Divide and Protect: Capping Gold Nanoclusters with Molecular Gold

of Applied Physics, Chalmers University of Technology, SE-41296 Göteborg, Sweden ..... Structural Information on the Au–S Interface of Thiolate...
0 downloads 0 Views 349KB Size
J. Phys. Chem. B 2006, 110, 9927-9931

9927

Divide and Protect: Capping Gold Nanoclusters with Molecular Gold-Thiolate Rings Hannu Ha1 kkinen,*,† Michael Walter,† and Henrik Gro1 nbeck‡ Nanoscience Center and Department of Physics, UniVersity of JyVa¨skyla¨, FI-40014 JyVa¨skyla¨, Finland, and Competence Center for Catalysis and Department of Applied Physics, Chalmers UniVersity of Technology, SE-41296 Go¨teborg, Sweden ReceiVed: March 30, 2006

Density functional theory calculations are used to explore phosphine- and thiolate-protected gold nanoclusters, namely, Au39(PH3)14Cl6 and Au38(SCH3)24. For Au38(SCH3)24, a novel structural motif is predicted, consisting of ringlike (AuSCH3)4 units protecting a central Au14 core. The calculated optical spectrum of this species features a large optical gap (about 1.5 eV) and a prominently peaked structure, correlating with experimental findings of “molecular-like spectra” of thiolate-protected 1.1 nm gold nanoparticles. Ligand-ligand interactions and steric effects in the ligand shell are suggested as possible driving forces toward an ordered gold core structure. A novel mechanism for ligand-exchange reactions on gold clusters is proposed.

1. Introduction Synthesis, characterization, and functionalization of sizecontrolled, ligand-stabilized gold nanoparticles are long-standing issues in the chemistry of nanomaterials.1,2 Ligand-protected gold nanoparticles offer an intriguing possibility to economically fabricate building blocks for potential applications in catalysis, sensing, photonics, biolabeling, and molecular electronics. These building blocks are expected to have unique size-dependent physical and chemical properties that have been documented in studies of size-selected gold clusters in the gas phase.3,4 Collective efforts by several groups have established a series of Au core sizes in 1-3 nm range that are predominantly formed in the process of reducing gold from metal salt in the presence of phosphines (P-R3) or thiols (HS-R).5 A profound understanding of the growth mechanism or of the physical reasons for the observed “magic” core sizes is, however, still lacking due to the complexity of the growth process and the heavy numerical burden of modeling the ligated gold clusters by stateof-the-art electronic structure calculations. Additional complication is brought by the fact that ligand-exchange reactions involving phosphines and thiolates (S-R) have been shown to modify the growth pattern and to lead to stabilization of core sizes different from that of the original compound.6-9 Moreover, high sulfur-to-gold ratios result in etching of the gold core.10 These experimental facts point to the drastically different nature of the gold/phosphine and gold/thiolate interfaces. Here we present large-scale density functional theory (DFT) calculations for Au39(PH3)14Cl6 (1) and Au38(SCH3)24 (2) (Figure 1). The explored systems have core sizes of about 1.1 nm and are chemical homologues to experimentally isolated cluster compounds. The calculations reveal a novel form of gold core protection, namely, by gold-thiolate tetraunits (AuSCH3)4; thus, compound 2 can be written as Au14[(AuSCH3)4]6 (2a). The interaction between (AuSCH3)4 and the remaining Au14 core in 2a is comparable to the phosphine-Au bond strength in 1, which is suggested to have important consequences for the * Corresponding author. Fax: +358 [email protected]. † University of Jyva ¨ skyla¨. ‡ Chalmers University of Technology.

14

260

4756.

E-mail:

Figure 1. (A-C) Structures of clusters 1, 2a, and 2b. (D) Au core of 1, shown by a 90° rotation about the horizontal axis on the left. (E) Au-S framework of 2a, with 45° rotation about the vertical axis on the left. (F) Au-S framework of 2b. Au, orange-brown; S, yellow; P, red; Cl, green; C, dark gray; H, white.

understanding of ligand-exchange reactions involving phosphines and thiolates. We also discuss the implications of different structure motifs regarding optical spectra of these species.

