J. Phys. Chem. C 2010, 114, 15645–15652
15645
Nanoscale Patterning by C60 Ordering on Pt(110) Xavier Torrelles,*,† Veronique Langlais,‡,§ Maurizio De Santis,§ Helio C. N. Tolentino,§ and Yves Gauthier§,| Institut de Ciencia de Materials de Barcelona, ICMAB-CSIC, 08193 Bellaterra, Barcelona, Spain, Physics Department, Autonomous UniVersity of Barcelona, 08193 Bellaterra, Barcelona, Spain, and Institut Ne´el, CNRS & UniVersite´ Joseph Fourier, BP 166, F-38042 Grenoble, France ReceiVed: January 27, 2010; ReVised Manuscript ReceiVed: July 1, 2010
Grazing incidence X-ray diffraction was used to determine the structure of one monolayer of C60 on Pt(110)(1×2) surface. The data reveal that a c(4×4) reconstruction was induced by fullerene on Pt with the creation of a regular array of nanopits. The partial refilling of the initial (1×2) missing row by Pt atoms involves long-range mass transport ending up with a 75% Pt occupancy of the topmost surface layer. The orientation of the molecule is compatible with cm local surface symmetry, with one of the C60 pentagons facing the surface and two adjacent hexagonal rings almost parallel to the (111) facets formed by the nanopit. The increase of the van der Waals repulsion forces induced by the short nearest neighbor C60 distances (9.6 Å) is compensated by the formation of 15 C-Pt bonds with lengths ranging from 2.1 and 2.6 Å that could be responsible for a charge transfer between the substrate and the molecule through a large contact area and a reduction of 6% of the C60 diameter. Extensive comparison of our model derived from experimental X-ray data with other models published in literature was carried out. 1. Introduction Interaction at the interface between large molecules (LMs) and metallic substrates has attracted considerable interest due to the development of new devices based on organic films and/ or functionalized inorganic molecules. Usually, adsorption and self-assembly of LMs on surfaces1-4 result in drastic changes of the interface chemical and/or physical properties which might be tailored by functionalizing the adsorbed molecules. Deeply modified interfaces are often observed as surface restructuring involving long-range atomic transport as well as structural deformation of the adsorbed molecules.1 In this context, C60 can be seen as a model system. The presence of strong directional π molecular states, together with their capability of being either electron acceptors or donors, allows fullerenes to establish different kinds of interaction with surfaces, ranging from van der Waals interaction to covalent bonding through the C atoms in direct contact with the substrate. Since the number of C atoms directly bonded to the surface is quite variable, it is therefore important to determine the orientation of the molecule with respect to the substrate as well as all details of the modified substrate structure due to adsorption. Final atomic restructuring depends on the substrate chemical nature and on its orientation. Upon fullerene adsorption, formation of nanopits has already been reported on Pd(110),5,6 Ag(110),7 Ag(111),8 Pt(111),4 Au(111),9 and Au(110).2 The driving force leading to nanopit formation is usually attributed to a strong bonding between C60 and the substrate resulting in a partial embedding of the fullerenes in the first layer(s) of the metal. This surface restructuring is in most cases accompanied by a change on the apparent size of the C60 molecules as observed by scanning tunneling microscopy (STM).10,11 †
Institut de Ciencia de Materials de Barcelona, ICMAB-CSIC. Physics Department, Autonomous University of Barcelona. § Institut Ne´el, CNRS & Universite´ Joseph Fourier. | Retired. ‡
Up to now, fullerene-surface interactions have been mainly studied by STM, on systems including isolated molecules and monolayers.5,12-18 In the case of C60 on Pt(111),19 the C-Pt interaction is particularly strong and has a predominant covalent character as deduced from the short bond lengths4 (1.9 Å). Onto Pd(110),5 nanopits of variable size (two or three lattice distances long) and depth (one or two layers), that arrange to form a regular net, host the fullerenes that are partly embedded. Such a strong covalent character is also expected for C60/Pt(110). Recently, combined STM and density functional theory (DFT) studies of C60 chemisorption on Pt(110)-(1×2) have been reported.10,20,21 Isolated molecules were found to bind to the surface with strong covalent bonds leading to immobile adsorbates at room temperature. The corrugation of the missing row reconstruction may help to achieve ordered molecular film, but thermal treatment is the crucial parameter favoring surface diffusion to reach final surface restructuring and optimum C60 packing density. Thermal C60 self-assembly yields two distinct adsorption phases: a quasi-hexagonal C60 lattice forms under annealing at 700 Ksas well obtained by direct deposition at 810 K as observed in our workswhereas an oblique lattice is observed for annealing at 850 K. The quasi-hexagonal phase exhibits a c(4×4) unit cell as on Ag(110).7 In contrast with Au(110) and Pd(110) where the substrates are shown to rearrange forming nanopits (i.e., partially empty/filled rows), other adsorption mechanisms as molecular adsorption down the troughs (i.e., empty rows) are proposed by Orzali and coworkers,21 while C60 adsorption on a fully relaxed Pt(110)-(1×1) surface is proposed for the oblique phase on DFT basis. Therefore, the particular behavior of fullerenes on Pt(110) has motivated a new analysis using surface X-ray diffraction (SXRD) to decipher the detailed interface structure of the c(4×4) quasi-hexagonal phase. As already shown,4 X-rays allow determination of not only the relative orientation of C60 with respect to the substrate but also the type of the bonds deduced from the corresponding bond length. The Pt(110) interface
10.1021/jp1008175 2010 American Chemical Society Published on Web 08/26/2010
15646
J. Phys. Chem. C, Vol. 114, No. 37, 2010
Figure 1. Directional scans along the [11j0] direction of the reciprocal space for the (0, 1/2, 0.05) reflection. The narrower peak corresponds to the fwhm measurement of the clean Pt(110)-(1×2) missing row surface, whose intensity has been reduced by a factor of 50 for clarity. The broader peak corresponds to the same reflection after C60 deposition with the optimum c(4×4) superstructure preparation.
