J . Phys. Chem. 1993,97, 11611-11616
11611
Structural Investigations of Au55 Organometallic Complexes by X-ray Powder Diffraction and Transmission Electron Microscopy W.Vogel,' B. Rosner, and B. Tescbe Fritz-Haber-Institut der Max-Planck-Gesellschaft, 141 95 Berlin, Germany Received: May 1 1 , 1993: In Final Form: August 17, 1993'
Two samples of Au~s(PPh3)lzCl6organometallic clusters prepared by different methods were investigated by X-ray powder diffraction (XRD). The measured diffraction curves were fitted by a set of Debye functions for AuNwith the "magic" numbers N = 13,55,147, ... for cuboctahedra and icosahedra; N = 54,181 for decahedra; and N = 13, 57, 154 for hexagonal close-packed clusters. It was clearly shown that most A U ~ ~ ( P P ~ ~ ) ~ ~ C clusters do not have an fcc structure, but a good fit was achieved using the icosahedral model. The nearestneighbor distance is 0.278 nm, Le. -3% smaller than that of the bulk gold material (0.288 nm); however, the center of the first coordination shell is uncontracted within the limits of error (0.5%). In the sample that was not filtered by an anotop filter, an additional maximum in the small-angle range was observed. It is attributed to a disordered fcc-type supercluster of (Au55)55 with a cluster-cluster distance of 2.50 nm. The diffraction curve could only be explained by a nonuniform cluster size, verified by T E M micrographs. About 40 wt % of the material belongs to clusters with nuclearities N larger and smaller than 55. Occasionally, TEM reveals large (-40 nm), two-dimensional, densely stacked arrays of clusters. The lattice fringes of 2.14 nm are in very good agreement with the value 2.50 X (3/4)'/* expected from XRD.
Introduction In the present work, X-ray powder diffraction (XRD) combined with simulationcalculations using Debye functions is applied for the first time to theA~ss(PPh~)~2C&cluster molecule. Previously, this 'Debye function analysis" (DFA) has been successfully used for the structural characterization of silica-supportedplatinum particles in the standard catalyst EUROPT-1.' The ligandstabilized AUSScluster is an ideal candidate for DFA on account of the negligible contribution of the ligand shell to the scattered X-ray intensity. Theoretical calculations favor an icosahedral arrangement of the naked cluster. In conflict with this, most experimentalwork published so far has been interpreted in terms of a cubic close-packed structure of the AUSSnucleus with cuboctahedral morphology. Gold organometallic complexes of different sizes have been intensively studied by various methods (EXAFS, Fairbanks et a1.,2Marcuset al.;aTEM, Fauthet a1.,4Schmidet al.;sMBssbauer optical spectroscopy, Fauth et al.;4 spectroscopy, Smit et SIMS, Feld et al.;' PDMS, Fackler et a1.8). Mingosg carried out a general theoretical analysis of bonding in cluster compounds of gold and predicted the possibility of synthesizingicosahedral one-shell gold clusters (13 atoms). In 1981 such clusters ([A~I~(PP~M~~)~OC~I](PF~)~) were prepared by Briant et a1.I0 as very dark red crystals, which were suitable for single-crystal X-ray analysis. An icosahedralstructure was found, but showed some significant deviations from the idealized icosahedral geometry. Two-shell gold clusters ( 5 5 atoms) were first synthesized and described by Schmid et al." in 1981. In contrast to the A~13organometallic complex, the two-shell clusters cannot be crystallized, and therefore a single-crystal X-ray analysis is not possible. Schmid et al.11 deduced a cuboctahedral structure for this cluster from Mbsbauer spectra and geometrical considerations (number of ligands). Later investigationsalso indicated an fcc structure (TEM, Wallenberg et a1.;12 EXAFS, Fairbanks et al.,2'Marcuset a 1 3 The main argument used for the deduction of an fcc structure from the results of EXAFS measurements is the good correlationof the determined coordinationnumber with the theoretical coordination number of an Au55 cuboctahedron. *Abstract published in Advance ACS Abstracts. October 1, 1993.
