Survey of structural and electronic properties of C60 on close-packed metal surfaces
- 1.8k Downloads
The adsorption of buckminsterfullerene (C60) on metal surfaces has been investigated extensively for its unique geometric and electronic properties. The two-dimensional systems formed on surfaces allow studying in detail the interplay between bonding and electronic structures. Recent studies reveal that C60 adsorption induces reconstruction of even the less-reactive close-packed metal surfaces. First-principles computations enable access to this important issue by providing not only detailed atomic structure but also electronic properties of the substrate–adsorbate interaction, which can be compared with various experimental techniques to determine and understand the interface structures. This review discusses in detail the ordered phases of C60 monolayers on metal surfaces and the surface reconstruction induced by C60 adsorption, with an emphasis on the different types of reconstruction resulting on close-packed metal surfaces. We show that the symmetry matching between C60 molecules and metal surfaces determines the local adsorption configurations, while the size matching between C60 molecules and the metal surface lattice determines the supercell sizes and shapes; importantly and uniquely for C60, the number of surface metal atoms within one supercell determines the different types of reconstruction that can occur. The atomic structure at the molecule–metal interface is of crucial importance for the monolayer’s electronic and transport properties: these will also be discussed for the well-defined adsorption structures, especially from the perspective of tuning the electronic structure via C60–metal interface reconstruction and via relative inter-C60 orientations.
KeywordsScanning Tunneling Microscopy Scanning Tunneling Spectroscopy Adsorption Configuration Scanning Tunneling Microscopy Study Interface Reconstruction
The adsorption of large molecules on surfaces is gaining increasing attention because of the multitude of possible adsorption configurations, their numerous internal degrees of freedom, and the opportunities afforded by adsorbate-induced substrate reconstruction [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]. Buckminsterfullerene (C60) adsorbed on metal surfaces, in particular, has been extensively studied for its unique geometric and electronic properties, as well as for its potential applications as basic building block for molecular scale electronic devices [14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24]. The two-dimensional systems formed on surfaces allow studying in more detail the interplay among bonding, electronic structure, and molecular orientation [25, 26, 27, 28, 29, 30, 31]. Indeed, through careful control of the number of metal atoms bonding to C60, it was demonstrated that the interface of C60 with metal surfaces is of crucial importance for its performance in molecular electronics and optoelectronics [27, 28].
Recent studies reveal that C60 can reconstruct not only the relatively reactive “open” (110) metal surfaces but also the less-reactive close-packed fcc(111) and hcp(0001) metal surfaces [18, 19, 20, 21, 29, 30, 31, 38, 39, 40, 41]. In this review, we discuss in detail the ordered phases of C60 monolayers on close-packed metal surfaces and their frequent C60-induced interface reconstruction, with an emphasis on monolayers that have one C60 molecule per supercell. (The monolayer is here defined as the saturation coverage of the first adsorbed layer.) We will explain how the various surface lattice spacings of different metals, such as Cu(111), Pt(111), and Ag(111), result in different supercell sizes and shapes, and cause different types of reconstruction upon C60 adsorption. The adsorption on other low-Miller-index surfaces will also be mentioned. Structures containing several or many inequivalent molecules per supercell (i.e., molecules in different adsorption configurations) will be discussed as well, although their interface structure is generally less well known [19, 42, 43, 44, 45]. The well-ordered monolayer phases usually must be prepared experimentally at higher temperatures or through sufficient post-growth annealing, which could induce not only interface reconstruction, but also polymerization or even decomposition of the C60 molecules [36, 40, 46], and this aspect will also be discussed (note that there may be unreconstructed structures before such annealing, e.g., for C60 on Cu(111), even though there is no detailed structure determination for these ). Investigating the interfacial atomic structure is very important since the structural details determine electronic and the transport properties [26, 27, 28, 29, 47], which will also be addressed. More discussions of the electronic properties of C60 monolayers, including with alkali doping, may be found in the reviews of refs.  and .
