Volume 192, number 1 CHEMICAL PHYSICS LETTERS 24 April 1992

Electronic shell structure in the ionization potentials

of copper clusters
Mark B. Knickelbein
Chemistry Division, Argonne National Laboratory, Argonne, IL 60439, USA

Received 3 1 December 199 1

The ionization potentials ( IPs) of copper clusters (Cu,) up to 150 atoms have been bracketed using laser photoionization.
Large decreases in IP at n = 2, 8, 20, 30, 34, 40 and 60 atoms are observed, as well as a strong odd-even alternation extending
beyond 100 atoms. These features correspond to electronic shell and subshell closings predicted for systems of n valence electrons
confined to spherical or spheroidal potentials. The decreases in IP expected to accompany shell closings at n= 58 and 92 are
absent, however, and a new set of local extrema are observed, including local maxima at n=57 and 88-91 and minima at n=49
and 92. The possibility of perturbations to the electronic energy level structure due to geometric packing constraints is discussed.

1. Introduction ization study of cold copper clusters from which pre-

The success of the electronic shell model in pre-
dicting discontinuities in the size-dependent prop-
erties in clusters of monovalent free electron metals
is now well established [ 1,2 1. This model specifies
the ordering and degeneracies of levels occupied by
delocalized electrons within a structureless, square
(or nearly square) potential well and predicts spher-
ical geometries accompanying electronic major shell
closings at ne=2, 8,20,40, 58, 92 ... , where n, is the
number of delocalized electrons obtained from va-
lence considerations. Between (spherical) closed
shells, ellipsoidal geometries are predicted to be the
most stable [ 21, leading to the removal of orbital de-
generacies and to additional electronic subshells. The
simple shell model successfully accounts for much of
the observed line structure (i.e. “magic numbers”)
observed in the abundance [ 3-5 ] and ionization po-
tentials [ 6-8 ] of neutral alkali clusters and in the
stability of ionic clusters of alkali [ 51 and coinage
metal [ 9,10 ] atoms. Calculations based on more so-
phisticated potentials, such as the self-consistent jel-
lium approach [ 11,121 and effective medium theory
[ 131 provide more quantitative results, but still pre-
dict most of the important shell-like features given
by the simple models [ 1,2] .
In this Letter, we present the results of a photoion-
cise estimates of the vertical ionization potentials
(IPs) are obtained. The IPs obtained in this study
display local extrema characteristic of electronic shell
filling as well as some deviations from the expected
shell filling pattern for larger clusters. In a previous
photoionization study, Powers et al. employed three
discrete wavelengths to obtain coarse IP brackets for
copper clusters up to CuZ9 [ 141. In the present study
we have extended the size range investigated to 150
atoms, employing ionization energy intervals of 0.05-
0.07 eV. As shown by Alameddin et al. in an accom-
panying paper [ 15 1, the ionization potentials of sil-
ver clusters display many of the same electronic shell
features displayed by alkali and copper clusters.
2. Experimental methods
Copper clusters were generated by laser vapori-
zation of copper atoms from an isotopically pure 63Cu
target using a cooled (77 K) flow tube source em-
ploying continuous helium flows [ 161. After for-
mation, the cold clusters expanded into vacuum
through a 0.074 cm diameter orifice in the end of the
flow tube. The free jet was skimmed, forming a mo-
lecular beam, which passed through a differentially
pumped vacuum stage into a detection chamber

0009-2614/92/$ 05.00 0 1992 Elsevier Science Publishers B.V. All rights reserved. 129
Volume 192, number 1 CHEMICAL
maintained at lop6 Torr. The clusters were pho-
toionized within the ion extraction region of a time-
of-flight (TOF) mass spectrometer using tunable ul-
traviolet light from a dye laser which had been fre-
quency doubled (Aov = 206-247.5 nm) or tripled
(A,,= 200-206 nm) by second harmonic genera-
tion and sum frequency generation, respectively. The
mass separated cluster photoions were detected us-
ing a microchannel plate ion detector, the output of
which is amplified and digitized. TOF spectra were
recorded at 2.5 nm intervals from 200 to 247.5 nm
by summing 103-1 O4 spectra in a microcomputer.
