A Complete Structural Study of the Coverage Dependence of the

Department of Chemistry, UniVersity of Glasgow, Glasgow, G12 8QQ, U.K., and CLRC Daresbury Laboratory,. Warrington, WA4 4AD, U.K.. ReceiVed: June 16 ...
0 downloads 0 Views 114KB Size
140

J. Phys. Chem. B 2001, 105, 140-148

A Complete Structural Study of the Coverage Dependence of the Bonding of Thiophene on Cu(111) P. K. Milligan,† B. Murphy,‡ D. Lennon,† B. C. C. Cowie,‡ and M. Kadodwala*,† Department of Chemistry, UniVersity of Glasgow, Glasgow, G12 8QQ, U.K., and CLRC Daresbury Laboratory, Warrington, WA4 4AD, U.K. ReceiVed: June 16, 2000; In Final Form: October 11, 2000

The influence of coverage on the structure and bonding of thiophene within chemisorbed overlayers on Cu(111) has been investigated. Normal incidence X-ray standing wavefield absorption (NIXSW) and nearedge X-ray absorption fine structure spectroscopy (NEXAFS) were used to determine the adsorption site and orientation of thiophene. In the low coverage phase the thiophene molecule forms a π-bonded species, adopting an atop adsorption site with a Cu-S separation of 2.62 ( 0.03 Å with the ring inclined by 26 ( 5°. On increasing coverage, thiophene undergoes a compressional phase transformation to a more weakly bound sulfur lone-pair-bonded species. In this new phase the thiophene still occupies an atop site but the Cu-S separation has increased by 0.2 Å and the ring is now inclined at 44 ( 6° to the surface.

Introduction Studying the structure of adsorbed atoms and molecules has been an area of great interest and activity since the very beginnings of surface science. Structural investigations of model interfaces have been greatly facilitated over the years by the development of synchrotron-based structural probes. Early structural studies concentrated on relatively simple adsorbates, such as atomic and diatomic species; however, in contemporary work the orientation and adsorption site of more complex molecular species can be readily quantitatively determined with synchrotron-based techniques. It is surprising therefore that in the literature there are few examples where in the same study both the orientation and local registry of a molecule are probed using synchrotron techniques. A reason for the dearth of studies in which both the orientation and local registry are determined in a single study is that for most adsorbates the synchrotron techniques used for determining orientation require a completely different range of X-ray energies from those used to provide information on local registry. For example X-rays with energies 2000 eV. Unfortunately these two energy ranges cannot be covered with a single synchrotron beamline. However, in this study we were able to determine the local registry and orientation using a single station, because of the chosen combinations of adsorbate and structural techniques. The objective of this investigation is to determine the effects of increasing coverage on both the local registry and orientation of a heterocyclic molecule which is weakly chemisorbed to a metal surface. Specifically, we have investigated the coverage dependence of the adsorption structure of thiophene on Cu(111). This present work expands on an earlier study1 in which we * Author to whom correspondence should be addressed. Tel: +44 141 330 4380. Fax: +44 141 330 4888. E-mail: [email protected]. † University of Glasgow. ‡ CLRC Daresbury Laboratory.

have determined the adsorption structure of thiophene at a fixed coverage. In this present investigation we have determined the local registry of the thiophene molecule using normal incidence X-ray standing wavefield absorption (NIXSW). The orientation of the thiophene molecules is found using S K-edge NEXAFS. Uniquely in this study we have monitored changes in both the orientation and local registry of molecules within an overlayer which undergoes a coverage-driven structural transformation. The bonding of aromatic heterocyclic molecules to metal surfaces has long interested surface scientists. These model systems are intrinsically interesting from a fundamental science viewpoint because the adsorbate molecule can bond to a surface either via its π-system or the lone pair of electrons associated with the heteroatom. Consequently, the molecule can either bond to a metal surface either in a flat or an upright geometry, or somewhere in between. In recent years, studies of model aromatic heterocycle/metal adsorption systems have been motivated by the potential technological importance of these type of interfaces. For example, polythiophenes/pyrroles are the basis of conducting organic materials which have applications in LED fabrication and sensors. Since the electronic behavior of these materials is related to their structure and ordering, an understanding of the physical and chemical factors that influence these two parameters is necessary to optimize performance. Several previous studies2-7 have investigated the effects of coverage on the bonding of a chemisorbed overlayer of aromatic heterocycles. However, all of these studies have relied on techniques such as ARUPS,2,6 HREELS,2,4 Raman spectroscopy,3 SHG,5 and RAIRS,7 which provide only qualitative information on bonding geometry. There have been no previous studies where both orientation and bond distances have been monitored with increasing surface coverage. Demuth and co-workers2 performed one of the earliest studies in which the effect of coverage on the bonding of a chemisorbed heterocycle was investigated. In this work the coverage dependence of the bonding of chemisorbed pyridine to Ag(111) was probed with ARUPS and HREELS. Pyridine was found to bond to the Ag(111) via its π-system at low coverages, while at higher coverages a compressional phase transformation occurred to give

10.1021/jp002186u CCC: $20.00 © 2001 American Chemical Society Published on Web 12/09/2000