Before discussing the adsorption of C60 on metal surfaces, we will in the following subsection briefly summarize the relevant properties of the free molecule and of its bulk condensed phase. Then, we can address how the monolayer C60 structures result from the competition between inter-C60 forces and C60-substrate interactions. In particular, for C60-substrate interactions, two aspects affect the detailed atomic structures: (1) the symmetry matching between C60 and substrate, which determines which part of the C60 molecule binds to the surface; and (2) the size matching between C60 and different metal lattice constants, which determines the various supercell sizes and shapes. These two aspects will be discussed in detail in two further subsections below.
Free and bulk buckminsterfullerene C60
The free C60 molecule is highly symmetric (as shown in Fig. 1a–c), with a shape similar to a European football, consisting of 12 identical pentagons and 20 identical hexagons with carbon atoms at each corner . The carbons are connected through alternating ‘double’ bonds between two adjacent hexagons and ‘single’ bonds between a hexagon and a pentagon; thus there are two different C–C bond lengths, resp. 1.40 and 1.45 Å. The interaction between C60 molecules, both for solid C60 and for C60 monolayers on surfaces, is similar to the inter-layer interaction of graphite, i.e., of Van der Waals (VdW) type. Under pressure, C60 molecules undergo deformation, polymerization, and relative intermolecular orientation changes [49, 50]. In the bulk phase, at room temperature the C60 molecules adopt an fcc lattice and rotate freely so that the molecules are on average equivalent; below 260 K the C60s adopt a simple cubic lattice: the four molecules in an fcc cube become orientationally ordered but in different orientations, such that the electron-rich 6:6 bonds face the electron-deficient pentagon rings of adjacent C60s to minimize the electrostatic repulsion between adjacent molecules [51, 52]. C60 may act as either an electron acceptor or an electron donor . With alkali metal doping C60 3− becomes a superconductor, while C60 4− becomes an insulator accompanied by drastic geometric structure changes because of the Jahn–Teller effect [54, 55, 56]. Recent studies have discovered a series of “super-atom molecular orbitals” of C60 for its unoccupied energy levels that lie more than 3.5 eV above the Fermi level; these orbitals are bound to the spherical potential of the cage shell rather than the central potentials of the individual carbon atoms [16, 17].
For C60 deposited on metal surfaces, under thermal activation, the molecules can move to find the optimum adsorption sites, and can rotate to find the best orientation: the latter involves changing the polar angle to determine what part of the cage bonds to the surface, and changing the azimuthal angle to further optimize the bonding geometry. During this process, the matching of the high symmetry sites between C60 and the substrate plays an important role in lowering the total energy of the C60–metal composite system. Figure 1 shows the symmetries of C60 and those of a metal surface at different sites using fcc(111) surfaces as an example.
As seen in Fig. 1, the C60 molecule and the close-packed metals surfaces are highly symmetric. For C60 there are three main symmetries, after projection onto a surface: (1) 5-fold axes and mirror planes through the centers of two opposing pentagons (Fig. 1a); (2) 2-fold axes with pairs of mirror planes perpendicular and parallel to the edges between two hexagons (Fig. 1b); and (3) 3-fold axes and mirror planes through the centers of two opposing hexagons (Fig. 1c). For the ideal clean close-packed fcc(111) and hcp(0001) metal surfaces, there are two main kinds of symmetries (Fig. 1d): (1) if we ignore the second and deeper metal layer, there are two orthogonal mirror planes through the bridge site; (2) there are also 3-fold axes through atop and hollow (hcp hollow and fcc hollow) sites. While Fig. 1d uses the fcc(111) surface as an example, the hcp(0001) surface has the same symmetries. For the bridge site in Fig. 1d, if we consider also the deeper layers (shown as circles and crosses), there is only one mirror plane and no 2-fold rotational axis; however, the deeper layers are expected to have little influence on site preference by C60, so that the bridge site in effect has the higher symmetry of the single layer (2-fold axis and two mirror planes).