The fluence of the ionization laser was controlled us-
ing a continuously variable optical attenuator, and
was normally limited to 6 175 pJ cm- * to minimize
the effects due to multiphoton processes.
3. Results and discussion
Time-of-flight mass spectra for 63Cu, recorded at
201.7, 222.5 and 240.0 nm are shown in fig. 1. The
fluence dependence of cluster ion intensity was in-
$
3 240.0 nm
Fig. 1. Photoionization time-of-flight mass spectra of copper clusters. A portion
to demonstrate the odd-even alternation in ion intensities.
130
PHYSICS LETTERS 24 April 1992
vestigated at 227.5 nm (5.45 eV). Those clusters
which could be ionized at this wavelength displayed
ion intensity which varied linearly with laser fluence
up to the maximum fluence employed, 800 pJ cm-‘.
Those clusters which were absent in the mass spec-
trum at low fluences ( G 100 pJ cm-*) remained so
at 800 pJ cmp2. Because fluences were normally lim-
ited to < 175 pJ cmp2, it is concluded that the clus-
ter mass spectra reported here are the result of one
photon ionization and are not distorted by higher or-
der ionization or fragmentation processes.
3.1. Ionization potentials and the electronic shell
model
Vertical ionization potentials, bracketed for Cu,
(n= 3-120) from a series of mass spectra obtained
at discrete 2.5 nm intervals, are plotted in fig. 2.
While Cug, Cu12 and Cu14 can be ionized with one
photon at 193 nm (6.42 eV) [ 161, no ion signal was
observed for these species at 200 nm (6.20 eV), es-
tablishing their IPs at 6.3 + 0.1 eV. Similar brack-
eting arguments place the IPs of Cu,, Cug, and Cus
46
57 61
of the spectrum recorded at 240.0 nm has been expanded
Volume 192, number 1 CHEMICAL PHYSICS LETTERS
6.4
6.2
6.0
S-
s 5.6
B
5.4
5.2
5.0
4.6
20 40 60 60 100
n ter cations [ 91 and anions [ lo], in which enhanced
Fig. 2. Vertical ionization potentials for copper clusters. Filled
circles mark the center of the appearance interval, while the error
bars span the interval.
between 6.42 and 7.90 eV [ 171.
Using fixed wavelength (excimer laser) photoion-
ization sources in conjunction with fluence depen-
dence studies, Powers et al. [ 141 concluded that all
clusters from Cug to CuZ9possess IPs lower than 5.58
eV. While our mass spectrometric measurements
confirm this conclusion for some of the odd n clus-
ters in this size range, we find that other odd n clus-
ters (e.g. CU,~, CulS) and all even n clusters except
for CuZ2,Cu14 and CuZ8possess IPs higher than 5.58
eV. The source of this discrepancy may be due to re-
sidual multiphoton ionization not accounted for in
the previous study or, in cases where the cluster IPs
lie close to the laser photon energy, to photoioni-
zation of internally excited clusters. In a followup in-
vestigation of CUE,Morse et al. [ 18 ] bracketed the
IP of this species at 5.48 f 0.5 eV using resonant two-
photon ionization (R2PI) spectroscopy, in reason-
able agreement with the 5.8OkO.04 eV value ob-
tained in the present study.
The large decreases in copper cluster IPs following
n= 2, 8, 20, and 40 and the odd-even alternation
pattern are similar to the behavior displayed by al-
kali metal clusters [ 6-81 and silver clusters [ 15 1,
and are signatures of spherical electronic shell clos-
ings expected to occur assuming n, = n as for neutral
alkali clusters. In additon, we observe significant IP
decreases leading to local minima following n= 30,
34, and 60, predicted to accompany the subshell
24 April 1992
closings for distorted (spheroidal) geometries
[ 2,11,12]. While the drops in IP following n= 10,
14, 18,26 and 36 can also be correlated with known
electronic subshell closings [ 2 1, these features do not
stand out from the strong odd-even alternation pat-
tern which characterizes the size range up to n = 4 1.
The local IP maxima reported here for neutral Cu,
are mirrored in the electron affinity (EA) measure-
ments reported by Pettiette et al. [ 191 for Cu; spe-
cies to n = 4 1, in which local minima in EA were ob-
served for these same values of n.
