You are on page 1of 9

Optical spectroscopy of copper clusters: Atom to bulk

M. Moskovits and J. E. Hulse

Citation: J. Chem. Phys. 67, 4271 (1977); doi: 10.1063/1.435365


View online: http://dx.doi.org/10.1063/1.435365
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v67/i9
Published by the AIP Publishing LLC.

Additional information on J. Chem. Phys.


Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
Optical spectroscopy of copper clusters: Atom to bulk
M. Moskovits and J. E. Hulse
Lash Miller Laboratories and Erindale College, University of Toronto, Toronto, Canada M5S lAl
(Received 14 June 1977)

Copper clusters containing two-five atoms and large colloidal particles of copper were formed within
matrices by co-condensing metal vapor with argon, methane, and dodecane onto a low-temperature
substrate. The electronic absorption spectra of the resulting solids were recorded and absorptions
belonging to individual aggregates of copper identified using a previously published calculation which
predicts the cluster distribution within matrices of given metal-to-host ratios. The "birth" of the band
structure of copper is clearly visible on going from the dimer upward, with Cu. already possessing many of
the features of the bulk metal.

Interest in the physical and chemical properties of one of us (M. M. ) with Ozin (Cr 2, 19 Cu 2, 20 V 2 21 ). In addi-
small atomic and molecular aggregates, especially metal tion, early gas phase work data are available for CU 2,22
clusters, has increased almost catastrophically in the Ag2,23 and AU2' 24 Luminescence from CU225 has also
past five years, partially as a result of theoretical and been reported recently. An ESCA study by Mason and
experimental innovations which have allowed quantum Baetzold on Ag aggregates 26 has also appeared. The
mechanical calculations to be performed for moderately preceding is only a partial list. The spectra of aggre-
large clusters and the uv-visible and photoelectron gates greater than dimers but smaller than colloidal par-
spectra of aggregates to be measured. ticles have been reported by US 27 for Nix and recently by
Schulze28 for Ag 3. In this paper we report the spectra of
It has been recognized for some time that small clus-
copper clusters in the range Cu-Cu, and that of copper
ters play key roles in such areas as the homogeneous
colloid in solid argon and paraffin matrices.
nucleation and growth of macroscopic particles in the
gas phase, 1 where the early stages of nucleation are of- EXPER 1M ENTAL
ten undetected, the photographic process2 for which one
Details of the apparatus design have been given else-
requires silver clusters of a minimum size before a
where. 7 Briefly, copper vapor was generated by resis-
usable latent image is formed, and heterogeneous cata-
tively heating a tantalum filament on which fine copper
lytic reactions 3 which in many cases proceed on small
wire was wrapped. The copper was melted and outgassed
metal aggregates supported in a refractory oxide ma-
before the deposition substrate was cooled. Copper de-
trix. Surface SCience, which nowadays is primarily
position rates were monitored with a piezoelectriC mi-
concerned with the physics and chemistry at single-
crobalance. 29 Research grade Ar and methane (Mathe-
crystal surfaces, has also turned its attention towards
son) were used without further purification and spectro-
the properties of metal clusters as a result of the use
scopic purity dodecane was deposited from the vapor
of small aggregates by theoreticians to model surfaces
above the vacuum degassed liquid. Vapor deposition
in either surface density of states calculations4 or in
rates were regulated by means of a fine needle valve and
quantum mechanical calculations of chemisorptive bond-
measured by noting the pressure drop in a bulb of known
ing. 5 In this manner interest was generated in such
volume. The deposition substrate was a LiF plate cooled
questions as the transition from the "molecular" to the
to a bottom temperature of lOoK by means of an Air
bulk metallic state 6 and the onset of bulk properties.
Products Displex closed-cycle refrigerator. Spectra
The recognition that homogeneous and heterogeneous were recorded on a Varian Techtron uv-visible spectrom-
catalysis have certain underlying similarities sparked eter with 1. O-nm slits.
interest in the progress of the chemisorptive bond on
RESULTS
going from the case of ligands bound to single metal
atoms to that of a molecule bound to large metal clus- The deposition conditions under which matrices con-
ters. 7 This in turn resurrected such ancient queries taining small copper clusters could be formed lay in a
as whether or not the chemisorptive bond is local. narrow range of gas-to-metal ratio and gas depOSition
rates. Outside these confines one obtained either well-
Molecular orbital calculations on clusters were pion-
isolated copper atoms or such large aggregates that the
eered by Blyholder 5 and Baetzold8 using the extended
spectra were dominated by that of colloidal copper. In
Huckel technique. With the development of the scattered
argon, efficient small-cluster production was achieved
wave Xex technique by Slater and Johnson 9 the number of
in the metal-to-argon ratio range 3-10 parts per 1000,
calculations multiplied rapidly, notably through the work
and for argon deposition rates between 0.04 and 0.1
of Messmer, 10 Johnson, 10 and Rhodinl l ; and via Ander-
son's12 refinement of the extended Huckel technique which mmole/h. With methane the range was somewhat
combines two-body electrostatic atomic repulSions and broader, extending between approximately 1 and 25
one-electron orbital energies. Simultaneously, uv-visi- parts per 1000 Cu/CH, and gas deposition rate of approx-
ble absorption spectra of metal dimers and unidentified imately 0.01 and 0.1 mmole/h. In dodecane only the
higher aggregates were reported by Pimentel (Li 2) , 13 spectrum of colloidal copper was obtained, even at the
DeVore (Mn2 , Fe2, Ni 2),14 Brewer (Pb2), 15 Andrews (Li 2, largest feasible paraffin-to-metal ratio.
Na2) , 16 Knight and Evener (Mg2), 17 Jansson (Pt2), 18 and The concentration range 1-10 parts per 1000 is a use-