From the point of view of symmetry matching, the highest combined symmetry is obtained when a 3-fold axis of the molecule coincides with a 3-fold axis of the surface while the azimuthal orientation of the molecule makes mirror planes coincide as well: this corresponds to a hexagon of the molecule being centered and aligned with an atop or hollow site of the surface. Indeed, this local adsorption geometry with a 3-fold symmetry seems to be the most commonly observed, and is confirmed by DFT calculations [5, 18, 29, 30, 31, 38, 39, 40, 41, 57, 58, 59, 60]. For instance, with STM the molecular bonding symmetry and orientation can be determined from the symmetry seen in the images, in particular using the C60 lowest unoccupied molecular orbitals (LUMO), which show a three-lobed clover shape . The binding configuration with 2-fold symmetry with a C–C bond oriented down toward the close-packed metal surface is less frequently observed [42, 61].
The symmetry analysis also applies to fcc(100) surfaces with a 4-fold rotation symmetry: the adsorption with 2-fold symmetry (C–C bond down) matches the substrate symmetry best and is most observed in experiments [22, 23, 25, 26, 36, 62]. Although the symmetry analysis from Fig. 1d is based on the unreconstructed surfaces, it is still valid for reconstructed surfaces if the reconstruction does not change the local symmetry, as appeared to be the case for the missing-atom reconstructions induced by C60 on both fcc(111) and (100) surfaces [29, 30, 31, 36]. By contrast, this symmetry analysis does not directly apply to reconstructed fcc(110) surfaces, as the C60-induced rearrangement of (110) surface atoms may change their local symmetry [37, 63].
We now turn to the question of the optimum monolayer structure of C60. Being nearly spherical, C60 tends to form hexagonal or nearly hexagonal monolayer structures on surfaces. Forming hexagonal structures on a closed-packed fcc(111) or hcp(0001) surface is much easier than on fcc(100) and fcc(110) surfaces due to the difference in surface symmetries between these surfaces. On closed-packed metal surfaces, C60 monolayers can form hexagonal structures with one molecule per supercell, while keeping the intermolecular distance close to that in its bulk C60 phase (about 10 Å); but on (100) and (110) surfaces the adsorption structures can be very complicated because of the symmetry mismatch of the surface (4- or 2-fold) and of the C60 monolayer (which prefers 3-fold): as a result, the structures may be highly inhomogeneous and aperiodic, with C60 in a non-hexagonal arrangement, and each supercell containing several molecules in different, often inequivalent adsorption configurations [35, 36, 37, 63, 64]. In the following discussions of the size matching between C60 and metal surfaces, we focus on C60 on the close-packed metal surfaces, which is relatively simple, while there are many more possibilities for C60 on (100) and (110) surfaces.
Since the inter-C60 VdW interaction is relatively weak compared with the C60–metal chemical bonding, the inter-C60 separations can be changed to some extent on metal surfaces: usually one observes C60 monolayers with a molecular center-to-center distance in the range of about 9.5–11 Å [22, 29, 30, 31, 32, 65]. As an example of a slightly “compressed” C60 monolayer, on Pt(111), in addition to the (√13 × √13)R13.9° phase, the denser (2√3 × 2√3)R30° phase can also be formed, with an inter-C60 separation of 9.6 Å, reduced by a factor √(12/13) = 0.96. The reason why two phases can be formed on Pt(111) will be discussed in the “Further structural aspects” section.
Two types of reconstruction
Depending on the substrate lattice constant of different metal surfaces, different supercells with different numbers of metal atoms per supercell are formed to fit the size of C60, as discussed above (Fig. 2). Different supercells with different numbers of metal atoms per cell lead to different types of reconstruction: namely, (1) for a supercell with a smaller number of metal atoms, e.g., on Ag(111) a (2√3 × 2√3)R30° cell with 12 surface atoms per unreconstructed cell, the reconstruction forms a smaller 1-atom hole with C60 sitting on the hole (Fig. 3a); by contrast, (2) for a supercell with a larger number of metal atoms, e.g., on Cu(111) a (4 × 4) cell with 16 surface atoms per unreconstructed cell, the reconstruction forms a larger 7-atom hole with C60 sinking into the hole (Fig. 4). The 1-atom hole reconstruction is also formed for C60 on Pt(111) for the (√13 × √13)R13.9° cell with 13 metal atoms per cell (Fig. 3b). The reason why different types of reconstruction depend on the number of atoms per cell can be probed through a simple counting of the number of C60–metal bonds, as will be discussed later.