0022-3654/93/2097-11611%04.00/0
However, this is not direct evidence that the cluster is cubic (Marcus et al.3). Another type of cluster structure, namely, vertex-sharing 13atomicosahedra, is proposed by Teo et al.I3 Multiples of (Au13),+ and (A~13),,-have been shown to appear in the SIMS spectra of but have been controversially discussed in refs 7 and 8. On the other hand, the formation of densely packed [(M1p)l3],,(clusters of clusters) with M = Au, Rh, Ru, and Pt was suggested after electrochemical decomposition of larger, ligand-stabilizedM ~molecules s (Schmid et aI.I4). Recently, Teo et al.I5 have synthesized crystals of an Au39 organometallic complex, which is built up of two subunits with a modified hexagonal close-packed (hcp) structure. From a theoretical point of view, the energetics and structures of metal clusters have been studied by Yi et a1.'6 using ab initio Car-Parinello calculations for 13 and 55 atoms with AI chosen as a paradigm. For the 55-atom cluster several inequivalent but energetically nearly degenerate structures are found. It is important to note that the calculated structure factors of the ideal radially relaxed cuboctahedron and icosahedron are very different from those of the fully annealed structure. The annealing always led to icosahedral structures with very similar structure factors. Debiaggi et al.I7have studied the atomic configurations of A1147 in a semiempirical approach and have obtained results analogous to those of Yi et al. Sawada and SuganoI8 performed lifetime calculations of AUSSand AU147 clusters by use of the transition-state theory. They conclude that the AUSScuboctahedral structure has too short a lifetime to be observed around room temperature.
Theoretical Method The so-called Debye function analysis (Gnutzmannand Vogell) was used to determine the size distribution and structure of the clusters. The Debye function, N
D(b) =
fJm sin(2ubrm)/2rbrm
(1)
n.m- 1
(b, length of the diffraction vector, b = 2 sin 8 / h ; 8, diffraction angle; A, wavelength;f, atomic scattering amplitude of an atom; r,,,,,;distance between the atoms n and m ) , describes the average 0 1993 American Chemical Society
Vogel et al.
11612 The Journal of Physical Chemistry, Vol. 97, No. 45, 1993
structure factor of a large number of identical, randomly oriented clusters of N atoms in the kinematical approximation. The functionDN,db) was used to calculate the average structure factors of monodisperse clusters of full-shell cuboctahedral, icosahedral, decahedral, and hcp clusters. Nearly globular hcp clusters may be constructed by stacking 3, 5,7, ... (1 11)-netplanes with (3,7, 3), (7, 12, 19, 12, 7), (12, 19, 27, 37, 27, 19, 12), .,. atoms, respectively. Thermal vibrations, statistical deviationsfrom the average atom position, and a constant background intensity must, in addition, be taken into account for the simulation of the diffracted intensity. The theoretical intensity function I t h of a mixture of clusters can then be calculated by
- Au,,
a)
----
icosahedron vertex-sharing icosahedron
(Au,,),
A
.-Y
- Au,, cuboctahedron
$’-;*
C
----
Au,, Au,,
----
Au,,
a
4
icosahedron decahedron
v
.-b In ~ f b ) ’ ( 1 - exp(-~M)))
+ c (2)
C
P)
.-C
((YN,~, factor describing the number fraction of the cluster; N,
number of atoms of the cluster, 13,55, 147,309,561,923, 1415 (cuboctahedra and icosahedra), 54,18 1 (decahedra), and 13,57, 154(hcp); T, typeof cluster (cuboctahedra, icosahedra,decahedra, hcp); M(b) = Bb2/4, Debye parameter B = 8r2(u 2 ) ; ( u 2 ) ,mean square linear amplitude corresponding to thermal vibrations and static statistical deviations from the average atom position; C, background intensity). The parameters CYN,T,B, and Care used as free parameters. They are determined by a comparison of theoretical and experimental diffraction curves in order to minimize the R factor (R= - Zexp)’/n) using a nonlinear least-squares fit routine. For simplicity, and to reduce the number of free parameters, only full-shell clusters were used for the simulation. Generally speaking, the weight factors obtained by DFA serve as settling points of the real size distribution. Discontinuous and continuous sizedistributions are practically indistinguishablein the diffraction pattern (Gnutzmann and Vogel’). The calculation of the Debye function of vertex-sharing icosahedra was only performed with three Au13 complexes in order to simplify the calculations. It is very difficult to find typical differences between the Debye functions of the Au13 icosahedra and vertex-sharing icosahedra (see Figure la). Therefore, it is not possible to recognize Aul3 vertex-sharing icosahedra by the DFA method, and it is sufficient to use normal Aul3 clusters for the fit. The Debye function of the Auss(PPh3)&16 cuboctahedral cluster was calculated with and without ligands in order to test theinfluenceof theligands on thediffractionp a t t e d 9(see Figure Ib). Significant differences could only be found in the range of small angle diffraction (Figure 2a). Therefore, the approximation of naked clusters is satisfactory for simulation of the wide angle diffraction pattern. The Debye functions of the AUSScuboctahedra and icosahedra show some typical