10.1021/jp0619787 CCC: $33.50 © 2006 American Chemical Society Published on Web 04/29/2006

9928 J. Phys. Chem. B, Vol. 110, No. 20, 2006 2. Computational Methods DFT calculations included self-consistent gradient corrections to the exchange-correlation energy, scalar-relativistic normconserving pseudopotentials (valence 11 for gold), and symmetry-unconstrained optimizations to locally stable energy minima. The atomic and electronic structure of the ligandprotected gold clusters were calculated within the DFT in combination with Born-Oppenheimer (BO) molecular dynamics (MD) (BO-DFT-MD)11 including self-consistent gradient corrections via the so-called PBE-GGA functional (PBE-GGA is a shorthand notation for Perdew-Burke-Ernzerhof Generalized Gradient Approximation).12 Au(5d106s1), S(3s23p4), C(2s22p2), and P(3s23p3) electrons were included in the valence, and the interaction to the (frozen) ion cores was described by scalarrelativistic nonlocal norm-conserving pseudopotentials devised by Troullier and Martins.13 For hydrogen a local potential was used. The Kohn-Sham (KS) states were expanded in a plane wave basis set, where convergence was found by using a kinetic energy cutoff of 62 Ry. The BO-DFT-MD method deals with an effectively nonperiodic computational cell where the periodic images of the Hartree potential are removed via a convolution method;11,14 consequently, neutral and charged systems as well as those that develop internal charge transfer or multiple moments can be treated on equal footing and accuracy. The same method was used by one of us previously to study the truncated-octahedral-symmetric Au38(SCH3)24 nanocrystal.15 In this work, the essential difference is the inclusion of the selfconsistent PBE functional during reoptimization of the structure reported in ref 15, which led to the novel structure 2a. We compare this structure to a previously suggested structure having a disordered gold core and ligand shell.16 Clusters 1, 2a, and 2b were optimized without any symmetry constraints, and the structure was deemed relaxed when the maximum force component to an atom was less than 0.001 hartree/a0. Reference calculations were performed using the BeckeLee-Yang-Parr (BLYP) approximation17 to the exchange correlation functional. This set of calculations was performed with the CPMD code18,19 using the same computational procedure and pseudopotentials as in ref 20. This implies, scalar relativistic pseudopotentials according to the method proposed by Troullier and Martins,13 plane-wave expansion of the valence orbitals, structural and electronic optimizations with direct inversion in the iterative space (DIIS),21 and periodic image decoupling.11,14 The optical spectra were calculated in the BO-DFT-MD framework from the linear-response time-dependent DFT following the formulation given in ref 22 as implemented and tested in ref 23. Recently, we applied this method to calculate optical spectra of small magnesia-supported gold clusters.24 Briefly, to get the weights FI and energies ωI of optical transitions I, one solves an eigenvalue problem ΩFI ) ωI2FI. The matrix elements contain a diagonal term that gives the energy difference of KS single-particle energies and a nondiagonal term that mixes KS particle-hole transition (ij) with (kl). Mixing is due to a coupling matrix that describes the linear response of the electron density to the single-particle-single hole excitations in the basis spanned by the ground-state KS orbitals. The convergence of the spectra (oscillator strengths and energies of the major peaks) depends on the extent of the KS particle-hole basis and on the size of the grid representing the wave functions. We have carefully tested that the spectra presented here are converged up to about 3.5 eV. In practice we used the KS particle-hole basis of about 2000 transitions. For more details see refs 22-24.

Ha¨kkinen et al. TABLE 1: Structural Information for AuCl Molecule and Gold-Phosphine Clusters (Distances in Angstroms) 〈d(Auo-Au)〉 d(Au-Au) AuCl AuPH3 AuPMe3 AuPPh3 Au39(PH3)14Cl6 a Au39(PPh3)14Cl6 b

d(Au-P)

d(Au-Cl) 2.27

3.13 c 3.04 c

2.79-3.24 d 2.69-3.04 d

2.33 2.35 2.37 2.27-2.31 2.16-2.28

2.43 2.33

a Calculation (this work, PBE functional). b Experiment for the phosphine-passivated Au39 cluster.25 c Average of distances from the central Au atom (Auo) to its 12 neighbors. d Range of the other nearestneighbor Au-Au distances.

