Endo M., Iijima S., Dresselhaus M.S. (eds.) Carbon nanotubes
Подождите немного. Документ загружается.
METAL-COATED FULLERENES
U.
ZIMMERMANN,
N.
MALINOWSKI,"
A.
BURKHARDT,
and
T.
P.
MARTIN
Max-Planck-Institut fur Festkorperforschung, Heisenbergstr.
1,
70569
Stuttgart, Germany
(Received
24
October
1994;
accepted
10
February
1995)
Abstract-Clusters
of
c60
and
C,,
coated
with
alkali or alkaline
earth
metals are investigated using photo-
ionization
time-of-flight
mass
spectrometry.
Intensity anomalies in
the
mass spectra of clusters
with
com-
position
C,M,
and C70Mx
(x
=
0.
.
.500;
ME
f
Ca,
Sr,
Bal)
seem to be caused
by
the
completion of
distinct metal layers around a central fullerene molecule.
The
first layer around Cs0 or C,o contains
32
or
37
atoms, respectively, equal to the number
of
carbon rings constituting
the
fullerene cage. Unlike the
alkaline earth metal-coated fullerenes, the electronic rather than the geometric configuration seems to be
the
factor determining
the
stability
of
clusters
with
composition
(c60)"Mx
and
(C70),M,,
M
E
(Li, Na,
K,
Rb,
Cs].
The
units
CsoM,
and
C70M6 are found
to
be
particularly stable building blocks of the
clus-
ters. At higher alkali metal coverage, metal-metal bonding and an electronic shell structure appear.
An
exception
was
found for C60Li12, which is very stable independently of charge. Semiempirical quantum
chemical calculations support that
the
geometric arrangement of atoms is responsible for the stability
in
this case.
Key
Words-Fullerenes,
mass
spectrometry,
clusters,
electronic
shells,
icosahedral layers.
1.
INTRODUCTION
In their bulk intercalation phase compounds of
c60
and alkali or alkaline earth metals have been studied
intensively, spurred particularly by the discovery of
su-
perconductivity in several
of
these metal fullerides,
such as C6&, cs0Rb3, C60Ca,, etc.[l-5]. However,
despite the wealth
of
information that could yet be ex-
tracted from these fullerene compounds, we would still
like to return briefly to looking at some interesting ex-
periments that can be done by bringing just one sin-
gle fullerene molecule in contact with atoms of the
metals commonly used for the doping bulk fullerite.
The properties of these very small metal-fullerene sys-
tems then termed clusters, can be studied quite nicely
in the gas phase[6]. We observed that, in the gas phase,
such a single fullerene molecule can be coated with lay-
ers of various alkali and alkaline earth metals[7,8]. In
lhis contribution, we will focus primarily on the struc-
ture, both electronic and geometric, of this metal coat-
ing
of
the fullerenes
c60
and
C70.
The method we use to study these metal-fullerene
clusters
is
photoionization time-of-flight mass spec-
trometry (section
2).
The clusters are produced by
coevaporation of fullerenes and metal in
a
gas aggre-
gation cell. By ionizing and, in some cases, heating the
clusters with
a
pulsed laser, various features appear in
the mass spectra that contain the information neces-
sary to suggest a geometric or electronic configuration
for the cluster investigated.
When building clusters by coating the fullerenes
with metal, features similar
to
the electronic and geo-
metric shells found in pure metal clusters[9] are ob-
served in the mass spectra. In the case of fullerene
molecules coated with alkaline earth metals (section
3),
we find that a particularly stable structure is formed
*Permanent address: Central Laboratory
of
Photopro-
cesses, Bulgarian
Academy
of
Sciences,
1040
Sofia, Bulgaria.
each time a new layer of metal atoms has been com-
pleted around
a
central fullerene molecule, the stabil-
ity of these clusters seeming to be purely geometric in
origin. The first layer contains exactly the same
num-
ber of metal atoms as there are rings in the fullerene
cage. In growing additional layers, the metal might be
expected to prefer the icosahedral shell structure ob-
served in pure alkaline earth cIusters[lO,l
I].
However,
our measurements suggest a different growth pattern.
Coating the fullerenes with alkali metals (section
4),
the resulting structures seem to be primarily governed
by the electronic configuration. For example, the
charge transfer of up to
6
electrons to the lowest
un-
occupied molecular orbital
(LUMO)
of
C60
observed
in bulk alkali fullerides[5] is also observed
in
our ex-
periments, leading to the very stable building block
C&&
for clusters, where
M
is any alkali metal. An
exception to this
is
the cluster C60Li12. Supported by
semiempirical quantum chemical calculations, we find
the high stability of C60Li,2 to be caused by the geo-
metric arrangement of the metal atoms rather than by
the electronic configuration[l2].