differences (see Figure 1b), and therefore both structures can clearly be distinguished by their diffraction pattern. The differences can be generalized for other cluster sizes, e.g. the Debye function of the fcc structure has two maxima between b = 6.5 nm-I and b = 9 nm-1 corresponding to the reflections (220) and (31 1,222) (see Figure lb) in contrast to only one maximum of theicosahedral structure. On theother hand, it is moredifficult to distinguish the Debye functions of icosahedra and decahedra (Figure lb). The Debye function of an icosahedral supercluster consisting of 55 Auss icosahedral clusters shows typical oscillations (produced by a so-called superlattice factor) around the Debye function of a simple AUSScluster (shape factor) (Figure IC). Disorder in the supercluster damps out these oscillations with increasing length of the diffraction vector (Figure IC). Therefore, the Debye function in the wide-angle range equals that of the simple Auss cluster (Figure IC). In the small-angle range, however, the first
icosahedron supercluster (Au,,),, ;; f\ Y\,, - supercluster (AI+,),,
c)
9 2
6 b = 2sin Olh (nm-l)
4
- ideal - statistica
deviations
8
Figure 1. Debye function of different clusters in the wide-angle X-ray
scatteringrange (WAXS):(a) Aul, icosahedron, (Aul3)3vertex-sharing icosahedron; (b) Auss(PPh3)&& cuboctahedron, Auss cuboctahedron, Auss icosahedron, Au54 decahedron; (c) Auss icosahedron, supercluster icosahedron of 55 Aus~icosahedral clusters with a nearest-neighbor distance between the clusters of 2.5 nm, supercluster with a statistical modulated nearest-neighbordistance.
-
Au,,(PPh,),,CI, cuboctahedron - Au,, cuboctahedron
---- (Au,,),, - (Au,,), **** (Au,,),
0.2
icosahedron icosahedron cuboctahedron
0.4 0.6 0.8 b = Psin O/h (nm-1)
1.0
Figure 2. Debye function of different clsuters in the small-angleX-ray scattering range (SAXS): (a) Auss(PPhs)l~C&cuboctahedron, Auss cuboctahedron;(b) different superclusterswith a nearest-neighbordistance ~ between the clustersof 2.5 nm, that is, icosahedronof 13 A U Sicosahedra, icosahedronof 55 Auss icosahedra,cuboctahedronof 55 Au55 icosahedra. supercluster peak remains (Figure 2b) and is amplified by the increasing shape factor of the simple cluster (Figure 2a).
Experimental Section All X-ray diffraction patterns weremeasured with a commercial Guinier diffractometer (HUBER) in the 45O transmission
Auss Organometallic Complexes
The Journal of Physical Chemistry, Vol. 97, No. 45, 1993 11613
Figure 3. TEM micrograph of the Auss clusters. (a) Inset showing the particle sizes counted from this figure (open bars) and the sizes obtained from XRD (shaded bars). (b) Inset showing an area of planar ordered clusters (circle) and the corresponding optical diffraction of this area.
geometry combined with a Johansson-type Ge monochromator to produce a focused Cu Kal primary beam. Two different types of Au55(PPh3) clusters were investigated.20 Both samples were synthesizedaccordingto the Schmid method (Schmid et al.ll), where ( C ~ H ~ ) ~ P AisUreduced C ~ by B2H6 in benzol. Sample 2 was additionally filtered by an anotop filter. While sample 1 is the raw material, sample 2 consists of gold clusters embedded in a foil of polystyrene (thickness 0.1 mm) with 5 wt % gold. The sample 1 powder was fixed during most measurements between two foils of polyethylene (thickness 3 rm). Two new specimenswere prepared from sample 1 by contacting it to an organic solvent: (i) dissovled in dichloromethane followed by immediate drying onto a 7.5-pm Kapton foil (sample la); (ii) dissolved in pyridine (50 mg/g) and sealed into a Mark capillary (sample lb). Low-temperature measurements and some of the roomtemperature measurements of sample 1 were carried out in high vacuum, and the powder was fixed between two thin aluminum foils in order to obtain good thermal contact. The measurements at low temperatures were performed with a continuous-flow cryostat (Oxford Instruments). In each case, two measurements were performed: one measurement on the supported gold clusters and the other measurement on the pure support material only. The diffraction curves of the pure gold clusters were obtained by subtracting the support measurement using a suitable scaling factor. The data have been corrected by the angular-dependent absorption, geometry, and polarization factors. Transmission electron microscopy (TEM, Siemens Elmiskop 102) was performed on a suspension of sample 1 in dichloromethane after immediate drying on an amorphous carbon foil with a thickness of 5 nm supported on a 400 mesh/inch copper grid. In some experiments, uranyl acetate (2% in CH2C12) was used to stabilize the cluster distribution. Results TEM Measurements. A typical TEM micrograph of clusters in sample 1 is shown in Figure 3. In Figure 3a, a histogram of the cluster sizes obtained from this micrograph is given (broad
-6
;
:
6
6
10
b (nm-l)
Figure 4. XRD pattern (without angular-dependent correction factors) of (a) sample 2, anotop filtered, 5 wt ?6 supported in polystyrene; (b) sample 1, raw material (powder); (c) sample la, sample 1 dissolved in CH2C12 and dried on Kapton. Curves b and c are plotted with an offset of 100 and 200 units, respectively.
