Preparation, characterisation and performance of encapsulated copper-ruthenium bimetallic catalysts derived from molecular cluster carbonyl precursors

Article abstract of DOI:10.1002/(SICI)1521-3765(19980710)4:7<1214::AID-CHEM1214>3.0.CO;2-E

The advantages of producing high-performance, bimetallic nanoparticle catalysts from their precursor metalcluster carbonylates anchored inside the mesoporous channels of silica (MCM41) are described. In situ X-ray absorption and FT-IR spectroscopies as well as ex situ high-resolution scanning transmission electron microscopy were used to chart the progressive conversion, by gentle thermolysis, of the parent carbonylates to the denuded, bimetallic nanoparticle catalysts. Separate copper and ruthenium K-edge X-ray absorption spectra yield a detailed structural picture of the active, approximately 15 A diameter catalyst: it is a rosette-shaped entity in which twelve exposed Ru atoms are connected to a square base composed of relatively concealed Cu atoms. These in turn are anchored by four oxygen bridges to four Si atoms of the mesopore lining. The bimetallic catalysts exhibit no tendency to sinter, aggregate or fragment into their component metals during use. The nanoparticles perform well in the catalytic hydrogenation of hex-1-ene-a detailed kinetic study at 373 K and 20 bar H₂ is presented here (maximum TOF in [(mol(substr)) (mol(cluster))^-1 h^-1] 51200, average TOF 22 400)-and also in the hydrogenations at 65 bar H₂ and 373 K of diphenylacetylene, phenylacetylene, stilbene, cis-cyclooctene and D-limonene, the average TOFs being 17, 610, 70, 150 and 360, respectively.

Full text of DOI:10.1002/(SICI)1521-3765(19980710)4:7<1214::AID-CHEM1214>3.0.CO;2-E

FULL PAPER
Preparation, Characterisation and Performance of Encapsulated
Copper± Ruthenium Bimetallic Catalysts Derived from Molecular Cluster
Carbonyl Precursors
Douglas S. Shephard, Thomas Maschmeyer, Gopinathan Sankar, John Meurig Thomas,*
Dogan Ozkaya, Brian F. G. Johnson,* Robert Raja, Richard D. Oldroyd, and
Robert G. Bell
Abstract: The advantages of producing
high-performance, bimetallic nanoparti-
cle catalysts from their precursor metal-
cluster carbonylates anchored inside the
mesoporous channels of silica (MCM41)
are described. In situ X-ray absorption
and FT-IR spectroscopies as well as ex
situ high-resolution scanning transmis-
sion electron microscopy were used to
chart the progressive conversion, by
gentle thermolysis, of the parent carbon-
ylates to the denuded, bimetallic nano-
particle catalysts. Separate copper and
ruthenium K-edge X-ray absorption
spectra yield a detailed structural pic-
ture of the active, approximately 15 Š
diameter catalyst: it is a rosette-shaped
entity in which twelve exposed Ru
atoms are connected to a square base
composed of relatively concealed Cu
atoms. These in turn are anchored by
four oxygen bridges to four Si atoms of
the mesopore lining. The bimetallic
catalysts exhibit no tendency to sinter,
aggregate or fragment into their compo-
nent metals during use. The nanoparti-
cles perform well in the catalytic hydro-
genation of hex-1-eneÐa detailed kin-
etic study at 373 K and 20 bar H2 is
presented here (maximum TOF in
1
[(molsubstr)(molcluster
)
1 h ] 51200, aver-
age TOF 22400)Ðand also in the hydro-
genations at 65 bar H2 and 373 K of
diphenylacetylene,
phenylacetylene,
stilbene, cis-cyclooctene and d-limo-
nene, the average TOFs being 17, 610,
70, 150 and 360, respectively.
Keywords: bimetallic nanoparticles
´ clusters ´ copper ´ heterogeneous
catalysis ´ hydrogenations ´ ruthe-
nium
Introduction
With the advent of mesoporous materials, the science and
catalytic potential of oxide supports (especially siliceous ones)
has undergone a revolution.[1±3] These silicas, the surface areas
1
of which often exceed 1000 m2 g , possess a highly regular
structure composed of channels the diameter of which may be
controllably arranged to fall within the range 15 to 80 Š
(Figure 1). The larger pore sizes of these silicas (lined with
pendant silanol groups), in contrast to the smaller ones of
high-silica faujasitic[4] and UTD-1[5, 6] zeolites or metal alumi-
nophosphates like DAF-1[7] and STA-1,[8] offer the possibility
[*] Prof. Sir J. M. Thomas, Dr. T. Maschmeyer, Dr. G. Sankar, Dr. R. Raja,
Dr. R. D. Oldroyd, Dr. R. G. Bell
Davy Faraday Research Laboratories
The Royal Institution of Great Britain
Figure 1. Computer graphic view along the axis of a single mesopore in
MCM-41 silica (pore opening ꢀ 30 Š). The profusion of pendant silanol
groups lining the pore function as the anchoring points for the adsorbed
species. (Colour code: yellow, silicon; red, oxygen; white, hydrogen).
21 Albemarle Street, London, W1X 4BS (UK)
Fax: (‡44)171-629-3569
Prof. B. F. G. Johnson, Dr. D. S. Shephard, Dr. T. Maschmeyer
The University Chemical Laboratories
Lensfield Road, Cambridge, CB2 1EW (UK)
of creating isolated, catalytically active sites in a struc-
turally well-defined manner[9±11] at the inner walls of the
mesopores.
Prof. Sir J. M. Thomas, Dr. D. Ozkaya
Department of Materials Science, University of Cambridge
New Museum Site, Cambridge, CB2 3QZ (UK)
1214
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1214 $ 17.50+.50/0
Chem. Eur. J. 1998, 4, No. 7

1214±1224
Bimetallic nanoparticles themselves possess catalytic prop-
erties that differ sharply from those of solid solutions of the
bulk metals. This was one of the key conclusions to emerge
from the early work of Sinfelt,[12] and its veracity has been
repeatedly demonstrated by subsequent workers. Among the
many studies of bimetallic catalysts some are of great
technological importance (e.g. Pt ± Re, Pt ± Ir and Pt ± Rh) in
processes such as the reforming of naptha and the control of
auto emissions.[13] Of late, the advantages of using mixed-
metal carbonyl clusters as starting materials for the produc-
tion of efficient bimetallic nanoparticle catalysts have been
appreciated.[14±21] For example, in Re ± Ir[19] and Ag ± Ru[20]
systems, it is possible to follow the gradual conversion of the
cluster precursor to the active catalyst using a variety of in situ
(X-ray absorption and FTIR spectroscopies)[22, 23] and gravi-
metric techniques.
