8864
J. Phys. Chem. A 2010, 114, 8864–8872
Theoretical Investigation of the Binding Process of Corannulene on a Cu(111) Surface† Laura Zoppi,‡ Alberto Garcia,§ and Kim K. Baldridge*,‡ UniVersity of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, and Institut de Ciencia de Materials de Barcelona, ICMAB-CSIC, Campus UAB, 08193 Bellaterra, Spain ReceiVed: March 24, 2010; ReVised Manuscript ReceiVed: June 4, 2010
DFT-GGA calculations, enhanced to include effects of dispersion, are used to investigate the adsorption process of corannulene on a Cu(111) surface. In accord with experiments, we consider the dynamics of corannulene approaching the surface in a tilted fashion, concave side-up, enabling interactions between one of the sixmembered rings and the surface over a 3-fold hollow site. Electronic structure analyses, including projected density of states and detection of work function modification, are used to aid in the understanding of the specific nature of the interaction between the corannulene and the metal surface in the complex system. Results show substantial charge rearrangement at the interface, the net effect being a large interface dipole that, added to the intrinsic molecular dipole, causes a significant decrease of the surface work function. Despite the charge rearrangement, no appreciable charge transfer occurs, and the general orbital structure of the individual components is retained. The analysis suggests that the adsorption of corannulene on Cu(111) is not a chemisorption process. Increased packing of corannulene on the surface leads to progressively smaller adsorption-induced interface dipoles, due to the depolarizing field created by the molecules. Introduction The interaction of organic molecules with metal surfaces has attracted considerable interest due to the fundamental physical and chemical phenomena involved,1-3 ranging from adsorption driven by van der Waals forces (physisorption), to binding involving charge transfer and/or covalent interaction (chemisorption). In the former process, the basic orbital structure of the organic molecule is retained upon adsorption, while in the latter there is a considerable mixing of states. It is often difficult to clearly differentiate the exact nature of the process with either experimental or theoretical means,1,4 since even in absence of covalent interactions, a structural distortion within the molecule can be accompanied by a strong charge rearrangement and shifts in energy levels.3 Organic molecules are convenient choices as they can serve as building blocks for electronic devices designed with specialized functionality, for example, application in light emitting diodes (OLEDs),5,6 field effect transistors (OFETs),7,8 and photovoltaic cells (OPCVs).9,10 From a more fundamental view, deposition of organic conjugated molecules on metal surfaces has opened important routes for investigating molecular selfassembly and molecular recognition processes.11-14 Due to the very weak nature of the interactions between noble metal surfaces and conjugated molecules (mainly dispersive), these surfaces often serve as excellent substrates to experimentally probe intermolecular interactions.1 It is clear that a thorough understanding of the complex processes at the molecule-metal interface calls for expertise that spans across domains, encompassing both experiment in organics and materials chemistry and theoretical and computational treatments. The present investigation is a continuation of efforts carried out with our experimental collaborators,15 where we focus on conformational aspects of corannulene, †
Part of the “Klaus Ruedenberg Festschrift”. * To whom correspondence should be addressed: E-mail:
[email protected]. ‡ University of Zurich. § Institut de Ciencia de Materials de Barcelona.
C20H10,16 on a copper surface. The relative ease of preparation of the 5-fold symmetric molecule and the ability to tailor its properties,17 including access to chiral derivatives,12 enables the exploration of a wealth of phenomenology when molecules are deposited on a surface. Using STM techniques, our collaborators have shown corannulene undergoing interesting supramolecular aggregation processes when deposited on either Cu(110)14 or Cu(111) surfaces.15 On the former surface, corannulene adsorbs by spontaneously forming enantiomorphous 2D lattice structures.14 While the exact positioning of the molecule on the surface is very difficult to determine experimentally, it is relatively well established14,15 that the geometry of the adsorbate viewed in the STM images corresponds to corannulene interacting with the metal with its convex side (e.g., bowl up). In the case of the Cu(111) surface, corannulene adsorbs also bowl up, but with one of the six member rings closer to the surface, with a substantial tilt angle.15 Experimental STM images also show a temperature reversible phase transition where, upon cooling, the room temperature phase contracts in a denser crystal phase, followed by rearrangements into yet another phase at lower temperatures. Successive heating restores the high-temperature phase. A greater understanding of the full dynamical process not only is desirable from a fundamental point of view but also will enable insight into how to tailor the properties of potential molecular devices. While corannulene is not the first conjugated organic molecule to be investigated on a metallic surface, the special nature of its electronic structure, including the high intrinsic dipole, as well as its dynamic properties, make it an ideal system for investigating the nature of its interaction with the metal surface in more detail.16,18-21 The large surface area of delocalized and polarizable electron densities offered by the π system, however, requires a particular sensitivity to methodology for modeling this complex process.22 As such, we employ a hybrid DFTGGA electronic structure approach, enhanced for the treatment of vdW interactions,22,23 to treat the approach of corannulene onto a six-layer Cu(111) slab. This approach minimizes, as much