As
predicted by
ab
initio
calculations, this arrangement most likely has
perfect icosahedral symmetry[
131.
At higher alkali
metal coverage, the coating becomes increasingly me-
tallic and an oscillating structure caused by the suc-
cessive filling of electronic shells shows up in the mass
spectra, if photon energies near the ionization thresh-
old are used.
Note that we always speak of
c60
and
C70
as
a moi-
ecule and not
a
cluster. We reserve the word 'cluster'
to
refer to units composed of several fullerenes and
metal atoms.
2.
EXPERIMENTAL
Figure
1
shows a schematic representation
of
the
experimental setup used to study the metal-fullerene
169
170
U.
ZIMMERMAN
et
al.
REF
7
C
LERATION
REOlON
"
CLUSTER PUMPING
TOF MASS
CONDENSATION
STAGE
SPECIXOMETER
CELL
Fig.
1.
Experimental setup: the clusters are emitted from the
cluster condensation cell, passing as a particle beam through
a differential pumping stage into the focus of a time-of-flight
mass spectrometer, where they are ionized by a laser pulse.
clusters. The cluster source on the left is
a
low-
pressure, inert gas condensation cell filled with ap-
proximately
1
torr He gas and cooled by liquid nitrogen
flowing through the outer walls of the cell. Inside the
cell, two electrically heated ovens, one containing
a
fullerene and one containing
a
metal, produce inter-
penetrating vapor clouds of the two materials. This
mixture is cooled by collisions with the He gas, thereby
supersaturating the vapor and causing clusters to con-
dense. The size distribution of the clusters thus pro-
duced is rather broad, but the mean composition of
the clusters depends on the relative density of the va-
por components and can be adjusted by the tempera-
tures of the ovens. However, the range of cluster
compositions that can be studied using mass spectrom-
etry is limited, despite the high resolution of the mass
spectrometer employed: Due to the various natural iso-
topes exhibited by most of the metals studied, the mass
spectra become increasingly confused with rising metal
content, making exact identification of the peaks im-
possible. For metals with more than one significant
isotope we can, therefore, only study clusters with ei-
ther small metal and high fullerene content or high
metal content and just one fullerene per cluster. The
formation of pure metal clusters has to be avoided for
the same reason. By keeping the temperature of the
metal oven below the threshold for formation of pure
metal clusters and introducing only small amounts of
fullerenes as condensation seeds into the metal vapor,
it is possible to generate cluster distributions consist-
ing almost completely of compositions C60Mx or
C70Mx with
M
E
(Li, Na,
K,
Rb, Cs, Ca, Sr, Ba) and
x=o
...
500.
After condensation, the clusters are transported by
the He-flow through
a
nozzle
and
a
differential pump-
ing stage into
a
high vacuum chamber. For ionization
of the clusters, we used excimer and dye laser pulses
at various wavelengths. The ions were then mass an-
alyzed by
a
time-of-flight mass spectrometer, having
a
two-stage reflector and a mass resolution of better
than
20,000.
The size distribution of the clusters produced in the
cluster source is quite smooth, containing no informa-
tion about the clusters except their composition.
To
obtain information about, for example, the relative
stability of clusters, it is often useful to heat the clus-
ters. Hot clusters will evaporate atoms and molecules,
preferably until
a
more stable cluster composition is
reached that resists further evaporation. This causes
an increase in abundance of the particularly stable spe-
cies (Le., enhancing the corresponding peak in the
mass spectrum, then commonly termed 'fragmentation
spectrum'). Using sufficiently high laser fluences
(=50
pJ/mm2), the clusters can be heated and ion-
ized simultaneously with one laser pulse.
3.
COATING WITH ALKALINE EARTH METALS
In this section, we will investigate the structure of
clusters produced when the metal oven is filled with
one of the alkaline earth metals Ca,
Sr,
or
Ba.
A mass spectrum of C60Bax is shown in Fig.
2.
The mass peaks corresponding to singly ionized clusters
have been joined by
a
connecting line. Note that the
series of singly ionized clusters shows
a
very prominent
peak at the mass corresponding to C60Ba32, implying
that this cluster is particularly stable. In searching for
an explanation for the high stability of this cluster, we
can obtain
a
first hint from looking at the doubly ion-
ized clusters also visible in Fig.