Figure 3 sketches 1-atom hole structures for the Ag(111)-(2√3 × 2√3)R30°  and Pt(111)-(√13 × √13)R13.9°  cases, in which C60 sits above the 1-missing-atom vacancy hole. In the 1-atom hole structures, the local bonding structures are similar for both supercells (and for both metals): C60 bonds with a hexagonal face down and with its mirror planes parallel to those of the substrate (note that for the periodic structure there is an azimuthal misalignment in the (√13 × √13)R13.9° case due to the 13.9° rotation of the superlattice results in the absence of the overall mirror symmetry). This local orientation of single C60s relative to the substrate maximizes the numbers of C-metal bonds since there is a carbon atom pointing toward each of the metal atoms surrounding the vacancy site. C60 forms six C-metal bonds with the six lower-coordinated metal atoms surrounding the vacancy site, and these bonds are much stronger than the C-metal bonds in the unreconstructed structures, where the silver atoms are surrounded by the maximum number of metal neighbors in the (111) surface. The stronger C-metal bonds in the reconstructed adsorption can compensate for the energy cost of forming holes, so that the reconstructed adsorption becomes more favored than the unreconstructed adsorption [29, 34].
Several C–metal bond lengths compared with the sum of covalent radii of C and metal atoms in the 1-atom hole (Ag, Pt) and 7-atom hole (Cu) structures, (all in Å); the bond lengths are from DFT local density approximation results
C–metal bond lengths
Sum of covalent radii
C60/Pt(111), 2√3 cell
C60/Pt(111), √13 cell
Figure 5 shows the 7-atom hole structure on Cu(111) after structural relaxation by DFT  (as confirmed by detailed LEED analysis [Ref. , and the article of Xu G et al., in preparation]). The top Cu layer is significantly distorted, resulting in alternating rotation of the hexamers and strong distortion of the hexagonal hole edge. For each C60, the six strong C–Cu bonds (to red Cu atoms) are now complemented by three weaker C–Cu bonds (to the nearest yellow Cu atoms).
The intact C60 molecule causes extensive reconstruction of Cu(111), structural stability of which derives from an intricate matching of the Cu lattice spacing, the (4 × 4) ordering, the C60 orientation, and the top Cu layer distortion. For other surfaces with a similar lattice constant as for Cu, such as the Ni(111) surface, the 7-atom hole reconstruction is also expected from the STM measured apparent height profile . More recently, the 7-atom hole reconstruction was also observed for individually adsorbed C60 molecules (at low coverage) on the so-called elbow site of the “herringbone reconstruction” of the Au(111) surface  (note that at each Au(111)'s “elbow site,” the clean reconstructed surface already has a vacancy due to a single missing Au atom ).
STM studies on Cu(111) have provided evidence for the 7-atom hole reconstruction model from the apparent-height difference of a sample prepared with both reconstructed and unreconstructed interface areas [30, 41]. Namely, in the experiment, the so-called core–shell islands were prepared with a two-stage method by first depositing C60 at 500 K (for the ‘‘core’’ with a reconstructed interface) and then C60 at 250 K (for the ‘‘shell’’ with an unreconstructed interface) . The STM topographic height of the core region is about 2 Å lower than that of the shell region. This large height difference suggests that the interface must have large enough vacancies to let a C60 ‘‘sink’’ by about ~2 Å. The number of missing Cu atoms in a (4 × 4) supercell can be estimated from a mass-flow analysis by in situ STM monitoring of C60 growth, yielding seven missing Cu atoms per cell .
The 7-atom hole structure is significantly distorted relative to bulk atomic positions, as shown in Fig. 5; this is different from the 1-atom hole reconstruction, in which only very small displacements of the metal atoms from their bulk positions are detected for both Ag(111) and Pt(111) [18, 29, 30, 31]. This “nanopatterning” of substrates is essential to the stability of the ordered structures of molecules and can critically influence their electronic structure, which is an important aspect in the design of molecular electronic devices. The influence of interface reconstruction on electronic structure will be discussed in a later section.