shows a regular distribution of biatomic type vacancies where the C60 molecules lie down with one of the pentagonal rings of the molecule almost parallel to the surface. Moreover, we find a distribution of C-Pt bonds between 2.1 and 2.6 Å and an apparent reduction of the C60 diameter of about 6%. 2. Experimental Details The missing row Pt(110) reconstructed surface was prepared by argon sputtering (0.8 keV) and annealing (T ) 970 K) cycles, followed by cooling in oxygen (pO2 ) 1.5 × 10-7 mbar) down to 850 K. The sample was heated by electronic bombardment, and its temperature was controlled with an infrared pyrometer. The resulting (1 × 2) domain size was found to be typically 2000 Å (Figure 1) as determined from the width of the scan crossing the (0, 1/2, 0.05) reflection along the H direction with H, K, L parallel to the [11j0], [001], and [110] directions of the Pt(110) surface, respectively.22 Then highly pure C60 (purity 99.9%) was sublimated, from an alumina crucible kept at 750 K, on the Pt substrate maintained at the optimum temperature of ∼810 K. The fullerene coverage was calibrated recording the C273/Pt247 Auger peak-to-peak ratio. After C60 deposition, the intensity of the (1×2) fractional order reflections of the clean Pt decreased on average by a factor of 50 suggesting that the missing row corrugation of the original (1×2) reconstruction as well as its domain size have changed drastically as a consequence of its interaction with C60 (Figure 1). Pseudohexagonal close-packed domains are formed, and before saturation, they may coexist with uncovered surface regions still showing the (1×2) missing row reconstruction of the clean Pt.20 As determined from the width of the (0, 1/2, 0.05) and (1/2, 0, 0.05) reflections, c(4×4) domains are typically 200 Å wide (Figure 1). One of the c(4×4) reflections was monitored during deposition to determine the C60 film completion after the sample was cooled back to room temperature: an extensive fractional order data set was then collected, with 0.05 e L < 2.5 in reciprocal lattice units. The spectra were collected at the surface beamline BM32 of the European Synchrotron Radiation Facility, ESRF, in Greno-
Torrelles et al. ble.23 The X-ray beam was generated by a bending magnet, monochromatized with a water-cooled double crystal Si(111) monochromator, and sagitally focused. The energy and size at the sample of the incident beam were set to 18.0 keV and 0.5 mm × 0.1 mm (horizontal × vertical dimensions at the sample surface), respectively. The Pt(110) single crystal [miscut e0.1°] was mounted in the ultrahigh vacuum diffraction chamber (base pressure 1 × 10-10 mbar) coupled with a six circle diffractometer. The incidence angle of the diffractometer was set to the critical angle of 0.25° and maintained constant during the whole experiment. The integrated intensities were recorded by setting the correct diffraction condition and then by rocking the crystal around its surface normal. A total of 1775 reflections, specific to the c(4×4) structure, were measured, which reduced to 1221 after averaging the equivalent reflections24 belonging to 45 FORs (fractional order rods) and 9 CTRs (crystal truncation rods). The new lattice parameters corresponding to this c(4×4) cell are expressed as (a′1, a′2, a3) ) (4*a1, 4*a2, a3).22 The standard deviations σHKL of the structure factors FHKL were close to 11%, evaluated by the square sum of a systematic error estimated from the measurements of several equivalent reflections and a statistical error.25 From the inspection of the experimental data, the c2mm symmetry group was determined. 3. Substrate Structure Determination Calculations were performed with a modified version of the structure refinement ROD software26 based on optimizing the χ2-goodness factor. The analysis was performed “ab initio” i.e. without assumption, using the CTRs and FORs, and starting from the initial (1×2) reconstruction of the clean surface, i.e., an ordered arrangement of dense rows followed by empty ones in the [001] direction which fulfills the observed c2mm experimental symmetry. The atomic density of all sites, including the initially empty row sites, in the three topmost layers was then allowed to vary, but constrained to fulfill the observed symmetry, to find out which atomic positions were actually occupied by Pt atoms or, in other words, to determine the shape and size of the nanopits, if any, supposed to host the molecule. The substrate structure is a first-order parameter compared to the actual orientation of the molecule and/or to its registry with respect to the substrate due to the atomic number difference. The molecules were then considered in a second step, with C60 supposed to reside in the optimum substrate configuration derived from the first stage of the analysis and compatible with the c2mm symmetry. The final analysis included 48 independent Pt atoms distributed over five layers and one fullerene molecule. We determine a total of 62 positional coordinates for bulk atoms simultaneously with the height of the C60 center and three Euler angles for its orientation. Seven isotropic thermal vibration parameters were also optimized, one for the C atoms of the fullerene molecules, two for the topmost Pt layer, and one for each of the four deeper Pt layers. Thermal vibrations parameters were assumed identical for all C atoms of the molecule (BC ) 14 Å2). Meanwhile, different values were assigned to Pt atoms of both types of rows in the top layer: BPt-top-part ) 9 Å2 (partially filled missing row) and BPt-top-filled ) 6 Å2 (filled row). Deeper layers were given the following values: BPt-2 ) 2 Å2, BPt-3 ) 0.7 Å2, BPt-4 ) 0.3 Å2, and BPt-5 ) 0.3 Å2. The root-mean-square (rms) vibration amplitude u (defined by B ) 8π2〈u2〉) is 0.4 Å for C atoms and 0.2 Å for the (averaged) top layer. The thermal vibrations contribute to a reduction by almost 2 of the agreement factor, and another significant additional decrease is obtained when the positional coordinates of the Pt atoms in the surface cell were also considered in the last refinement step process.