Coverage Dependence of Bonding of Thiophene on Cu(111) a weakly bound, nitrogen lone-pair-bonded species with a more inclined orientation. In a later study Sexton4 investigated the adsorption of furan, pyrrole, and thiophene on Cu(100), using HREELS and TPD measurements. Like Demuth and co-workers, Sexton found that thiophene underwent a compressional phase transition, from a π-bonded species, to a weakly bound, lone-pair-bonded species which was more inclined. However, this behavior contrasted with that found for furan and pyrrole, where bilayer formation was observed. In the proposed bilayer model an upright, lonepair-bonded species forms a second layer above a more strongly bound, flat π-bonded overlayer. In a more recent work by Bent and co-workers it has been suggested that benzene also adopted a similar bilayer structure on Cu(111).8 In a recent NEXAFS study of thiophene on Ag(111) a compressional phase transformation was also observed,13 this observation being confirmed in a STM study of the same system by Chen and co-workers.14 The principal objective of the current study was to discover whether thiophene underwent a compressional phase transition or formed a bilayer structure. Experimental Section Experiments reported in this paper were performed in two separate UHV systems. Initial characterization experiments were performed in a UHV system at Glasgow. This system was equipped with a retarding field analyzer (RFA) that was used both to collect electron-induced Auger (AES) spectra and to obtain low-energy electron diffraction (LEED) patterns. Residual gas analysis and temperature-programmed desorption (TPD) were performed with a triple filter (0-300) amu quadrupole mass spectrometer (QMS) (Hiden Ltd). The QMS was attached to a linear drive, which allowed it to be reproducibly positioned within 1 mm of the face of a crystal during TPD. To prevent the detection of gas molecules which may have desorbed from the sample holder and not the crystal surface, a shield with a 3 mm diameter entrance aperture was placed over the head of the QMS. A heating rate of 0.5 K/s was used to collect TPD spectra. Two methods of dosing thiophene were used in TPD experiments. In one method, referred to as “line-of-sight” dosing, the Cu(111) crystal was positioned 30 mm in front of a stainless steel dosing pipe (dia. 1.0 mm) while thiophene was allowed into the chamber. During line-of-sight dosing, thiophene was admitted into the chamber until a pressure rise of 3 × 10-9 mbar was observed. The actual pressure of thiophene in the region of the crystal surface was larger than the 3 × 10-9 mbar rise detected by the ion gauge. In the second dosing method, referred to as “back-filling”, the crystal was positioned 20 mm above the dosing pipe with the crystal face rotated so it did not face the dosing pipe, and thiophene was admitted into the chamber with a pressure rise of 2 × 10-8 mbar. From calibration desorption experiments we have established that for the same ion gauge measured “exposure” line-of-sight dosing deposits approximately 20 times more thiophene on to a surface, than does the back-filling method. Line-of-sight dosing was used when large coverages were required, because with this method the crystal could receive a large dose of thiophene without exposing the rest of the chamber to excessive quantities. The exposures quoted in the text of this paper refer to those obtained using the back-filling method. NIXSW and NEXAFS experiments were performed at the Central Laboratories of the Research Councils (CLRC) laboratory at Daresbury, at station 6.3 of the Synchrotron Radiation Source (SRS). This station has been described in detail

J. Phys. Chem. B, Vol. 105, No. 1, 2001 141 elsewhere.9 The associated UHV chamber is equipped with the usual sample heating and argon ion bombardment facilities in addition to LEED optics. A double-pass cylindrical mirror analyzer is used for electron AES, and a concentric hemispherical analyzer (CHA) is used for photon-stimulated AES and X-ray photoelectron spectroscopy (XPS). In both systems the Cu(111) sample crystal was cleaned by cycles of Ar+ bombardment followed by annealing to 773 K. After Ar+ sputtering and annealing the cleanliness and surface quality of the crystal was verified with electron-stimulated AES and LEED. A Brief Introduction to NIXSW. NIXSW has proven to be an easy and flexible tool for the determination of the structure of molecular adsorbates. In this section we will briefly outline the NIXSW technique (a more detailed description can be found in ref 14). In an NIXSW experiment X-ray radiation with the energy required to generate a Bragg reflection at normal incidence strikes a surface of a solid. A standing wavefield is generated within the solid by the interference of the incident and backscattered waves. The nodal planes of the standing wavefield lie parallel to the Bragg scattering planes, and their separation is equal to that of the scattering planes. When the photon energy is scanned through a range of reflectivity associated with the standing wavefield, its phase changes in a way that can be modeled. If the X-ray absorption of an atom is monitored as the photon energy is scanned, the resulting profile can be used to determine its location relative to the bulk scattering planes. The X-ray absorption profiles are fitted using two variable parameters, the coherent fraction (fco) and the coherent height (D). D is the height of an atom above a bulk reflecting plane, and takes a value between 0 and 1 lattice spacings. A measure of the uncertainty in the position of an atom with respect to the scattering planes is given by fco, which takes a value between 0 and 1, with 0 representing a completely ill-defined position, and 1 a position with no uncertainty. The local registry of an element can be unambiguously determined by triangulating its position with respect to two different sets of reflecting planes, (111) and (1h11) in the present case. Generally Auger and photoelectrons are used to monitor X-ray absorption profiles, because of their inherent surface sensitivity. In the present study the intensities of the S(1s) photoelectrons and the Cu(LVV) Auger electrons were monitored to obtain NIXSW profiles. From fitting the copper profiles the energy spread of the incident beam is determined, this is then subsequently used as a nonadjustable parameter in fits of the sulfur profiles. Results AES, LEED, and TPD Characterization. Prior to detailed structural studies of thiophene overlayers with NEXAFS and NIXSW at the Daresbury SRS, characterization experiments were performed with TPD, AES, and LEED in a separate UHV system at Glasgow. From these experiments it was found that thiophene reversibly adsorbs on Cu(111). The lack of any decomposition of thiophene was confirmed by AES spectra collected from surfaces which had undergone 4 cycles of adsorption and desorption of multilayers, which showed no detectable amounts of sulfur or carbon. No ordered overlayers of thiophene were detected with LEED. The characterization data that provided most information on the effects of coverage on the bonding of thiophene were obtained from TPD measurements. Spectra were collected from surfaces that had been exposed to sequentially higher doses of thiophene. In Figure 1 are nested TPD spectra that show

142 J. Phys. Chem. B, Vol. 105, No. 1, 2001

Figure 1. Nest TPD spectra, obtained by monitoring the parent ion signal (84 amu), collected from Cu(111) surfaces that had been exposed, using the “back-filling” (a) and “line-of-sight” (b) methods to sequentially larger amounts of thiophene. The exposure of thiophene for each TPD spectrum in graph (a) is as follows: 0.6 (solid line), 1.2 (squares), 2.4 (circles), 3.6 (up triangles), 4.8 (down triangles), 6.0 (diamonds), 7.2 (+), and 12.0 × 10-6 mbar s (×). While in graph (b) the exposure of thiophene, as measured by the ion gauge, for each TPD spectrum is as follows: 0.9 (solid line), 1.8 (squares), 3.6 (circles), 5.4 (up triangles), 7.2 (down triangles), 9.0 (diamonds), and 13.0 and 12.0 × 10-7 mbar s (+). In the case of “line-of-sight” dosing, the actual exposure at the crystal is approximately 20 times that measured at the ion gauge. Desorption peaks associated with defects, R-, β-, and multilayer states are labeled. Each spectrum was collected using a heating rate of 0.5 K/s.