The Journal of Chemical Physics, Vol. 67, No.9, 1 November 1977 Copyright © 1977 American Institute of Physics 4271

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
4272 M. Moskovits and J. E. Hulse: Spectroscopy of copper clusters

ful one in studying cluster formation since, as we have


shown elsewhere, 30 the absorbance of a species Cu,
when divided by that of Cu atoms will increase with total
metal concentration Mo approximately as MQ-1. The re-
sults in methane were found to be particularly amenable
to analysis in terms of this rule.
The uv-visible spectra of the products of cocondensa-
tion of Cu vapor with Ar in various concentrations are
shown in Fig. 1. An analogous set of results for Cu in
CH 4 is shown in Fig. 2. In both sets of figures the spec-
tra are dominated by an intense absorption triplet near
300 nm when the Cu/gas ratio is low. Weak features on
either side of this absorption grow rapidly as the metal
content of the matrix increases. The observed absorp-
tions are summarized in Table I. The low-energy lines
which we will later assign to CU z seemed under some
deposition conditions to be disproportionately large. _ _L -_ _ _ _ _ L--_----L-____ ~~ ____- L I
This, we feel, is due to the fact that under some circum- 200 300 400 500 600 nm
stances a large fraction of the copper vaporizes as the FIG. 2. Portions of the uv-visible absorption spectra of
dimer. This results in spectra such as the one at the methane matrices containing copper. Metal to gas ratios in-
bottom of Fig. 1 in which CU z dominates the spectrum, crease top to bottom. Ordinate is absorbance in arbitrary
dwarfing even Cu lines. The conditions under which this units. Some lines belonging to Cu are suppressed for clarity.
happened could not be easily reproduced. In general,
inordinate quantities of CUz, were formed when the metal
was vaporized at low temperatures. A mass-spectro-
scopic study is underway to help clarify this phenomenon.
The spectrum obtained after cocondensing copper va-
por with dodecane is shown in Fig. 3. This spectrum is
different from those shown in Figs. 1 and 2. No trace
remains of the bands near 300 nm nor of the plurality
of absorptions on either side of that wavelength. The
spectrum slopes gently upward towards high energies
with two broad features at approximately 210 and 600
nm.
The spectral changes resulting from warming a Cui
CH 4 matrix to various temperatures are shown in Fig.
4. Warming Cu/Ar matrices caused a decrease in the
lines belonging to Cu (see later) and eventually produced
colloidal copper with little growth in the other spectral
bands. For temperatures near the boiloff point of the
I matrix, however, all bands in the spectrum dropped
sharply in absorbance to be replaced by those of col-
I 1
loidal copper and two new features, as shown in Fig. 5.
I \

~u
DISCUSSION
Copper atoms
The uv-visible spectrum of Cu atoms isolated in solid
argon has been reported by several workers. 31-33 That

!\

-'---
300 400 500 600 nm
---L -------- .~~------
200 300 400 500 FIG. 3. Portions of the spectrum of Cu in dodecane (upper
nClinometers curve). The spectrum of copper colloidal particles in dodecane
FIG. 1. Portions of the uv-visible absorption spectra of argon calculated using the method outlined in Ref. 27 and Johnson and
matrices containing copper. Metal to gas ratios increase top Christy's44 values for the optical constants of bulk copper
to bottom. Ordinate is absorbance in arbitrary units. (lower curve).

J. Chem. Phys., Vol. 67, No.9, 1 November 1977

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
M. Moskovits and J. E. Hulse: Spectroscopy of copper clusters 4273

TABLE I. Wavelengths and assignments of the absorptions observed in argon and methane matrices con-
taining copper.