10.1021/jp102662t 2010 American Chemical Society Published on Web 06/22/2010
Binding Corannulene on a Cu(111) Surface as possible, methodological compromises necessary for modeling such a large complex, while enabling accurate predictability of the electronic structure as gauged by comparison with experiments. For systematic investigation, we first focused on the case of an isolated molecule on the surface, to study the adsorption phenomenon without influence deriving from other molecules. We consider the electronic structure of the full complex during the adsorption process, the effect of electronic charge rearrangement at the interface, the formation of the interface dipole, and the modification of the surface work function. Our analysis points to a physisorption scenario for the binding of corannulene to the surface. Finally, we consider the effect on the interface dipole and the work function of an increased density of molecules on the surface. Methodological Details Surface Representation and Basis Set Issues. There are two common models used in the literature to represent a surface, the cluster or the periodic slab. A major limitation in the use of the former is that the resulting system is not periodic, but finite, and typically quite limited in size. In addition, the atoms are typically kept fixed at their bulk positions, which hampers a proper accounting of typical surface behavior such as relaxation. To avoid these problems, a periodic slab within a supercell approach is a significative improvement, provided that basic issues of slab thickness and extent of the vacuum region are properly addressed. In the present work, a six-layer slab of Cu atoms has been found to be optimal, where the two central layers are kept fixed at their ideal bulk positions, and two layers on each side are allowed to relax, minimizing artifacts in the calculation due to asymmetries in the free surfaces. The dimension of the supercell along the direction perpendicular to the slab is 38.1 Å. This corresponds to a vacuum space of about 25 Å in the z direction, large enough to accommodate the adsorbed molecule without direct interaction with the periodic replicas of the slab. The Cu(111) slab geometry in the periodic surface directions is based on appropriate repetitions of the basic 1 × 1 surface unit cell, corresponding to a theoretical bulk lattice constant a0 ) 3.665 Å. Our largest calculation involves 408 atoms, 378 of which correspond to the slab. Prior tests with a reduced three-layer slab revealed inadequacies in the ability to describe the charge rearrangement upon adsorption of the molecule. The asymmetry between the two sides of the slab due to the presence of the molecule on only one of the sides, and the intrinsic dipole moment of corannulene, results in a unit cell with a sizable net dipole moment. Since periodic boundary conditions are used, the system is subjected to a field of a periodic array of dipoles, the effects of which have to be removed (by the application of a compensating field24) before proceeding to the analysis of the electronic structure, particularly the electrostatic potential profile. Computation of the electronic structure employs the SIESTA method25 based on DFT, normconserving pseudopotentials for the description of the ion-electron interaction, and finite-range numerical pseudoatomic orbitals that are explicitly generated for each atomic species. Due to the large size of the corannulene-on-surface complex, it is necessary to strike a balance between quality vs cost of the computational description. In SIESTA, this balance includes the use of proper cutoffs for the real-space description of the charge density and potentials, as well as the choice of basis set. The latter involves considerations of size (e.g., number of orbitals per atom) and range (e.g., actual extent of orbitals in space). Our guiding principles for choice of basis sets have been the correct description of the individual pieces (i.e., the isolated
J. Phys. Chem. A, Vol. 114, No. 33, 2010 8865 TABLE 1: Details of the Basis Sets Useda subsystem molecule
atom C H
slab
Cu
Cu(surf)
basis A
basis B
DZP 2s (4.09) 2p,3d (4.87) DZP 1s,2p (4.71) DZP 4s (5.46) 4p (6.05) 3d (5.13) DZP+diff 4s (5.46), 5s (10.0) 4p (6.05) 3d (5.13)
TZP 2s (5.80) 2p,3d (7.45) TZP 1s,2p (7.02) (same as A)
DZP+diff 4s (9.3), 5s (10.0) 4p (9.3) 3d (5.8)
a For each atom, the cutoffs of the largest orbitals of each shell in bohr are given in parentheses. DZP and TZP stand for double- and triple-ζ polarized, respectively, and “diff” indicates an extra diffuse orbital.
Figure 1. Classification of corannulene C-C bonds.