desorption of intact thiophene molecules from surfaces at 110 K which had been exposed to sequentially higher doses of thiophene. Exposure of the Cu(111) surface to thiophene initially leads to the population of a desorption state at 297 K, which is rapidly saturated. We ascribe this state to desorption from defect sites. This assignment was confirmed in a separate experiment where a thiophene TPD spectrum was taken from a roughened surface which was produced by Ar+ sputtering a surface at room temperature. In the TPD spectrum from the roughened surface there is a significant increase in the amount of desorption at 297 K. With increasing exposure a second desorption feature, labeled R, develops and the defect desorption feature saturates. Initially for low coverages of thiophene (spectrum with exposure of 6.0 × 10-7 mbar s), the R-state has a desorption maximum at 234 K. However, with increasing coverage of thiophene the R-state broadens and shifts to lower temperatures. By the time the R-state has saturated after an exposure of 6 × 10-8 mbar s, its maximum desorption temperature has shifted to 211 K. Once the R-state has been saturated two other features develop concurrently at 157 and 173 K, the development of these with increasing coverage is shown most clearly in Figure 1b. The desorption peak at 173 K, labeled β, grows rapidly and is saturated after an exposure of 14.4 × 10-6 mbar s. The feature at 157 K develops at a slower rate than the β-state, and in contrast to the other features does not saturate with increasing thiophene exposure. The behavior of the 157 K state is consistent

Milligan et al. with its origin occurring from desorption of condensed multilayers of thiophene. The relative amounts of thiophene desorbing from a surface that gave rise to R- and β-desorption-states compared to that from a surface which only gave rise to the R-state was determined experimentally. This was achieved by comparing the areas of two TPD spectra which were collected from surfaces, which had been exposed to large doses of thiophene and then subsequently annealed to 157 and 173 K, respectively. For the surface annealed to 157 K both R- and β-desorptionstates were seen, while only a R-state was observed from the surface annealed to 173 K. A R + β/R ) 1.7 ( 0.2 was obtained from these experiments. A quantitative measure of the amounts of adsorbed thiophene on the surfaces annealed to 157 and 173 K was found by calibrating their sulfur Auger intensities against a surface of known sulfur concentration. A (x7×x7)R 19° sulfur overlayer with a coverage of 0.43 ML19 was used for coverage calibration. This surface was prepared by exposing the Cu(111) to 4.0 × 10-4 mbar s of H2S at room temperature. We should state that 1 ML is defined to be one adsorbate atom per substrate atom. Using this calibration method, it was found that surfaces which had been annealed to 157 and 173 K have sulfur coverages of 0.08 ( 0.03 and 0.14 ( 0.03 ML, respectively. These values agree well with the relative coverages determined with TPD. Structural Measurements. The objective of synchrotronbased NIXSW and NEXAFS experiments was to determine the coverage dependence of the adsorption structure of thiophene. In both NIXSW and NEXAFS experiments two methods were used to prepare thiophene overlayers with differing coverages. For both preparation methods the amount of sulfur within an overlayer was calibrated against a (x7×x7)R 19° surface with XPS. One method involved the annealing of thiophene multilayers to obtain overlayers with differing coverages. In particular, multilayers were annealed to temperatures which were known to be sufficient to desorb (i) condensed multilayers only, and (ii) condensed multilayers and the β-state from TPD characterization. The second method used to prepare overlayers involved the dosing of aliquot amounts of thiophene at 100 K to produce surfaces with sequentially higher coverages. Between each aliquot dose either a NIXSW or a NEXAFS measurement was made. The sequential dosing method had two advantages over annealing; first overlayers with wide ranges of coverages could be produced in a controlled manner. Second, significant changes in NIXSW and NEXAFS data caused by the dosing of a single aliquot were readily apparent. (a) NIXSW. The coverage dependence of the adsorption site of thiophene was studied using NIXSW. A copper profile collected using the (111) planes which is representative of those collected during the study is displayed in Figure 2. As expected for a well-ordered Cu(111) surface, all copper NIXSW profiles collected during this work were fitted with a D value of 0.00 Å and with fco values that fell in the range 0.85-0.90. A typical set of nonadjustable parameters used to fit a Cu NIXSW profile are an energy spread (∆E) of 0.8 eV and a mosaic spread of 0.1°. Sulfur NIXSW data will be initially considered from overlayers prepared using the annealing method. In these experiments, annealing produced overlayers that spanned two distinct coverage regimes. In one series of experiments overlayers with coverages in the range 0.08-0.11 ML were investigated, while a second coverage range spanned 0.14-0.17 ML. Since the uncertainty associated with these coverage measurements is

Coverage Dependence of Bonding of Thiophene on Cu(111)

Figure 2. A copper NIXSW profile (dotted line) collected using the (111) and by monitoring the intensity Cu(LVV) Auger. The fit to the experimental profile is also shown (solid line), the parameters used to generate it are D ) 0.00 Å, fco ) 0.90, an energy spread (∆E) ) 0.90 eV, and a mosaic width ) 0.1°.