Origin of different reconstructions
A counter-argument in the foregoing reasoning is that with shrinking unit cell, each remaining surface metal atom loses more metal neighbors, so that the remaining metal–metal bonds become stronger with shrinking unit cell, thus favoring the 7-atom hole; however, the C–metal bonds are stronger than metal–metal bonds, which tips the balance in favor of more C–metal bonds and against more metal–metal bonds.
We further point out that the reconstruction types depend neither on the substrate electronic structure nor on the C60–metal bonding types (covalent or ionic). Indeed, the substrate electronic structures and C60–metal bonding types are similar for Ag and Cu, while they are different for Pt. The bonding types are mainly covalent for C60 on Pt, and mainly ionic for C60 on Ag and Cu. Both Ag and Cu have full d bands, while the d bands of Pt are not full; therefore, compared to Ag and Cu, Pt has a d band much closer to the Fermi level. As a result of the above comparison, we must conclude that the reconstruction types do not depend on the substrate electronic structure; we think this is because the interface reconstruction is thermally activated by higher annealing/depositing temperatures, which can change both the electronic and the structural properties of the substrate substantially, while the lattice spacing (supercell size) remains largely unchanged.
Further structural aspects
In the preceding two sections, we have emphasized the general character of C60 adsorption and metal interface reconstruction. In this section, we address several related issues, including the two C60 structures that exist on Pt(111), coexisting rotational overlayer domains with 0° versus 180° orientation for both the 1- and 7-atom hole structures, more complex structures with several or many inequivalent C60 molecules per cell, and decomposition or polymerization of C60 on metals surfaces.
Two phases on Pt(111)
C60 on Pt(111) can form two phases because of a unique combination of factors, even while the local metal–molecule binding structure remains the same: the particular value of the Pt lattice constant, 3.92 Å, coupled with a rotation by 30 − 13.9 = 16.1° of the superlattice cell from (√13 × √13)R13.9° to (2√3 × 2√3)R30° and the consequent relative rotation of the C60 molecules. These factors result in two structures that have nearly identical binding energies per surface area in spite of different inter-C60 geometries . Detailed structural analysis shows that the 2√3 phase can exist because the shortest C–C distance between two C60 molecules is reduced by only 0.1 Å relative to the √13 phase because of a better and denser packing (Fig. 7c, d), even though the supercell length (equal to the intermolecular center-to-center distance) is reduced by 0.39 Å (from 10.0 to 9.61 Å). For pure C60, the relative intermolecular orientation does not vary at low temperatures; but the orientation can change for C60 in the C60/Pt combined system, because the C60 molecules are fixed to orientations imposed by the Pt substrate via covalent bonding with Pt, while forming differently oriented superlattices in the √13 and 2√3 phases, as seen in Fig. 7.
At higher temperatures, the denser 2√3 phase will be more favored, since a higher density of holes is entropically favored at higher temperatures. This may be the reason why, at higher deposition temperature (500 K), the 2√3 phase alone (without √13) can be produced . It is suspected that when depositing at 500 K, the 1-atom hole reconstruction occurs immediately, and we get the denser 2√3 phase only, thanks to good parallel alignment of the molecules; note that the √13 unit cell length fits the bulk C60 lattice better than does the 2√3 phase . By contrast, after deposition at RT, the C60 molecules will initially have various orientations , and so they cannot easily fit in the tighter 2√3 unit cell (Fig. 7c, d). It has moreover been observed that annealing at 500 K cannot partially change the √13 phase to the 2√3 phase, while such changes may occur at even higher temperatures .