Electronic shell effects have also been observed in
1
120
the intensity distributions of sputtered copper clus-
stability are observed for Cuz+ 1 at n = 2, 8, 20, 34,
40, 58 and 92 and for CUT-,at n = 18,34 and 58 as
predicted by the shell model. Electronic shell clos-
ings in copper clusters are expected to be accom-
panied by sudden decreases in IP at the same values
of n, for which enhanced ion stabilities are observed,
as is the case for alkali clusters [ 5,6]. While we do
observe IP drops signifying shell/subshell closings at
n = 8, 14, 20, 30, 34, and 40, we do not observe sim-
ilar features at n = 58 or 92 as expected. Instead, lo-
cal maxima are observed at n = 57 and 88-9 1 with
local minima at n = 49,6 1 and 92, none of which are
predicted within the framework of the spherical/el-
lipsoidal shell models.
Because the simplest one-electron models are based
on structureless square well [ 1,2] or jellium [ 11,121
potentials in which the positive charge due to the ion
cores is assumed to uniformly fill the volume of the
cluster, they are unable to account for the influence
on the electronic level structure by the specific lo-
cations of the atomic cores, i.e. geometric structure
effects. It is possible that beyond 41 atoms, copper
clusters adopt rigid geometries (e.g. icosahedral )
rather than the spherical or spheroidal structures
predicted to minimize the electronic energy. The
equilibrium structures of neutral copper clusters from
3 to 29 atoms have been calculated by Christensen
et al. using the Monte Carlo method in conjunction
with a potential calculated using effective medium
theory [ 131. In general agreement with the simple
shell models [ 1,2], these calculations predict
(nearly) spherically symmetric geometries for Cus
and CuZo, with all other clusters assuming distorted
structures. Moreover, these calculations demon-
131
Volume 192, number 1 CHEMICAL
strate that the equilibrium geometries of copper clus-
ters in this size range are determined by a compe-
tition between two competing factors: ( 1) the
tendency of the ion cores to achieve the minimum
energy close packing configuration and (2) the ten-
dency of the valence electronic charge distribution to
achieve the shape which minimizes the electronic
energy as predicted, for example, in the simple el-
lipsoidal shell model [ 21. The cessation of the elec-
tronic shell filling and odd-even alternation patterns
between Cu.+, and Cuho may be a result of these two
competing energy factors becoming similar in
magnitude.
An alternative explanation for these apparent de-
partures from the shell model predictions is obtained
by considering the details of the photoionization
process itself. For (cold) clusters in which the geo-
metric structures of the neutral and ion are nearly
the same, the measured vertical IPs (IP,) will be very
good approximations to the adiabatic IPs (IP,) and
will be expected to display the fine structure asso-
ciated with electronic shell and subshell closings. In
general, however, the equilibrium geometry of a neu-
tral cluster Cu, may be different from that of the cor-
responding ion Cu,‘, so that the Franck-Condon
factors will be largest for ionizing transitions to ex-
cited vibronic states of the ion rather than to the
ground state. In this case, IP, will exceed IP, by an
amount which varies with the magnitude of the ge-
ometry change. It is conceivable that the IP fine
structure associated with an electronic shell closing
may be obscured if the deviation of IP, from IP, var-
ies significantly from one cluster size to the next in
the region of that shell closing. This may account for
the anomalous IP behavior displayed by Cu42_60 and
Cug2 and may be the source of the discrepancies be-
tween the local IP extrema observed in this study and
the shell closing patterns observed in the stabilities
of copper cluster ions [ 9, lo]. However, it is ex-
pected a priori that deviations of IP, from IP, will
be largest for small clusters and diminish as cluster
size increases, which is not consistent with our ob-
servations. Whether the IP anomalies observed for
n = 42-60 and n = 92 are due simply to the effects of
large geometry changes on the Franck-Condon en-
velopes or due to significant geometry-dependent
shifts in energy levels themselves is uncertain. Cal-
culations employing realistic potentials may be help-
132
PHYSICS LETTERS 24 April 1992
ful in distinguishing between these two possibilities.