(0.02 ML all the overlayers within each range can be assumed to have the same coverage within experimental error. The coverages obtained in the first range would only produce the R-state in TPD experiments, while in the second coverage range overlayers would give rise to both R- and β-desorption-states. NIXSW profiles collected from the overlayers with coverages in the R-state coverage range were identical within experimental error. The profiles obtained from the β-state overlayers were also identical to each other within experimental error, however they were significantly different to those collected from the R-state surfaces. In Figures 3 and 4 representative sulfur NIXSW profiles obtained using (111) and (1h11) planes from R- and β-state surfaces are shown. The average values of D and fco using the (111) and (1h11) reflecting planes for both R- and β-states are listed in Table 1. In Table 2 fco and D values determined from (111) NIXSW profiles that were collected from a series of sequentially dosed surfaces are listed. The first four measurements were from overlayers with coverages (0.02, 0.04, 0.07, and 0.10 ML) that are within the R-state range. For all four coverages, high fco values (0.90) were determined from NIXSW data, which is indicative of a single well-defined adsorption site. The D values obtained for 0.04 (0.45 ( 0.05 Å), 0.07 (0.47 ( 0.05 Å), and 0.10 ML (0.50 ( 0.05 Å) overlayers are within or just outside experimental error of those obtained from annealed surfaces (0.54 ( 0.03 Å). This would indicate that the adsorption site of thiophene in these surfaces is identical to that found for annealed R-state overlayers. It is interesting to note that the D values determined for the 0.02 ML surface of 0.40 ( 0.05 Å is smaller than the value obtained for the annealed R-state overlayers by 0.14 ( 0.08 Å. Although this is a small difference the 0.02 ML does have a shorter D value than those obtained for the annealed R-state. We believe that this difference can be assigned to a shorter Cu-S distance for thiophene observed at defect sites. A shorter bond distance is consistent with the more strongly bound nature of the defect species.

J. Phys. Chem. B, Vol. 105, No. 1, 2001 143

Figure 3. Sulfur (111) and (1h11) NIXSW profiles (dotted line) collected from an R-state thiophene overlayer prepared by annealing are displayed here. Also shown are the corresponding fits (solid lines) to the experimental-derived profiles. The (111) fit was obtained using a D ) 0.55 Å and a fco ) 0.95, while the (1h11) profile fitted to D ) 0.80 Å and a fco ) 0.85.

Figure 4. Sulfur (111) and (1h11) NIXSW profiles (dotted line) collected from a β-state thiophene overlayer prepared by annealing are displayed here. Also shown are the corresponding fits (solid lines) to the experimenta-derived profiles. The (111) fit was obtained using a D ) 0.75 Å and a fco ) 0.50, while the (1h11) profile fitted to D ) 0.85 Å and a fco ) 0.60.

The 5th and 6th doses produce coverages of 0.13 and 0.15 ML, respectively, which are both in the β-state regime. The increase in coverage from 0.10 to 0.13 ML results in a rapid decrease in fco from 0.90 to 0.60, in contrast there is no change in D. With a further increase in coverage to 0.15 ML the fco

144 J. Phys. Chem. B, Vol. 105, No. 1, 2001

Milligan et al.

TABLE 1: The Average Values of D and fco Determined from (111) and (1h11) Profiles for r- and β-State Surface Prepared by Annealing D (111)

fco (111)

D( 1h11)

fco (1h11)

R-state (annealed) 0.54 ( 0.03 0.89 ( 0.05 0.78 ( 0.05 0.78 ( 0.03 β-state (annealed) 0.75 ( 0.05 0.50 ( 0.10 0.80 ( 0.05 0.60 ( 0.10

TABLE 2: Listed Here Are the (111) D and fco Determined for Overlayers That Were Prepared by Sequential Dosing coverage/mL

fco

D/Å

R-state

0.02 0.04 0.07 0.10

0.90 0.90 0.90 0.90

0.40 0.45 0.47 0.50

β-state

0.13 0.15 0.22 0.25

0.60 0.45 0.25 0.15

0.50 0.50 0.35 0.30

decreases further to 0.45, and once again the D value remains unchanged. The last two experiments correspond to multilayer surfaces, as would be expected these give rise to profiles that are fitted to low fco. (b) NEXAFS. The dependence of the orientation of the aromatic ring on thiophene coverage was monitored using sulfur K-edge NEXAFS. Spectra were obtained by monitoring the yield of sulfur KLL Auger electrons, as the photon energy of the light was scanned through the K-edge. The experimental geometries used to collect NEXAFS spectra were identical to those used in NIXSW experiments. Grazing (19.5°) NEXAFS spectra were collected in the (1h11) NIXSW geometry, while normal spectra were collected using the (111) geometry. There have been several previous experimental10,11 and theoretical10 NEXAFS studies involving thiophene, which have greatly facilitated the quantitative analysis of spectra in the current study. Stohr and co-workers performed the most comprehensive study of S K-edge NEXAFS, which involved theoretical calculations and gas- and condensed-phase K-edge measurements. While in a recent study, the adsorption geometry of thiophene on Pd(100) was investigated using S K-edge NEXAFS by Ohta and co-workers. In the study by Stohr and co-workers NEXAFS spectra were interpreted with the aid of multiple scattering calculations, with experimental data being compared to theoretical models. From these calculations it was found that NEXAFS spectra consisted of five resonances. The first two resonances, which are most intense, originate from transitions to π* and σ* states, the subsequent three resonances were assigned to Rydberg-like transitions to 4s, 4p, and 5p states, respectively. Ohta and co-workers analyzed their data using a more empirical approach than Stohr. They performed curvefitting analysis of NEXAFS spectra, which is based on the superposition of several resonances and a step function. NEXAFS spectra were fitted to four resonances and a step. The two most intense resonances were once again assigned to transitions to π* and σ* states, the relative positions of these features being identical to those calculated by Stohr. Ohta and co-workers used a single broad peak to fit the region were Stohr believed transitions to 4s, 4p, and 5p states were located. Ohta and coworkers included a pre-π* feature in their curve fitting, a feature not included in the modeling of Stohr and co-workers. Pre-π* features are found at lower photon energies than a main π* resonance, and have been observed in NEXAFS studies of adsorbed aromatic molecules, their origin is attributed to an electronic transition from the substrate to the adsorbate π* state.12 In the present study curve-fitting analysis was performed on NEXAFS spectra. All resonances were fitted with symmetric

Figure 5. A comparison is made in this plot between the positions of resonances and ionization threshold used to fit thiophene NEXAFS spectra in this work, and those used in two previous studies by Stohr and co-workers10 and Ohta and co-workers.11 Peaks a, b, and c were assigned to transitions to 4s, 4p, and 4d by Stohr and co-workers.