Coexisting 180° rotated domains
When a C60 molecule is rotated by 180° (or 60°) around an axis joining two opposing hexagons (such as around the surface normal in the adsorption structures discussed here and shown in Figs. 3, 4, 5, and 7), its other hexagons and pentagons are mutually exchanged: so it does not have 180° rotational symmetry. The 3-fold rotational adsorption sites on fcc(111) and hcp(0001) also do not possess 180° rotational symmetry (due to the stacking arrangement of the second and deeper layers). Therefore, rotating the molecules by 180° around the surface normal gives a new inequivalent adsorption structure. However, the energy difference between these orientations is very small, since the symmetry is broken only by atoms that are more distant from the interface. As a result, for both the 1-atom hole and 7-atom hole reconstructions, there are coexisting “domains” with 0° versus 180° orientation of the C60 molecules around their axis perpendicular to the surface. For the 1-atom hole structure, the C60 mirror planes are parallel to those of the substrate; for the 7-atom hole structure, the C60 mirror planes are perpendicular to those of the substrate: in each case, there are two nonequivalent such orientations of the C60 molecules, one as shown in Figs. 3 and 4, and one rotated by 180°. A 180° rotation of C60 around its axis perpendicular to the surface does not change the stability of the C60–metal composite structure, since the overall C60–metal chemical bonding and the inter-C60 VdW interactions do not significantly change with 180° rotation for both the 1- and the 7-atom hole structures. Therefore, DFT calculations predict that two domains with 180° rotation coexist with essentially equal probability. Evidence from STM images and LEED calculations confirms these two coexisting domains for both the 1-atom hole and the 7-atom hole structures [5, 29, 30, 31, 41]. In addition to the above inequivalent domains with 180° rotations, in the case of the (√13 × √13)R13.9° structure on Pt(111), equivalent mirrored overlayer domains with a mutual 27.8° rotation coexist, namely, rotation clockwise by 13.9°, and rotation anticlockwise by 13.9° [31, 69].
More complex structures
In addition to the ordered phases of a supercell with only one C60 molecule, more complicated or less-ordered structures are also observed, in which several or many inequivalent C60 molecules exist with different orientations within each supercell. The adsorption structures depend not only on the thermal history of the sample but they are also sensitive to the initial deposition coverage [33, 70, 71, 72]. For example, on Cu(100), STM reveals that for an initial coverage of 0.5 monolayer, C60 orders in a one-bright-and-one-dim row sequence along the  direction, whereas for a coverage close to one monolayer, C60 orders in a two-bright-and-one-dim sequence .
For C60 on Al(111), STM studies have shown that one molecule out of three protrudes from the surface upon reconstruction, and DFT calculations suggested an irreversible (6 × 6) reconstruction of the system from the (2√3 × 2√3)R30° phase [19, 20]. In the (6 × 6) structure, it is proposed that two of the three C60s in the supercell induce the 1-atom hole reconstruction, while the two removed Al atoms form an ad-dimer in the interstice below two C60s and the third C60 adsorbs without causing a reconstruction. The (6 × 6) reconstruction explains the different heights observed in STM studies, while avoiding compression of the C60 overlayer (note that in the Al(111)-(2√3 × 2√3)R30°-C60 phase, there is a 1.1% compression in the intermolecular spacing).
For C60/Ag(100), the C60 monolayer is highly inhomogeneous and aperiodic; combined STM and DFT studies reveal that the bright–dim contrast has a definite geometric origin and that there are two types of dim C60 molecules, one a monomer and the other a dimer, i.e., polymerization of C60s occurs . For C60 on Ag(111) and Au(111), in addition to the adsorption with a hexagon down, co-adsorption with a C–C 6:6 bond down is reported [42, 73]. For C60/Au(111), in addition to the 7-atom hole at the “elbow sites”  and the mix of molecules with either hexagon down (with 1-atom hole reconstruction) or C–C 6:6 bond down , many more complicated structures are reported [43, 44, 45], all based on qualitative STM experiments: for example, a (7 × 7) supercell with four C60s, and a (√589 × √589)R14.5° supercell containing 49 molecules which adopt 11 different orientations, while the monolayer structure can be divided into a faulted half and an unfaulted half similar to the (7 × 7) reconstruction of Si(111) .