bars). Besides seeminglyunstructured bulky aggregates (arrows), accidentally large (=40 nm) ordered areas of clusters are seen in the TEM pictures (Figure 3b, circle) with lattice fringes of 2.14 nm. Most probably the former are three-dimensional, and the latter are two-dimensionally ordered domains of cluster molecules. The lattice fringesappear to be structured like “strings of pearls”, and individual cluster molecules can be seen. The optical diffraction of this area (inset to Figure 3b) shows a poorly developed6-fold symmetry. This pattern is expected for a planar, dense packing of spheres. Similar observations have been made in previous TEM ~ o r k . However, ~,~ the 6-fold symmetry has not been proven so far. For larger gold colloid particles, such hexagonal 2D structures are well established.21 Wide-Angle X-rayScattering (WAXS). XRD intensityprofiles are shown in Fiugre 4 for sample 2 (curve a), sample 1 (curve b), and dichloromethane-contacted and dried sample 1a (curve c). The errors are on the order of the size of the dots. No major differencesare seen for samples 1 and 2, except for some structure appearing in the range of the second maximum of sample 2. For
Vogel et al.
11614 The Journal of Physical Chemistry, Vol. 97, No. 45, 1993 100
ICOSAHEDRA
8 0 4 -
6on
s
v
80
-
3 h
3
-
CUBOCTAHEDRA
s
Y
5 .-
60
2
40
E
20
I
I
0 '
I
n
IC)
60-
v
0 80
.i3 v)
C
E .-
40-
20
4
DECAHEDRA
0
-
J
' 1
I
I
3
4
5
I
I
J I
1
6 7 8 9 b (nm-') Figure 6. Corrected XRD pattern (dots) of sample 1b, dissolved in pyridine and DFA-fit (solid line). Inset: Mass frequencies used for the fit (open bars = cuboctahedra; shaded bars = icosahedra). HCP
40
20
0
4
5 h
6 0-z-
nn
7 I..-.I\
8
0 9
-r-r
0:5 110 1.5 2.0 2.5 diameter (nm)
Figure 5. Comparison of the experimental diffraction curve of sample 1 (including angular-dependent corrections) and simulated diffraction curves in the WAXS range (room temperature) and the corresponding mass fractions of the clusters: (a) only icosahedra, R = 0.067, B = 1.1 ( 1 t 2 nm2); (b) only cuboctahedra, R = 0.252, B = 2.4 ( l e 2 nm2); (c) only decahedra, R = 0.085, B = 1.2 ( 1 t 2 nm2); (d) only hcp clusters, R = 0.094, B = 1.5 ( 1 t 2 nm2).
