Structures of Neutral Au7, Au19, and Au20 Clusters in the Gas Phase

See allHide authors and affiliations

Science  01 Aug 2008:
Vol. 321, Issue 5889, pp. 674-676
DOI: 10.1126/science.1161166


The catalytic properties of gold nanoparticles are determined by their electronic and geometric structures. We revealed the geometries of several small neutral gold clusters in the gas phase by using vibrational spectroscopy between 47 and 220 wavenumbers. A two-dimensional structure for neutral Au7 and a pyramidal structure for neutral Au20 can be unambiguously assigned. The reduction of the symmetry when a corner atom is cut from the tetrahedral Au20 cluster is directly reflected in the vibrational spectrum of Au19.

Haruta et al.'s finding that dispersed gold nanoparticles show pronounced catalytic activity toward the oxidation of CO has triggered a gold rush in cluster chemistry (1). Although bulk gold is a classic example of chemical inertness (2), many later studies have confirmed the size-dependent reactivity of deposited gold clusters (36). Small particles of gold differ from the bulk because they contain edge atoms that have low coordination (7) and can adopt binding geometries that lead to a more reactive electronic structure (8). Thus, the secret of the catalytic properties of gold nanoparticles lies at least partly in their geometric structure. Structural information for deposited gold mono- and bilayers on titania has been obtained by using high-resolution electron energy–loss spectroscopy on CO adsorbates (9). Determining the three-dimensional (3D) structure of deposited gold nanoparticles is more challenging, but recently has been achieved for clusters containing around 310 atoms by means of aberration-corrected scanning transmission electron microscopy (10).

The geometry of nanoparticles can also be studied in the gas phase. The advantages of this approach are the exact knowledge of the clusters' size and the absence of any interaction with the surrounding environment. The properties of such well-defined species can thus be modeled very precisely with quantum-mechanical calculations. Different experimental techniques exist for the study of free clusters. By measuring the mobility of size-selected gold anions and cations in helium, a transition from 2D to 3D structures has been found (11, 12). This transition appears at different cluster sizes for cations and anions and is yet to be determined experimentally for neutral species. A combination of photoelectron spectroscopy and quantum-mechanical calculations has revealed fascinating structures of anionic gold species—for example, cages for clusters containing 16 to 18 atoms (13), a tetrahedral pyramid for Au20 (14), and a possibly chiral structure for Au34 (15). These structural motifs have been confirmed by measurement of the electron diffraction pattern of size-selected trapped anions (15, 16). Although ion mobility measurements, photoelectron spectroscopy, and trapped-ion electron diffraction have all substantially added to the understanding of the geometric properties of free nanoparticles, these methods are restricted to the investigation of charged species.

We investigated neutral gold clusters in the gas phase by means of vibrational spectroscopy, which is inherently sensitive to structure. In infrared (IR) absorption spectroscopy, the number of allowed transitions is restricted by selection rules, and thus directly reflects the symmetry of the particle. Far-IR multiple-photon dissociation (FIR-MPD) spectroscopy is a proven technique for obtaining the vibrational spectra of gas-phase metal clusters and, hence, by comparison with calculated spectra, their geometries (17, 18). It is the only technique for determining the structure of free metal clusters that is not limited to charged species. We explored three representative sizes of neutral gold clusters. With Au7, we investigated the structure in a size region in which the anions and cations are known to adopt planar structures, and we thereby addressed a controversy in theoretical studies. With Au20, we confirmed that the neutral cluster retains the symmetrical pyramid geometry established for the anion. With Au19, we directly observed the reduction of symmetry when one of the corner atoms is removed from Au20.