2
(the peaks not con-
nected by the line correspond to doubly ionized
clusters). Again, the peak at
x
=
32
is particularly
strong. This seems to indicate that the stability of
CmBa,, is not caused by
a
closed-shell electronic con-
figuration. Instead, the high stability is expected to be
of geometric origin. Remembering that the total
number of faces or rings constituting the cagelike
structure of the
c60
molecule is
32
and, thus, equal to
the number of Ba-atoms required to form this highly
800
-
:
9
-3
-2
8
Y
c
1
0
1000
3000
5000
7000
mass [amu]
Fig.
2.
Mass spectrum of photoionized C,Ba, clusters con-
taining both singly and doubly ionized species: the solid line
connects peaks
of
singly ionized clusters. The sharp edge oc-
curs at
32
metal atoms, equal to the total number of hex-
agonal and pentagonal rings
of
the C60 molecule.
Metal-coated fullerenes
171
stable cluster, the arrangement of metal atoms in this
cluster becomes obvious. By placing one Ba
atom
onto
each of the 12 pentagons
and
20 hexagons of the
c60
molecule, a structure with full icosahedral symmetry
(point group
I,,)
is obtained that can be visualized as
an almost close-packed layer of 32 Ba-atoms coating
the
c60
molecule. It seems reasonable that this struc-
ture exhibits an unusually high stability, somewhat
similar
to
the geometric shells observed in pure alka-
line earth clusters[lO,l
I].
Any additional metal atoms
situated
on
this first metal layer are likely to be only
weakly bound to the layer underneath and, thus, evap-
orate easily, causing the mass peaks of C60Bax with
x
greater than
32
to disappear almost completely. The
small peaks at
x
=
35,38,
and
43
might signal the com-
pl’etion of small stable metal islands on the first metal
layer. We can, however, presently only speculate on
the nature of these minor structures.
For a rough estimate of the packing density of this
first metal layer, assume the atoms to be hard spheres
having the covalent radii of the respective atoms
(0.77
A
for
C;
1.98
A
for Ba[14]). Placing the carbon spheres
at
the appropriate sites of the Cm structure with bond
lengths 1.40
-4
and
1.45
A[
151
and letting the Ba spheres
rest on the rings formed by the carbon atoms, the Ba
spheres placed on neighboring hexagons will almost
touch, spheres on neighboring pentagons and hexa-
gons will overlap by
a
few tenths of an hgstr~m. The
distance of the metal atoms to the center of the mol-
ecule is almost equal for atoms on hexagonal and pen-
tagonal faces. In this simple picture, the packing
of
the metal layer is almost perfectly dense, the Ba atoms
having an appropriate size. Incidentally, this argument
also holds in a similar manner for Sr- and Ca-atoms.
Of course, this simple picture constitutes only a
crude approximation and should be valued only for
showing that the completion of a metal layer around
C60 with
32
Ba-atoms is, indeed, plausible. More pre-
cise predictions would have to rely on
ab
initio
calcu-
lations, including a possible change in bond lengths of
C60, such as an expansion of the double bonds of
C,jo
due to electron transfer
to
the antibonding LUMO (as
was found in the case of C60Li,2[12,13]).
The significance of the magic number
32
found in
the experiment may also be stated in
a
different man-
ner. If
a
cluster containing Ba and a fullerene molecule
will be stable and, thus, result in
a
clearly discernible
structure in the mass spectra every time there is exactly
one Ba-atom situated on each of the rings of the ful-
lerene molecule, this property might be used
to
‘count
the rings’
of
a
fullerene. Of course, such
a
proposal
has
to
be verified
using
other
fullerenes, for example,
C70 which is available in sufficient quantity and pu-
rity for such an experiment.
In investigating the metal coating of C70, we will
also replace Ba by Cain the data presented. The coating
of the fullerenes with the latter material is basically iden-
tical but exhibits additional interesting features that
will be discussed below. Figure
3
shows two mass spec-
tra, the upper one of C,,Ca:, the lower of C70Ca;,
both obtained under similar conditions as the spec-
I
x
=
32
300
1
-
104
0
50
100
130
X
Fig.
3.
Mass spectra of photoionized C,,Ca; (top) and
C7&a: (bottom): the lower
axis
is
labeled by the number of
metal atoms on the fullerene molecule. The peaks at
x
=
32
for CmCa, and
x
=
37
for C&a, correspond to a
first
metal layer around the fullerenes with one atom located at
each of the rings. The edges at
x
=
104
and
x
=
114,
respec-
tively, signal the completion
of
a second metal layer.
trum in Fig.