the dissolved and dried sample la, some tiny extra peaks (marked by arrows) indicate the presenceof a few bulk-like gold particles.22 First, we shall discuss the simulations carried out for sample 1. All of the fits are shown in Figure 5. The best fit was achieved using icosahedra only. The simulated curve compared with the experimental curve and the corresponding cluster distribution are shown in Figure 5a. The mass fraction is maximal for the two-shellcluster. However, according to this analysis, about half of the gold atoms are not in the AUSSclusters. The situation changes if we are looking for the number frequencies. Here the fractions of the larger three- and four-shell clusters are less than 5%. In all cases, however, the number fraction of the one-shell A~13cluster is largest (=60%) and drops nearly exponentially for the higher-shell clusters. (The number fractions obtained by XRD are plotted in Figure 3a as shaded bars for comparison.) If, for instance, theAu13cluster is excluded from the fit, deviations from the experiment are far beyond the experimental limits of errors. The same is true if the two-shell AUSScluster is forced to the fcc symmetry, the others being of any type of local symmetry tested in this work. In any case, a good fit was achieved using five parameters, namely, a1, ..., a4 for icosahedra, and B (C = 0, cf. eq 2). As already mentioned above, the applied DFA is not proof for a discrete size distribution, but clusters with intermediate nuclearities other than 13, 55, and 147 may also be present. Independent of this, it can be clearly seen that it is impossible to obtain a good fit using only cuboctahedra (Figure 5b). The use of both icosahedra and cuboctahedra for the simulation improves the R factor but not the quality of the fit. In addition, we have performed simulations using decahedra and hcp clusters
(Figure 5c, d). The simulated intensities of both structures are comparable, but the fits are inferior to those using icosahedra. The fit in Figure 5a was achieved with a nearest-neighbor distance of 0.278(1.5) nm, which is smaller than the nearestneighbor distance of the bulk material (0.288 nm). The Debye parameter is B = 1.16( 1.5) (1 t 2nm2). The published values for bulk gold at room temperature range from B = 0.55 to B = 0.64 ( nm2). It is interesting to note that the fit using hcp clusters uses a larger nearest-neighbor distance (0.286 nm) compared to the icosahedral model, and its size distribution is nearly monodisperse (57-atom clusters; mass fraction 86%). A low-temperature experiment was performed for sample 1 in vacuo at 8 K and, as a reference, at 298 K. The same structural results were obtained (best fits using only icosahedra with a size distribution similar to the measurements in air). The thermal contraction of the cluster, estimated from the shift of the first peak of the diffraction curve, is 4 . 4 ( 1)% and is comparable to the value expected from bulk gold. However, the Debye parameters are different: using icosahedra only, at room temperature in vacuo, B = 1.78(1.5) (10-2 nm2), and at 8 K, B = 1.43(1.5) ( nm2). The bulk Debye parameter of gold at nm2). 10 K is B = 0.08 Figure 6 shows the diffraction curve of the pyridine-dissolved sample 1b after background and intensity correction (dots). Evidently, the primary cluster molecules are destroyed and most of the material (87% from DFA) consists of larger fcc particles. For the simulated curve (solid line in Figure 6), icosahedra and/ or cuboctahedra have been used. An improvement of the fit could be achieved by introducing twin faults to the fcc particles. The corresponding size distribution obtained from DFA is displayed in the inset of the figure. The fcc lattice spacing and the Debye parameters are a0 = 0.404 nm and B = 0.57 (le2 nmz), respectively, and approach the value of bulk gold. Small-Angle X-ray Scattering (SAXS). In the range of smallangle X-ray scattering, an additional maximum was observed, corresponding to the diffraction of superclusters (see inset to Figure 7). This peak is most developed in the original powder of sample 1. The peak becomes broader and less intense after dissolution/dryingtreatment in CH2C12. No SAXS peak appears, however, in the anotop-filteredsample 2 or for sample 1b, dissolved in pyridine. From the position of this maximum a nearestneighbor distance of 2.50(3) nm was determined, which is a little greater than the diameter of the Auss(PPh3)12C16complex of 2.2 nm given by Schmid et al.5 This secondary structure is highly disordered, as demonstrated in Figure 7. Three typical simulations are performed for the
The Journal of Physical Chemistry, Vol. 97, No. 45, 1993 11615
Au55 Organometallic Complexes
the intensity rises steeply at a certain temperature to a saturation value. The first derivative of the signal (dashed line) is related tothegrowthrate. It hasamaximumat 147 OC. Thediffraction curve after this treatment (maximum temperature 175 "C) and then rapid cooling shows broadened lines indicative of fcc gold particles. From the analysis of the width of the first five lines accordingto the method described in ref 22, the average crystallite size is 15.5 nm. The gold particles are distorted by stacking faults (1.65%) and by internal strains (317 N/mm2). We also observed a series of narrow lines at low angles. This phase could be identified as crystalline Ph3PAuCl according to the singlecrystal data given by Baenziger et After heating to 235 "C, these lines disappear completely. 0
1.5
1
0.5
2
Discussion and Conclusions
b (nm-')
Figure 7. Three SAXS simulations of sample la (solid lines) in a
momentum plot (dashed lines represent the respective shape factors): (a) shape factor of the cuboctahedral (PPh,)l2AussCl6, lattice factor of ideal cuboctahedron; (b) shape factor for a log-normal distribution of spheres,lattice factor of ideal cuboctahedron, (c) shape factor for a lognormaldistributionof spheres,latticefactor of a disordered cuboctahedron with 20%local displacements;dots represent experimentaldata of sample la. Curves a and b are plotted with an offset. The inset shows a comparison of the SAXS of sample 1 (solid line) and sample la (dots).
12-
s
0
z
8-
____-.-I I -.a'
4 1 100
120
I
140 160 T ("C)
*-.-
I
180
Figure8. Open-slit intensity measurement of sample la around the Au(1 11) Bragg peak versus temperature (solid line). Rate = 20 OC/min. The derivative of this signal (dashed line) is related to the growth rate of Au particles.