TABLE 3: A Comparison Is Made between the Experimentally Observed (1h11) D and fco Values Obtained for the r-State and Those that Would Be Expected for Atop, Bridging, and Three-Fold Sites (FCC and HCP) Given the Observed (111) Values D/Å fco

atop

bridging

FCC

HCP

expt

0.87 0.89

0.87 0.30

1.56 0.89

2.26 (0.18) 0.89

0.78 0.78

TABLE 4: A Comparison Is Made between the Experimentally Observed (1h11) D and fco Valves Obtained for the β-state and Those That Would Be Expected for Atop, Bridging, and Three-Fold Sites Given the Observed (111) Values D fco

atop

bridging

FCC

HCP

expt

0.94 0.50

0.94 0.17

1.63 0.50

2.33 (0.25) 0.50

0.80 0.60

TABLE 5: The Areas and Positions of the 6 Gaussian Peaks Used to Fit Representative r- and β-state Surfaces Shown in Figures 6 and 7 resonance position/[eV] grazing R normal R grazing β normal β Pre-π π σ a b c

2466.5 2467.8 2468.7 2470.7 2471.7 2472.8

0.69 3.47 1.76 1.69 0.02 0.03

0.19 0.44 2.52 1.21 0.21 0.03

0.35 1.44 1.71 1.70 0.04 0.02

0.29 0.76 1.73 0.81 0.03 0.01

Gaussians, and a Gaussian-broadened step was used to model the absorption threshold or edge jump. The degree of broadening of the step was determined from thiophene multilayer NEXAFS spectra, where the step edge was more discernible because of the relatively weak intensities of the resonances. Prior to curve fitting all NEXAFS spectra were normalized to the edge jump. In the curve-fitting procedure 6 peaks were used, which have been labeled pre-π*, π*, σ*, a, b and c. The areas and positions of the 6 peaks used to fit R- and β-state surfaces are given in Table 5. A comparison between the positions of peaks and step used to fit our data, with those used by Ohta et al. and Stohr et al. is shown in Figure 5. There is good agreement between the positions found in our study and those determined in the previous two studies. The positions of π* and σ* are within experimental error identical to those found by Stohr and Ohta, while the pre-π* feature is in the same position to that determined by Ohta. Like Stohr we have used three peaks, a,

Coverage Dependence of Bonding of Thiophene on Cu(111)

Figure 6. Grazing and normal incidence spectra (dotted lines) collected from an R-state surface prepared by annealing are plotted in this figure. Also shown are fits (solid lines) which were derived from the addition of 6 Gaussians and an ionization step (dashed lines).

b, and c, to fit a broad feature at 2470.7 eV. It should be noted that peak a is significantly more intense than both peaks b and c, however the presence of these two small components did improve the quality of fits. There are some differences between the positions of these peaks found in our study and those of Stohr. We believe that this can be ascribed to the difference in the position of the steps in the present study and Stohr’s work. After the curve-fitting procedure the orientation of the thiophene ring was determined from the polarization dependence of the intensity of the π* resonance.12 Although the polarization dependence of the σ* could have been used to determine orientation this approach was rejected because the intensity of this resonance is sensitive to the absolute position of the step edge and the resonances a, b and c. Clearly, the intensity of the σ* resonance could have a systematic error due to the positioning of the step edge. A similar experimental approach to that used for NIXSW was employed in NEXAFS experiments, with surfaces being produced by annealing and sequential dosing methods. We will first consider the annealed surface data. Annealing formed overlayers in two distinct coverage ranges; these were 0.08 ( 0.03 to 0.11 ( 0.03 ML and 0.14 ( 0.03 to 0.18 ( 0.03 ML. The first of these coverage regimes is in the R-state region, while the second covers the β-state. Representative NEXAFS spectra of R- and β-surfaces prepared by annealing are displayed in Figures 6 and 7. The average angles of inclination to the surface determined for the R- and β-states to be 32 ( 6° and 45 ( 6° respectively. In sequential dosing experiments, overlayers with coverages from 0.03 to 0.23 ML were investigated. Examples of NEXAFS spectra collected from these overlayers are shown in Figure 8. The angles of inclination of the thiophene ring to the surface determined for sequentially dosed overlayers are plotted in Figure 9 and shown in Table 6. Of the surfaces prepared three (0.03, 0.05, and 0.10 ML) are in the R-state coverage range and two (0.12 and 0.14 ML) are in the β-state coverage region.

J. Phys. Chem. B, Vol. 105, No. 1, 2001 145

Figure 7. Grazing and normal incidence spectra (dotted lines) collected from a β-state surface prepared by annealing are plotted in this figure. Also shown are fits (solid lines) which were derived from the addition of 6 Gaussians and an ionization step (dashed lines).

Figure 8. Plotted here are grazing and normal incidence NEXAFS spectra that were collected from a Cu(111) surface that was sequentially dosed with thiophene. The coverage of thiophene for each pair of NEXAFS spectra is as follows: 0.03 (solid line), 0.05 (squares), 0.10 (circles), 0.12 (up triangles), 0.14 (down triangles), 0.18 (diamonds), and 0.24 (+) ML. The positions of the π* and σ* resonances are indicated.

The orientations of thiophene within the 0.03, 0.05, and 0.10 mL overlayers are 12 ( 2°, 20 ( 3°, and 25 ( 4°. Clearly the orientation angle of thiophene within the 0.03 ML overlayer is

146 J. Phys. Chem. B, Vol. 105, No. 1, 2001

Milligan et al.

Figure 9. The angles calculated from the NEXAFS data collected from sequentially dosed overlayers have been plotted against thiophene coverage. Also shown are the coverage ranges over which defect, R-, β-, and multilayer desorption states were observed in TPD experiments.