Nanoscale Patterning by C60 Ordering on Pt(110)
J. Phys. Chem. C, Vol. 114, No. 37, 2010 15647
Figure 3. The different structural models proposed in the literature that have been checked for the substrate: (1) (1×2) missing row; models (2, 3, 4, 5) show noncentered cells; (6, 7, 8, 9, 10, 11, 12, 13) are centered models satisfying the c2mm symmetry except (3, 5, 6, and 10) with cm symmetry; (13) our model arising from X-ray diffraction analysis. The model number (14) corresponds to the unreconstructed Pt-(1×1) surface.
TABLE 1: GIXRD Results for the Various Models of Figure 3 with χ2 Given for C60 Optimum Configuration (see below)a Figure 2. Top view of the optimum surface slab ordering over an extended area showing the geometry and distribution of the nanopits (two layers) on the Pt(110)-c(4×4) surface (marked unit cell). Note the shift of some atoms perpendicular to the dense rows. Darker color atoms correspond with deeper surface atoms.
These two sets of parameters can be refined together when the structure is close to the one giving the lowest χ2 (Table 2). Figure 2 shows the result of the first step of the procedure described above. The best fit is obtained for a substrate configuration where the initial missing row is partially filled yielding 75% occupancy in the first layer and with regularly distributed double vacancies (nanopits) along one of the atomic rows in the [11j0] direction. This restructuring of the Pt(110) surface upon adsorption with the formation of a regular array of nanopits is consistent with what has been already observed on Pd(110)5 and on Au(110)2 that have interatomic distances 0.7% smaller and 4% larger than Pt(110), respectively: the molecules break the surface structure to produce nanopits optimizing so the number of bonds with the substrate. In both cases, these nanopits consist of 2- or 3-fold vacancy one or two layers deep. C60s tend to “dig out” hexagonal-like nanopits to mimic the (111) orientation case. Indeed, on Au(111)9 and Ag(111),8 a hexagonal landing base is provided to the C60 by removing one substrate atom. In the present system, C60 on Pt(110), a 2-fold vacancy is necessary to mimic locally (111) facets. In contrast, for much smaller interatomic distances, C60 increases the number of C-metal bonds by producing a onedimensional corrugation in the [001] direction as observed on Cu and Ni.27 Since Pt has electronic and physical properties as well as interatomic distances in between those of Pd and Au, similar formation of nanopits could be reasonably expected in contrast with the Cu or Ni cases. Nevertheless, this substrate structure is quite at odds with the STM results and DFT calculations on the same Pt(110) surface.21 Here, we compare the published models with the one derived from the present work. Figure 3 shows the different models that have been checked. The main differences between them reside in the first layer filling. For most of them, the top layer is a regular alternation of full and half-filled rows with a 50% total occupancy: the basic one being the (1×2) missing row structure of clean Pt(110). This substrate arrangement has been proposed as a twodimensional (2D) template to host the C60 molecules.21 Even if electron and X-ray diffraction patterns indicate that the (1×2)
model (1) (2) (3) (4) (5) (6) (7)
(1×2) not c(4×4) not c(4×4) not c(4×4) not c(4×4)
compatible compatible compatible compatible
χ2
model
χ2
7.5 5.9 4.7 4.3 3.7 4.0 4.4
(8) (9) (10) (11) (12) (13) (14) (1×1)
7.2 4.2 4.5 5.4 2.2 1.5 7.6
a Note that all checked models give high χ2 values, with the exception of the 3- (12) or 2-fold (13) vacancy models. The best model (13) corresponds to the highest Pt fraction in layer 1 and a hexagonal-like, 2-fold, vacancy, i.e., the optimum Pt-C60 bonding configuration.
periodicity may be preserved after growth, the important changes in FORs intensity reveal that the substrate structure has been drastically modified upon adsorption. Table 1 summarizes the χ2 agreement factor for all models given in the previous figure. All these models have been checked using the optimum C60 configuration obtained from our best model 3.13 (model number 13 of Figure 3), as described in the next section. Two or four domains were included to simulate the experimental data symmetry depending on the presence or absence in each model of a mirror plane parallel to the [11j0] direction, respectively, i.e., models 3.3, 3.5, and 3.6 do not have this symmetry plane. Table 1 clearly points out to our model 3.13 that results in a χ2 value lower than any other configuration. The second best model is precisely the 3-fold vacancy nanopit (3.12) that occurs simultaneously with the 2-fold vacancy nanopit in the case of C60 adsorption on Pd(110). Both configurations allow for high packing density of the molecules. Indeed, we note that, in the case of Pd(110), the rotated-stripe phase is the densest one (88.2 2 Å /molecule). In the present case, Pt(110), the area/molecule is 86.9 Å2, i.e., hardly smaller. 4. Molecular Orientation Determination From the data analysis and the previous discussion, we established that the c(4×4) unit contains one fullerene and a Pt double vacancy; as a consequence, in this dense quasi-hexagonal phase, all C60 should have the same orientation and the same nanopit adsorption site. First, the C60 were assumed to adsorb with an orientation compatible with c2mm symmetry, i.e., the configuration C2 of Figure 4. From now on, the configurations labeled Cn (n ) 1, 2) refer to our X-ray analysis, while the Mn (n ) 1-8) labeling refers to Casarin’s molecular configurations.10
15648
J. Phys. Chem. C, Vol. 114, No. 37, 2010
Figure 4. Bottom view of C60 balls showing the C1 and C2 configurations. The fullerenes interact with the substrate via a pentagon almost parallel to the surface (in the page plane) or via a 6-6 bond respectively.