According to Maxwell et al. , the C60–substrate interaction can be divided into three categories: strong predominantly covalent, intermediate predominantly ionic or covalent, and weak predominantly Van der Waals. Under thermal activation, in the case of strong C60–metal interaction, such as on Ru(0001), Pt(111), and a nickel thin film, C60 decomposes and graphene is formed [40, 46, 74]; with intermediate C60–metal interaction, such as on Cu(111) and Ag(100), C60 undergoes polymerization instead of decomposition [36, 46]. We thus expect that the strength of the C60–substrate interaction determines the final fate of the adsorbed C60 under thermal activation (i.e., annealing to higher temperatures): for weak interaction, C60 may desorb; for intermediate interaction, C60 may polymerize thanks to sufficient C60 mobility on the surface; for strong interaction, C60 may decompose since strong interaction reduces the mobility of C60 and can break the C–C covalent bonds .
The local atomic arrangements at the C60–metal contact affect not only the stability of the structure and the strength of the bonding but also the electronic structure and transport properties of the supported molecule. Many factors affect the electronic structures of the supported C60: interface reconstruction, fixed inter-C60 orientations due to bonding with surface atoms, co-adsorption with alkali and magnetic atoms, C60 polymerization, layer dependence in multilayer adsorption, etc. [29, 30, 31, 46, 55, 75, 76]. For example, by co-adsorption with alkali and magnetic atoms, the composite overlayer can be transformed from semiconducting into metallic, insulating, superconducting, and magnetic phases depending on the concentration and the nature of the dopant atoms [55, 76]. In the following, the effects of two factors on the electronic properties will be discussed in detail, namely, those of the structure of the metal–C60 interface and those of the inter-C60 interactions modified by surface bonding, since these two factors are most closely related to the structural properties discussed above. Details of the effect of interface contact on the transport properties will also be described in the following subsection.
Effects of the structure of the metal–C60 interface
C60 adsorbed on metal surfaces normally involves metal-to-C60 electron transfer which partially occupies the C60 LUMOs. For C60/Cu(111), the charge transfer has been measured to range from about 1–3 electrons per C60 by photoemission spectroscopy , while DFT calculations based on an unreconstructed interface model give a much smaller amount of charge transfer, ~0.8 electrons . From the previous discussions, we already know that the C60/Cu(111) interface structure can be either reconstructed or not depending on the thermal history of the sample. These different interface structures are of crucial importance to solve the above puzzle of various electron transfers. It emerges that the electron transfer amount does not simply depend on the substrate work function, but can depend strongly on the interface structures as well [30, 60].
The C60–metal interface structure also affects the transport properties of the adsorbed C60. For example, through careful control of the number of metal atoms (1–5 atoms) bonding to C60, it was demonstrated that the conductance increases with the number of bonding atoms . Thus, the concepts of “good” and “bad” contacts, commonly used in macro- and mesoscopic physics, can also be applied at the atomic scale. Furthermore, theoretical calculations show that the unreconstructed adsorption of C60 (coupling less strongly to the substrate) favors the appearance of a negative differential resistance compared with the reconstructed adsorption . At low coverages, due to the lack of intermolecular interactions, a larger number of local C60–metal contact configurations is observed than at monolayer coverage [33, 62, 72]: a low-temperature STM study of C60 on Cu(111) and K/Cu(111) at low coverages showed that the spatial distribution and the STS spectral width of the LUMO-derived states are particularly sensitive to the adsorbate–substrate contact ; and the spatial mapping of the spectral density of single C60 molecules on Ag(100) demonstrated the need for explicitly including the substrate interaction and the STM tip trajectory to understand the observed C60 local electronic structures .