We are concentrating our discussion in this area on
large carbonyl clusters in which the metals have many
metal ± metal connectivities since our aim is to form bi-
metallic particles with uniform size and composition from
such a species. Hence, in our approach, we are deliberately
avoiding the less readily controllable surface-mediated
synthesis, which relies on the aggregation of smaller cluster
precursors as exemplified by the work of, for example,
Gates et al.[16, 17]
The current interest in the use of bimetallic catalysts is
largely due to the enhanced activity and selectivity that may
be achieved by two metals working synergistically. However,
despite considerable effort, the preparation of these systems
by metal salt deposition and subsequent calcination followed
by reduction to the active low-valent species is not always
reliable and has many drawbacks, the greatest of which is the
lack of control over size, morphology and homogeneity of the
resulting bimetallic nanoparticle. These factors and the rather
random nature of the support lead to loss of selectivity and/or
activity of the catalyst, thereby rendering it in many instances
inefficient.
When considering the more precise route of depositing and
thermolysing carbonyl clusters for the production of support-
ed nanoparticles, several important criteria concerning the
choice of bimetallic cluster precursor have to be borne in
mind. First, the protective sheath surrounding the organo-
metallic precursor must be readily removable (we find that
mild thermolysis suffices). Second, to ensure a uniform
distribution of the precursor over the surface of the support
the ligand sheath must interact favourably with the functional
groups on the surface, the solvent and the counterion.
Moreover, interactions with the surface must be stronger
than those involved in solvation or between the precursor
species so that aggregation into small molecular crystallites
and subsequent sintering on the surface is suppressed upon
removal of the CO sheath. Anionic cluster carbonyl species,[24]
typified by [Ru6C(CO)16Cu2Cl]2[PPN]2 and [Ag3Ru10C2
Scheme 1. The deposition and thermolysis of carbonyl clusters for the
production of supported nanoparticles.
Over the last few years great progress has been made in the
synthesis of large (>10 metal atoms) bimetallic cluster com-
pounds of Group VIII [Fe,Ru,Os] and XI [Cu,Ag,Au] metals
at the Cambridge laboratories.[24] Clusters composed of
different metal ratios (M1/M2 ˆ 1 to 4) and nuclearities
(M1 ‡ M2 ˆ 8 to 30) have been synthesised. These anionic
molecular cluster species have all been structurally charac-
terised by single-crystal X-ray diffraction and so their precise
nuclearities and metal ratios are known.
Three multicentred mixed-metal molecular cluster carbon-
yls that have recently been used as precursors for heteroge-
neous catalysts are those studied by Nasher et al.[14, 19] and
Shephard et al.[20] In the former instance, where the clusters
were [PtRu5C(CO)16] and [Re6C(CO)18{m3-Re(CO)3}{m3-
Ir(CO)2}]2 , it was found that in the case of the Re ± Ir cluster
the metals segregated into separate entities upon thermolysis,
possibly because of the nature of the oxide support used. In
the case of the Ru ± Pd cluster aggregation rather than
segregation occurred, producing relatively uniform particles
(reported as between 8 ± 23 Š in size) with bimetallic
composition and a fcc packing arrangment. When the
cluster anion [Ag3Ru10C2(CO)28Cl]2 (1) was used,[19] there
was evidence neither for segregation nor for aggregation/
sintering upon thermolysis, and we believe this to be for three
reasons. First, the mesoporous silica support onto which
the cluster carbonylate is initially anchored is replete with
silanol groups that interact strongly with the carbonylate
anion. Second, the anionic character of the clusters keeps
them apart during the initial stage of deposition. Third, the
relative oxophilicity of the silver atoms makes them an ideal
(CO)28Cl][PPN]2
(PPN ˆ bis(triphenylphosphonyl)immin-
ium), fulfil these criteria as their interaction with the silica
(MCM41) surface is of the Si ± OHd‡ ´´´ d O ± C ± M type[25]
and intermolecular Coulombic repulsion prevents their ag-
gregation prior to thermolysis. A representation of this
approach is given in Scheme 1.
Chem. Eur. J. 1998, 4, No. 7
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1215 $ 17.50+.25/0
1215

J. M. Thomas, B. F. G. Johnson et al.
FULL PAPER
bonding centre securing the bimetallic cluster to the oxide
support.
(molSalt/molMCM41) 5:6.35 for [Ru12C2Cu4Cl2(CO)32][PPN]2/
MCM41 by weight, giving rise to an approximate coverage
of about one cluster per 50000 Š2.
In this paper, extending this latter philosophy, we report the
detailed synthesis of the bimetallic Cu ± Ru carbido carbonyl
[Ru6C(CO)16Cu2Cl]2[PPN]2 (2) and its adsorption onto the
inner walls of well-defined[26, 27] mesoporous silica specimens
known as MCM41. The resulting Cu ± Ru bimetallic system is
found to be potentially a more powerful heterogeneous
catalyst than its Ag ± Ru analogue. Apart from gaining deep
insights into the nanostructures and morphology of the
resulting catalysts using a combination of annular dark-field
(ADF) and bright-field (BF) high-resolution scanning trans-
mission electron microscopy,[20, 28] we tracked the precise
structural details of the conversion of this precursor material
into its active catalytic state principally by using the in situ X-
ray absorption techniques developed at the Davy Faraday
Research and Daresbury Synchrotron Radiation Laborato-
ries.[22, 23, 29±32] We were able to gain accurate local structural
information for the catalyst by the application of these
element-specific techniques (i.e. for Cu and Ru). The X-ray
absorption near-edge structure (XANES) investigation pro-
vides the electronic state and qualitative local structural
information, whereas the extended X-ray absorption fine
structure (EXAFS) study establishes the precise local struc-
tural details, thus revealing the internal structure of the
nanoparticle. Confirmation of the loss of the protective
carbonyl sheath during thermolysis was found by in situ FT-
IR and further corroborative evidence was established by
thermogravimetric analysis.
This catalytic precursor material was characterised by FT-
IR, revealing that the carbonyl cluster was still intact. The IR
spectroscopic profile of the CO stretching vibrations of the
MCM41-contained clusters is considerably more complex
than that of the same clusters in DCM solution (absorp-
1
tion maxima at 2060, 2021 and 1845 cm
for the
[Ru12C2Cu4Cl2(CO)32]2 cluster). Furthermore, the new ab-
sorptions (in both the terminal and bridging sections of the
spectrum) show a shift to lower energy, suggesting interaction
with the surface silanol groups (Figure 2). These observations
Results
Catalyst preparation: The bimetallic cluster [Ru6C(CO)16-
Cu2Cl]2[PPN]2 (2) was synthesised by a similar, but improved,
route to that described in the literature[33] in which the
hexaruthenium carbido carbonylate cluster is reacted with the
CuI salt [Cu(MeCN)4BF4] and subsequently with bis(triphe-
nylphosphonyl)imminium chloride (PPNCl). We found that
direct reaction of an excess of CuCl in refluxing tetrahydro-
furan (THF) gave better yields and was simpler to work up.