A chemical reactivity investigation by this group
[ 201 has shown that copper clusters in the same size
range studied here display reactivity patterns in which
both geometric and electronic shell effects are man-
ifested. In the Cu, + O2 reaction, clusters associated
with electronic shell or subshell closings at n = 14, 20,
34,40, 58 and 92 were found to possess anomalously
low reactivities. By contrast, the Cu,+H,O
SCu,H,O equilibrium distribution displays no cor-
relation with electronic closed shells but rather is
characterized by Cu,-HZ0 binding energy maxima
at n = 57 (a local IP maximum, but not a known
closed electronic shell) and at n=72, 75, 78, 81, 84,
87, 93, 96 .... features which have been shown to be
signatures of the geometric (specifically icosahed-
raf) subshell closings at n - 1 atoms [ 2 11. Thus, the
icosahedral structure inferred for CugZ is at odds with
the spherical structure predicted in the shell picture.
As shown in the accompanying paper by
Alameddin et al. [ 15 1, the ionization potentials of
silver clusters display the same shell-like features at
n=20, 30, 34 and 40 and strong odd-even altema-
tion observed for copper clusters as well as unex-
pectedly large IP decreases at n = 48 and 60 not pre-
dicted in the simple shell models. In contrast to
copper clusters, however, silver clusters display the
expected shell closing features at n = 58 and 92.
Honea et al. [ 81 have shown that the IPs of cold
sodium clusters display IP dips characteristic of elec-
tronic shell closings at 91 (and 137) atoms rather
than at 92 (and 138) as expected. The similarity of
these deviations from the shell closing pattern near
n = 92 for clusters of sodium and copper atoms im-
plies that perturbations of the electronic levels arise
from similar geometric constraints.
3.2. An electrostatic model
The size dependence of metal cluster IPs are con-
veniently compared to the predictions of the clas-
sical conducting droplet model [ 22-25 1, according
to which the IP of a small conducting sphere exceeds
the work function (WF) of the corresponding infi-
nite, flat surface by an electrostatic term propor-
tional to the reciprocal of the sphere’s radius R,
Volume 192, number 1 CHEMICAL PHYSICS LETTERS
I . I . 8 I
0.0 0.1 0.2 0.3
I/R (A-')
Fig. 3. Comparison of measured Cu, ionization potentials (n= j-
150) with the preditions of the conducting spherical droplet model
(solid line) with WF=4.35 eV.
e2
IP(R) =WF+ 2R.
Proceeding under the assumption that copper clus-
ters are roughly spherical in shape, we have calcu-
lated their “classical” radii R,, using the atomic den-
sity of bulk copper, to which an electron spillout
factor 6 of 0.74 8, has been added as prescribed by
Perdew [ 22 ] such that R = R,,+ S. Using these esti-
mates of copper cluster radii and WF=4.35 eV, we
have plotted in fig. 3 IP(R) - WF versus R --I for
clusters up to 150 atoms. This value of WF (which
is 0.30 eV smaller than the polycrystalline value
[ 261) was chosen so that the points at n = 20,40 and
91 lie on the line of slope fe’ as given by eq. ( 1).
Engel and Perdew have recently shown [ 251 that
(neglecting shell oscillations) the IPs of jellium
spheres may approach the bulk work function with
a slightly weaker radius dependence than predicted
in the purely classical model, namely as ( 1 +c)e2/R
where cx - 0.07. While copper cluster IPs decrease
with increasing R in rough agreement with the pre-
dictions of the simple electrostatic expression (eq.
( 1) ), a critical comparison between these two alter-
native R - ’ scaling factors and the IP data set is com-
plicated by the significant odd-even oscillations and
discontinuities due to electronic shell structure
throughout this size range.
24 April 1992
4. Summary
Most of the major features in the IPs of copper
clusters containing 41 or fewer atoms can be satis-
factorily explained by the electronic shell model. For
clusters larger than 42 atoms, the IP extrema ex-
pected to accompany known shell closings are not
observed, suggesting that larger clusters may assume
geometries which are significantly different from the
minimum energy configurations predicted in the
0.4
simple shell theories. A better understanding of the
complex interaction between cluster geometric and
electronic structures will undoubtedly emerge as
knowledge of cluster geometries are obtained
(through chemical probe [ 20,211 or other tech-
niques) and as additional size-dependent spectro-
scopic and physical properties are measured.