Cu
Relative
A nm Gas phase A (argon) A (CH() intensity Assignment
Assignment (gas) (gas)~ intensity (nm) (nm) (CH 4) (refers to Ar)
2PI/2 (3d I 04pl) 327.4 0.31 306.2 320.0 1.0 X2~ 112 -A 2713/2 of ArCu*
2P3/2 (3d 104pl) 324. 8 0.64 301. 8 314.3 0.9 2
-B 71 1/ 2
298.0 310.1 0.7 -C2~1'2
305.5 0.3
301.3 0.05
4p 3/2 (3,jl4s 14p l) 249.2 0.029 238.3 242 O. 02
233.6
4Pl/2 (3Jl4s14pl) 244.2 0.0057 229.3
4D1/2 (3Jl4s14pl) 222.6 0.0078 220.7
2PI/2 (3Jl4s14pl) 218.2 0.0160 218.0
2P3/2 (3,jl4s 14p l) 217.8 0.022 211.8 216 0.2
207.3 211 } 0.5
210

Cuz CU3 CU4


Anm Anm A nm Anm A nm A nm
(Ar) (CH 4) Assignment Calculated" (Ar) (CH 4) Assignmentb (Ar) (CH 4) Assignmentb
400 380} {d7f 6 -SU. 460 500
376
533l
404)
I-II
402 424
272
I-II
I-III
SU -SU
6 U 490
355 du6 -suU 350 335
283 SU6 -P71 u 280 262
236 256 dUU -pu6 240 234 280l
260
I-III
221 dUU -P7r6 200 230
221 227

"Reference 31. bSee text. "Reference 43.

of Cu in CH 4 has not, to our knowledge, been reported. Our own spectra of Cu in Ar are similar to those previ-
For Cu in Ar, the spectra consist of a prominent triplet ously reported except for improved resolution of the
centered around 300 nm, a slightly smaller doublet at small features. Occasionally, weak lines are seen on
about 210 nm and several rather small peaks in between. the low-energy slope of the 300 nm triplet. 31 These, we
find, disappear upon annealing and probably arise from
copper isolated in nonequilibrium phases of solid argon.
(A similar "red" triplet is reported for isolated alkali
atoms. See, for instance, Ref. 13.)

~ __ L~ ____
·~~,o nm
200 300 400 500 600 nm
FIG. 5. Portion of the spectrum of a Cu in Ar matrix after
FIG. 4. Spectra showing the effect of warmup on the cluster prolonged warmup at a temperature close to the matrix boil-
distribution within a Cu in methane matrix. Warmup time in- off point. The spectrum before warmup was similar to that at
creases top to bottom. the top of Fig. 1.

J. Chern. Phys., Vol. 67, No.9, 1 November 1977

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
4274 M. Moskovits and J. E. Hulse: Spectroscopy of copper clusters