Torrelles et al. neither in terms of registry relative to the substrate nor as concerns the substrate configuration. As shown in Figure 5, the agreement is significantly poor for the optimum DFT M1 model (where a 5:6 bond hexagon is almost parallel to the surface R ∼ -34°). The situation is even less favorable for the orientation M2 (a hexagon facing the surface at a bridge site R ∼ -26°) and is worst for the M3 model (6:6 bond at or very close to the bridge site). The lowest χ2 value (≈1.5) was obtained with the C1-configuration model (Figure 4a) after allowing the molecule to rotate and to shift with respect to the initial adsorption site and considering two mirror domains to reproduce the experimental c2mm symmetry. The improvement of χ2 is associated with changes in the interatomic distances in the first and second layers. 5. Geometrical Arrangement and Bond Length
Figure 5. χ2T ()∑iχ2i the summation is extended to the whole set of reflections) evolution versus polar tilt angle R (rotation axis parallel to [001] direction) from the initial refined C2 configuration model (equivalent to M3). Two minima are observed for equivalent configurations C1(≡5:6 bond) and C1′(≡6:5 bond; 90° difference with respect to C1). C1 and M8 are similar.10
Since with this molecular orientation (C2), the goodness factor (χ2 ≈ 2.1) was still relatively high, the surface symmetry was released to cm preserving the mirror plane parallel to the [11j0] direction. Therefore, two mirror domains perpendicular to the [001] direction were considered. In order to test all configurations compatible with the cm symmetry, we have calculated the χ2 as a function of the rotation of the C60 around an axis parallel to [001] in a large angular range. We thus scan all configurations including the most favorable ones: orientations M1 (5:6 bond), M2 (six-membered ring or hexagon), and M3 (6:6 bond). The starting point of the variation is close to the model M310 (i.e., our C2 configuration) where a 6:6 bond lies parallel to the surface above the center of the nanopit. This C2 configuration (compatible with c2mm symmetry) is considered as the initial C60 orientation, i.e., zero rotation angle. The variations of the χ2 goodness fit value with the rotation show two pronounced minima for each mirrored domain (180° apart, i.e., symmetrical arrangement) and correspond to a situation (rotation R ) -54.4°; see Figure 5) where a pentagon faces the substrate but with some marked angle, both atoms of a 5:6 bond being deepest into the nanopit and in the vicinity of a short bridge in the second Pt layer. The C1 optimum orientation is quite close to M8 configuration,10 in terms of bonding (i.e., molecule orientation) but
The final configuration of the system is shown in Figure 6. The corresponding atomic coordinates of the Pt atoms in the three outermost layers and those of the C atoms of the C60 in the cm asymmetric unit cell are given in Table 2. The C60 molecules are NOT embedded in the nanopits and lay ∼0.2 Å above the topmost Pt surface layer, close to the center of the nanopit. The first Pt-Pt interlayer distance -1.30 Å is significantly shorter than the bulk distance (1.375 Å) but much larger than for the clean surface28 (∼1.1 Å). This is additional evidence of the C60 impact on the substrate structure through electronic transfer. This relaxation is certainly due to the presence of “extra atoms” in the original (1×2) missing row. As in ref 10, the number of bonds is determined by the closest contact points between adsorbate and substrate atoms and a cutoff distance of 2.6 Å is chosen. Our most stable C60 configuration yields 15 C-Pt bonds with lengths ranging from 2.1 to 2.6 Å for an average bond distance of 2.4 ( 0.1 Å. Accounting for the Pt and C covalent radii (1.28 and 0.77 Å, respectively) the resulting C-Pt bond length is ∼2.05 Å very close to the experimental C-Pt bond length value of 1.9 ( 0.1 Å obtained for the C60/Pt(111) system. On Pt(110), we found that four C atoms have C-Pt bond lengths lower than 2.15 Å: two of them are linked to atoms of the partially filled row (Pt1-1; see Table 2 and Figure 6c) while the other two (those of the 5-6 bonds) are linked to Pt atoms in bridge site of the second layer (Pt2-1). These strong bonds between C60 and Pt1-1 are responsible of a lateral shift of 0.5 Å of the molecule along the [11j0] direction, that breaks the c2mm symmetry and helps to stabilize the final configuration of the molecule,4 i.e., to compensate the high number of C-Pt bonds on the right side of the molecule visualized as darker C atoms in Figure 6c. The