The MCM41 silica was loaded with cluster 2 by making a
slurry of the two components in a mixture of ether and a small
amount of a second solvent, dichloromethane (DCM), in
which the cluster salt is highly soluble. This general method
and solvent combination was found to be the most effective of
several different preparative routes. The small amount of
DCM is thought to act as a carrier, ferrying the cluster salt
through the liquid phase to the silica surface, making use of
entropic and enthalpic driving forces for the anions to enter
the channels of the MCM41 by substitution for loosely held
solvent molecules. On contact, the solvation sphere (predom-
inantly of DCM) is substituted by the silanol-rich silica
surface, allowing the diffusion of solvent molecules from the
mesopores, thereby freeing them to solvate more cluster and
to repeat the deposition cycle. We found this method allowed
maximum dispersion (vide infra) as more conventional
approaches yielded blocked pores and/or crystallites of 2.
The loading we selected for the catalysts tested here was
Figure 2. a) Temperature-resolved FT-IR spectra of MCM41-supported
catalyst during thermolysis; b) expansion of carbonyl stretching region.
are consistent with an adsorbed [Ru12C2Cu4Cl2(CO)32]2
cluster lying in a siliceous mesopore, in which both bridging
and terminal COs (see crystal structure in Figure 3) are in
contact with the surface. This interaction may be taken to be
the dominant binding force felt by the cluster anions.[34]
In situ studies: The catalyst precursor was activated by
thermolysis under high vacuum (0.01 Torr). Removal of the
protective sheath of carbonyls by this method yields naked
bimetallic nanoclusters which are highly active hydrogenation
catalysts. The process of activation and the associated
structural changes were principally followed by tempera-
1216
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1216 $ 17.50+.50/0
Chem. Eur. J. 1998, 4, No. 7

Cu±Ru Nanoparticle Catalysts
1214±1224
1
indicated by stretches at ca. 1430 ± 1390 cm , was surpris-
ing.[36] It would appear that these species need an extended
surface or a step site to exist.[37]
b) X-ray absorption spectroscopy (XAS): A systematic in situ
XAS investigation provided a great deal of information on the
nature of the production of the bimetallic catalyst described
here. The catalyst precursor was heated up to various
temperatures prior to room-temperature XAS measurements.
The results of these investigations are discussed below.
In the system under discussion, two edges (Ru K and Cu K)
are accessible and complementary information can be ob-
tained (Figure 4). The XANES region of the Cu K-edge (cf.
Fig.4a) shows very clear changes upon thermolysis, the most
Figure 3. Molecular crystal structure diagram of the precursor carbonyl
cluster anion [Ru12C2(CO)32Cu4Cl2]2
b) excluding carbonyls.
, a) including carbonyl ligands,
ture-resolved in situ studies by means of FT-IR spectroscopy
and X-ray absorption techniques (XANES and EXAFS)
using near-identical treatment protocols. The results of these
investigations are discussed below.
a) FT-IR spectroscopy: A pressed pellet was heated inside an
evacuated, purpose-built FT-IR cell[35] and its IR spectra
recorded at various temperatures. From the IR trace after
heating for 2 h at 2008C we can deduce that there are no
carbonyl species remaining in the material. It is of interest to
note that absorptions due to the phenyl stretching modes of
1
the PPN cations (ca. 1440 cm ) may still be identified after
heating at this temperature (vide infra). We can also see a
significant rearrangement of the adsorbed species between 80
and 1308C (as shown in the XANES CuK and RuK spectra,
vide infra). It is tempting to suggest that this transition is the
anchoring process of the bimetallic clusters, wherein Cl
ligands on the Cu atoms are replaced by Si ± O groups and
the two Ru octohedra condense over the nascent copper raft.
Futher examination over a range of higher temperatures then
show these cluster condensates being slowly denuded of their
CO sheath. The true nature of these carbonyl-deficient
intermediates is difficult to determine. Indeed, the broad
appearance of the CO stretching absorptions in the latter
stages of the experiment suggests there are different species
present in several stages of decarbonylation. Above 1308C the
terminal CO region displays a maximum at approximately
Figure 4. a) XANES spectra and b) first derivatives of Cu K-edge data and
c) XANES spectra and d) first derivatives of Ru K-edge data taken during
thermal activation, of i) catalyst as prepared, ii) heated to 808C, iii) heated
to 1308C and iv) heated to 1808C.
striking feature being a gradual decrease of a peak in the near
edge which is clearly seen in the first derivative shown in
Figure 4b and results from changes in the types of near
neighbours. However, the overall position of the absorption
edge does not differ markedly, indicating no major changes in
the electronic state of copper atoms. The XANES data of the
Ru K-edge reveal only moderate changes upon heating. These
modifications in the XANES region occur at temperatures as
1
1974 cm . This shows a shift to lower energy for the CO
stretch as compared to the precursor and may indicate the
electron-rich nature of the condensed bimetallic carbonyl
cluster by increased Md !p* donation (Dewer± Chatt ±
Duncanson model). During the final stages of decarbon-
ylation of the metal particles the absence of m4-CO species,
Chem. Eur. J. 1998, 4, No. 7
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1217 $ 17.50+.25/0
1217

J. M. Thomas, B. F. G. Johnson et al.
FULL PAPER
low as 808C, possibly indicating alterations in the electronic
structure due to changes in the local environment (which is
more apparent in the EXAFS results) around the copper,
owing to increasingly strong interactions with the surface
silanols.
was relatively uncomplicated for the Ru K-edge; however, the
Cu K-edge data proved intractable. Although it is possible to
analyse the Cu K-edge EXAFS data by considering a Cu ± O
shell at approximately 1.9 Š and either Cu ± Cu or Cu ± Ru
shell in the range of 2.5 to 2.65 Š, both the bond lengths and
coordination numbers obtained after refinements were found
to be considerably lower than what one would expect for
either monometallic (if there is elemental/phase separation)
or bimetallic Cu ± Ru systems. This is likely to be caused by
the presence of both types of neighbours (Cu and Ru) around
the central Cu atom, a chemical postulate consistent with our
previously reported study of Ag ± Ru nanoparticles.[20] Hence,
the reason for the low coordination number and chemically
unreasonable bond lengths obtained when considering only
one type of Cu ± Cu or Cu ± Ru neighbours can be found when
examining the opposing nature of the EXAFS oscillations
generated for these two scattering pairs. In Figure 6 we depict
the EXAFS oscillations resulting from a Cu ± Cu and Cu ± Ru
pair (Figure 6a,b) and from the addition of these two
(Figure 6c,d), the last showing a marked de-
crease in amplitude due to phase cancellation.