SAXS of sample la:(a) shape factor of the Au55(PPh3)12C16 molecule (dashed line) times the superlattice factor of a perfect cuboctahedron consisting of 55 simple clusters (solid line); (b) modified shape factor taken from a log-normal distribution of spheres with an average diameter larger than the naked Au55 cluster; and (c) same as b but the superlattice factor calculated for 20% local displacements of the spheres from an ideal fcc arrangement. The last curve gives good agreement with the experimental curve (dots). The shape factor used in curve a has a second maximum at the wrong position, and consequently the higher-order peaks of the superlattice factor are unreasonably amplified. The latter peaks are suppressed, however still present in curve b but damped out in curve c. For sample 1, a larger averagesizeof thesuperclusters must be assumed, but theclustercluster distance remains unchanged. Thermal Stability. We have observed the thermal decay of the cluster molecules by heating sample l a at a constant rate (20 OC/min) in a 200-mbar He ambient. In Figure 8 the total scattered intensity at tJ = 19.1O measured within the angular window of l o around the Au(ll1) peak is plotted versus temperature. As a result of a nucleation-and-growth process,
The X-ray measurements have shown that the majority of Au55(PPh3)&16 clusters cannot be simulated by an fcc structure but a good fit could be achievedusing icosahedra. For comparison, we refer to an XRD study using the DFA method on silicasupported platinum (standard catalyst EUROPT- 1):l Here the most frequent cluster size is in the range of Pt55, and the majority show clear fcc symmetry (less than 10%icosahedra). This result is in contrast to earlier considerations (e.g. Schmid et al.11): (i) The first argument raised in favor of the cuboctahedral structure was the surface topography which allows for a highly symmetric arrangement of the ligands, i.e. 12 PPh3 bonded to the vertices and six chlorine bonded to (100) microfacets. However, for the Rh complex Rh~5[P(?-Bu)3]12Cl20,stoichiometric arguments do not conflict with an icosahedral cluster nucleus with 12 P(?-Bu)3 bonded to the vertices and 20 chlorine at 3-fold hollow sites of (1 11) microfacets. In this context, the question may be raised as to whether well-defined adsorption sites are necessary for the stabilityof these high-nuclearityorganometalliccomplexes. In an NMR study, Schmid et al. observed a high mobility of the ligands and/or the rhodium at0ms.~5 Briant et a1.lO have shown that for the icosahedral gold cluster compound [Aul~(PMezPh) &lz] (PF& in the crystallinestate the chlorideligands occupy para vertices. However, dissolved in CD2Cl2 at room temperature the para isomer isomerizes into ortho and/or meta isomers. For steric reasons, the nearly globular shape of the icosahedronwould allow for rotational freedom of the metal nucleus against the ligand shell as a whole. (ii) From Mossbauer studies performed on Au55,6Jl~~~ good agreement is achieved with the cuboctahedral model originally proposed by Schmid." Unfortunately, no attempts have been made to describe the spectra in terms of icosahedral Au55. In a recent publication,26 three different Au55 cluster compounds are compared with surprising changes in the quadrupole and isomer shifts of the various Au sites (13 "bulk"-sites, 24 unbonded and 18 ligand-bonded surface sites). The cited Mossbauer investigations have not discussed the influence of a possible -40% fraction of the gold located in clusters smaller or larger than Au55. (iii) The same restrictions are valid for EXAFS.2*3 Indeed, the coordination number K = 7.8(1) found by Marcus et ala3is close to the expected value for a cuboctahedron of 55 atoms. But a non-monodisperse size distribution will modify this number. The coordination number averaged over a variety of sizes then gives no conclusive arguments with respect to a cubic (K= 7.85) or icosahedral (K = 8.51) structure of Au55, respectively. In a review article of the title molecule, Thiel et al.27 argued that the EXAFS Fourier transforms show only one single peak and therefore the Auss nucleus must be cubic: For a closed-shell icosahedron, the EXAFS peak should be split, since the radial nearest-neighbor distances (84 for Mess) are smaller than those oriented tangentially (150 for Mess).2*.29 From the view of practical EXAFS evaluation, however, the observation of a single peak in the Fourier transform does not necessarily exclude the