The spectrum of Cu in CH 4 is similar to that of Cu in The Jahn-Teller explanation was revived by Belyaeva
Ar except that the triplet at around 300 nm is replaced et at. 39 to explain the fluorescence spectra of isolated
with a quintet whose structure persists even on anneal- Li atoms. These show Stokes shifts of approximately
ing. We shall discuss Cu in Ar first. 4000 cm- 1 and no triplet structure. The separation be-
tween components in the 2S - 2P triplet of all group IA
The spectra of isolated atoms are usually discussed
and IB metals in Ar exceeds 200 cm- 1, We do not be-
by correlating the absorption lines of the matrix with
lieve that vibronic coupling with the host can give rise to
those of the gas phase atom. This procedure has been
such a large split nor is a spin-orbit split of the ob-
formalized by Gruen 31 and named AMCOR. Broadly, the
served magnitude feasible in Li interacting nonspecifical-
spectrum of isolated atoms is expected to be similar to
ly with its host. We therefore reject the Jahn-Teller
that of the gaseous atom, as regards both line intensity
explanation and propose the observed triplet to be caused
and position, provided that one corrects for the "matrix
by exciplex formation between the excited Cu (or alkali)
shift" arising from nonspecific interactions between
atom and one of its rare-gas neighbors. The excited
host and guest and for the degeneracy of the excited state
compound so formed may nevertheless be fluxional, with
(if the intensity is obtained from emission). Although
the Cu bonded to only one rare- gas atom at anyone time.
this procedure yields a good overall understanding of the
A Cu*Ar bond of mainly Cu-4p and Ar-3d character (4d
spectra of isolated atoms, it fails quantitatively and even
for Kr and 5d for Xe) gives rise to three molecular
qualitatively in some important ways. So, for instance,
states: 22:: 1/ 2, 2nl/2' and 2n 3/ 2. Such exciplexes are well
the line at 207 nm of Cu in argon, usually assigned to the
2S1 /2 - 2P3 /2 (3cf4s14pl) transition, is roughly 60% as in-
known in alkali metal/rare gas systems. 40 Moreover,
the potential energy calculations of Baylis 41 and experi-
tense as those at 298, 302, and 306 nm which are as-
mental work due mainly to Gallagher 40 indicate that the
signed to the2S1/2-2Pl/2,3/2 (3d 10 4p 1) transition. In the
gas phase, on the other hand, the line associated with 22::1/2 state of alkali-rare-gas complexes is dissociative
the former transition is less than one-tenth as intense while the two (spin-orbit) 2rr states are bound. Vibra-
as that of the latter. The fact that the 2S1/2-2P 1/ 2,3/2 tional fine structure has even been detected in both the
transition forms a triplet rather than a doublet, as in absorption spectrum to and emission spectrum from the
21T3/2 state of CsAr. 40
the gas phase, is another feature of matrix-isolated
atoms whose origin is not yet fully agreed upon and Our proposa~, therefore, is that the Cu 2S_2P transi-
which is not conveniently obtained from AMCOR. The tion in argon is in fact three transitions cooresponding
problem of 25 - 2P triplets is not a new one. Similar to (Cu) 25 (formallyX 22::1/2)-(Cu*Ar)22:: 1/ 2, 2n3/2' 2nl/2'
triplets are obtained for alkali atoms in rare gase ma- and the components of the resulting triplet are split even
trices. 13.34 (With the alkalis one often obtains more than farther by the noncubic environment left behind as a re-
one triplet or sometimes even quintets. These, how- sult of the unique interaction between the Cu and one of
ever, reduce to a single, well-defined triplet on anneal- its argon neighbors. Assuming this model to hold for
ing. ) both Group IA and IB metals, we can now understand
For a 25_ 2p transition to be a triplet one must either both the large Stokes shift and the single band structure
trap Cu atoms in more than one matrix site or remove, observed in the fluorescence spectrum of isolated Li. 39
somehow, the degeneracy of the p orbitals of the Cu. Presumably, nonradiati ve processes depopulate all but
The cubic environment around all the usual sites in a the lowest vibrational level of only one of these states
regular, face-centered cubic crystal such as that of (perhaps the 2n3/2 state), which radiates to the dissocia-
solid argon or solid methane does not lift the degeneracy tive ground state.
of p orbitals. Several proposals have been made to ex- In the gas phase the 22::1/2 state of the alkali-noble-gas
plain the observed triplet structure of 25_ 2p absorptions. complexes is highest in energy. Matrix shifts are suffi-
Although, in the main, these have been directed towards ciently large, however, to place that state below or be-
isolated alkali atoms, similar arguments apply to group tween the other two in energy. We feel, however, that
IB atoms as well. Among these are (i) multiple trapping spin-orbit coupling is not large enough in Cu*Ar to war-
sites; (ii) a static distortion of the matrix ariSing from, rant assigning the outer two components of the triplet to
for instance, a local change in crystal structure from
the 2n1 / 2,3/2 states.
ccp to hcp,33 or as a result of a vacant argon site next
to the metal atom 35 ; (iii) Jahn-Teller effect arising from A complete assignment, then, amounts to deciding
the interaction of the excited metal atom with normal vi- whether the highest or lowest energy transition involves
brations of the host of the required symmetry36; and the 22::1/2 excited state. Our data cannot resolve this
(iv) non-nearest-neighbor, guest-guest interactions. 13 dilemma. Interpreting recently published data of Kolb
The last was disproved by Brewer and Kin~2 by codepos- et al. 33 in terms of our model, however, can simulta-
iting more than one group IB metal together and noting neously provide us with an assignment and help ration-
that the 2S - 2 P triplet remained unchanged. Moreover, alize their data, Kolb found that the peak positions of
the ESR studies of Kasai 37 indicate that Cu in solid argon the two low-energy components of- the triplet belonging
occupies a single, highly symmetric site, precluding to Cu in Kr and Xe decrease in energy with increasing
static, ground-state perturbations of the atom and also matrix temperature while the highest energy component
multiple sites. Kasai's work constrasts with ESR studies increases with increasing temperature. For Cu in Ar
on trapped alkali atoms which indicate the presence of the highest energy state decreases slightly while the
several sites as borne out by the uv-visible spectra lower two decrease rapidly with increasing temperature.
which show several 2S _ 2P triplets. 38 By assuming that the interaction between the excited

J. Chern. Phys., Vol. 67, No.9, 1 November 1977

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
M. Moskovits and J. E. Hulse: Spectroscopy of copper clusters 4275

pend, of course, on the magnitude of the Franck-Condon


factors pertaining to the transition and good correlation
between the intensities of the electronic bands of the
matrix-isolated atoms and those in the gas phase would
not necessarily be expected.
The frequencies and assignments of bands associated
with Cu isolated in Ar and CH 4 are given in Table I.