Effect of inter-C60 interactions
Figure 9a–d shows the PDOS for the two phases both without (Fig. 9a, b) and with (Fig. 9c, d) Pt substrate, which help us understand the effects of C60–C60 and Pt–C60 interactions. Figure 9a and b for the PDOS between the two phases without Pt illustrate that the differences between the inter-C60 interactions of the two phases are reflected in differences in the peak shapes, including different peak broadening and splitting. These differences are due to the different inter-C60 orientations and distances, especially for the PDOS near the C60 “equator” (blue lines in the plots). Comparison of Fig. 9a with c [and Fig. 9b with d] for the PDOS without and with Pt shows that the PDOS is also considerably affected by the Pt–C60 interactions: the peaks are broadened and split (or partially merged in the case of the unoccupied states in the 2√3 phase). The peak positions also shift slightly. Thus, the inter-C60 interactions change the peak shape, including broadening and splitting; these changes are enhanced by Pt–C60 interactions. The different splitting of different molecular orbitals can be attributed to the different couplings for different orbitals: namely, the interactions are different for different orbitals . The strength of the C60–Pt interactions is shown most clearly through the broadening of the bottom carbon PDOS (green lines). The covalent interactions result mainly in orbital hybridization, but no significant energy shift and hence very small charge transfer: The charge transfer is small and similar for both phases, as ~0.25 electrons are transferred from the Pt surface to each C60; this contrasts with a charge transfer of about three electrons in the C60/Cu(111) 7-atom hole reconstructed system. The strong covalent Pt–C60 interactions turn the C60 monolayer into a metal (states appear just below the Fermi level), which is another difference compared with the charge transfer-induced metallization of C60/Cu(111). Also, the simulated PDOS features using the reconstructed model better fit the experimental spectroscopic data (ultraviolet photoemission spectroscopy and metastable atom electron spectroscopy data in Ref. ) than with the unreconstructed model adopted in that study.
The determination of structural and electronic properties of the interface between large molecules and metal surfaces include at least the following three important issues: (1) finding the geometrical arrangement and bonding of each atom at the interface for samples prepared under various experimental conditions; (2) accounting properly for the Van der Waals interactions between the molecules and the substrate metals; and (3) developing a quantitatively reliable description of the energy level alignment between the molecule and the substrate. Regarding the first issue: The interface structures depend on the thermal history and the kinetic progress of the sample, which should correspond to different local minima of total energy calculations. However, different local minima may have similar total energies. For example, for C60/Al(111), a vacancy-adatoms model is proposed, while for C60/Ag(111) and C60/Pt(111), the proposed models assume that the missing atoms have moved far away (e.g., to step edges). However, vacancy structures with the expelled metal bonding as an adatom with both C60 and the reconstructed surface cannot be ruled out, for both structures on Ag(111) and on Pt(111); the difficulty exists for both total energy and diffraction methods (because the adatom’s effect on the diffraction spectra is very small ). Regarding the second issue: Van der Waals density functionals with different exchange–correlation functionals can separately treat either molecules or substrate metals quite well. However, they do not perform well for the composite molecule–metal system. Recent studies demonstrate that with an appropriately chosen exchange–correlation functional, it is possible to solve this problem . Regarding the third issue: The methods need to be improved that allow correcting for the underestimation of the DFT band gap while still being applicable to the large systems which are of practical relevance .
This study was supported in part by the City University of Hong Kong Grant 9610059; the High Performance Cluster Computing Centre, Hong Kong Baptist University, which receives funding from the Research Grants Council, University Grants Committee of the HKSAR and Hong Kong Baptist University; and the CityU Centre for Applied Computing and Interactive Media.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
- 3.Van Hove MA, Watson PR, Hermann K (2002) In: Landolt-Börnstein, Group III—condensed matter, Vol 42—physics of covered solid surfaces, subvolume A—adsorbed layers on surfaces, part 2—measuring techniques and surface properties changed by adsorption. Springer, Berlin, Chapter 4.1, pp 4.1-43–4.1-164Google Scholar
- 6.Gavioli L, Cepek C (2008) Ultra-thin fullerene-based films via STM and STS. In: Bushan B, Fuchs H (eds) Applied scanning probe methods V. Springer, BerlinGoogle Scholar
- 7.Cheng HP (2010) Supported fullerenes. In: Sattler KD (ed) Handbook of nano-physics, volume of Clusters and fullerenes. CRC, Taylor & Francis, LondonGoogle Scholar
- 14.Ōsawa E (2002) Perspectives in fullerene nanotechnology. Kluwer Academic, New YorkGoogle Scholar
- 66.Somorjai GA (1981) Chemistry in two dimensions: surfaces. Cornell University Press, IthacaGoogle Scholar