects, and have carded out a detailed analysis of the p (x/ff × x/if)R19.2°-structure. We find that the molecule is bound on the hcp-site in C3v(o'v) orientation, at an average distance from the metal of 2.12/~. The distortion of the molecule as compared to its free state is very small: the molecular radius is almost identical to that in the gas phase, but there is a small crown-like deviation from planarity which leads to good adjustment to the threefold symmetry of the site. We argue that this site selection and slight distortion of the molecule might be connected selfconsistently with the decreased ~--character of the chemisorbed molecule. Very small deviations from the gas phase value of the C-C bond length and of the distances of the C-atoms from the molecular center are found which are not outside the error bars, as is a possible small rotation of the ring in response to the neighbouring molecules. The main adsorbate induced modification of the substrate is a very large almost uniform contraction of the distance between the first and the second Ru layer of 0.05/~ compared to the clean surface which is the largest found on this surface so far. The very similar experimental/V-curves of the other ordered benzene structures on this surface suggest that they have basically the same local geometries. Keywords: Aromatics; Chemisorption; Low energy electron diffraction (LEED); Low index single crystal surfaces; Ruthenium

The geometry o f benzene layers on close-packed transition metal surfaces, both pure and with coadsorbed CO, is somewhat controversial. While all spectroscopic and geometric investigations agree that the benzene molecule lies fiat on these surfaces and is ~-bonded, the conclusions about the degree of modification introduced by this bonding into the molecule vary widely. Most spectroscopic works, both from electronic and vibrational studies, agree that this modification is small (in many cases the maximum expected C6v symmetry is maintained ( R u ( 0 0 1 ) [ 1], R h ( l l l ) * Corresponding author. Fax: +49 89 3209 2824; E-mail: [email protected] l Present address: University of Cambridge, Department of Chemistry, Lensfield Road, Cambridge CB2 1EW, UK.

[2,3], weak symmetry breaking with resulting C3v has been seen for N i ( l l l ) [4,5] ), while clear deviations from C6v symmetry have been found in STM and LEED-IV work, in particular for benzene with coadsorbed CO. For instance, STM-patterns o f benz e n e + C O / R h ( 1 1 1 ) show strongly threefold molecular images [6], and LEED-/V for the same surface has resulted in extremely distorted benzene (two C C bond types with d = 1 . 8 1 / 1 . 3 3 / ~ for one structure [7,8] ). On the other hand, for the p ( 3 x 3) benz e n e + 2 C O layer on P d ( l l l ) essentially no distortion has been derived from L E E D - / V [9]. The only ( D ) L E E D structure determination up to now for a (disordered) pure benzene layer on Pt ( 111 ) has found a boat-like distortion with C2v symmetry and C - C

0039-6028/95/$09.50 (~) 1995 Elsevier Science B.V. All rights reserved SSDI 0 0 3 9 - 6 0 2 8 ( 9 4 ) 00842-6

C. Stellwag et al./Surface Science 325 (1995) L379-L384

bona lengms of 1.45 and 1.55/~, respectively [ 10]. It is important to note that the ordered phases of benzene on these surfaces are quite difficult to obtain, while even small amounts of coadsorbed CO often lead to well-ordered layers which are more easily accessible for a LEED structure determination. It is not clear at present whether coadsorbed CO is also instrumental in inducing the mentioned strong modifications of the benzene molecule. Also, as always with molecular adsorption, the possibility of radiation damage by the electrons used for the investigation has to be envisaged for these surfaces as well. A careful geometry determination for a well-characterized pure ordered benzene layer, under conditions of negligible radiation damage, appeared to be desirable, therefore. We have chosen the Ru(001) surface for this purpose, since ampie experience with the adsorption of benzene on it as well as on its coadsorption with various species (CO, NO, O, H [ 11,12] ) exists in this laboratory, and since pure benzene layers on this surface lead to a number of ordered structures [ 1,13 ]. Furthermore, in chemical terms one would expect the strongest effect on the benzene structure among the 4d-metals for ruthenium, therefore the distortions found in this investigation can be considered as an upper limit for distortions to be expected on other 4d-metals. The measurements were carded out in our video LEED system specifically designed to minimize the total electron doses necessary for data accumulation of complete/V spectra. As described in Ref. [ 14], a CCD video camera is used for this purpose which allows to rapidly store the complete information present on the screen at a certain electron energy for easy offline analysis. Together with provisions to leave the electron beam on the surface only during data accumulation and deflect it during data transfer and parameter readjustment, a complete/V data set can be accumulated with only a few electrons per surface molecule; such a run takes about 35 min with the irradiation of the sample lasting only about 20% of this time. A single image can be taken within 1 s at beam currents of 0.3/xA. In the present case it was not necessary to go to very stringent conditions, as has been checked by examining the constancy of the spot intensities with exposure to electrons. Of great importance, however, were the purity of surface and adsorbate which was ensured by low background pressure ( 5 x 10-1 xmbar) and dosing of the crystal only through a dosing sys-