C60 stability is achieved by the combination of both a high number of C-Pt bonds that help to redistribute the substratemolecule charge via a large contact area and the coexistence of different types of covalent bonds:19 the different bond lengths stick the molecule on the surface by minimizing the surface stress at the interface. This fact is supported by the atomic shifts off the “ideal” bulk positions of the Pt atoms at the reconstructed surface, as observed in Table 2. The effect of this strong bonding shows up as an apparent ∼6% reduction of the C60 diameter, as obtained from the refinement procedure using a deformation parameter that allows for a uniform reduction/increase of the molecular size. Anisotropic deformations of the C60 cage at constant volume were also tested and discarded since the spherical shape of the molecule remained undistorted. This apparent change of the molecular size has already been observed by STM.20 The interatomic distance in bulk Pt is 2.77 Å, whereas the shortest and longest nearest-neighbor Pt-Pt dis-
Nanoscale Patterning by C60 Ordering on Pt(110)
J. Phys. Chem. C, Vol. 114, No. 37, 2010 15649 TABLE 2: Atomic Coordinates (in normalized lattice units) of Atoms Included in the cm Asymmetric Unit Cell for the Best Bridge C1 Modela
Figure 6. (a) Top view of the C60/Pt(110)-c(4×4) superstructure showing the best and most plausible orientation of the C60 molecules according to the least-squares refinement procedure. The C60 molecule has been cut to better visualize their bonds with the Pt substrate. The C60 molecule is shifted by 0.5 Å long [11j0] direction. (b) Lateral view. (c) Surface slab including the C60 molecule and the two Pt topmost layers viewed from down showing the C1 configuration: the C60 molecule approximately places a 5:6 bond (pentagon almost parallel to the surface) on a bridge site of the second Pt layer, respectively. Black or gray atoms indicate C atoms with shorter or larger C-Pt distances than 2.4 Å, respectively. Topmost or second surface Pt layers are indicated with light (whiter) or (darker) sky blue colors, respectively. The black C bonds (pentagon) help to identify the pentagon almost parallel to the substrate surface.
tances in our final structure are 2.54(3) and 2.96(3) Å, respectively. The minimum value represents a bond contraction of 8% as observed in similar systems29 while the maximum corresponds to a 7% increase, which is quite reasonable in regards to the large tolerance range for bond lengthening. As usual for metallic surfaces, the largest distortions occur for vertical atomic displacements and they gradually decrease with depth. The absolute average atomic distortions 〈|Z|〉 of the five surface layers, from top to bulk, are 0.09((2), 0.04(2), 0.05(2), 0.04(2), and 0.02(2) Å, respectively, and correspond mainly to a global contraction of the surface slab with respect to the situation in the bulk. The overall quality of the data fit can be seen in Figure 7. 6. Discussion Our X-ray results shed new light on the C60 interaction with the Pt(110) surface. Indeed, while we conclude to similarsalbeit
element
X (0.001
Y (0.001
Z (0.004
Pt1-1 Pt1-2 Pt1-3 Pt1-4 Pt2-1 Pt2-2 Pt2-3 Pt2-4 Pt3-1 Pt3-2 Pt3-3 Pt3-4 Pt3-5 Pt3-6 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 C18 C19 C20 C21 C22 C23 C24 C25 C26 C27 C28 C29 C30 C31 C32
0.375 + 0.024 0.125 + 0.014 0.125 + 0.012 0.375 + 0.012 0.000 + 0.001 0.000 + 0.014 0.250 + 0.005 0.250 + 0.014 0.125 + 0.007 0.375 + 0.010 0.125 + 0.003 0.375 + 0.004 0.125 + 0.019 0.375 + 0.019 0.113 -0.076 -0.006 0.199 -0.034 -0.177 0.301 0.055 0.173 -0.133 -0.205 0.329 0.243 0.010 -0.106 -0.251 0.357 0.206 0.080 -0.152 -0.225 0.296 0.224 0.034 -0.080 -0.197 0.271 0.123 -0.108 0.096 -0.020 0.176
0.000* 0.500* 0.250 + 0.012 0.250 + 0.007 0.125 + 0.000 0.375 + 0.002 0.125 - 0.002 0.375 + 0.010 0.000* 0.000* 0.500* 0.500* 0.250 - 0.000 0.250 + 0.001 0.041 -0.005 0.067 0.084 0.137 -0.004 0.045 0.191 0.155 0.137 0.068 -0.090 0.154 0.213 0.183 0.042 0.046 0.192 0.212 0.158 0.087 -0.068 0.139 0.184 0.159 0.044 0.003 0.141 0.089 0.074 0.047 0.005
0.000 - 0.033 0.000 + 0.012 0.000 - 0.029 0.000 - 0.046 -0.500 - 0.029 -0.500 + 0.020 -0.500 + 0.002 -0.500 + 0.012 -1.000 + 0.003 -1.000 + 0.053 -1.000 - 0.023 -1.000 - 0.025 -1.000 - 0.010 -1.000 + 0.013 0.166 0.216 0.190 0.382 0.434 0.484 0.587 0.628 0.637 0.723 0.745 1.018 1.043 1.092 1.131 1.181 1.442 1.468 1.481 1.554 1.580 1.862 1.870 1.911 1.956 2.000 2.134 2.157 2.231 2.432 2.460 2.479
a Only the three topmost Pt surface layers are included. Pti-j indicates Pt atom j located in layer i.