CU2. CU3, and CU 4


Lines in the spectrum of matrix-isolated Cu other
than those belonging to single Cu species have been re-
ported by several workers 42 • 2o and attributed to Cu
clusters. One of us (M. M.), for instance,20 assigned
all but two of the "extra" lines (similar to those shown
" in Fig. 1) to Cu2. That assignment is incorrect, how-
ever; several of the bands in the spectra belong to Cu 3 •
Moreover, bands belonging to CU 4 (and perhaps even
R CUs ) grow in on warmup.
FIG. 6. Schematic representation of the potential for the
ground state and three excited states of CuAr vs internuclear The rationale that we use to determine the size of a
separation. The solid curves are similar to what one would cluster giving rise to a particular band was outlined by
expect in the gas phase, the dotted curve represents the re- us in Ref. 30. Broadly, the absorbance of a line be-
pulsive interactions with the surrounding host suffered by the longing to CUn when divided by that of a line belonging
two constituent atoms of the CuAr molecule at large separation. to Cur, will depend on Mo, the metal-to-gas ratio ap-
proximately as

copper atoms and their remaining noble gas neighbors AcuJAcu ex M~ ,


(with which they interact nonspecifically) is attractive, where Cl is a number between x - 2 and x-I in value,
and making reference to Fig. 6, one sees that as the and closer to the latter, Mo is the metal-to-gas ratio,
temperature is raised, increasing the host-host distance, and A denotes absorbance. So, a plot of 10gA cu/A Cu
the nonspecific interaction between Cu* and its neighbors versus 10gMo will be a straight line of slope approxi-
decreases. This, in turn, increases the strength of the mately 0, 1, and 2 for, respectively, an absorption of
Cu*-noble gas-atom bond in the exciplex and decreases Cu, Cua, and Cu 3 • The above expression refers to spec-
its bond length. If the highest energy component of the tra obtained before warmup and was obtained by assum-
triplet is assigned to the excited 2.0 1/2 repulsive state, ing the clusters to form from monatomic copper in the
then, depending on the initial internuclear distance, one yet unfrozen layers of the matrix or the cooled gas phase
can have the peak positions of the lower two components immediately above it. (One would have to modify the
decrease in frequency while the upper increases. analysis somewhat should sizeable quantities of CU 2 ex-
ist in the gase phase. )
Extrapolating these arguments to the Cu/CH 4 system,
a similar exciplex may be formed whose bond has mainly The spectra of Cu clusters in CH 4 show fewer over-
Cu 4p and CH 4 IT2 character. The latter state is even laps than those in Ar. We will therefore proceed by dis-
closer in energy to the Cu 4p state than the 3d state of cussing that system. Analyzing the Cu in CH 4 spectra
Ar, implying even larger separation between the compo- in the manner outlined above, we see, for instance (Fig.
nents of the multiplet of Cu *CH 4 than for Cu* Ar, as is 7), that the line at 221 nm belongs to Cu, those at 256
observed. In contrast with the latter, the complex with and 380 nm to Cu2 , and those at 227, 260, and 533 nm
methane may have several forms according to the align- to Cu3 • Because some bands overlap, clear-cut assign-
ment of the hydrogen atoms with respect to the Cu-C ments of the features near 280 and 400 nm could not be
axis, resulting in more than one triplet. If, for instance, made using the above method alone and warmup data had
only two are preferred; a C 3V and a C 2v complex, say, to be used to help separate bands belonging to aggregates
then, in principle, a sextet would result made up of two of different sizes. Presumably all lines belonging to a
triplets: 2Al ZEI/Z' 2ES/2 for the former and a zA h ZBr, single species would grow or decay together on warmup.
2 B z for the latter. (Actually, owing to spin-orbit cou- A typical warmup result is shown in Fig. 8. From it and
pling, each of these components is a Kramers doublet results of similar experiments were deduced that a line
whose degeneracy can only be removed magnetically. at 260 nm (which forms part of a doublet) belongs to CU 3
The zA l of the C3v complex is formally a 2E 1 /2' So too as does a line centered near 280 and 404 nm. In addi-
are the three states of the C2v complex.) The 2Al states tion, the weak lines centered at approximately 335 nm
in the two forms would presumably be dissociative as in probably also belong to CUs • On prolonged warmup two
the C.. v case and may therefore coincide in energy, lines at approximately 272 and 424 nm grow in, attaining
yielding the observed quintet. By similar arguments considerable size (Fig. 4). These we attribute to Cu 4 •
to those above, the high-frequency lines observed in the For clarity of presentation and subsequent discussion,
spectra of Cu in Ar and CH 4 also belong to the respective we show the spectra of the various clusters separately
exciplexes. The intensities of the transitions would de- in Fig. 9. Other lines have undoubtedly been missed in