TABLE 6: The Angle Plotted in Figure 10 Are Listed coverage/ML

angle

R-state

0.03 0.05 0.1

12 ( 2° 20 ( 3° 25 ( 4°

β-state

0.12 0.14

41 ( 6° 45 ( 6°

mutlilayers

0.18 0.23

55 ( 8° 54 ( 8°

significantly smaller than the other two surfaces. The angles determined for the two β-state coverages of 0.12 and 0.14 ML are 41 ( 6° and 46 ( 6°, respectively. This clearly demonstrates that molecules in the β-state are more inclined with respect to the surface than those in the R-state. Discussion Three questions must be addressed in conjunction with the results of this study: (a) what is the adsorption structure of thiophene within R-state overlayers, (b) what is the origin of the β-state, and (c) what is the adsorption structure of thiophene adopted within β-state overlayers. The first of these questions has been partially addressed in our earlier preliminary study,1 and the results obtained in this study agree well with that work. From NIXSW data collected from annealed surfaces the adsorption site of thiophene within R-state overlayers can be readily determined. For the R-state surfaces using the (111) data gave D ) 0.54 ( 0.03 Å and a fco ) 0.89 ( 0.05, while from (1h11) data D ) 0.75 ( 0.05 and fco ) 0.78 ( 0.03 were obtained. These results are very similar to those we obtained in an earlier NIXSW study of thiophene on Cu(111) at a coverage corresponding to the R-state.1 The (111) D value of 0.54 Å is much too small to be the distance of the sulfur atom from a (111) plane that would pass through the unreconstructed Cu(111) surface, because it would give an unfeasibly short Cu-S separation. Consequently, the D (111) of 0.54 Å must be the distance of the sulfur atom from the next hypothetical

(111) scattering planes. If this was the case the sulfur atom would be 2.62 ( 0.03 Å (0.54 Å + a (111) lattice spacing (2.08 Å)) above an unreconstructed Cu(111) surface. The high (111) fco of 0.89 is indicative of the sulfur atom residing in a single well-defined site on the surface. The adsorption site of the sulfur can be unambiguously determined using the (1h11) data. Given that the height of the sulfur above a unreconstructed Cu(111) surface is 2.62 Å, the expected D and fco (1h11) values for atop, bridging, and 3-fold sites can be calculated and compared with experimental values. This comparison is made in Table 3, the experimental (1h11) data (D ) 0.75 Å and fco ) 0.78) are in closest agreement with an atop adsorption site for the sulfur (D ) 0.87 Å and fco ) 0.89). The experimental (1h11) fco is slightly lower than is predicted from the (111) data, this suggest that there is a higher degree of uncertainty in the position of the sulfur with respect to the (1h11) planes, than there is with respect to the (111) planes. There are two possible reasons for this, either the sulfur is slightly displaced from a perfect atop site or the uncertainty in its position is due to vibrational motion parallel to the surface. From the current data we cannot distinguish between these two possible sources, indeed it is likely that both play a role in introducing uncertainty in the position of the sulfur. Apart from the surface with lowest coverage, which has a significant contribution from defects, NIXSW data collected from sequentially dosed surfaces are identical within experimental error to those obtained from surfaces prepared by annealing. The NEXAFS data from annealed surfaces are also consistent with the sequential dosing results. The average inclination angle for the 0.08-0.11 ML range (R-state) annealed data was 32 ( 6°, which is slightly larger than the average tilt angle determined from the sequentially dosed overlayers. However, it should be noted that the coverages of the annealed overlayers are at the upper range of the R-state regime. Consequently, some of the overlayers may have small regions of β-state thiophene, which would have the effect of increasing the observed average tilt angle. Once again only with the lowest coverage sequentially dosed surface is the data significantly different from that obtained from the annealed surfaces. For the lowest coverage overlayer (0.03 ML) the average orientation determined by NEXAFS was 12 ( 2°, which is less than the average of 26° obtained from the other sequentially dosed and the annealed R-state surfaces. Once again we believe that the significant difference between the 0.03 ML surface and the other R-state overlayers is due to the fact that adsorption on this surface will be predominantly at defect sites. From TPD we can estimate that the defect density on Cu(111) is in the region of 0.01 to 0.02 ML, which means that at a coverage of 0.03 ML, 3366% of thiophene molecules are adsorbed at defect-related sites. Since the 12 ( 2° found for the 0.03 ML overlayer is the average angle of orientation of thiophene adsorbed at defects and on terraces, we determine that molecules at defect sites make an angle of between 0° and 7° with respect to the surface, assuming a 26° orientation for terraces. Therefore we have a complete description of the adsorption structure of R-state thiophene molecules. Within these overlayers thiophene adsorbs on terraces at atop sites with a Cu-S separation of 2.62 Å and with an average orientation of 26°. However thiophene molecules adsorbed at defect sites have a shorter Cu-S separation and are less inclined to the surface, with an estimated orientation of between 0 and 7°. Finally we must consider the two closely related questions of the origin of the β-state and the adsorption structure of thiophene molecules within it. Previous studies on Cu(100)3 and