Examination of the changes in the associated Fourier
transforms (shown in Figure 5) of the corresponding Ru K-
edge EXAFS data corroborates the loss of carbonyls, reveal-
ing profound structural changes even at 808C (this is also
borne out by the FT-IR studies, vide supra), culminating in the
formation of the metal nanoparticle at 1808C. Interestingly,
the Cu K-edge EXAFS data suggest that its structural
environment undergoes significant modification only between
1308C and 1808C (Figure 5). Although it is possible to analyse
the EXAFS data recorded at various temperatures, our
primary interest was to gain quantitative information on the
final bimetallic cluster. This was achieved by a detailed
analysis of the EXAFS data in terms of a structural model
based on the structure of the precursor material. The analysis
The detailed analysis of the data taking into
account both Cu ± Cu and Cu ± Ru neighbours
did indeed yield a superior fit compared with
that obtained using a single Cu ± Cu or Cu ± Ru
shell. The associated coordination numbers as
well as bond lengths are found to be chemically
much more sensible values, the model being
comprised of coordination environments of
4.3Ru at approximately 2.66 Š and 1.1Cu at
approximately 2.59 Š for the Ru K-edge data
and 0.9O at ꢀ 1.95 Š, 3.1Cu at ꢀ 2.53 Š and 2.1
Ru at ꢀ 2.59 Š for the Cu K-edge data (see
Table 1), demonstrating the bimetallic nature of
the particles. However, in the Cu K-edge data
analysis the heavy scattering of the Ru atoms still
dominated. To overcome potential problems
with correlation effects, we made use of the
recently developed strategy of analysing both
the EXAFS data sets simultaneously. This meth-
odology has several advantages; in particular,
bond lengths of the two scattering pairs (in this
case Cu ± Ru from the Cu K-edge and Ru ± Cu
from the Ru K-edge) are refined together to
achieve identical values and any distortions in
the coordination numbers resulting from corre-
lations between closely similar bond lengths can
be resolved. The simultaneous refinement of
both the data sets was performed using XFIT-
(WIN95) and one common local environment
around both Cu and Ru atoms (see Figure 7 and
Table 2 below).
In the model shown above, Cu(0) is an
absorber surrounded by scatterers Ru(0),
Ru(2), Cu(1) and O and Ru(0) is the other
absorber, surrounded by Ru(1) and Cu(0). The
refinement of the environment around either
absorber is performed in turn, but taking ac-
count of both data sets simultaneously. This
Figure 5. Fourier transforms of EXAFS data taken during thermal activation: a) Ru K-edge
data, b) Cu K-edge data.
1218
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1218 $ 17.50+.50/0
Chem. Eur. J. 1998, 4, No. 7

Cu±Ru Nanoparticle Catalysts
1214±1224
Figure 7. Structural model used in dual edge EXAFS data refinement,
where Ru(0) and Cu(0) are the absorbers.
Table 2. Combined K-edge (Cu and Ru) refinement.
Absorber ± scatterer
N
R [Š]
s2 [Š2][a]
R factor
16%
Ru(0) ± Cu(0)
Cu(0) ± Cu(1)
Cu(0) ± Ru(0)
Ru(0) ± Ru(1)
Cu(0) ± O
1.1
2.7
0.9
3.9
1.4
1.4
2.59
2.54
2.59
2.66
1.97
2.59
0.003
0.003
0.003
0.003
0.006
0.003
Cu(0) ± Ru(2)
[a] Held equal for all metals during refinement.
restrains were imposed, such as the Debye ± Waller factors
being set equal for Ru(0) and Ru(2) or setting shell
occupancies to lie within the limits of between 1 ( Æ 1) to 8
( Æ 1) to minimise any correlation effects between coordina-
tion numbers and Debye ± Waller factors during the refine-
ment process. The results obtained by using this approach (fits
and raw data shown in Figure 8) demonstrate and confirm the
bimetallic nature of the particles.
Scanning transmission electron microscopy (STEM) investi-
gations: In order to evaluate the particle size and understand
its morphology, we used a dedicated STEM (probe size ꢀ
0.5 nm).[20] The key point with high annular dark-field imaging
is that the intensity of the (Rutherford) scattered beams is
directly proportional to Z2, where Z is the atomic number of
the scattering element. Thus, heavy atoms (such as Ru and
Cu) stand out very clearly on a light background of Si and O.
Analysis by energy dispersive X-ray emission (EDX) further
confirmed the relative average atomic proportions of the
bimetallic particles. Moreover, electron-stimulated X-ray
maps further confirm the uniformity of the supported
bimetallic nanoparticles (compare with the figures in
ref. [20]).
Figure 6. Figure 6. EXAFS simulations of single shells: a) ruthenium shell,
b) copper shell, c) superposition of both shells and d) summation, solid line,
and individual components, broken lines.
Figure 9 shows the annular dark-field images from the
Cu4Ru12 bimetallic particles (EDX analysis yields a Ru:Cu
ratio of 2.9) before and after an electron exposure of 12 Â
Table 1. Individual K-edge analysis.
2
Absorber Scatterer
N
R [Š]
s2 [Š2]
R factor
[%][a]
1025 electronsm respectively. In particular, Figure 9a pres-
ents a MCM41 cylindrical bundle of 4 channels across, curved
at either end. Four distinct particles inside one of the channels
have been marked. In Figure 9b the mesoporous structure of
the MCM41 is vitrified owing to beam damage. However, the
position of the 4 nanoparticles marked in the previous image
has not changed and there is no sign of aggregation, in spite of
the fact that they are very close to one another under an
electron beam so intense as to destroy the structural integrity
of the support. Comparison of the two images shows that this
also holds for the other particles, thus further confirming the
strength of the anchoring on the MCM41 surface of the Cu ±
Ru nanoparticles.
Ru
Cu
Ru
Cu
4.3
1.1
2.66
2.59
0.005
0.006
23.27
O
Cu
Ru
0.9
3.1
2.1
1.95
2.53
2.59
0.012
0.009
0.009
12.69
[a] Reliability factor.
improves the ratio of data points to variables, and hence
provides a more conservative approach. The shells were
allowed to vary in terms of occupancy, Debye ± Waller factor
and absorber± scatterer distance. Several constraints and
Chem. Eur. J. 1998, 4, No. 7
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1219 $ 17.50+.25/0
1219

J. M. Thomas, B. F. G. Johnson et al.
FULL PAPER
b)
a)
c)
Figure 9. High-angle annular dark-field images of Cu4Ru12 bimetallic
particles inside MCM4 a) before and b) after vitrification due to electron
beam damage; c) histogram of particle size distribution from image (a).
accurate determination. If a first approximate beam-broad-
ening behaviour is considered (for a probe diameter of 5 Š
and a beam convergence of 7 mrad on a 200 Š thick sample)
the beam broadening is likely to increase the beam size by a
minimum of 7.8 Š and the difference is more than 40%, which
is more than sufficient to explain the variations of FWHM in
Figure 9c. An extensive discussion of these considerations has
been previously published by Ozkaya et al.[39]
Figure 8. a) Cu K-edge EXAFS data, b) associated Fourier transform,
c) Ru K-edge data, d) associated Fourier transform; solid lines exper-
imental data, broken lines calculated spectra.
Thermogravimetric studies: To determine the weight loss on
heating the sample under activation conditions, the bulk
sample was heated and the weight loss recorded periodically,
giving the results shown in Figure 10. The graph shows two
regions of different but near-constant rates. The first [A to B]
is believed to show the weight loss resulting from the
carbonylate clusters and silanol condensation, whilst the
second part [B to C] is due to silanol condensation alone.