not identicalsconfiguration for the C60 orientation with respect to the surface, the outcome of the diffraction analysis is in striking contrast with DFT conclusions on the basis of STM comparisons.21 We now discuss some of the models proposed by previous authors. Basically there are two preferred surface configurations to account the ordering of C60 on Pt(110) for two different phases appearing at the so-called low and high temperatures studied using STM or combined with DFT techniques,21 respectively. (i) The low-temperature phase (∼700 K) considers a c(4×4) superstructure keeping the original missing row structure (model 3.1). However, and according to X-rays analysis, the c(4×4) surface substrate arrangement is more compatible with a partially filled (1×2) missing row (biatomic vacancy model). Similar types of vacancy models to that obtained from X-rays, as the 3.2, 3.3, and 3.4, were nevertheless disregarded on the
15650
J. Phys. Chem. C, Vol. 114, No. 37, 2010
Figure 7. Fractional and crystal truncation rods of the C60/Pt(110)c(4×4) structure. The continuous lines correspond to the calculated data from the structure shown in Figure 6. Two rods by graph defined each one in the negative or positive L region, respectively.
grounds of STM images showing antiphase domains supposed to exclude the possibility of such types of nanopits. More specifically, model 3.4 includes a regular distribution of biatomic Pt vacancies in the top layer, similar to those we find; however, it is not compatible with the observed c(4×4) periodicity but with a p(4×4) instead. Moreover, the surface coverage of the topmost Pt layer is 62.5% compared to 75% in our model, so this model would contain three biatomic vacancies per unit cell and three possible sites for C60 adsorption (one per nanopit) which would result in inconsistent C60 intermolecular distances. Models 3.2 or 3.3 would result in the coexistence of different types of nanopits in the (4 × 4) unit cell and consequently would lead to some corrugation within the C60 domains: such a corrugation has not been detected by STM ! (ii) The reason why the biatomic Pt vacancy models come out less favorable than the original quasi-hexagonal c(4×4) C60arrangement phase on the (1×2) missing row remains unclear. From the X-ray point of view, the latter model would produce two different sets of fractional order data with very different average intensities. Reflections to which the (1×2) missing row and C60 molecules contribute simultaneously should be more intense by at least 1 order of magnitude than the c(4×4) reflections due to the molecules only. The experimental results show that these two data sets have similar average intensities allowing no room/possibility for this latter model to occur. (iii) The models tested by DFT to compare with STM images for the high temperature phase ignore the intermolecular interaction and consider one isolated molecule on various Pt local arrangements (bi- and triatomic vacancies and (1×1)-Pt flat layer). The DFT calculations discard any type of vacancies when compared with the STM images. The best agreement is obtained for a model in which a local flat (1×1) unreconstructed Pt slab and one C60 molecule are considered. The fact of neglecting the intermolecular interaction allows extrapolating
Torrelles et al. the DFT results to the low temperature phase, we report in this paper, permitting the conclusion that the vacancy models would be excluded by DFT for any of the phases. However, the existence of biatomic type vacancies at the interface with fullerenes established from our X-ray results demonstrates the limitations of DFT for the considered models to distinguish between vacancy and unreconstructed types of surfaces. A local arrangement with vacancies for the high temperature phase would be more consistent within the framework of C60 systems studied by diffraction techniques. At low coverage, on Pd(110), C60 is expected to adsorb in bi- and triatomic vacancy nanopits, one and two layers deep, respectively.6 Additionally, XPD illustrates continuity in the shape and contrast of the measured spectrashence, the orientation and site adsorptionsat various coverages. In the densest phase, two types of nanopitssincluding those observed in less dense phases, i.e., those we proposesare regularly distributed. Similar type of nanopits are observed on Au(110) upon C60 adsorption.2 Pt has a lattice constant which lies in between those of Au and Pd, and thus it is not surprising that our diffraction analysis concludes to similar behavior with respect to molecular adsorbates on close fcc(110) substrates. The present resultssbiatomic vacancy modelshave the advantage to give a quite coherent picture of the evolution of C60 adsorption on Pt(110) with the coverage and, in addition, a more comprehensible overall scheme for the fullerene-substrate configuration on large lattice constant face centered cubic (fcc) metals whether on dense (111) or less dense (110) orientations of Pt, Pd, and Au: (quasi)hexagonal nanopits form upon fullerene deposition with increasing top-layer atomic density at the expense of the (1×2) when present. A clear outcome of the 2- or 3-fold vacancy nanopit formation is that it allows maximum coordination as compared to the (1×2) case or any other substrate