J. Chern. Phys., Vol. 67, No.9, 1 November 1977

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
4276 M. Moskovits and J. E. Hulse: Spectroscopy of copper clusters

"
f\
~
~\ (U4
211 424

A
W Lfl')
-~-
m
~-----............-
404
/~ CU 3
5'3

2'><, ,283 3BO

AJl ~-'-
200-'---300--406- ---sao' "---'--6-66
CU2
nm
FIG. 9. A schematic representation of bands assigned to Cu 2,
CU3 and Cu 4 •

ue of ~o, we arrive at a matrix shift (Vrnatrlx - Vga.) of


1960 cm- I •

-1 o As mentioned previously, no substantial growth of the


log Mo Cu cluster species in the range CU 2-CU 4 was observed
FIG. 7. A plot of -logA/A 216 vs 10gMo. Ax, A 216 , and Mo on warming the Cu in Ar matrices. On prolonged warm-
are the absorbance due to a cluster band, that of the Cu line up one obtained instead a broad band centered at about
absorbing at 216 nm and the total metal to gas ratio. The last 233 nm and a weaker band at 281 nm. These may belong
in arbitrary units. Open circles, filled circles, and triangles to CU 5 or higher species. The breadth and position of
indicate, respectively, the absorbances of the 221 nm band of
the former foreshadow the spectrum of colloidal copper.
Cu, the 283 nm band of CU2' and the 533 nIT! band of Cu 3.

SPECTRAL ASSIGNMENTS
Since molecular states are labeled with irreducible
the forest of bands occupying our spectral region. representations of the group to which the molecule be-
The spectrum of Cu in Ar is similar to that of Cu in longs, one must know the geometry of the clusters be-
CH 4 • The matrix shifts are, of course, different in the fore formal spectral assignments are possible. Our data
two media, causing the band positions and, at times, do not yield geometries. Molecular orbital calculations
the severity of overlap to differ. The correlation of by Baetzold for Agx 8 and Anderson for CUr 43 indicate,
bands for the two environments is given in Table I. however, that close-packed geometries seem to be less
There are no discrepancies except that in Ar it is clear stable than linear or zigzag forms of the clusters. The
that the band attributed to CU2 which comes at 380 nm in latter, for instance, finds that CU 3 and CU 4 have C2v and
methane is at least two bands which, in Ar, are centered C 2h structures, respectively, with very small deviations
at approximately 345 and 400 nm, respectively. The lat- from linearity. Because of the serious approximations
ter is quite likely the X -A progression reported by made in those calculations, one cannot take these shapes
Kleman and Lindkvist 22a for gaseous CU 2 with ~o coming to be conclusive. It is, however, extremely useful to
at 490 nm. This band has been assigned by Aslund et use energy-level diagrams, such as those calculated by
al. 22b to the I~; - Inu (d1T,- SGu ) transition. We concur Anderson on the basis of these geometries, in discuss-
with this assignment. The only other likely one, that of ing the spectra. Other quantum-chemical calculations
SG,- SGu , would not show such large vibrational fine have been performed on clusters of copper as well6 and
structure since in the excited state the Cu-Cu single will be referred to as required.
bond, which has mainly Cu 4s character, is essentially Fig. 10 shows the energy levels of Cu, Cu2 , Cu 3 , and
absent. On the other hand, the d1T,- SGu transition, and CU 4 as calculated by Anderson. 43 The eigenenergies of
for that matter any transition from the d manifold, is the clusters seem to fall into three groups of levels.
expected to weaken the Cu-Cu bond only slightly. De- The lowest energy group (1) consists of filled states hav-
spite many attempts, we have not been able to find vi- ing mainly sand d character. The middle set of states
brational fine structure in either of our two bands (or (II) is empty and possesses largely S character with some
indeed any of the Cll,; absorptions). If, however, we p as well. The highest set (III) consists of unfilled states
take the low-energy edge of the 400-nm band as our val- of p character with some S character. These rather
arbitrary divisions are shown in Fig. 10. The gap be-
tween the highest occupied molecular orbital (HOMO)
and the lowest unoccupied molecular orbital (LUMO) de-
creases as one proceeds from CU 2 to Cu 4 • (In the last,
FIG. 8. The change in ab- in fact, this transition is predicted to lie in the near ir. )
sorbance of the 227 nm (filled This gap should, 'in fact, reduce to a very small quantity
circles), 553 nm (open cir- as bulk copper is approached (as shown by Baetzold for
cles) , and 256 nm (triangles) Ag), with regions I and II forming the filled and empty
bands vs warmup time. The
halves of the conduction band of copper. Likewise, re-
first two bands belong to CU3;
the last to CU2'
gion III forms the empty sp band of the bulk metal.