TABLE 2: HOMO-LUMO Gaps (HLG), Optical Gaps (OG), and Local Charges of the Constituents of 1, 2a, and 2b (Calculations Done with the PBE Functional)

1 2a 2b

HLG (eV)

OG a (eV)

Au b

PH3 c

Cl d

0.8 ∼0 0.1

0.8, 1.2 1.2, 1.5 0.5, 1

+0.9 +4.4 (-0.1, +4.5) 5.8 (+0.4, +5.4)

+1.6

-3.5

SCH3 e -4.4 -5.8

a See insets to Figure 3. Two values are given, with the first one corresponding to the first (weakest) feature and the second one to the edge of a “continuous” (in the scale of the insets) absorption band. b Total for all Au atoms. Unit is the electron charge e. For (2a) and (2b), the numbers in parentheses refer to the total for the inner 14 and outer 24 Au atoms, respectively. The outer 24 atoms belong to the six (AuSCH3)4 units. c Total for all of the 14 phosphines. d Total for all of the 6 Cl atoms. e Total for all of the 24 thiolates. Note that the local charges add up to -1 in the case of the anionic cluster (1).

3. Results and Discussion The initial structure of the gold core and the positions of the P and Cl ligands of 1 were taken from an early X-ray data25 for the crystallized [(Ph3P)14Au39Cl6]Cl2 compound. Remarkably, we could not find any theoretical characterizations of its electronic structure in the literature, although the X-ray structure was reported already in the early 1990s. For the theoretical modeling, we replaced the triphenylphosphine ligands, used in the experiments, with simple phosphines. Calculations with R ) H, Me, Ph have shown that the Au-PR3 bond does not appreciably depend on the nature of the radical (Table 1; see also ref 26). Cluster 1 was optimized in its anionic state as we did not wish to consider explicit spin effects (the neutral cluster would have an odd number of valence electrons). Given the large number of valence electrons (584) in the system, the charge state is not expected to affect the atomic structure significantly (contrary to the case of a few-atom metal cluster). The X-ray structure of 1 was characterized25 as having an approximate D3(32) point group symmetry, with the gold atoms forming a hexagonal antiprismatic cage, filled by one bulkcoordinated gold atom. The optimized structure of 1 (see Figure 1) compares well with the experiment (Table 1), with all the calculated interatomic distances consistently elongated by about 0.1 Å, which is a typical result, e.g., for the Au-Au bond with state-of-art DFT functionals.27 All the ligands bind at terminal positions. Analysis of the local charge composition28 of this compound (Table 2) reveals that the phosphines act as weak electron donors to gold (σ-donation akin to CO), whereas the chlorines strongly retract electrons from the gold core. As a result, the gold core is weakly positively charged (+0.9e). Passivation by PH3 and Cl ligands opens a clear HOMOLUMO gap (HLG) of 0.8 eV, which can be contrasted to the

Capping Gold Nanoclusters with Gold-Thiolate Rings

Figure 2. Visualization of electron density difference induced upon bonding of the six (AuSCH3)4 units to the central Au14 core in cluster 2a. The cut plane includes the center of mass of the cluster. Red shows electron accumulation and blue depletion. The scale is in units of e/a03. In the ball-stick model of the cluster, Au is purple, S yellow, and C green.