11616 The Journal of Physical Chemistry, Vol. 97, No. 45, 1993
existenceof two close-lying distances with the proper coordination number ratio. Their separation is even rather unlikely.30 (iv) Wallenberg and Bovin12 have imaged a microcrystal approximately 1.2 nm across, resembling the proposed cuboctahedral model viewed along (1 10). Particles of this size are, however, instable when imaged with 200-kV electrons and subjected to changes in size and morphology. With less energetic electrons, it has so far not been possible to definitely exclude an icosahedral model of the Au55 cluster molec~le.~’ However, it could be shown that for certain preparations cuboctahedra of this size were found.32 There may be alternative reasons for the differences; for example, we used ideal cuboctahedral, icosahedral, decahedral, and hcp structures. However, the external shell of the Au55 clusters is influenced by the ligands, and therefore distortions are probable.10 An indirect proof of this is the large Debye parameter which must be used for the fit, even at 8 K, because its timeindependent part describes static statistical deviations from the ideal structure. The ratio between the Debye parameters of the clusters and the bulk material (=l:2) qualitatively agrees with earlier EXAFS results (Marcus et 1:1.6). From the temperature dependence of their EXAFS amplitudes, a 40% reduced thermal vibration of the cluster gold atom was deduced by these authors. The gross vibrational properties have been explained in terms of a combination of surface-induced compression and softeningdue to surface modes. In our measurements at 300 and 8 K, we obtain a thermal contribution to the Debye parameter of AB = 0.35 (10-2 nm2) compared to the bulk value of AB = 0.47 (10-2 nm2). This is an additional argument for the “stiffening” of the cluster relative to the bulk.3 The lowtemperature measurements prove the nonexistence of a lowtemperature transition from icosahedral to fcc symmetry. The measurements in air and in vacuo showed nearly the same results aside from a slight increase of the Debye parameter. We conclude from this that the Schmid clusters in the original powder are stable in vacuum, i.e. desorption of ligands is negligible. Thermal desorption of the ligands, followed by a rapid growth, was observed by in situ XRD. After heating to a maximum temperature of 175 OC in a 200-mbar He ambient, bulk Au particles 15.5 nm in size are formed, plus an additional phase identified as mononuclear Ph3PAuCl. The rate of growth has a sharp maximum at 147 O C . 3 3 This result is in accordance with differential scanning calorimetry performed by Benfield et al.,34 who observed a narrow exothermic peak at 156 OC. In samples 1 and la, but not in the anotop-filtered sample 2, the presence of a three-dimensional supercluster could be detected by SAXS. After short exposure to CH2C12 (sample la), these secondary structures still remain but decrease in size. They can be assigned to distorted cubic densely packed (Au55)55 superclusters. Thesuperlatticeconstant is A0 = 2.50 X 2l/2nm = 3.54 nm. The observed SAXS clustersluster distance of 2.50 nm agrees nicely with the two-dimensional lattice observed by TEM: If the same near-neighbor distance is assumed for a twodimensional superstructure, a fringe distance of 2.50 X (3/4)1/2 nm = 2.165 nm would be expected. The observed TEM value is 2.14 nm. The measured contraction of the clusters (nearest-neighbor distance 0.278 nm, accuracy 0.5%) in relation to the bulk material (0.288 nm) agrees reasonably well with earlier EXAFS results (Fairbanks et al.,z 0.276 nm, 0.278 nm; Marcus et al.,3 0.2803 nm). As mentioned above, for an icosahedral cluster the secondnearest-neighbor distance is only 5% apart, and therefore both distances contribute to the first coordination shell. More reasonably, the center of gravity of both distances has to be compared with the bulk value. It is about 3% larger than the nearest-neighbor distance (3.6% for Me13,2.97%for Me3w),i.e. 0.287 nm. For the much smaller icosahedral A ~ 1 organometallic 3 cluster, the two nearest-neighbor distances obtained from single-
Vogel et al. crystal X-ray analysis1° range from 0.2716 to 0.2789 nm and from 0.2852 to 0.2949 nm, respectively. Here the weighted average is 0.286 nm, a value 1%smaller than that of bulk gold. It has been pointed out by Hansen et al.35that EXAFS may give rise to artificially contracted interatomic distances for small metal particles.