TABLE 2: Calculated Geometry of Corannulene Compared to Experiment,18 Using the (revPBE-D) Level of Theory and Two Different Basis Sets, A and B (See Text) C-C hub (Å) C-C spoke (Å) C-C flank (Å) C-C rim (Å) bowl depth (Å)
basis A
basis B
experiment
1.432 1.406 1.457 1.406 0.9
1.436 1.413 1.467 1.415 0.86
1.42 1.38 1.44 1.39 0.87
molecule and the clean Cu(111) surface) with a proper equilibrium between the quality of these descriptions, so that we introduce no unwanted asymmetries that could compromise the study of the adsorbed system. For the single molecule, corannulene, we began with a standard double-ζ polarized (DZP) basis set (labeled A), consisting of 5 orbitals per H atom and 13 orbitals per C atom, using a moderate range (Table 1) for these orbitals. The geometry of the isolated molecule was optimized using a unit cell large enough (30 × 30 × 30 Å) to minimize interactions with replicas. As shown in Figure 1 and Table 2, the internal geometry of the relaxed structure is in very good agreement with experimental values. In particular, the bowl depth, which is directly related to the dynamical bowl-to-bowl interconversion process, is well reproduced at 0.9 Å compared to the experimental value of 0.87 Å.19 Moreover, the ground state electronic structure of corannulene shows the greater electron density at the inner carbons (e.g., hub of the bowl), as is known to be the case. The structure, therefore, has a relatively substantial dipole moment, experimentally found to be 2.1 D.26 However, while
8866
J. Phys. Chem. A, Vol. 114, No. 33, 2010
TABLE 3: Calculated and Experimentally Measured Properties of Isolated Cu(111) Surfacea basis set
work function (eV)
surface energy (eV)
(A) (B) exp
4.4 4.9 4.9457
0.55 0.5
a The work-function calculation used the same slab, k-point sampling ((122) k-point mesh) and mesh cutoff (200 Ry) as in the corannulene-surface complex. The surface energy has been calculated with a 1 × 1 slab over a range of increasing thickness according to the method of Fiorentini et al.57 (where the maximum slab thickness considered corresponds to 18 layers with a (1,14,14) k-point mesh).
the broad features of the electronic structure of the molecule are well represented with this basis set, the total dipole (1.0 D) is significantly lower than the experimental value. As the dipole is an essential feature of the molecule and thus important for proper description of the adsorption process, we improved the basis set, settling on a triple-ζ polarized (TZP) basis with longerrange orbitals (basis B, Table 1). This leads to a dipole moment of 2.08 D, in very good agreement with the experimental value. Other structure and property predictions are not substantially different from those computed with the original basis. For the metallic slab, we followed earlier work of Garcı´aGil et al.27 on noble metal surfaces. They found that augmentation of the basis set of the surface atoms with a shell of diffuse orbitals, or the addition of one or two layers of floating orbitals in the vacuum next to the surface, dramatically improves the surface energy, the work function, and the wave function decay shape, all of which are critically involved in the electronic structure of the surface. Accordingly, we have employed a moderate-range DZP basis for the Cu atoms (4s, 4p, and 3d orbitals, considering the 3d states in the valence), with an extra 5s diffuse shell with a range of 10 Bohr for the atoms on the topmost layers of the slab (Table 1). This basis set (A) gives acceptable values for the work function and surface energy, as shown in Table 3. One further check is warranted to validate these basis sets for their use in the description of the adsorbed system. Of particular concern with systems of this type are effects due to basis set superposition error (BSSE), which manifests itself when two monomers of a complexed system (the “complex”) have incomplete basis sets, resulting in some sharing of each others’ basis functions. This error becomes particularly important when analyzing the charge transfer and other details of the electronic structure involved in bonding. Therefore, our analysis of the BSSE focuses on the degree of charge imbalance (spurious dipole) introduced by the presence of the basis orbitals of the other monomer. Our tests use the relative positions of corannulene and slab in the actual binding arrangement. A calculation of the net dipole of the slab augmented with the corannulene orbitals gives a sizable 1.5 D, caused by the tendency of electrons to spill out in the direction of the extra orbitals. This indicates that the slab basis set, despite being sufficient to describe clean-surface properties, would be inadequate to analyze the changes in the distribution of charge brought about by adsorption. Consequently, we found it necessary to improve the slab basis set. In this case, the cost of a triple-ζ basis set is prohibitive; however, an increase in the range of the orbitals of the Cu atoms in the topmost layer is actually enough to bring the spurious dipole down to 0.02 D. This new basis set for the slab (B in Table 1) is sufficient to study the charge rearrangement upon adsorption and also leads to better agreement with the experimental value of the work function (Table 3).
Zoppi et al. In addition to the above BSSE considerations, conventional evaluation of BSSE has been carried out using the Boys-Bernardi counterpoise correction.28 With this procedure, the BSSE estimated value for our system corresponds to 0.7 eV. The analysis of basis set convergence presented above highlights the kinds of issues and trade-offs that appear when trying to investigate large systems near the limits of current computational practice. In what follows, our analysis will be carried out using the balanced B bases. All calculations are carried out with a real-space grid cutoff of 200 Ry, and a k-point sampling equivalent to a cutoff of 18 Å,29 which results in just two k-points for this large system. Empirical Dispersion Corrections. We have employed a DFT-D method for including vdW interactions into a standard density functional. In particular, the RevPBE30 exchangecorrelation potential as implemented in the SIESTA code is supplemented with the empirical dispersion correction of Grimme,23 given by
Edisp ) -s6∑
Cij6
i