tem (at the same time providing a homogeneous coverage), and the quality of the crystal. The latter may well be the cause (in addition to impurities) if ordered phases are suppressed for pure benzene. It was ensured by using procedures developed in this laboratory over the years which demonstrably (LEED profiles [ 15], rare gas adsorption [ 16] ) lead to step densities below 1%, and by mapping the crystal surface with respect to the (1,0)-beam reflectivity to select the best area (for further details see e.g. Ref. [ 17] ). Provided the purity of the benzene layer is better than 1% of CO (the main potential impurity), several ordered benzene structures have been seen earlier on this surface which are stable to 270 K [ 1 ] : a (4 2 ) or p(2x/3 × 2v/3)R30 ° structure ("2v/3 '') /

at a coverage of O = 1 relative to Ru surface atoms (50% of saturation), a (31 2) 3 ° r p ( v / f f x v/if)R19"2° structure ("x/if") at O = 1, and two incommensu0.57)at rate structures Yl 1.57 3.15j and y2 2.12 2.69 0.134 and at saturation (O = 0.163), respectively. They have all been reproduced in the present work.

(3.15 1.57"~

(2.69

~9=

In addition a (31~13 structure (or c(2x/~ x 4)rect) has been found in our investigation at a relative coverage of ¢9 = 1 which disorders already at 230 K. /V-curves for all the commensurate phases have been recorded; their similarities suggest that the site geometry of the benzene molecules is essentially the same in all these phases. We only mention this aspect here and will elaborate on it in a separate publication [ 13]. For the v/ff and the 2v/'3 phases, full dynamical LEED calculations have been carried out, with the most extensive variation of parameters for the v/if, because of its smallest surface mesh and, therefore, less - but still very large - computational effort. These have yielded unambiguous results which we will describe briefly for the v/-ffphase. Because of the undebated information on the fiatlying orientation and on the coverage which shows that the x,fff surface mesh can only contain one benzene molecule [ 1 ], we always assume these two features. Apart from these, a rather complete search has been conducted. 12 inequivalent beams (obtained by averaging of all equivalent beams) in the energy range between 40 and 260 eV with a total energy span of 1884 eV were used for comparison between experiments

C. Stellwag et al./Surface Science 325 (1995) L379-L384

and calculations (the energy range was restricted by the CPU time and the memory requirements of the calculations rather than by the available experimental data set). The analysis was carded out in three steps. In the first step, the molecule was assumed to remain undistorted, and the site, orientation, vertical distance from the first substrate layer dA1 (range: 1.05 to 2.55 /~, depending on the adsorption site), and distance between first and second substrate layer dl2 (range: 2.00 to 2.20/~), were varied. Adsorption sites with the molecule centered on top, bridge, fcc, and hcp hollow positions were considered, each with the two nonequivalent, high symmetry, rotational orientations of C-atoms pointing along the [ 110] or [ 1]-0] directions. The parameter space was scanned using fully dynamical LEED programs of the Van Hove/Tong type (RFS, combined space method, etc. [ 18] ) on DEC workstations. For comparison with the experimental spectra, the R-factors Rp [19] and RB1 [20] were used, which are sensitive to different aspects of the spectra; the energy span given above leads to an RRfactor of 13%. The clear answer from this scan was that the molecule is centered on the hcp site, with C-atoms occupying ontop and fcc sites (symmetry C3v(O'v)) leading to R-factors of Rp = 0.412 and Rm = 0.315. Even at this stage, all other geometries are made very improbable by their high R-factors (see Table 1 ). A second step tested the influence of molecular distortions such as lateral and vertical deformations (4-0.2/~) of the boat-(C2v), crown-(C3v(O-v)), and Kekule (C3v(trd)) types and variations of the ring radius (1.40-1.65/~) on the R-factors for all sites. In all cases the R-factor changes due to this optimization were small. This clearly selects the hcp (C3v(O'v)) geometry, and at the same time shows that the distortion of the molecule must be small. The best Rpfactor achieved in this search step was 0.367, which is a reasonably good value for a molecular structure of this complexity. The good agreement of the parameters optimized by the two R-factors Rp and Rm makes the conclusion even safer. In these two steps only an uniform relaxation of the first Ru layer distance was allowed, but no buckling of individual Ru atoms within the v/ff unit cell. This degree of refinement was added in the final step maintaining the C3v(O'v) symmetry of the structure. This time the optimization was performed by a computer program utilizing the simplex algorithm [ 21 ] together