-
Acknowledgment. The authors are indebted to G. Schmid, Institut fur Anorganische Chemie der Universitlt Essen, for the kind provision of the specimens and critical discussions. We gratefully acknowledge Prof. J. Urban, Fritz-Haber-Institut, for delivering the atomic coordinates of the icosahedra and decahedra. We also thank Mrs. U. Klengler, for her skillful assitance with the microscopy. Dr. B. Herrschaft, Freie Universitlt Berlin, helped in the identification of the X-ray structure of PhSPAuCl. References and Notes (1) Gnutzmann, V.; Vogel, W. J . Phys. Chem. 1990,94, 4991. (2) Fairbanks, M. C.; Benfield, R. E.; Newport, R. J.; Schmid, G.Solid State Comm. 1990, 73, 431. (3) Marcus, M. A.; Andrews, M. P.; Zegenhagen, J.; Bommannavar, A. S.; Montano, P. Phys. Rev. 1990, BIZ, 3312. (4) Fauth, K.; Kreibig, U.; Schmid, G. 2.Phys. 1991,020,297;Z. Phys. 1989,012, 515. ( 5 ) Schmid, G.; Klein, N.; Korste, L.; Kreibig, U.; Schbnauer, D. Polyhedron 1988, 7, 605. (6) Smit, H. H. A.; Thiel, R. C.; de Jongh, L. J.; Schmid, G.;Klein, N. Solid State Comm. 1988, 65, 915. (7) Feld, H.; Leute, A.; Rading, D.; Benninghoven, A.; Schmid, G. Z. Phys. 1990, D17, 73. (8) Fackler, J. P.; McNeal, C. J.; Winpenny, R. E. P.; Pignolet, L. H. J. Am. Chem. SOC.1989, 111, 6434. (9) Mingos, D. M. P. J. Chem. Soc., Dalton Trans. 1976, 1163. (IO) Briant, C. E.; Theobald, B. R. C.; White, J. W.; Bell, L. K.; Mingos, D. M. P.; Welch, A. J. J. Chem. Soc., Chem. Comm. 1981,201. (1 1) Schmid, G.;Pfeil, R.; Boese, R.; Bandermann, F.; Meyer, S.; Calis, G . H. M.; v.d. Velden, J. W. A. Chem. Ber. 1981, 114, 3634. (12) Wallenberg, L. R.; Bovin, J. 0.;Schmid, G. Sur/. Sci. 1985, 156, 256. (13) Teo, B. K.; Hong, M. C.; Zhang, H.; Huang, D. B. Angew. Chem., Int. Ed. Engl. 1987, 26, 897, (14) Schmid, G.; Klein, N. Angew. Chem. 1986, 98,910. (15) Teo, B. K.;Shi, X.; Zhang, H. J. Am. Chem.Soc. 1992,1/4,27432745. (16) Yi, J. Y.;Oh, D. J.; Bernholc, J. Phys. Rev. Lett 1991, 67, 1594. (17) Debiaggi, S.; Caro, A. Phys. Reo. 1992, 846, 7322. (18) Sawada, S.; Sugano, S . 2.Phys. 1991, D20, 259. (19) We gratefully acknowledge Prof. G. Schmid for providing the coordinates. (20) The two samples were prepared by G. Schmid, Essen. (21) Giersig, M.; Mulvaney, P. J. Phys. Chem. 1993, 96,6334. (22) It must be noted that sample l a was prepared from sample 1 after 20 months of storage in air at room temperature. (23) Vogel, W.; Tesche, B.; Schulze, W. Chem. Phys. 1983, 74, 137. (24) Baenziger, N. C.; Bennett, W. E.; Soboroff, D. M. Acta Crysrallogr. 1976. B32. 962. (25) Schmid, G.; Giebel, U.; Huster, W.; Schwenk, A. Inorg. Chim. Acta 1984, 85, 97. (26) Mulder, F. M.; v. d. Zeeuw, E. A.; Thiel, R. C.; Schmid, G.Solid State Commun. 1993, 85, 93. (27) Thiel, R. C.; Benfield, R.E.; Zanoni, R.; Smit, H. A. A.; Dirken, M. W. Struct. Bonding 1993,81, 1. (28) The authors refer to an EXAFS work on an Aull-cluster compounds where the two distances are resolved as separate peah. The structure of Aul can be assigned as an incomplete icosahedron. From single-crystal XRD it is knwon that the average tangenital and radial nearest-neighbor distances of this cluster are different by the much greater amount of 12% as compared to 5% for closed-shell icosahedra. (29) Cluskey, P. D.;Newport,R. J.;Benfield, R. E.;Gunnan,S. J.;Schmid, G. Z . Phys. D, in press. (30) Haase, J. Private communication, Berlin. (31) Urban, J. Private communication, Berlin. (32) Urban, J.; Sack-Kongehl, H.; Weiss, K.; Goyhenex, C. Electron Microsc. (EUREM 92, Granada, Spain) 1992, 2, 679. (33) In the XRD experiment. the temperature of the He ambient is measured, which is slightly lower than the specimen temperature, depending on the rate of heatinn. (34) Benfield, R . k Creighton, J. A.; Eadon, D. G.;Schmid, G. 2.Phys. 1989, DI2, 533. (35) Hansen, L.B.; Stotze, P.; Nerskov, J. K.; Clausen, B. S.; Niemann, W. Phys. Rev. Lett. 1990, 64, 315.