(1)
Acknowledgements
The author thanks Professor Manfred Kappes for
several informative and interesting discussions and
Dr. W.J.C. Menezes for his assistance in the early
stages of the experiments. This work is supported by
the US Department of Energy, Office of Basic En-
ergy Sciences, Division of Chemical Sciences, under
Contract W-31-109-ENG-38.
References
[ 1] W.A. de Heer, W.D. Knight, M.Y. Chou and M.L. Cohen,
in: Solid state physics, Vol. 40, eds. H. Ehrenreich and D.
Tumbull (Academic Press, New York, 1987) p. 93.
[ 21 K. Clemenger, Phys. Rev. B 32 ( 1985) 1359.
[ 31 W.D. Knight, K. Clemenger, W.A. de Heer and W.A.
Saunders, Phys. Rev. Letters 52 ( 1984) 2 14 1.
[4] M.M. Kappes, M. Schar, P. Radi and E. Schumacher, J.
Chem. Phys. 84 ( 1986) 1863.
[ 51 C. Brechignac, Ph. Cahuzac and J.Ph. Roux, J. Chem. Phys.
87 (1987) 229.
[6 ] W. A. Saunder, K. Clemenger, W.A. de Heer and W. D.
Knight, Phys. Rev. B 32 (1985) 1366.
[7] T. Bergmann and T.P. Martin, J. Chem. Phys. 90 ( 1989)
2484.
[8] E.C. Honea, M.L. Homer, J.L. Persson and R.L. Whetten,
Chem. Phys. Letters 17 1 ( 1990) 147.
[9] I. Katakuse, T. Ichihara, Y. Fujita, T. Matsuo, T. Sakurai
and H. Matsuda, Intern. J. Mass Spectrom. Ion Processes
67 (1985) 229.
133
Volume 192, number 1 CHEMICAL PHYSICS LETTERS
[lo] I. Katakuse, T. Ichihara, Y. Fujita, T. Matsuo, T. Sakurai
and H. Matsuda, Intern. J. Mass Spectrom. Ion Processes
7’4(1986) 33.
[ II] W. Ekardt and Z. Penzar, Phys. Rev. B 38 (1988) 4273.
[ 121 Z. Penzar and W. Ekardt, Z. Physik D I7 ( 1990) 69.
[ 131 O.B. Christensen, K.W. Jacobsen, J.K. Norskov and M.
Manninen, Phys. Rev. Letters 66 ( 199 1) 22 19.
[ 141 D.E. Powers, S.G. Hansen, M.E. Geusic, D.L.
Michalopoulos and R.E.Smalley, J. Chem. Phys. 78 ( 1983)
2866.
[ 15 ] CLAlameddin, J. Hunter, D. Cameron and M.M. Kappes,
Chem. Phys. Letters 192 ( 1992) 122.
[ 161 M.B. Knickelbein, S. Yang and S.J. Riley, J. Chem. Phys.
93 (1990) 94.
[ 171 B.J. Winter, E.K. Parks and S.J. Riley, unpublished results.
[ 181 M.D. Morse, J.B. Hopkins, P.R.R. Langridge-Smith and
R.E. Smalley, J. Chem. Phys. 79 (1983) 5316.
134
24 April 1992
[ 19 ] C.L. Pettiette, S.H. Yang, M.J. Craycraft, J. Conceicao, R.T.
Laaksonen, 0. Cheshnovsky and R.E. Smalley, J. Chem.
Phys. 88 (1988) 5377.
[20] B.J. Winter, E.K. Parks and S.J. Riley, J. Chem. Phys 94
(1991) 8618.
[21] E.K. Parks, B.J. Winter, T.D. Kl0tsandS.J. Riley, J. Chem.
Phys. 94 (1991) 1882.
[22] J.P. Perdew, Phys. Rev. B 37 (1988) 6175.
[23] G. Makov, A. Nitzan and L.E. Brus, J. Chem. Phys. 88
( 1988) 5076.
[24] W.A. de Heer and P. Milani, Phys. Rev. Letters 65 ( 1990)
3356.
[25] E. Engeland J.P. Perdew, Phys. Rev. B43 (1991) 1331.
[26] D.E. Eastman, Phys. Rev. B 2 (1970) 2.