Details of the technique of FIR-MPD have been described elsewhere (17, 19). Neutral gold clusters are produced by means of laser vaporization from a gold rod in a continuous flow of He and Kr (1.5% Kr in He) at 100 K. Under these conditions, complexes of the bare metal clusters with one or two Kr ligands are formed. The molecular beam is overlapped with a pulsed FIR beam delivered by the Free Electron Laser for Infrared eXperiments (FELIX) (20). The neutral complexes are ionized by an F2-excimer laser (7.9 eV per photon) and mass-analyzed in a time-of-flight mass spectrometer. Resonance of the FIR light with an IR-active vibrational mode of a given neutral cluster may lead to the absorption of several photons. The subsequent heating of the complex results in the evaporation of a loosely bound Kr ligand and a depletion of the corresponding mass spectrometric signal. Recording the mass-spectrometric signal while scanning the wavelength of FELIX leads to depletion spectra, from which absorption spectra Math are reconstructed (21).

Figure 1A shows the vibrational spectrum of neutral Au7 obtained with FIR-MPD of its complex with one Kr ligand. A number of bands were found in the region between 47 and 220 cm–1, usually having a full width at half maximum of less than 4 cm–1 (21). This is close to the spectral bandwidth of FELIX, which is about 2 to 3 cm–1 and nearly constant over the whole tuning range. The number of peaks implies a rather nonsymmetric structure for neutral Au7. The geometry of Au7 was established by comparing the experimental spectrum with the calculated vibrational spectra for multiple isomers predicted by density functional theory (DFT) calculations within the generalized gradient approximation (21, 22). We found a planar edge-capped triangle with Cs symmetry (Iso1) to be lowest in energy. This structure has been previously reported as the global minimum (23). A hexagonal planar structure (Iso2) (24) and a 3D structure (Iso3) (25) have been proposed as lowest-energy structures as well, but were computed to be higher in energy in the present study.

Fig. 1.

Vibrational spectra of neutral Au7. (A) The FIR-MPD spectrum of Au7Kr. The red dots represent relative cross sections, Embedded Image, the average of up to ∼1000 single laser shots at a fixed frequency, whereas the black line interconnects a binomially weighted five-point running average, thus accounting for the spectral bandwidth of the IR laser. It is compared with the spectra of four low-lying isomers (Iso1 to 4) (C to F) obtained by DFT calculations. The peak positions in the experiment are in best agreement with a planar structure of Cs symmetry (Iso1). Including the Kr ligand in the calculation does not substantially change the structure of the cluster or the positions of the resonances (Iso1·Kr) (B), but has an effect on the IR intensities, which become very similar to those in the experimental spectrum.

The experimental vibrational spectrum unambiguously tested the reliability of the theoretical methods. The calculated spectra were distinctive in the range between 150 and 220 cm–1. The peak positions of Iso1 fit with experimental absorptions at 165, 186, and 201 cm–1. Only the relative intensities of the bands did not agree completely; the central band at 186 cm–1 was much more pronounced in the experiment. Figure 1B shows the calculated absorption spectrum of the complex Iso1·Kr, in which Kr is bound to the energetically most favorable position of the Iso1 cluster with a bond dissociation energy of 0.09 eV. The positions of the resonances were not changed, but the relative intensities were substantially affected. The three bands were then in excellent agreement with the experiment, and, furthermore, all absorptions between 50 and 150 cm–1 became more pronounced (26). The calculations show that all of these vibrational modes are highly delocalized and involve the motion of all atoms in the cluster [see normal mode displacement vectors for the three highest-energy vibrations of Au7 (all in-plane) in fig. S3].

In principle, multiple isomers can be present in the molecular beam, in which case the spectrum would represent a superposition of their individual contributions. Iso2 has one strong absorption at 185 cm–1, and its presence in the molecular beam could also explain the pronounced intensity of the central peak in the experimental spectrum. Ion mobility measurements, however, excluded the presence of major amounts of additional isomers for cationic and anionic Au7. For experimental modes that were not present in the calculated spectrum of hexagonal Iso2—for example, at 165 cm–1 and 201 cm–1—the mass-spectrometric intensity of Au7Kr went down to below 30% of its original value when irradiated by FELIX, which set the upper limit for the abundance of other isomers. Therefore, although a minor contribution of Iso2 could not be ruled out, the capped triangle could be assigned as the dominant structural isomer of neutral Au7 present in our experiment.