with the same fully dynamical LEED programs as already used in the first two steps of the search and using Rp and Rm. An alternative search was performed using the automated tensor-LEED program (ATLEED) by Rous et al. [ 22 ] and Wander et al. [ 23 ], not assuming any symmetry. The most important parameters of the results of these three searches (Rp = 0.334/0.348, Rm = 0.270) are tabulated in Table 2. The parameters of all three optimum geometries agree very well within the error bars determined in the conventional way using the RR-factor [ 19]. Therefore, we use the averaged values (also tabulated in Table 2) for a description and discussion of this structure. The topmost substrate layer is deformed in such a way that the Ru atoms Rul and Ru2 neighbouring to the C-atoms of the ring are slightly lifted with respect to Ru3 (0.02/0.03/~) and relax away from the center of the molecule. The amounts of these shifts, however, are not significant considering the error bars of 0.05 and 0.10/~, respectively, while the contraction of the first-to-second Ru layer distance (2.05/~) by almost 5% with respect to the bulk value (2.14/~) or 3% with respect to the relaxed clean surface (2.10/~) is indeed striking as it is one of the largest found for a close packed surface so far. The strongest contractions found for this surface up to now are induced by the p(2 x 2)Cs [24] and the p ( v ~ x x/3)D20 bilayer [14] structures (2.04/~); in both these structures, however, not all atoms of the first Ru layer are at the same height so that the overall contraction is smaller than in the present case. A common feature of all three systems is that electrons are donated from the adsorbate to the substrate, as indicated by the work function change and the probable adsorbate bond. We note that the clean Ru(001) surface possesses a contraction of about 2% of the first,to-second layer distance [ 17,25,26], and that for electronegative adsorbate systems such as O, H, or O-t-CO, this contraction is removed or even slightly expanded overall (see Ref. [26] ). We therefore believe that the large contraction in our present case, as for other electropositive adsorbates, is due to the adsorbate-induced increase of electron density in the surface layer of the metal. A quantitative analysis would appear interesting. The slight deformation of the benzene ring follows the corrugation of the substrate at an average height of 2.12/~ above the first Ru layer: the "on top" C1 atoms are more elevated than those on "fcc" sites (C2). With

C. Stellwag et al./Surface Science 325 (1995) L379-L384

Table 1 Geometries tested in the first step of the search Geometry model

Range dl2 (/~)

Range dAl(/k)

Opt. d12 (/~)

Opt. dA1 (/k)

Best RB1

Best Rp

C6v (ontop) C ---+ [ 110] C ~ [ 1i0] C2v (bridge) C ~ [ 110] C --~ [ 1 i0] C3v (O'd) (fcc) C3v (o'v) (fee) C3v (trd) (hcp)

2.00-2.20 2.00-2.20

1.25-2.25 1.05-2.25

1.98 2.15

2.15 2.10

0.771 0.655

0.889 0.798

2.00-2.20 2.00-2.20 2.00-2.20 2.00-2.20 2.00-2.20

1.40-2.25 1.35-2.20 1.30-2.15 1.35-2.20 1.30-2.55

2.05 2.00 2.13 2.05 2.08

2.25 2.20 2.15 2.15 2.20

0.515 0.571 0.653 0.469 0.424

0.686 0.614 0.909 0.652 0.709

C3v (try) (hcp)

2.00-2.20

1.35-2.20

2.13

2.15

0.315

0.412

The notation C ~ [ l l 0 ] / [ l i 0 ] means that the vectors between the C-atoms and the center of the molecule axe parallel to the given (or an equivalent) direction, dA1 = vertical distance between adsorbate and first substrate layer; d12 = vertical distance between first and second substrate layer. The optimum parameters given are those for Rp.

Table 2 Comparison of the best fit parameters arrived at with tensor-LEED (using RF) and with full dynamical calculations using Rp and RB1, respectively (the labels of the atoms correspond to Fig. 1) Parameter