Extrapolation from BC to the ordinate gives the % weight loss
excluding that due to siloxane formation (assuming the rate of
this process is constant at low loading). The figure of 10.5%
thus obtained is entirely consistent with the weight loss due to
complete decarbonylation of the cluster species, FT-IR and
microanalysis [C, H, N, Cl]. Using this result we assumed that
thermal activation for 2 hours at 1908C gave a 11.5% weight
loss for the purposes of calculation of catalytic turnover
frequencies (TOFs).
Figure 9c shows a histogram outlining the particle size
distribution of the sample. The particle size is taken as the
FWHM (full width half maximum) intensity in this image.
Although this seems to be a widely used convention in particle
size determination,[14] the particle intensities in the image are
very much affected by the differences in defocus, especially so
because of the narrow depth of field in the high-angle annular
dark-field images. This means that the particles that are at
different heights within the sample have different intensity
distributions, that is, different FWHM values. The size of a
particle in the image can be thought of as a convolution of the
main probe intensity profile and the scattering potential of the
particle (a function of Z2 and density distribution within the
particle).[38] The effect of the noise further complicates an
1220
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1220 $ 17.50+.50/0
Chem. Eur. J. 1998, 4, No. 7

Cu±Ru Nanoparticle Catalysts
1214±1224
conducting competition experiments. The hex-1-ene conver-
sion is practically linear with time after the initial induction
period and the product distribution does not change signifi-
cantly until after about 360 mins. Interestingly, although the
major product (as expected) is n-hexane, significant amounts
of cis- and trans-2-hexene are also formed, with the ratio of
both isomers remaining constant around 1.5:1 for the major
part of the reaction. The increase in selectivity for n-hexane
after 480 mins, is probably a result of the subsequent hydro-
genation of the cis- and trans-hex-2-ene isomers.
The hydrogenation of an internal olefin (cis-cyclooctene)
was also studied at ambient temperature and H2 pressure.
Kinetic data obtained for this reaction (Figure 12) shows
Figure 10. Graph used in the weight-loss extrapolation of gravimetric
measurements.
Catalytic evaluation: The catalytic performance of the
supported bimetallic nanoparticles in the hydrogenation of
unsaturated molecules was tested on a wide variety of
unsaturated species: hex-1-ene, phenylacetylene, diphenyl-
acetylene, trans-stilbene, cis-cyclooctene and d-limonene. The
highly efficient hydrogenation of hex-1-ene was accompanied
by the isomerisation reaction to cis- and trans-hex-2-ene,
which were subsequently hydrogenated (albeit at a much
slower rate) as the reaction continued. Phenylacetylene is
completely converted to ethylbenzene under the reaction
conditions used. No hydrogenation of the phenyl group was
detected. Hydrogenation of diphenylacetylene gave both
stilbene (predominently trans-) and dibenzyl.[40]
Figure 12. Plot of the hydrogenation of cyclooctene at 1 bar versus time,
showing total conversion.
Painstaking kinetic studies at 20 bar hydrogen and 3738C
show an induction time of 60 minutes and an overall turnover
smooth conversion to cyclooctene. d-limonene, which con-
tains both internal and pendant olefin groups, also readily
underwent hydrogenation, and analysis of the final product
mixture suggested a preference for hydrogen addition at the
pendant olefin.[41] Little or no selectivity over the formation of
cis/trans-1-methyl,4-isopropylcyclohexane (29% conversion
to the cis and 34% conversion to trans isomer) could be
observed. This perhaps mirrors the small size of the catalyti-
cally active species, in that double binding of the substrate
might not be possible.
1
frequency of 25700 mol[hex]mol[Cu4Ru12] 1 h . The kinetic
details are summarised in Figure 11. The observed induction
time has several possible causes, for example in situ activation
of the catalyst particles by further reduction in hydrogen,
diffusion of hydrogen into the mesoporous structure, diffu-
sion-induced lag of product emerging from the mesopores or
effusion of PPNCl. These different aspects are currently being
investigated by varying the silica support structure and by
Discussion
The present study introduces a novel method of preparing
bimetallic catalyst anchored to high-surface-area mesoporous
silica (MCM41). The FTIR spectroscopic results identify the
nature of the loss of carbonyl ligands associated with the
thermal activation of the cluster. However, these results alone
cannot provide further information about the nature of the
final bimetallic cluster, and so we turned to the powerful
combination of STEM, XANES and EXAFS studies. Qual-
itatively, both XANES and EXAFS results corroborate the
IR results, in that there are substantial structural modifica-
tions during the thermolysis. The final detailed EXAFS data
analysis of both the Ru and Cu K-edge data, in particular
using the recently developed methodology of simultaneously
refining both K-edge data sets, quantitatively revealed the
presence of bimetallic clusters, with the Ru ± Cu distance lying
between those of bulk Ru ± Ru and Cu ± Cu.[42] Neither the
Figure 11. Plot of the hydrogenation of hex-1-ene at 20 bar versus time,
showing total conversion, product distributions and hydrogen pressure
variation (1-hexene: 44.7 g, Ru12Cu4/MCM-41: 20 mg, T: 373 K).
Chem. Eur. J. 1998, 4, No. 7
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1221 $ 17.50+.25/0
1221

J. M. Thomas, B. F. G. Johnson et al.
FULL PAPER
average Ru ± Ru distances nor the Cu ± Cu distances are
significantly smaller (as one would expect for small metallic
clusters) than that of twelvefold coordination in the bulk
metal (although the low coordination numbers clearly suggest
the presence of small particles, as does STEM). This is
surprising, but may be explained by the cluster± surface
interaction stretching the structure and/or by the presence of
carbido atoms which would be undetectable by EXAFS. The
argument that the carbido atom remains in the centre of the
Ru octahedron is a strong one, as these structural fragments
are chemically very stable and our activation conditions are
very mild. Iwasawa and Izumi have reported[43] the contrac-
tion and expansion (inhaling and exhaling in the presence of
CO) of Ru ± Ru bond lengths (corresponding closely to our
data) in ruthenium and rhodium clusters when deprived of, or
exposed to, CO under conditions similar to those employed
here. That there is anchoring of the bimetallic particle to the
surface by the formation of Cu ± O bonds is evident from the
EXAFS results, since only one Cu ± O shell is present and
there is no direct indication for any Ru ± O shells. Consistent
with these results, we may construct a detailed structural
picture of the nanoparticle, the principal characteristic of
which is an exterior dome of twelve Ru atoms, linked by a
base of four Cu atoms which are, in turn, covalently bonded to
oxygens that, by inference, are themselves bound to the
silicons of the support. Hence, we propose the model shown in
Figure 13, derived from EXAFS and incorporating carbido
atoms. On the basis of these arguments, we suggest that the
K-edge EXAFS spectrum, the Fourier transform envelope of
which was comparable to the experimentally observed ones in
terms of shape and amplitude.