HLG ≈ 0 for the Au39- gold core, where the Au positions are taken from the passivated complex. Although the “s-electron count” of 40 for Au39- would imply an electronically closedshell system, we point out that gas-phase photoelectron spectroscopy studies of gold clusters do not indicate shell closing at 40 electrons.4a The HLG for 1 is thus a collective effect due to the stabilizing ligands, the induced gold-core geometry, and the detailed charge-balance effects. We next discuss thiolate-protected cluster 2, of which structural characterization has remained experimentally elusive. The early DFT calculations, using the local density approximation (LDA), yielded a compact truncated octahedral (TO) structure as a stable isomer.15 Later, an alternative, disordered compact structure was suggested.16 In the present work, we fully reoptimized structures reported in refs 15 and 16 by using the PBE functional. A dramatic change was observed for the TO structure,15 which spontaneously opened up forming an Oh gold core of 14 atoms and six planar, ringlike gold-thiolate tetraunits (Figure 1, structure 2a). The reoptimization of the structure of ref 16 resulted in 2b with only a minor relaxation of interatomic distances. Within PBE, the two structures are very close energetically, with 2b favored by 0.5 eV; in terms of thermal vibrational energy this is only 12 K. As pointed out in ref 20, gold-thiolate systems represent a challenge for DFT calculations, because different types of chemical bonds have to be described by one single functional. The total energy of Au14[(AuSCH3)4]6 is a balance between energy terms originating from the metal-metal cohesion and the intramolecular bond strengths of the (AuSCH3)4 units. In fact, reoptimizing 2a and 2b with the BLYP functional changes the result qualitatively: With this functional, 2a is stabilized by 3.0 eV, indicating that 2a as well as 2b represent relevant competing structure motifs. It is interesting to note that the gold-thiolate tetraunits are recognizable also in structure 2b. Consequently, bonding in thiolate-protected gold clusters can be viewed as a competition between (i) maximizing the gold-gold cohesion and (ii) forming molecular ringlike gold-thiolate units on the surface of the cluster, in the expense of reducing the number of gold-gold bonds in the core. Examination of the electron density difference plot of 2a shown in Figure 2 shows that these ringlike units are

J. Phys. Chem. B, Vol. 110, No. 20, 2006 9929 bound to the central Au14 core with weak metallic bonds between the Au atoms in the units and the vertex atoms of the Au14 core. The binding energy of a thiolate radical to a gold atom, cluster or the (111) surface has been calculated to be about 2.5 eV via state-of-the-art gradient-corrected DFT methods.20,29 We calculate the Au-SCH3 bond strength to be 2.45 eV. The interaction between phosphines and gold is much weaker; for cluster 1 we calculate 0.93 eV per PH3. This difference is the underlying reason for the disparity of the structural and relaxation patterns of the metal core observed in 1 and 2. The strong Au-S interaction and the possibility to form stable (AuSCH3)4 units provide an alternative picture for the understanding of bonding in thiolate-capped Au nanoparticles. We find that the Au14-(AuSCH3)4 binding energy is 1.1 eV. This interaction is of a similar strength as binding of phosphines to the gold core (0.93 eV for the Au-PH3 bond (1); for an isolated Au-PPh3 bond we calculate 1.25 eV; see also ref 12), which points to a novel mechanism for the phosphine-thiolate ligandexchange reactions. It is often observed that the number of gold atoms in the “core” is changed during such reactions.6-9 As the binding energies of phosphines and (AuSCH3)4 units to the core in 1 and 2a, respectively, are similar, our calculations suggest that PH3 could be exchanged with (AuSCH3)4 resulting in a natural change of the number of gold atoms in this process. In particular, this qualitatively explains both the observed increase (phosphines replaced by gold-thiolate units6,7) and decrease (gold-thiolate units replaced by phosphines9) of the nuclearity of gold. The existence of stable ringlike (AuSCH3)4 structures has been previously discussed in gold chemistry.30-32 Our calculations show that these units are relevant for thiolate-protected gold nanoclusters as well. It should be noted that a successful synthesis of thiolate-capped gold nanoparticles from goldthiolate polymeric structures recently has been reported.32 We have separately studied the stability of polymeric ringlike structures (AuSCH3)N for 2 e N e 12 and found out that the cohesion per AuSCH3 building block as well as the HOMOLUMO gap (about 4 eV) are converged already at N ) 4. This implies that thiolate-capped complexes AuM-X(AuSR)X could form for different values of X. Charge analysis shows that 2a contains Au atoms in two distinct charge states: the 14 inner “core” atoms are essentially neutral, whereas the 24 Au atoms within the gold-thiolate units are positively charged. The analysis yields an electron deficiency of 0.19e per atom, resulting in a total charge transfer of about 4.5 electrons to the thiolates. A similar charge-transfer from the outermost 24 gold atoms to the thiolates takes place in 2b. We note that the total oxidation state of the gold atoms is controlled by the choice of ligand: thiolates oxidize while phosphines reduce gold. Figure 3 displays the optical absorption spectra of clusters 1, 2a, and 2b, calculated via the linear-response time-dependent DFT. The optical gaps, determined from the thresholds, are reported in Table 2. Inspection of Figure 3 and Table 2 reveals that the cluster 2a, which has the most symmetric atomic structure, also has the most distinct optical spectrum and the largest optical gap of 1.5 eV. In contrast, the disordered structure 2b has a featureless spectrum with a smooth absorption edge between 0.5 and 1 eV. “Molecular” charging and optical absorption signatures have been observed in many experiments for 1.1-1.4 nm thiolate-protected gold nanoparticles.33-39 These “molecular” spectra feature fairly large optical gaps (even up to 1.6 eV) and quite distinct spectral structures; this has been