The HOMO-LUMO transition should therefore recede


warmup time quickly into the infrared, broadening as it goes, until

J. Chern. Phys., Vol. 67, No.9, 1 November 1977

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
M. Moskovits and J. E. Hulse: Spectroscopy of copper clusters 4277

o Cu CU2 CU4
CU3

-5
~lT'-

pu-
~n-
p~ - ) ;)1
"·-1 '" 15(, •-I'+ " :lJ7
1100
535 .180
111. }
-s 380
s
"'1.1:
-
~cr

-10
-d
.11" ...
.(;
.r=
"to
45- ! \
eV ,IlT
40"

FIG. 10. The eigenvalue spectra calculated by Anderson43 for CU2' Cu3, and Cu4. The VOIP's of the Cu 3d, 4s, and 4p used in
the calculations are also shown. The highest filled orbital is indicated by either a single or double vertical arrow for hall and
fully filled orbitals. The spd labels next to the levels of CU2 indicates their major character. In fact most levels marked d and p
contain substantial s character and the su and su* states contain a great deal of d and p character, respectively. Large vertical
arrows indentify possible transitions giving rise to the observed bands indicated beside each.

in the bulk it is replaced by the spectrum of the free many of the features displayed by bulk copper, many
electrons that extend from the far ir to the visible. This aspects of these clusters such as the energies and spd
picture agrees very well with our experimental results compositions of its valence orbitals vary dramatically
if we identify one of the broad bands of CU 2 in the vicinity with cluster size. This behavior has been put to good
of 400 cm-1 (all wavelengths will refer to the Cu in use in heterogeneous catalysis where it is known that
methane system) with the sa, - sau transition (i. e., a catalysts containing particles of different mean radius
region I - II transition). In the context of bulk copper, often behave quite differently. 45 One would expect,
this absorption is an intraband transition. (The other moreover, that the change in electronic properties per
band of CU 2 in that region is probably the sa- pa* transi- atom would be most dramatic in very small clusters
tion.) The CU3 band at 533 nm is the HOMO- LUMO such as those we described. Mass filtering these ag-
transition in that molecule which begins the trek to low gregates and isolating them in an appropriate support
energies on its way to forming the Drude region of the so that all the clusters within a sample are of uniform
spectrum of bulk copper. CU 4 has two states in region size could yield a spectrum of catalysts each containing
II. The HOMO- LUMO transition is already in the near the same kind of metal but exhibiting totally different
ir and the band at 424 nm arises from a transition to catalytic properties. Such a prospect is close at hand.
what may be termed an incipient conduction band upper
boundary. ACKNOWLEDGMENTS
The other observed transitions are assigned in a gen- The authors are indebted to the National Research
eral way in Fig. 10 and Table I. Because there are Council of Canada, the Research Corporation, and the
many closely spaced levels in the energy diagrams, firm Atkinson Foundation for financial support. One of us
assignments are unjustified. Briefly, with the exception (J. E. H. ) wishes to thank the N. R. C. C. for a scholar-
of the 355 nm transition in CU 3 which is still a region I- ship. The technical assistance of K. Molnar, T. Hig-
n transition, all others are region I-III transitions, i. e. son, A. Campbell, and M. Mittelstaedt is also greatly
incipient interband transitions. These transitions re- appreciated. Thanks are also due to professor A. B.
main essentially in the same region of the spectrum Anderson for allowing us to quote the results of his cal-
(200-300 nm) in all copper clusters from CU2 to the culations on copper clusters prior to their publication.
bulk.
A spectrum of bulk copper in colloidal form is shown IF. F. Abraham, Homogeneous Nucleation Theory (Academic,
in Fig. 3. The broad feature extending from 200 nm to New York, 1974); J. Hecht, J. D. Eversole, and H. P.
lower energies is due to interband transitions. The fea- Broida, J. Appl. Phys. 48, 1503 (1971) and references there-
ture at approximately 600 nm is the so-called optical in.
conduction resonance and is due to conduction electrons 2J. F. Hamilton, J. Vac. Sci. Technol. 13, 319 (1976),
3J . R. Anderson, Structure of Metallic Catalysts (Academic,
(Le., intraband processes). Its position and prominence New York, 1975).
(in fact the entire spectrum) can be calculated with con- 4M. C. Desjonqueres and F. Cyrot-Lackmann, Surf. Sci. 53,
siderable accuracy as described in Ref. 27 using the op- 429 (1975); R. O. Jones, P. J. Jennings and G. S. Painter,
tical constants of bulk copper given by Johnson and ibid., p. 409.
Christy.44 Such a calculated spectrum is shown in 5G. Blyholder and M. C. Allen, J. Amer. Chern. Soc. 91,
Fig. 3. 3158 (1969); G. Blyholder, Surf. Sci. 42, 249 (1974).
GR. P. Messmer, S. K. Knudson, K. H. Johnson, J. B. Dia-
Although small copper aggregates already possess mond, andC. Y. Yang, Phys. Rev. B 13, 1396 (1976).