Annular dark-field electron microscopy is able to pinpoint
the rather uniform spatial distribution of these particles, and
establishes (as does the EXAFS analysis) that there is no
fragmentation or sintering during catalysis, further ration-
alising the remarkable catalytic performance of these bimet-
allic nanoparticles. From a very careful and conservative
analysis of our STEM images we estimate the maximum
particle size to be around 23 Š when allowing for an error of
‡40%. Thus, the EXAFS structural model (being about 14 Š
across), further refined by forcefield minimisations, would be
perfectly consistent with the observeables and with general
chemical considerations.
Together these results demonstrate that our initial strategy
of removing the stabilising CO sheath from a mixed-metal
cluster to produce a well-defined metal nanoparticle and
anchoring the more oxophilic second metal to the MCM-41
surface has met with success. This work also reveals that there
is abundant scope for further exploitation of bimetallic metal-
cluster carbonylates as precursors for other supported nano-
particle catalysts. Moreover, a wide range of catalytic
reactions besides hydrogenation awaits study.
Experimental Section
Materials and methods: All reactions were carried out under exclusion of
air with solvents freshly distilled under an atmosphere of nitrogen or argon.
Solution IR spectra were recorded on a Perkin ± Elmer 1600 series FTIR in
CH2Cl2 using NaCl cells or a Nujol mull. Negative fast-ion bombardment
mass spectra were obtained on a Kratos MS50TC spectrometer with CsI as
calibrant. Products were separated with Merck thin-layer chromatography
(tlc) plates as supplied; 0.25 mm layer of Kieselgel 60 F254. [Ru3(CO)12],
[Ru6C(CO)16][PPN]2 and MCM41 were prepared by the literature proce-
dures.[27, 45, 46] CuCl was purchased from Aldrich and used without further
purification.
Synthesis of [Ru12C2Cu4Cl2(CO)32][PPN]2: [Ru6C(CO)16][PPN]2 (0.5 g)
was refluxed in THF (20 mL) for 5 h with excess CuCl (1.0 g). The reaction
mixture was cooled and filtered and an excess of PPNCl added. The solvent
was then removed in vacuo and the mixture dissolved in CH2Cl2 and
layered with ethanol to give deep red crystals. These were washed with
ethanol and pentane to yield 385 mg of [Ru12C2Cu4Cl2(CO)32][PPN]2.
Elemental analysis: 35.60% C, 1.74% H, 0.69% N, 6.89% Cu (calcd 35.9%
C, 1.71% H, 0.79%N, 7.18% Cu); ‡FAB MS: m/z ˆ 2997
{[Ru12C2Cu4Cl2(CO)32][PPN]} , 2427 {[Ru12C2Cu4Cl2(CO)31] }; IR: nÄCO
ˆ
1
2059(m), 2021(vs), 2967(sh), 1947(sh), 1855(wsh), 1829(brw) cm
.
Preparation of MCM41/[Ru12C2Cu4Cl2(CO)32][PPN]2: The mesoporous
silica [MCM41] (120 mg) was dried under high vacuum (0.01 mmHg) at
473 K for 6 h, then washed with dry solvent and dried again under high
vacuum at 473 K. It was then slurried with dry ether (30 mL) and the
copper± ruthenium cluster salt [Ru12C2Cu4Cl2(CO)32][PPN]2 (50 mg cluster
for 57 mg MCM41, i.e. 47.0% cluster salt by mass) along with CH2Cl2
(0.1 mL) at ambient temperature in darkness for 72 h. The resulting pink
solid was washed with ether (10 mL) and dried under high vacuum
(0.01 mmHg) at ambient temperature for 4 h. Elemental analysis gave
16.85% C, 2.63% H and 0.2%N (calcd 16.8% C, 0.4% H, 0.80% N).
Spectroscopic data for MCM41/[Ru12C2Cu4Cl2(CO)32][PPN]2: IR (Nujol):
Figure 13. Energy-minimised model of naked metal particle attached to a
silica surface, based on EXAFS analysis and on the breathing carbido
model of Iwasawa and Izumi, showing the internal structure of the particle
and the Van der Waals radii.
particle is very small ( ꢀ 14 Š diam.), indeed, too small to use
EXAFS reliably to estimate its size. Further support for the
proposed cluster geometry was provided by computer simu-
lation: A model of the cluster consistent with the EXAFS data
was subjected to an energy-minimisation calculation with the
program Discover[44] and the esff forcefield. The minimised
structure is shown in Figure 13. Additionally, using the
minimised model it was possible to simulate a Cu and Ru
nÄCO ˆ 2058(s), 2011(vs),1997(vs), 1981(s), 1912(shm), 1958(w), 1948(w),
1
1968(wbr), 1835(mbr), 1822(wbr) cm
.
Determination of mass loss on catalyst activation by thermolysis of
MCM41/[Ru12C2Cu4Cl2(CO)32][PPN]2: A precise quantity of the mesopo-
rous silica loaded with the bimetallic cluster was placed into an Al-foil
bucket of known mass, which was loosely crimp-sealed. This was then
1222
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1222 $ 17.50+.50/0
Chem. Eur. J. 1998, 4, No. 7

Cu±Ru Nanoparticle Catalysts
1214±1224
heated in sand (453 K) to ensure even heat transfer under vacuum
(0.01 mmHg) for several periods of time (see graph) and weighed at each
interval after careful cleaning with an air brush. After a combined time of
12 h the sample did not exhibit any bands in the CO stretching region and
was submitted to elemental analysis, which gave 15.76% C, 2.74% H and
0.2%N.
EXAFS data collection and refinement: EXAFS data were recorded at the
Daresbury SRS facility on station 9.2. The station was equipped with a
Si(220) monochromator, and data were collected in fluorescence mode
using a thirteen-element Canberra fluorescence detector. A powdered
sample of each catalyst studied was pressed into a self-supporting wafer,
and mounted in a custom-built in situ cell.[47] The EXAFS data were
subsequently analysed with the XFIT (for Windows 95)[48] suite of
programs. In the double-edge refinement the S20 parameters were taken
from the single-edge refinements and held constant and the s2 parameters
were constrained to be equal for the copper and ruthenium atoms. Two E0
parameters were included in the model, one for copper and one for
ruthenium. All parameters (i.e. coordination numbers, distances, E0 and s2
values) were refined simultaneously. The precise fitting protocol and
theoretical underpinning will be published elsewhere.
Catalyst activation: thermolysis of MCM41/[Ru12C2Cu4Cl2(CO)32][PPN]2:
A precise quantity of MCM41/[Ru12C2Cu4Cl2(CO)32][PPN]2 was loaded
into an Al-foil bucket inside an argon-filled Schlenk tube. The ensemble
was then heated for 2 h at 453 K under vacuum (0.01 mmHg), after which
the apparatus was purged with Ar or H2 and the cooled catalyst transfered
to the reaction vessel.