9930 J. Phys. Chem. B, Vol. 110, No. 20, 2006

Ha¨kkinen et al. 4. Summary

Figure 3. Optical spectra of clusters 1, 2a, and 2b. The insets show the blow-ups of the regions close to the optical band edge. The spectra are composed of folded oscillator strengths (FOS) of individual optical lines. Folding is by a Gaussian with σ ) 0.05 eV.

In summary, we have shown that the different chemical nature of phosphine and thiolate ligands induces profoundly different effects on the structure and electronic properties of the gold core of ligand-capped gold nanoclusters. We have suggested a novel motif for bonding in thiolate-protected gold clusters, where AuM(SR)X is understood as AuM-X(AuSR)X, which opens new avenues to understand the properties of the small-core thiolateprotected compounds. Furthermore, it qualitatively explains mechanisms that would lead to changes of nuclearities of the gold core by (i) ligand-exchange reactions involving phosphines and thiolates or (ii) etching by thiolates. Finally, we expect that the structures published here will advance theoretical work aimed at understanding the complex problem of gold-ligand interactions.41 Acknowledgment. This work is supported by the Academy of Finland and the Swedish Energy Agency through the Competence Centre for Catalysis at Chalmers. Computations were performed at CSC in Espoo, Finland, and at PDC in Stockholm, Sweden. We thank P. Pyykko¨ and B.M. Quinn for helpful discussions. We thank Prof. I. L. Garzon for sending the coordinates of the structure described in ref. 16, which we used as the input for optimization of structure 2b. References and Notes

Figure 4. Structure of the hexanethiolate-protected Au38 cluster 3. The viewing angle is the same as that for cluster 2a in Figure 1.

suggested to imply symmetric structures (see, e.g., ref 39), which would well-correlate with our findings for cluster 2a. So far, theoretical studies of thiolate-protected gold nanoclusters have concentrated on deciphering the effects of the Au-S bond, which can be well-described by considering only the shortest possible alkylthiolate, namely, SCH3.29 However, using SCH3 as protecting ligand neglects chain-chain interactions which certainly could influence the structures formed experimentally. In the experiments, longer and/or “bulkier” protecting ligands are used, such as hexane- or dodecanethiolates (SC6 and SC12, respectively) or glutathionates. Combination of steric effects and chain-chain attraction should drive a certain degree of order within the ligand layer akin to self-assembled monolayers (SAMs),40 which should make structure motifs consisting of disordered gold cores, as in 2b, energetically unfavorable. Cluster 2a offers indeed an attractive structure motif for longer or bulkier thiolate ligands, as can be seen in Figure 4 of a model structure of the hexanethiolateprotected compound 3, Au38(SC6H13)24, now expressed as Au14[(AuSC6H13)4]6 built by attaching separately fully optimized (AuSC6H13)4 units to the central Oh Au14 core. The electronic structure of the gold core and the core-(AuSC6H13)4 bond strength are similar to those of 2a. The analysis of the effects of the longer chains on charging and capacitive properties will be published elsewhere.