Coverage Dependence of Bonding of Thiophene on Cu(111) Ag(111)13 have shown that the thiophene overlayers undergo a coverage-driven phase transformation. In these studies HREELS3 and NEXAFS13 complemented with TPD provided evidence for the phase transformation. Although the results of previous work on thiophene would suggest that a compression-driven phase transformation does occur on Cu(111), there is also an alternative model for the adsorption behavior. Sexton suggested that rather than undergoing compression, furan and pyrrole formed a bilayer structure. In this model upright molecules adsorb above more strongly bound flat-lying molecules. Two pieces of evidence were given by Sexton for bilayer formation. First, the observation of β-like peaks in TPD spectra of furan and pyrrole, in contrast to thiophene spectra where a single broad feature was observed. Second, the crystal structure of furan has alternate layers of furan adopting different orientations, a geometry referred to as “T” shaped, which closely mirrors a bilayer structure. Another example where the formation of bilayers has been postulated is in the case of benzene on Cu(111).8 Once again evidence for bilayer formation was the observation of a β-like desorption state in TPD, and the similarity between the “T” like geometry of bulk benzene and a proposed bilayer structure. A further argument used in favor of bilayer formation was that the β-like state was observed at a coverage that was larger than that required to saturate a surface with flat-lying benzene molecules. Although advocates of bilayer structures have used the presence of a well-defined β-like desorption state as evidence for their model, those who side with the compression model also cite it as supporting evidence. For instance TPD data obtained for thiophene from Ag(111) by Vaterlein et al.13 are almost identical to that found in this study, with the observation of a β-like state being assigned to desorption from a more inclined thiophene species formed by the compression of a flatlying overlayer. Clearly, the interpretation of TPD is open to debate, with qualitatively similar data being used to support completely opposing models. Since the observation of a β-state is not on its own sufficient evidence for an unambiguous assignment of a structural model, other pieces of evidence are required before assignment of either a compression phase transition or bilayer formation can be made. The first factor that should be considered is the coverage range over which the β-state is observed. This is important because in the study of benzene on Cu(111) β-like desorption peaks were only observed at coverages higher than those required to saturate a surface with flat-lying molecules, this was used for evidence of bilayer formation. For thiophene on Cu(111) we can estimate from van der Waal radii that a coverage of 0.24 mL would be required to saturate a surface with flat-lying molecules. By comparing this calculated value (0.24 mL) with the coverage at which the R-state, which is known to consist of flat-lying molecules, saturates (0.08 mL) it is clear that the β-state develops well before the surface is completely saturated with flat-lying molecules. This observation is clearly not consistent with the arguments used by Xi et al. to justify the formation of a benzene bilayer on Cu(111). A final piece of evidence for a phase transition is the coverage dependence of the orientation of the thiophene ring. The NEXAFS data clearly show that there is a significant increase in the orientation of the ring from 25 ( 4° to 41 ( 6° after the dosing of the forth aliquot, which increases the coverage from 0.10 to 0.12 mL. This large increase in inclination for a small coverage change is strong evidence for a phase transition. So we believe that by a combination of arguments based on the absolute coverages of the phases and the coverage dependence

J. Phys. Chem. B, Vol. 105, No. 1, 2001 147 of the orientation of the adsorbate we can assign the adsorption behavior of thiophene to a compression driven phase transition. Since we have established that thiophene undergoes a coverage-driven phase transition, in agreement with previous work on related systems, the NIXSW data collected from β-state overlayers can now be fully interpreted. To determine the adsorption site of the sulfur atom within β-state overlayers we must consider the annealed data. The D (111) of 0.75 ( 0.05 Å is clearly too small to be a realistic distance, as in the case of the R-state data we must add a lattice spacing (2.08 Å) to give a height of 2.83 ( 0.05 Å above the (111) plane that coincides with the unreconstructed surface. This height above the surface (111) plane is clearly longer than that determined for the R-state. This longer distance is entirely consistent with the TPD data, which clearly shows that the β-state is more weakly bound to the surface than the R-state. The local registry of sulfur within β-state overlayers can be determined using a similar procedure to that used for the R-state. In Table 4, (1h11) values for D and fco that have been calculated from (111) data for atop, 3-fold, and bridging sites are compared to the experimentally determined values. The best agreement with the experimental data is for the occupancy of an atop site. The (111) and (1h11) fco values for the β-state are lower than those found for the R-state. This indicates that there is greater uncertainty in the local registry of thiophene within β-state overlayers. The lower fco values for the β-state are consistent with previous NIXSW studies of other weakly bound molecular adsorbates, which display much lower fco values than are observed for more strongly bound chemisorbed species.15 The lower fco seen for weakly adsorbed species can be rationalized in terms of the weaker molecule-surface bond, giving rise to larger amplitudes of vibration both perpendicular and parallel to the surface, which produce a reduction in both (111) and (1h11) fco values. Apart from having a larger dynamic contribution to the uncertainty in local registry, more weakly bound species are more likely to be displaced from well-defined sites through adsorbateadsorbate interactions. The (111) D value obtained from sequentially dosed β-state surfaces (0.50 ( -0.05 Å) is significantly smaller than the corresponding annealed values (0.75 ( 0.05 Å). This can be readily rationalized by considering the intrinsic differences between the two methods of overlayer preparation. In contrast to the annealing method, sequentially dosed surfaces were prepared at 100 K. Consequently, the sites populated will be kinetically limited, hence some sites occupied may not be the most thermodynamically favorable. When a surface dosed at 100 K is annealed ordering of the overlayer could then take place, with thiophene molecules adopting more thermodynamically favorable adsorption sites. The (111) D value of 0.50 Å observed for sequentially dosed surfaces would be consistent with thiophene populating adsorption sites other than atop. For the annealed thiophene overlayers with atop adsorption and no relaxation of the substrate, the Cu-S separation is 2.83 Å. If the Cu-S separation remains the same in sequentially dosed overlayers then D (111) values of 0.34 and 0.45 Å would be expected for bridging and 3-fold sites. So the observed D (111) value for sequentially dosed overlayers could be rationalized by either occupation of bridging sites or a mixture of atop and 3-fold sites. Indeed some combination of occupancy of atop, bridging, and 3-fold sites would also give the observed D (111) value. By combining the results of the present study and those of previous work some general trends in the bonding of thiophene to metal surfaces can be developed. As in previous studies of

148 J. Phys. Chem. B, Vol. 105, No. 1, 2001

Milligan et al. adopts an atop adsorption site in both R- and β-states, and that as would be expected there is an elongation of Cu-S bond distance for the more weakly bound β-state. Clearly in both phases the most dominant molecule-surface interaction is that involving the S lone pairs rather than the π-system. Summary

Figure 10. The structures of thiophene within R- and β-states are shown schematically.