configuration with less dense surface layers or local environments. In this respect, nanopits are also much more favorable than a flat layer that might occur in the case of dominant molecule-molecule interaction and weak adsorbate-substrate interaction yielding a perfect hexagonal close packed (hcp) fullerene layer (never observed). The strong influence of the number of C-Pt bonds on the adsorption energy is already well documented.10 In our study, the maximum number of C-atoms bonded to Pt interface atoms is 15 (for configuration C1 and model 3.13), and reduces to 10 only for the C2 configuration.30 In comparison, the optimum number of C-Pt bonds is 14 at most in the various configurations (missing row adsorption in opposition to nanopits) of ref 10, and in particular for the M1 configuration retained in ref 21. This difference in the number of bonds will influence the stability of the molecule as already pointed out. Simultaneously, a reduction of the number of C-Pt bonds will concentrate the C/Pt charge in a lower spatial interfacial region yielding a stronger reduction of the apparent C60 diameter. GIXRD calculations yield a 6% (10%) contraction of the fullerenes for the optimum C1 configuration (respectively C2). This high number of bonds with strong covalent character for some of them would be enough to immobilize the molecule on the surface. This conclusion would be, moreover, supported by the following facts: (i) the carbon atoms have a certain degree of positional disorder as indicated by their relatively high rms vibration amplitude value (0.4 Å) but much lower than in the C60/Au(110)-(6×5) system with a value close to the C-C bond distance2 (∼1.0 Å). This latter indicates a strong delocalization of the C atoms which could be interpreted as a free rotation of the molecules at room temperature. On the Pt(110) substrate,
Nanoscale Patterning by C60 Ordering on Pt(110) the C60 are vibrating around their equilibrium position but could be either immobile on the surface or hopping between the two equivalent orientations (5:6 bond). Figure 5 shows four minima corresponding to two 5:6 and two 6:5 equivalent orientations (6:5 is the mirrored orientation of the 5:6) which arise from the consideration of the two domains. The molecules, however, cannot freely hop between these four orientations since this would be observed on high-resolution STM images with different density of states distribution on molecules within a single domain, and this has not been detected. (ii) The hopping between two equivalent orientations is moreover discarded due to the very short coherence time of the synchrotron radiation, ∼10-16 s. Intermediate orientations of C60 during the hopping process would be observed since the free rotation time of the C60 is in the picoseconds range.31 However, as we have pointed out above, the positions of the C-atoms in the molecule are well localized. Therefore, the C60 are immobile on the Pt surface. About the 6% reduction of the C60 apparent diameter, the associated displacement of the carbon atoms is lower than the rms vibration amplitude; thus, the uncertainty on the C atoms localization might be, in principle, included in the Debye-Waller factor. Nevertheless, the χ2 goodness factor remains sensitive to an isotropic contraction of the C60, which could be the signature of a partial charge transfer from the molecule to the substrate as pointed out by DFT calculations10 where C60 is found to be positively charged on Pt(110). Regarding the molecular orientation, we wish to emphasize the great similarity for the adsorption on Pd and Pt. Indeed, in both cases, the C60 balls stick to the surface via or close to a 5:6 bond. The polar tilt is null for the Pd case. In our case, if the molecular symmetry is completely relaxed, so that the cm symmetry is lost, the molecules suffer a 5° azimuthal rotation, which is close to the 4.5° observed in the Pd case,5 but without improvement of the χ2. Weckesser et al. (ref 5) wonder whether this is a more general rule, the tilt being driven by the chemical nature of the substrate and by the C60 local environment. The present results indicate that the preference of a 5:6 bond facing the nanopit is more general and extends to the Pt(110) situation. However, it seems, in the present state of our knowledge, that both the azimuth and tilt are substrate dependent. Meanwhile, this is at odds with Au(110) surface2 where the C60 are found to rotate freely, in striking contrast with Cu, Pd, and Pt onto which the molecule is immobile. The situation is also different on the (111) face due to the perfectly hexagonal nature of the surface layer: the balls bind to the substrate via a hexagon and not via a 5:6 bond as on the (110) surfaces. Conclusions The structural properties of the C60/Pt(110)-c(4×4) system have been studied by grazing incidence X-ray diffraction. The atomic arrangement of the Pt at the interface has been determined as well as the optimum C60 configuration. The interaction between the topmost Pt surface atoms and the fullerenes yields an ordered distribution of biatomic vacancies (nanopits) that forms a 2D template where the quasi-hexagonally packed C60 molecules are hosted. The experimental surface symmetry is c2mm while the local symmetry of the structure has been determined to be cm. The optimum molecular orientation, compatible with this latter symmetry, gives rise to 15 C-Pt bonds with a length distribution from 2.1 to 2.6 Å, indicating a charge delocalization between the metal and the molecule through a rather large contact area. The interaction between C60 molecules and the topmost Pt layers induces a