T-LEED

Rp

RB1

Average

Error

Vor

10.5 eV

10.4 eV

10.4 eV

10.4 eV

4-0.5 eV

0.08 A 1.46/~ 1.43 A 2.08 •

4-0.05/~ 4-0.10 A 4-0.10 A 4-0.04 /~

AZc-c Rcl RC2 dA1

0.07 1.45 1.41 2.11

A A ,~ A

0.05 1.45 1.43 2.10

A A /~ /~

0.12 A 1.48/~ 1.46/~ 2.02 /~

AZRul ARRul AZRu2 ARRu2 /~ZRu3 d12

0.00 0.02 0.01 0.08 --0.03 2.03

A A /~ A A A

0.00 0.04 0.00 0.08 0.00 2.04

h A h A A A

0.00 h --0.01 ,~ 0.02 .3, 0.06 A --0.03 A 2.06/k

R-factor

0.334

0.348

0.270

0.00 0.01 0.01 0.07 --0.02 2.04

A ,~ /~ A A ,~

4=0.05 h "4-0.I0 A 4=0.05 /~ 4-0.10 A 4-0.05 A 4-0.05/~ 13%

Vor = real part of the inner potential. AZc_c = buckling within the benzene ring (C1 versus C2). Rcl, Rc2 = ring radius for C1/C2, i.e., lateral distance from the molecular center. d M = vertical distance C2-1st Ru layer (center of mass). AZRul, AZRu2, AZRu3 = vertical distance of the atoms Rul/Ru2/Ru3 from the center of mass of the first Ra layer ( + = outwards). ARRul, ARRu2 = lateral shifts of Rul/Ru2 from their bulk positions (away from the molecular center). d12 = vertical distance 1st Ru layer (center of mass)-2nd Ru layer.

C. Stellwag et al./Surface Science 325 (1995) L379-L384

A

B

1.46 A~ 1.43 A

2_~.8A i ]

(t~

:

~

-

~

] 0.01 A

1.63 A

Fig. 1. Top (a) and side ~,iew (b) of the best fit geometry.The geometrical parametersare listed in Table 2. an error of 0.10/~ the small difference in the radii of the two sets of C-atoms is definetely not significant. This is also the case for a slight rotation of 2 ° about the molecular axis indicated by the tensor-LEED search, even though it would appear reasonable in view of the Pauli repulsion between the hydrogen atoms of neighbouring molecules. The final geometry is shown in Fig. 1. We mention that the site geometry in the 2x/3 structure is essentially the same; a more detailed description of this structure, together with further details, will be given in a later publication [ 13]. It is seen that the distortions are small so that it is easily understandable that photoemission and vibrational spectroscopy see an undistorted C6v symmetry [ 1,11 ]. Strictly speaking, the site symmetry is of course C3v, but this symmetry breaking can easily remain undetected in spectroscopies, in particular those mainly sensitive to molecular properties. The very small changes indicated by the last optimiza-

tion will not be discussed further, therefore, ,~L.~, ,,,~ shall address the main features of the structure found. Occupation of a threefold site has been found for the p(3 x 3)-(C6H6 + 2CO) structures on P d ( l l l ) and Rh(111), however, the fcc site is selected and the other ring orientation (C3v(O'a), [9] ) is realized. The local geometry in these latter cases is similar to that of most organometallic complexes involving benzene bound to small Ru clusters (e,g.: RU3(CO)9 (/£3 _~/2 : ~/2 : r/2 C6H6) [27]). It is a n interesting question why our system behaves differently. The common feature of all these structures is that 7r-bonding does not select the most symmetric situations which would be one of the ontop geometries. But even for a hollow site, the orientation found on Ru(001) has the lower symmetry in terms of the positions of the C atoms. In the C3v(o'd) orientation all six C atoms could lie on bridge sites while in the found Cav(O'v) orientation ontop and hollow sites are occupied. In the geometry found here, on the other hand, the connecting lines between the C atoms, i.e., the locations of largest 7r-electron density, are all in the equivalent situation, and this makes sense in view of the 7r-bonding mechanisms assumed to operate between benzene and the substrate. Within the overall symmetry C3v the buckling of the ring follows the corrugation of the substrate. It is suggestive that the adsorptive bond leads to a change of the electronic structure of benzene wh i ch favours this distortion, so that occupation of a site which would make all C-atoms equal and the ring planar would in fact cost energy. The reason for this effect can easily be seen when the benzene-surface bond is understood as a 7r-donor bond, as generally accepted. Although the decrease in the 7r-character of the molecule is a small effect, it will still favor a deviation from planarity towards the chair configuration of cyclohexane (the type of deviation which we have found). The sites occupied by the C atoms in the O-v orientation, which themselves offer the corresponding corrugation, can then lead to a strengthened bond as compared to the "planar" O'd configuration. Quantum chemical calculations would be helpful to test this suggestion. In conclusion, we have carried out a detailed investigation of the geometry of benzene on Ru(001). Great care has been taken to ensure that the observed ordered structures are free of contamination effects, and that no radiation damage is induced by the elec-