Hydrogenation of hex-1-ene (batch expt.): A 150 mL teflon-lined autoclave
equipped with a magnetic stirrer was charged with catalyst ( ꢀ 8 mg), hex-1-
ene (3.0 mL) and H2 (64 bar). The ensemble was heated to 373 K for 4 h,
after which the vessel was cooled to ambient temperature and the contents
High-resolution electron microscopy (HREM): Experiments were con-
ducted with a field emission gun dedicated scanning transmission electron
microscope (VG HB501) with a windowless EDX detector. MCM41
samples were prepared by crushing the particles between two glass slides
and spreading them on a perforated carbon film supported on a Ti grid. The
samples were briefly heated under a light bulb in the specimen preparation
chamber of the STEM prior to their introduction into the microscope.
1
analysed by H NMR to reveal ꢁ99% conversion to n-hexane.
Hydrogenation of hex-1-ene (kinetic study):
A 150 mL PEEK-lined
autoclave equipped with a magnetic stirrer was charged with catalyst
(10 mg), hex-1-ene (44.6 g, 66.27 mL) and H2 (20 bar). The ensemble was
maintained at 373 K for 16 h, after which it was cooled to ambient
temperature. Every 30 min samples were taken at high pressure and
analysed by GC. A typical mass balance shows minimal sample loss: moles
of 1-hexene taken: 0.532, product analysis: moles of n-hexane formed:
0.367, moles of cis-2-hexene formed: 0.122, moles of trans-2-hexene
formed: 0.039, moles of 1-hexene remaining: 0.002, total moles of products:
0.530, loss (moles): 0.002, loss (grams): 0.168 g, loss (%): 0.4%.
Acknowledgements: We thank the EPSRC for a rolling grant to JMTand a
regular one to BFGJ, the EU for support of TM, and the Comissioners of
the 1851 Royal Exhibition for an award to RR. GS was funded partly by
JMTꢁs rolling grant and the CCRL Daresbury Laboratory, to whom we are
grateful.
Hydrogenation of diphenylacetylene: A 150 mL teflon-lined autoclave
equipped with a magnetic stirrer was charged with of catalyst (7.1 mg),
diphenylacetylene (400 mg), ethanol (1 mL), pentane (10 mL) and H2
(65 atm). The ensemble was heated to 393 K for 6 d, after which the vessel
was cooled to ambient temperature and the contents analysed by 1H NMR
and GC/MS to reveal 66% conversion to trans-stilbene, >0.5% conversion
to cis-stilbene and 33% conversion to dibenzyl. >0.5% of unconverted
starting material was detected. From these results an overall TOFhydro of
16.8 mol[DPA]mol[Cu4Ru12] 1 h 1 was calculated.
Received: October 29, 1997 [F869]
[1] J. S. Beck, J. C. Vartuli, Curr. Opin. Solid-State Mat. Sci. 1996, 1, 76.
[2] J. M. Thomas, C. R. A. Catlow, R. G. Bell, in Handbook of Hetero-
geneous Catalysis (Eds.: G. Ertl, H. Koezinger, J. Weitkamp), VCH,
Weinheim, 1997.
[3] J. M. Thomas, W. J. Thomas, Principles and Practices of Heterogeneous
Catalysis, Ch. 5, VCH, Weinheim, 1997.
[4] J. Klinowski, J. M. Thomas, M. Audier, S. Vasudevan, C. A. Fyfe and
J. S. Hartman, J. Chem. Soc. Chem. Commun. 1981, 570.
[5] K. J. Balkus, A. Khanmamedova, A. G. Gabrielov, S. I. Zones, Stud.
Surf. Sci. Catal. 1996, 101, 1341.
[6] R. F. Lobo, M. Tsapatsis, C. C. Freyhardt, S. Khodabandeh, P. Wagner,
C. Y. Chen, K. J. Balkus, S. I. Zones, M. E. Davis, J. Am. Chem. Soc.
1997, 119, 8474.
Hydrogenation of phenylacetylene:
A 150 mL teflon-lined autoclave
equipped with a magnetic stirrer was charged with catalyst (12.0 mg),
phenylacetylene (1 mL), pentane (5 mL) and H2 (65 atm). The ensemble
was heated to 373 K for 24 h, after which the vessel was cooled to ambient
temperature and the contents analysed by 1H NMR and GC/MS to
reveal ꢁ99.9% conversion to ethylbenzene; no unconverted starting
material was detected. From these results an overall TOFhydro of
1
615 mol[PA]mol[Cu4Ru12] 1 h was calculated (assuming styrene is the
intermediate).
[7] P. A. Wright, R. H. Jones, S. Natarajan, R. G. Bell, J. Chen, M. B.
Hursthouse, J. M. Thomas, J. Chem. Soc. Chem. Commun. 1993, 633.
[8] G. W. Noble, P. A. Wright, P. Lightfoot, R. E. Norris, K. J. Hudson, A.
Krick, H. Graafsma, Angew. Chem. 1997, 109, 76; Angew. Chem. Int.
Ed. Engl. 1997, 36, 81.
[9] T. Maschmeyer, F. Rey, G. Sankar, J. M. Thomas, Nature (London)
1995, 32, 696.
[10] J. M. Thomas, Chem. Eur. J. 1997, 3, 1557.
[11] R. D. Oldroyd, J. M. Thomas, G. Sankar, Chem. Commun. 1997, 21,
2025.
[12] J. H. Sinfelt, Bimetallic Catalysts, Wiley, New York, 1983.
[13] J. H. Sinfelt, Int. Rev. Phys. Chem. 1988, 7, 281.
[14] M. S. Nasher, A. I. Frenkel, D. L. Adler, J. R. Shapley, R. G. Nuzzo, J.
Am. Chem. Soc. 1997, 119, 7760.
Hydrogenation of trans-stilbene: A 150 mL teflon-lined autoclave equip-
ped with a magnetic stirrer was charged with catalyst (8.0 mg), trans-
stilbene (400 mg), pentane (10 mL), ethanol (1 mL) and H2 (65 atm). The
ensemble was heated to 393 K for 48 h, after which the vessel was cooled to
ambient temperature and the contents analysed by 1H NMR and GC/MS to
reveal 79% conversion to dibenzyl. 21% of unconverted starting
material was detected. From these results an overall TOFhydro of
69.5 mol[stil]mol[Cu4Ru12] 1 h 1 was calculated.
Hydrogenation of cis-cyclooctene: A 250 mL Erlenmeyer flask equipped
with a magnetic stirrer was charged with catalyst ( ꢀ 10 mg), cis-cyclo-
octene (10 mL) and H2 (1 atm). The ensemble was kept at 298 K for 72 h,
during which time the contents were analysed by 1H NMR and GC/MS to
reveal
a steady conversion to cyclooctane. No unconverted starting
material was detected after 72 h. From these results an overall TOFhydro
[15] M. Ichikawa, Adv. Catal. 1992, 38, 283.
[16] B. C. Gates, Chem. Rev. 1995, 95, 511.
1
of 152 mol[cyclo-C8]mol[Cu4Ru12] 1 h was calculated.