thiophene on Cu(100) and Ag(111) we have found that a coverage-driven phase transition results in an increase in the orientation of the thiophene ring and a weakening of the thiophene surface bond, the structure of the two phases is illustrated in Figure 10. It has been previously suggested that the driving force for the coverage-driven reorientations of heterocycles is that the adsorbate-adsorbate interactions, such as π-π stacking or π-lone pair, are increased.2,4 This enhancement in adsorbate-adsorbate interactions outweighs the weakening of the overall molecule-surface bond. The ability of thiophene to undergo this compression behavior on Cu(111), Cu(100), and Ag(111) is correlated to the weak bonding of the molecule to the three substrates. On all three surfaces thiophene is not strongly bound, which is demonstrated by the observed reversible chemisorption. The formation of a weakly bound π-bonded species is clearly necessary for the phase transition to occur. If thiophene did form a strongly bound flat-lying species, then compression would not occur because adsorbateadsorbate interactions would not compensate for the loss of a strong molecule-surface interaction. These points are illustrated in the behavior of thiophene on the more reactive Pt(111),16 Ni(111)17, and Rh(111)18 surfaces. On all three surfaces no compression transition is observed, instead thiophene interacts very strongly with the substrates and undergoes desulfurization. Prior to desulfurization a π-bonded thiophene species has been identified on all three surfaces. On Rh(111) and Ni(111) the π-bonded thiophene species undergoes C-S bond scission by 150 K. In the case of Pt(111) thiophene forms a flat-lying species at cryogenic temperatures (170 K), however on heating the molecule begins to tilt. This tilting is not due to a phase transition, but is however associated with an increase in the Pt-S interaction. Upon further heating to 370 K the thiophene undergoes complete desulfurization. This behavior completely contrasts with that observed on Cu(111) where the tilted β-state molecules interact less strongly with the substrate. The relative weakness of the interaction between the π-system of thiophene and the Cu(111) substrate is illustrated by comparison with a previous photoelectron diffraction study of benzene on Ni(111).20 In this system the π-ring of benzene interacts very strongly with the substrate, and is found at 1.91 Å above the surface plane. In the present case the π-ring of thiophene, in both the R- and β-states, is significantly higher above the Cu(111) surface. This clearly demonstrates the weakness of the π-surface interaction for thiophene on Cu(111). In summary, this study has provided a more detailed description of the structure of the thiophene within two stable chemisorbed phases than that covered by previous work. Using NIXSW we have shown that, somewhat surprisingly, thiophene

We have performed the first complete structural study of the adsorption of thiophene on Cu(111). Uniquely the coverage dependence of both the local registry and orientation of an adsorbed molecule have been monitored. The main conclusions of the study are listed below. (i) Thiophene undergoes a coverage-driven phase transition. (ii) At low coverage the thiophene populates an atop site and has an average orientation of 26 ( 5°. A compression of the thiophene overlayer occurs that involves a lengthening of the Cu-S separation by approximately 0.2 Å, with a corresponding weakening of the thiophene surface bond, and an increase in the orientation of the ring from 26° to 44° with respect to the surface. (iii) Measurements made at low thiophene coverage (0.02 ML) indicate that molecules adsorbed at defect sites have shorter Cu-S separation and are less inclined to the surface with an estimated orientation between 0° and 7°. Acknowledgment. M.K. and D.L. gratefully acknowledge the support of the Royal Society for the award of an equipment grant. The EPSRC are thanked for the provision of a research studentship to PM. D.L. thanks Imperial Chemical Industries for the award of an ICI Lectureship in Heterogeneous Catalysis. References and Notes (1) Milligan, P.; McNamarra, J.; Murphy, B.; Cowie, B. C. C.; Lennon, D.; Kadodwala, M. Surf. Sci. 1998, 413, 166. (2) Demuth, J. E.; Christmann, K.; Sanda, P. N. Chem. Phys. Lett. 1980, 76, 201. (3) Eesley, G. L. Phys. Lett. A 1981, 81, 193. (4) Sexton, B. A. Surf. Sci. 1985, 163, 99. (5) Heskett, D.; Song, K. J.; Burns, A.; Plummer, E. W.; Dai, H. L. J. Chem. Phys. 1986, 85, 7490. (6) Cohen, M. R.; Merrill, R. P. Surf. Sci. 1991, 245, 1. (7) Haq, S.; King, D. A. J. Phys. Chem. 1996, 100, 16957. (8) Xi, M.; Yang, M. X.; Jo, S. K.; Bent, B. E.; Stevens, P. J. Chem. Phys. 1994, 101, 9122. (9) McDowell, A. A.; Norman, D.; West, J. B.; Campuzano, J. C.; Jones, R. G. Nucl. Instrum. Methods A 1986, 246, 131. (10) Hitchcock, A. P.; Horsley, J. A.; Stohr, J. J. Chem. Phys. 1986, 85, 4835. (11) Terada, S.; Yokoyama, T.; Sakano, M.; Imanishi, A.; Kitajima, Y.; Kiguchi, M.; Okamoto, Y.; Ohta, T. Surf. Sci. 1998, 414, 107. (12) Stohr, J. NEXAFS Spectroscopy; Springer: Berlin, 1992. (13) Vaterlein, P.; Schmelzer, M.; Taborski, J.; Krause, T.; Viczian, F.; Basser, M.; Fink, R.; Umbach, E.; Wurth, W. Surf. Sci. 2000, 452, 20. (14) Kadodwala, M.; Davis, A.; Scragg, G.; Cowie, B. C. C.; Jones, R. G.; Woodruff, D. P. Surf. Sci. 1997, 392, 199. (15) Chen, X.; Franks, E. R.; Hamers, R. J. J. Vac. Sci. Technol. B 1996, 14, 1136. (16) Land, J. F.; Masel, R. I. Surf. Sci. 1987, 183, 44. (17) Huntley, D. R.; Mullins, D. R.; Wingeier, M. P. J. Phys. Chem. 1996, 100, 19620. (18) Netzer, F. P.; Bertel, E.; Goldmann, A. Surf. Sci. 1988, 201, 257. (19) Domange, J. L.; Oudar, J. Surf. Sci. 1968, 11, 124. (20) Schaff, O.; Fernandez, V.; Hofmann, P.; Schindler, K. M.; Theobald, A.; Fritzsche, V.; Bradshaw, A. M.; Davis, R.; Woodruff, D. P. Surf. Sci. 1996, 348, 89.