On comparison with the experimentally determined structures of the corresponding ionic species, we found that Au7 is a cluster size, which changes its geometries for each charge state. Although the cation is highly symmetric and corresponds to the D6h structure Iso2 (12), the anion forms a threefold edge-capped square (11). We found this structure to be a saddle point in our calculations for neutral Au7 that relaxes into Iso1. The geometrical change as a function of cluster charge corresponds to a lessening of the average coordination as the electron density increases. Although the gold atoms in the cation have on average 3.43 nearest neighbors, this value decreases to 3.14 and 2.85 for the neutral and the anion, respectively. With additional electrons, the clusters favor more open structures.

Having shown that the experimental spectrum, in combination with theory, can be used to identify the geometry of the Au7 cluster, we moved on to bigger sizes. Photoelectron spectroscopy and quantum-mechanical calculations have shown that anionic Au20 is a pyramid and has Td symmetry (14). This structure has also been suggested to be the global minimum for neutral Au20 (14). The FIR-MPD spectrum we measured of the Au20Kr complex (Fig. 2A) was very simple, with a dominant absorption at 148 cm–1, which already pointed to a highly symmetric structure. The calculated spectrum of tetrahedral Au20 was in agreement with the experiment (Fig. 2C, and see fig. S4 for the IR spectra of less stable isomers). In the FIR-MPD spectrum of Au20Kr2, weaker features occurred at low frequencies and were well reproduced by theory when the Kr ligands are included in the calculations (fig. S1). The strong absorption at 148 cm–1 corresponds to a triply degenerate vibration (t2) in bare Au20 with Td symmetry.

Fig. 2.

Vibrational spectra of neutral Au19 and Au20. (A and B) The FIR-MPD spectra of Au20Kr and Au19Kr, respectively. (C and D) Calculated spectra for pyramidal geometries Au20 and Au19 are in excellent agreement with the experiment. The splitting of the degenerate resonance at 148 cm–1 for Au20 into two peaks at 149 cm–1 and 166 cm–1 for Au19 is due to the lowering of symmetry when going from the tetrahedron to the truncated pyramid.

Theory predicts a truncated trigonal pyramid to be the minimum energy structure for neutral Au19 (27), for which the removal of a corner atom of the Au20 tetrahedron reduces the symmetry from Td to C3v. As a direct consequence, the degeneracy of the t2 vibration of Au20 is lifted, and this mode splits into a doubly degenerate vibration (e) and a non-degenerate vibration (a1) in Au19. This splitting was observed in the vibrational spectrum of neutral Au19 (Fig. 2). The e vibration lies at 149 cm–1 and is hardly shifted as compared with the t2 mode of Au20. The a1 vibration is blue-shifted by 18 cm–1 relative to the t2 vibration in Au20. The truncated pyramidal structure of Au19 can thus be inferred directly from the IR spectrum. We also found the C3v structure of Au19 to be a minimum in our calculations, and the calculated vibrational spectrum fits the experimental one in terms of peak positions and relative intensities (Fig. 2D, and see fig. S5 for IR spectra of less-stable isomers). Again, the modifications in peak intensities induced by the Kr ligands agree well between theory and experiment (fig. S2).

We have shown that detailed structural information on small neutral gold nanoparticles can be obtained by means of vibrational spectroscopy. FIR-MPD is the only size-selective experimental technique available to date that allows for the structure determination of neutral metal clusters in the gas phase. It can be used to study the transition of 2D structures to 3D structures for neutral gold clusters as well as to study, for instance, ligand-induced geometrical modifications that are highly relevant in catalysis. With improved sensitivity for heavier masses, it will be possible to extend these measurements to larger clusters. Although spectral congestion may prohibit detailed analysis of non-symmetric structures, we can expect symmetric and near-symmetric structures to remain identifiable, as they do for Au20 and Au19. For instance, it should be possible to answer questions such as, is neutral Au55 icosahedral, or does it have a low-symmetry structure such as has been indicated for the anion (28)?

Supporting Online Material

Materials and Methods

Figs. S1 to S5

References and Notes

Stay Connected to Science

Navigate This Article