Hydrogenation of d-limonene: A 150 mL teflon-lined autoclave equipped
with a magnetic stirrer was charged with 13.0 mg of catalyst, d-limonene
(1 mL) and H2 (65 atm). The ensemble was heated to 393 K for 23 h, after
which it was cooled to ambient temperature and the contents analysed by
1H NMR and GC/MS to reveal 29% conversion to cis-1-methyl-4-
isopropylcyclohexane, 34% conversion to trans-1-methyl-4-isopropylcy-
clohexane, >0.5% conversion to 1-methyl-4-isopropenylcyclohexane and
35% conversion to 1-methyl-4-isopropylcyclohex-1-ene. >1% unreacted
starting material was detected. From these results an overall TOFhydro of
362 mol[d-lim]mol[Cu4Ru12] 1 h 1 was calculated.
[17] O. Alexeev, M. Shelef, B. C. Gates, J. Catal. 1996, 164, 1, and
references therein.
[18] J. M. Basset, J. P. Candy, A. Choplin, B. Didillon, F. Quigrard, A.
Theolier, in Perspectives in Catalysis (Eds.: J. M. Thomas, K. I.
Zamaraev), Blackwells/IUPAC, Oxford, 1992, pp. 125 ± 146.
[19] M. S. Nasher, D. M. Sommerville, P. D. Lane, D. L. Adler, J. R.
Shapley, R. G. Nuzzo, J. Am. Chem. Soc. 1996, 118, 12964.
[20] D. S. Shephard, T. Maschmeyer, B. F. G. Johnson, J. M. Thomas, G.
Sankar, D. Ozkaya, W. Zhou, R. D. Oldroyd, Angew. Chem. 1997, 109,
2337; Angew. Chem. Int. Ed. Engl. 1997, 36, 2242.
Chem. Eur. J. 1998, 4, No. 7
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1223 $ 17.50+.25/0
1223

J. M. Thomas, B. F. G. Johnson et al.
FULL PAPER
[21] M. S. Nasher, A. I. Frenkel, D. L. Adler, J. R. Shapley, R. G. Nuzzo, J.
Am. Chem. Soc. 1997, 119, 7760.
[35] G. Sankar, J. M. Thomas, C. R. A. Catlow, J. Turner, W. I. F. David,
Top. Catal., submitted.
[22] J. W. Couves, J. M. Thomas, G. N. Greaves, R. H. Jones, A. J. Dent,
G. E. Derbyshire, Nature (London), 1991, 354, 465.
[36] D. B. Brown, P. J. Dyson, B. F. G. Johnson, C. M. Martin, D. G. Parker,
S. Parsons, J. Chem. Soc. Dalton Trans. 1997, 1909.
[37] M. L. Colaianni, J. G. Chen, W. H. Weinberg, J. T. Yates, J. Am. Chem.
Soc. 1992, 114, 3735.
[38] M. M. J. Treacy, S. B. Rice, J. Microscopy 1989, 15, 211.
[39] D. Ozkaya, J. M. Thomas, D. S. Shephard, T. Maschmeyer, B. F. G.
Johnson, G. Sankar, R. D. Oldroyd, Inst. Phys. Conf. Ser. 1997, 153(9),
403.
[40] The overall TOF of this reaction was unexpectedly low; this may be a
result of deactivation by long-lived intermediates blocking reactive
sites on the bimetallic particles. We are currently attempting to
determine the cause of this loss of activity.
[41] The degree of substitution of a given olefin is known to be geometri-
cally inversely proportional to the rate at which it may be hydro-
genated on a metal surface; see Comprehensive Organic Synthesis,
Vol X, Pergamon, Oxford, 1988.
[23] J. M. Thomas, G. N. Greaves, Science, 1994, 265, 1675.
[24] M. A. Beswick, Ph.D Thesis, University of Cambridge, 1992, and
references therein; M. A. Beswick, J. Lewis, P. R. Raithby, M. C.
Ramirez de Arellano, Angew. Chem. 1997, 109, 303; Angew. Chem.
Int. Ed. Engl. 1997, 36, 291.
[25] S. J. Taverner, J. H. Clark, G. W. Gray, P. A. Heath, D. J. Macquarrie, J.
Chem. Soc. Chem. Commun. 1997, 1073.
[26] J. S. Beck, J. C. Vartuli, W. J. Roth, M. E. Leonowicz, C. T. Kresge,
K. D. Schmitt, C. T.-W. Chu, D. H. Olsen, E. W. Sheppard, S. B.
McCullen, J. B. Higgins, D. L. Schlenker, J. Am. Chem. Soc. 1992, 114,
10834.
[27] F. Rey, G. Sankar, T. Maschmeyer, J. M. Thomas, R. G. Bell, G. N.
Greaves, Top. Catal. 1996, 3, 121 ± 126.
[28] D. Ozkaya, J. M. Thomas, W. Zhou, unpublished results.
[29] I. J. Shannon, J. M. Thomas, G. Sankar, T. Maschmeyer, M. Sheehy, D.
Madil, R. D. Oldroyd, Catal. Lett. 1997, 44, 23.
[30] G. Sankar, P. A. Wright, S. Natarajan, J. M. Thomas, G. N. Greaves,
A. J. Dent, B. R. Robson, C. A.Ramsdale, R. H. Jones, J. Phys. Chem.
1993, 97, 9550.
[42] The precise theoretical treatment of this approach for these systems
will be published elsewhere, but is already available in its general form
from the XFIT documentation, Australian Synchrotron Research
Program, c/- ANSTO, Private Mail Bag 1, Menai, NSW, 2234
(Australia).
[31] J. M. Thomas, P. A. Wright, J. Chen, G. Sankar, G. N. Greaves, A. J.
Dent, L. Marchese, Angew. Chem. 1994, 106, 1922; Angew. Chem. Int.
Ed. Engl. 1994, 33, 1871.
[32] J. M. Thomas, Faraday Disc. Chem. Soc. 1996, 105, 1.
[33] M. A. Beswick, J. Lewis, P. R. Raithby, M. C. R. Dearellano, J. Chem.
Soc. Dalton Trans. 1996, 21, 4033.
[43] V. Izumi, Y. Iwasawa, ChemTech, 1994, 20.
[44] Discover 4.0.0, Molecular Simulations, San Diego, 1996.
[45] J. N. Nicholls, M. D. Vargas, J. Hriljac, M. Sailor, Inorg. Synth. 1989, 28,
280.
[46] A. A. Bhattacharyya, C. C. Nagel, S. G. Shore, Organometallics 1983,
2, 1187.
[34] Solvent extraction of this material with dichloromethane showed the
bimetallic clusters can be quantitatively desorbed and remain chemi-
cally unchanged.
[47] G. Sankar, J. M. Thomas, J. Chen, P. A. Wright, P. A. Barrett, G. N.
Greaves, C. R. A. Catlow, J. Nucl. Inst. Methods B 1998, 97, 37 ± 40.
[48] XFIT (for Windows 95), P. J. Ellis, H. C.Freeman, J. Synchroton
Radiat. 1995, 2, 190.
1224
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1998
0947-6539/98/0407-1224 $ 17.50+.50/0
Chem. Eur. J. 1998, 4, No. 7