Iv L384

C Stellwag et aL/Surface Science 325 (1995) L379-L384

tron bombardment necessary for the LEED investigation. We find that the molecule is bound in a C3v(trv) orientation with its center on the hep-site, at an average distance'of 2.12/~ from the metal. The distortion of the molecule as compared to its free state is very small: the molecular radius is almost identical to that in the gas phase, but there is a small crown-like deviation from planarity which leads to good adjustment to the threefold symmetry of the site. We have argued that this selfconsistent site selection and slight distortion of the molecule might be connected with the decreased 7r-character of the chemisorbed molecule. Very small deviations from the gas phase C - C bond length and of the distances of the C atoms from the molecular center would not be outside the error bars, as is a possible small rotation of the ring in response to the neighbouring molecules. The main adsorbate induced modification of the substrate is a very large almost uniform contraction of the distance between the first and the second Ru layer of 0.05/~ which is one of the largest found on this surface so far.

Acknowledgement This work was supported by the Deutsche Forschungsgemeinschaft through SFB 338. References [1] P. Jakob and D. Menzel, Surf. Sci. 201 (1988) 503. [2] B.E. Koel, J.E. CroweU, C.M. Mate and G.A. Somorjai, J. Phys. Chem. 88 (1984) 1988. [3] C.M. Mate and G.A. Somorjai, Surf. Sci. 160 (1985) 542. [4] W. Huber, H.-P. Steinriick, T. Pache and D. Menzel, Surf. Sci. 217 (1989) 103.. [5] W. Huber, P. Zebisch, T. Bornemann and H.-E Steinriick, Surf. Sci. 258 (1991) 16,

[6] H. Ohtani, R.J. Wilson, S. Chiang and C.M. Mate, Phys. Rev. Lett. 60 (1988) 2398. [7] M.A. Van Hove, R.E Lin and G.A. Somorjai, J. Am. Chem. Soc. 108 (1986) 2532. [8] R.F. Lin, G.S. Blackmann, M.A. Van Hove and G.A. Somorjai, Acta CrystaUogr. Sect. B 43 (1987) 368. [9] A. Barbieri, M.A. Van Hove and G.A. Somorjai, Surf. Sci. 306 (1994) 261. [10] A. Wander, G. Held, R.Q. Hwang, M.L. Xu, P. de Andres, M.A. Van Hove and G.A. Somorjai, Surf. Sci. 249 (1991) 21. [ 11 ] EA. Heimann, E Jakob, T. Pache, H.-E Steinriick and D. Menzel, Surf. Sci. 210 (1989) 282; P. Jakob and D. Menzel, Surf. Sci. 235 (1990) 15. [ 12] H. Ranscher, P. Jakob and D. Menzel, Surf. Sci. 234 (1990) 108; P. Jakob and D. Menzel, Surf. Sci. 235 (1990) 197; Langmuir 7 (1991) 134. [ 13] C. Stellwag, G. Held and D. Menzel, in preparation. [14] G. Held and D. Menzel, Surf. Sci. 316 (1994) 92. [15] H. Pfniir and D. Menzel, Surf. Sci. 148 (1984) 411. [ 16] H. Schlichting and D. Menzel, Surf. Sci. 272 (1992) 27. H. Schlichting, Dissertation, Technische Universit~it Miinchen, 1990. [17] G. Michalk, W. Moritz, H. Pfniir and D. Menzel, Surf. Sci. 129 (1983) 92. [ 18] M.A. Van Hove and S.Y. Tong, Surface Cristallography by I.F.ED (Springer, Berlin, 1979 ). [19] J.B. Pendry, J. Phys. C 13 (1980) 937. [20] G. Held, H. Pfniir and D. Menzel, Surf. Sci. 271 (1992) 21. [21] W.H. Press, B.P. Flannery, S.A. Teukolsky, W.T. Vetteding, Numerical Recipes in C (Cambridge University Press, Cambridge, 1988). [22] EJ. Rous, M.A. Van Hove and G.A. Somorjai, Surf. Sci. 226 (1990) 15. [23] A. Wander, M.A. Van Hove and G.A. Somorjai, Phys. Rev. Lett. 67 (1991) 626. [24] H. Over, H. Bludau, M. Skottke-Klein, G. Ertl, W. Moritz and C.T. Campbell, Phys. Rev. B 45 (1992) 8638. [25] P.J. Feibelman, J.E. Houston, H.L. Davis and D.G. O'Neill, Surf. Sci. 302 (1994) 81. [26] D. Menzel, Surf. Sci. 318 (1994) 437. [27] B.F.J. Johnson, J. Lewis, M. Martinelli, A.H. Wright, D. Braga and E Grepioni, J. Chem. Soc. Chem. Commun. (1990) 364.