J. Chern. Phys., Vol. 67, No.9, 1 November 1977

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
4278 M. Moskovits and J. E. Hulse: Spectroscopy of copper clusters

7J . E. Hulse and M. Moskovits, Surf. Sci. 57, 125 (1976). 27M. Moskovits and J. E. Hulse, J. Chern. Phys. 66, :3988
sR. C. Beatzold, J. Chern. Phys. 55, 4363 (1971). (1977).
9J . C. slater and K. H. Johnson, Phys. Rev. B 5, 844 (1972), 2BW. Schulze, H. U. Becker, and D. Leutloff, J. Phys. (Paris)
IOSee References in Ref. 6. (in press) (private communication),
lIN. Rosch and T. N. Rhodin, Phys. Rev. Lett. 32, 1189 29M. Moskovits, Ph. D. thesis, University Microfilm, Ann
(1974), Arbor, Michigan (1970).
12A. B. Anderson, J. Chern. Phys. 62, 1187 (1975). 30M. Moskovits and J. E. Hulse, J. Chern. Soc. Faraday Trans
13 L . Andrews and G. C. Pimentel, J. Chern. Phys. 47, 2905 2 73, 471 (1977).
(1967) . 31 0. H. W. Carstens, W. Brashear, D. R. Eslinger and D. 1\1.
14T. C. DeVore, A. Ewing, H. F. Franzen, and V. Calder, Gruen, Appl. Spectrosc. 26, 184 (1972).
Chern. Phys. Lett. 35, 78 (1975). 32L. Brewer and G. King, J. Chern. Phys. 53, 3981 (1970).
15 L. Brewer and C. -A. Chang, J. Chern. Phys. 56, 1728 33D. M. Kolb, O. Le utl off , and W. Schulze, J. Chern. Phys.
(1972). 66, 2806 (1977).
16L . Andrews, J. Mol. Spectrosc. 61, 337 (1976). 34C. B. Meyer, J. Chern. Phys. 43, 2986 (1965).
17 L. B. Knight and M. A. Ebener, J. Mol. Spectrosc. 61, 412 35 M. Brith and O. Schnepp, J. Chern. Phys. 39, 2714 (1963).
(1976). 36M. McCarty and G. W. Robinson, Mol. Phys. 2, 415 (1959).
18K. Jansson and R. Scullman, J. Mol. Spectrosc. 61, 299 37p. H. Kasai and D. McLeod, J. Chern. Phys. 55, 1566 (1971),
(1976). 38C. K. Jin, V. A. Bowers, E. L. Cochran, and S. N. Foner,
19 E . P. Kiindig, M. Moskovits and G. A. Ozin, Nature 254, Phys. Rev. 126, 1749 (1962).
503 (1975). 39A. A. Belyaeva, Yu. B. Predtechenskii, and L. O. Shcherba,
2oH. Huber, E. p. Kiindig, M. Moskovits, and G. A. Ozin, Opt. Spektrosk. 34, 40 (1973) [Opt. Spectrosc. 34, 21 (1973)].
J. Am. Chern. Soc. 97, 2097 (1975). 40C. G. Carrington, D. Drummond, A. Gallagher, and A. V.
ZIT. A. Ford, H. Huber, W. KlotzbUcher, E. P. Ktindig, Phelps, Chern. Phys. Lett. 22, 511 (1973) and references
M. Moskovits, and G. A. Ozin, J. Chern. Phys. 66, 524 therein.
(1977). 41W. E. Baylis, J. Chern. Phys. 51, 2665 (1969),
22(a) B. Kleman and S. Lindkvist, Ark. Fys. 8, 333 (1954). 42J. S. Shirk and A. M. Bass, J. Chern. Phys. 49, 5156
(b) N. Aslund, R. F. Barrow, W. G. Richards, and D. N. (1968), and Ref. 32.
Travis, ibid. 30, 171 (1965). 43A. B. Anderson (manuscript in preparation) (private com-
23Ibid. 9, 385 (1955). municated) .
24B. Kleman, S. Lindkvist, and L. -E. Selin, ibid. 8, 505 (1954). 44p. B. Johnson and R. W. Christy, Phys. Rev. B 6, 4370
25R . E. Steele, J. Mol. Spectrosc. 61, 477 (1976). (1972).
26M. G. Mason and R. C. Baetzold, J. Chern. Phys. 64, 271 45R . van HardeveldandF .. Hartog, Adv. Catal. 22, 75 (1972)
(1976). and references therein.

J. Chern. Phys., Vol. 67, No.9, 1 November 1977

Downloaded 16 Jul 2013 to 128.197.27.9. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

You might also like