(1) (a) Hayat, M. A. Colloidal Gold, Principles, Methods and Applications; Academic Press: New York, 1989. (b) Clusters and Colloids; Scmid, G., Ed.; VCH: Weinheim, Germany, 1994. (c) Nanoscale Materials in Chemistry; Klabunde, K. J., Ed.; Wiley: New York, 2001. (2) (a) Schmid, G. Chem. ReV. 1992, 92, 1709. (b) Daniel, M.-C.; Astruc, D. Chem. ReV. 2004, 104, 293. (c) For a recent comprehensive review on theoretical chemistry of gold, see: Pyykko¨, P. Angew. Chem., Int. Ed. 2004, 43, 4412; Angew. Chem. 2004, 116, 4512. (3) (a) Optical Properties of Metal Clusters; Kreibig, U., Vollmer, M., Eds.; Springer-Verlag: New York, 1995. (4) (a) Taylor, K. J.; Pettiette-Hall, C. L.; Cheshnovsky, O.; Smalley, R. E. J. Chem. Phys. 1992, 96, 3319. (b) Furche, F.; Ahlrichs, R.; Weis, P.; Jacob, C.; Gilb, S.; Bierweiler, T.; Kappes, M. M. J. Chem. Phys. 2002, 117, 6982. (c) Ha¨kkinen, H.; Yoon, B.; Landman, U.; Li, X.; Zhai, H.-J.; Wang, L.-S. J. Phys. Chem. A 2003, 107, 6168. (d) Ha¨kkinen, H.; Abbet, S.; Sanchez, A.; Heiz, U.; Landman, U. Angew. Chem., Int. Ed. 2003, 42, 1297; Angew. Chem. 2003, 115, 1335. (e) Ha¨kkinen, H.; Moseler, M.; Kostko, O.; Morgner, N.; Hoffman, M. A.; Issendorff, B.v. Phys. ReV. Lett. 2004, 93, 093401. (f) Schooss, D.; Blom, M. N.; Parks, J. M.; Issendorff, B. v.; Haberland, H.; Kappes, M. M. Nano Lett. 2005, 5, 1972. (5) (a) Early reports dealt with stabilization of the gold core by phosphines; see: Schmid, G.; Pfeil, R.; Boese, R.; Bandermann, F.; Meyer, S.; Calis, G. H. M.; van der Welden, J. W. A. Chem. Ber. 1981, 114, 3634. (b) For the so-called Brust-Schiffrin synthesis for thiolate-stabilized gold nanoparticles, see: Brust, M.; Walker, M.; Bethell, D.; Schriffin, D. J.; Whyman, R. J. Chem. Soc., Chem. Commun. 1994, 801. (c) Whetten, R. L.; Khoury, J. T.; Alvarez, M. M.; Murthy, S.; Vezmar, I.; Wang, Z. L.; Stephen, P. W.; Cleveland, C. L.; Luedtke, W. D.; Landman, U. AdV. Mater. 1996, 5, 428. (6) Balasubramanian, R.; Guo, R.; Mills, A. J.; Murray, R. W. J. Am. Chem. Soc. 2005, 127, 8126. (7) Shichibu, Y.; Negishi, Y.; Tsukuda, T.; Teranishi, T. J. Am. Chem. Soc. 2005, 127, 13464. (8) For a detailed study on kinetics of ligand-exchange reactions, see: Woehrle, G. H.; Brown, L. O.; Hutchison, J. E. J. Am. Chem. Soc. 2005, 127, 2172. (9) Wang, W.; Murray, R. W. Langmuir 2005, 21, 7015. (10) Schaaff, T. G.; Whetten, R. L. J. Phys. Chem. B 1999, 103, 9394. (11) Barnett, R. N.; Landman, U. Phys. ReV. B 1993, 48, 2081. (12) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. B 1996, 77, 3865. (13) Troullier, N.; Martins, J. L. Phys. ReV. B 1991, 43, 1993. (14) (a) Hockney, R. W. Methods Comput. Phys. 1970, 9, 136. (b) Eastwood, J. W.; Brownrigg, D. R. K. J. Comput. Phys. 1979, 32, 24. (15) Ha¨kkinen, H.; Barnett, R. N.; Landman, U. Phys. ReV. Lett. 1999, 82, 3264. (16) Garzon, I. L.; Rovira, C.; Michaelian, K.; Beltran, M. R.; Ordejon, P.; Junquera, J.; Sanchez- Portal, D.; Artacho, E.; Soler, J. M. Phys. ReV. Lett. 2000, 85, 5250.