J. Phys. Chem. C, Vol. 114, No. 37, 2010 15651 less stressed surface structure when the optimum C60 orientation is considered. The molecule is shifted by 0.5 Å in the [11j0] direction with respect to the nearest Pt bridge of the second layer with one of its pentagon rings almost parallel to the surface. In this molecular orientation, the hexagonal C-rings are almost parallel to the local (111) facets formed by the Pt-pits. The strong molecule-substrate covalent bonding with a definite preference for a single chemisorption site (bridge) leads to a close packed structure. The NN C60 distances (9.6 Å), smaller than on other metal surfaces such as Au(110)2 (∼10.6 Å) or Ag(100),32 ∼10.4 Å, increases the van der Waals repulsion that is partially compensated by the high number of C-Pt bonds and by the ∼6% apparent diameter reduction of the whole molecule which could be an indication of a charge transfer between the substrate and the molecule. Acknowledgment. X.T. thanks the Spanish MICINN agency for partially funding this project through projects CSD200700041 and MAT2009-09308. V.L. acknowledges the support of “Generalitat de Catalunya” through the 2009 SGR 1292 project and of the MEI through the MAT2007-66309-C02-02 project. We also thank the technical support received from the scientific staff of the ESRF (BM32 and BM25 beamlines) during the experiments. References and Notes (1) Langlais, V.; Torrelles, X.; Gauthier, Y.; De Santis, M. Phys. ReV. B 2007, 76, 035433. (2) Hinterstein, M.; Torrelles, X.; Felici, R.; Rius, J.; Huang, M.; Fabris, S.; Fuess, H.; Pedio, M. Phys. ReV. B 2008, 77, 153412. (3) Pedio, M.; Felici, R.; Torrelles, X.; Rudolf, P.; Capozi, M.; Rius, J.; Ferrer, S. Phys. ReV. Lett. 2000, 85, 1040. (4) Felici, R.; Pedio, M.; Borgatti, F.; Iannota, S.; Capozi, M.; Ciullo, G.; Stierle, A. Nat. Mater. 2005, 4, 688. (5) Weckesser, J.; Cepek, C.; Fasel, R.; Barth, J. V.; Baumberger, F.; Greber, T.; Kern, K. J. Chem. Phys. 2001, 115 (19), 9001. (6) Weckesser, J.; Barth, J. V.; Kern, K. Phys.ReV. B 2001, 64, 161403(R). (7) David, T.; Gimzewski, J. K.; Purdie, D.; Reihl, B.; Schlittler, R. Phys. ReV. B 1994, 50, 5810. (8) Li, H. I.; Pussi, K.; Hanna, K. J.; Wang, L.-L.; Johnson, D. D.; Cheng, H.-P.; Shin, H.; Curtarolo, S.; Moritz, W.; Smerdon, J. A.; McGrath, R.; Diehl, R. D. Phys. ReV. Lett. 2009, 103, 056101. (9) Pedio, M.; Torrelles, X.; Cepek, C.; Felici, R. Phys. ReV. B 2010 (submitted). (10) Casarin, M.; Forrer, D.; Orzali, T.; Petukhov, M.; Sambi, M.; Tondello, E.; Vittadini, A. J. Phys. Chem. C 2007, 111, 9365. (11) Pascual, J. I.; Go´mez-Herrero, J.; Rogero, C.; Baro´, A. M.; Sa´nchezPortal, D.; Artacho, E.; Ordejo´n, P.; Soler, J. M. Chem. Phys. Lett. 2000, 321, 78. (12) Rudolf, P.; Gensterblum, G.; Caudano, R. J. Phys. IV 1997, 7, 137– 149. (13) Futaba, D. N.; Chiang, S. Jpn. J. Appl. Phys. 1999, 38, 3809–3812. (14) Gimzewski, J.; Modesti, S.; Schlittler, R. R. Phys. ReV. Lett. 1994, 72, 1036. (15) Modesti, S.; Gimzewski, J. K.; Schlittler, R. R. Surf. Sci. 1995, 331-333, 1129. (16) Rogero, C.; Pascual, J. J.; Go´mez-Herrero, J.; Baro`, A. M. J. Chem. Phys. 2002, 116, 832. (17) Sakurai, T.; Wang, X. D.; Hashizume, T.; Yurov, V.; Shinohara, H.; Pickering, H. W. Appl. Surf. Sci. 1995, 87-88, 405. (18) Giudice, E.; Magnano, E.; Rusponi, S.; Boragno, C.; Valbusa, U. Surf. Sci. 1998, 405, L561. (19) Cepek, C.; Goldoni, A.; Modesti, S. Phys. ReV. B 1996, 53, 7466. (20) Orzali, T.; Petukhov, M.; Sambi, M.; Tondello, E. Appl. Surf. Sci. 2006, 252, 5534. (21) Orzali, T.; Forrer, D.; Sambi, M.; Vittadini, A.; Casarin, M.; Tondello, E. J. Phys. Chem. C 2008, 112, 378. (22) The Pt(110) surface is described by lattice vectors (a1, a2, a3) parallel to the [11j0], [001], and [110] directions, respectively, where a1 ) a3 ) a0/2 and a2 ) a0 (a0) bulk lattice constant). (23) Baudoing-Savois, R.; Renaud, G.; De Santis, M.; Barbier, A.; Robach, O.; Taunier, P.; Jeantet, P.; Ulrich, O.; Roux, J. P.; Saint-Lager, M. C.; Barski, A.; Geaymonf, O.; Berard, G.; Dolle, P.; Noblet, M.; Mougin, A. Nucl. Instrum. Methods Phys. Res., Sect. B 1999, 149, 213.
15652
J. Phys. Chem. C, Vol. 114, No. 37, 2010
(24) Feidenhans’l, R. Surf. Sci. Rep. 1989, 10, 105. (25) Robinson, I. K. Handbook of Synchrotron Radiation; North Holland: Amsterdam, 1991; Vol. 3. (26) Vlieg, E. J. Appl. Crystallogr. 2000, 33, 401. (27) Murray, P. W.; Pedersen, M. O.; Lægsgaard, E.; Stensgaard, I.; Besenbacher, F. Phys. ReV. B 1997, 55, 9360–9363. (28) Fery, P.; Moritz, W.; Wolf, D. Phys. ReV. B 1988, 38, 7275. (29) Copel, M.; Gustafsson, T. Phys. ReV. Lett. 1986, 57, 723.
Torrelles et al. (30) Torrelles, X.; Langlais, V.; De Santis, M.; Tolentino, H. C. N.; Gauthier, Y. Phys. ReV. B 2010, 81, 041404(R). (31) Rubtsov, I. V.; Khudyakov, D. V.; Nadtochenko, V. A.; Lobach, A. S.; Moravskii, A. P. Pis’ma Zh. Eksp. Teor. Fiz. 1994, 60 (5), 320. (32) Costantini, G.; Rusponi, S.; Giudice, E.; Boragno, C.; Valbusa, U. Carbon 1999, 37, 727.
JP1008175