Capping Gold Nanoclusters with Gold-Thiolate Rings (17) (a) Becke, A. D. Phys. ReV. A 1988, 38, 3098. (b) Lee, C. L.; Yang, W.; Parr, R. G. Phys. ReV. B 1988 37, 785. (18) Car, R.; Parrinello, M. Phys. ReV. Lett. 1985, 55, 2471. (19) CPMD, v3.9; Copyright IBM Corp., 1990-2001, Copyright MPI fuer Festkoerperforschung Stuttgart 1997-2001. Marx, D.; Hutter, J. Modern Methods and Algorithms in Quantum Chemistry; Forschungzentrum Juelich NIC Series, Vol. 1; FZ Juelich: Germany, 2000. (20) (a) Gro¨nbeck, H.; Curioni, A.; Andreoni, W. J. Am. Chem. Soc. 2000, 122, 3839. (b) Andreoni, W.; Curioni, A.; Gro¨nbeck, H. Int. J. Quantum Chem. 2000, 80, 598. (21) Hutter, J.; Luthi, H.; Parrinello, M. Comput. Mater. Sci. 1994, 2, 244. (22) Casida, M. E. In Recent DeVelopments and Applications of Modern Density Functional Theory; Seminario, J. M., Ed.; Elsevier: Amsterdam, 1996. (23) Moseler, M.; Ha¨kkinen, H.; Landman, U. Phys. ReV. Lett. 2001, 87, 053401. (24) Walter, M.; Ha¨kkinen, H. Phys. ReV. B 2005, 72, 205440. (25) Teo, B. K.; Shi, X.; Zhang, H. J. Am. Chem. Soc. 1992, 114, 2743. (26) Ha¨berlen, O. D.; Ro¨sch, N. J. Phys. Chem. 1993, 97, 4970. (27) Ha¨kkinen, H.; Landman, U. J. Am. Chem. Soc. 2001, 123, 9704. (28) It is clear that the absolute values of local charges depend somewhat on the method of analysis as there is no unambiguous way to assign volume to a given atom. We have used a simple scheme where the charge at a given point on the real-space electron density grid (grid spacing 0.2 a0 for all the systems discussed in this paper) is always assigned to the nearest atom analogous to defining a Wigner-Seitz primitive cell in solids. Relative changes of the local charge, e.g., upon adsorption of a molecule to a cluster, or after removing or adding an electron, are more meaningful than the

J. Phys. Chem. B, Vol. 110, No. 20, 2006 9931 absolute values. We have tested that the method used here produces similar patterns of intracluster charge distribution as other methods commonly used, e.g., Mulliken analysis. (29) Kru¨ger, D.; Fuchs, H.; Rousseau, R.; Marx, D.; Parrinello, M. J. Chem. Phys. 2001, 115, 4776. (30) (a) Bonasia, B. J.; Gindelberger, D. E.; Arnold, J. Inorg Chem. 1993, 32, 5126. (b) Bau, R. J. Am. Chem. Soc. 1998, 120, 9380. (31) (a) Pyykko¨, P. Chem. ReV. 1997, 97, 597. (b) GoldsProgress in Chemistry, Biochemistry and Technology; Schmidbaur, H., Ed.; Wiley: New York, 1999. (32) Corbierre, M. K.; Lennox, R. B. Chem. Mater. 2005, 17, 5691. (33) Chen, S.; Ingram, R. S.; Hostetler, M. J.; Pietron, J. J.; Murray, R. W.; Schaaff, T. G.; Khoury, J. T.; Alvarez, M. M.; Whetten, R. L. Science 1998, 280, 2098. (34) Quinn, B. M.; Liljeroth, P.; Ruiz, V.; Laaksonen, T.; Kontturi, K. J. Am. Chem. Soc. 2003, 125, 6644. (35) Donkers, R. L.; Lee, D.; Murray, R. W. Langmuir 2004, 20, 1945. (36) Jimenez, V. L.; Georganopoulou, D. G.; White, R. J.; Harper, A. S.; Mills, A. J.; Lee, D.; Murray, R. W. Langmuir 2004, 20, 6864. (37) Negishi, Y.; Nobusada, K.; Tsukuda, T. J. Am. Chem. Soc. 2005, 127, 5261. (38) Guo, R.; Murray, R. W. J. Am. Chem. Soc. 2005, 127, 12140. (39) Price, R. C.; Whetten, R. L. J. Am. Chem. Soc. 2005, 127, 13750. (40) Nuzzo, R. G.; Zegarski, B. R.; Dubois, L. H. J. Am. Chem. Soc. 1987, 109, 733. (41) The coordinate files for structures 1, 2a, and 2b are available through the Web at http://nano.jyu.fi/groups/compns/GoldMine or from the authors.