2
but with a higher metal vapor density.
A slight background caused by fragmentation of clus-
ters inside the drift tube of the mass spectrometer has
been subtracted. The lower axis
is
labeled with the
number of metal atoms
on
the respective fullerene.
Again, the coverage of C6,, with 32 Ca atoms leads
to a pronounced peak in the fragmentation mass spec-
trum. In the spectrum containing C70, a very strong
peak
at
C70Ca:7 is observed. Note that C70, just as
C60,
has 12 pentagons but
5
additional hexagons on
the equator around the remaining fivefold
axis,
totaling
37 rings. The ‘ring-counting’ thus seems
to
work for
C70
also. However, the applicability of this ‘counting
method’ to even higher fullerenes has
to
be verified as
these become available in sufficient quantities for per-
forming such an experiment.
If it
is
possible
to
put one layer of metal around a
fullerene molecule, it is tempting
to
look for the com-
pletion of additional layers also. In the spectra in
Fig.
3, the sharp edges
at
C60Ca:04 and C70Ca~,,
would be likely candidates for signaling the comple-
tion of a second layer.
As
we will see below, there is,
in fact, a very reasonable way of constructing such a
second layer with precisely the number of metal atoms
observed in the spectrum.
In proposing an arrangement of the atoms in the
second layer, we will focus first on the metal coating
of
C60.
Note that we speak of layers, not shells. The
term ‘shell’ implies self-similarity which, as we will see
172
U.
ZIMMERMAN
et
al.
later, does not apply in our case. In the following
paragraphs we will often specify the positions of the
metal atoms relative to the central CW molecule. This
is done for clarity and is not meant to imply any di-
rect interaction between the
c60
and the atoms of the
second layer.
In constructing the second layer, it seems reason-
able to expect this layer to preserve some of the char-
acteristic symmetry elements
of
the first layer (Le., the
fivefold axes). The second layer on
c60
contains
72
atoms, a number being indivisible by 5. This requires
that each of the five-fold symmetry axes passes
through two metal atoms. Consequently, in the sec-
ond layer there must be one metal atom situated above
each of the
12
pentagonal faces of
c60.
Let
us
first
assume that the second layer has the full icosahedral
symmetry
I,,
of the first layer. The remaining 60 at-
oms may then be arranged basically in two different
ways. The first would be to place the atoms such that
they are triply coordinated to the atoms of the first
layer (i.e., placing them above the carbon atoms of the
C6,
molecule as shown in Fig.
4
on the upper left).
The atoms above the pentagons of
c60
(black) consti-
tute the vertices of an icosahedron, the other atoms
(white) resemble the C,,-cage. This structure can also
be visualized as twelve caps, each consisting of a
5-atom ring around an elevated central atom, placed
at the vertices
of
an icosahedron. This structure, how-
ever, does not result in an even coverage: there are
20
large openings above the hexagonal faces of
Cm
while
neighboring caps overlap above the double bonds
of
C,,.
Pictured on the upper right in Fig.
4
is a second
way to arrange the 60 atoms with Ih symmetry, ob-
tained by rotating each of the caps described above by
Fig.
4.
Three possible geometries for arranging the
72
atoms
of the second layer: the atoms above the pentagons
of
Cs0
are shaded. The structure on the upper left can be trans-
formed into the more evenly distributed arrangement of
atoms on the upper right by
36"
turns of the caps around the
five-fold axes. From this, the structure on the bottom can be
obtained by rotating each triangular face of atoms
by
19".
one-tenth of
a
turn (36") around the 5-fold axis
through its center. The coordination to the atoms of
the first layer will then be only two-fold, but the cov-
erage will be quite even, making the latter of these two
structures the more probable one.
The latter structure could be described as an 'edge-
truncated icosahedron' with
20
triangular faces, each
face consisting of the three atoms at the icosahedral
vertices with
a
smaller, almost densely packed trian-
gle of three atoms set in between (exemplarily, one of
these triangles has been shaded). Note that this layer,
having no atoms right on the edges, is not identical to
a
Mackay icosahedron[l6] which is formed by pure al-
kaline earth metal clusters[lO,l
l].
However, in this
structure the two rows
of
atoms forming the truncated
edges are not close-packed within the layer. This might
be a hint that with the structure depicted on the up-
per right in Fig.
4
we have not yet found the most sta-
ble configuration of the second layer.
Up to this point, we have assumed that the second
layer of atoms preserves the full symmetry (Ih) of
the fullerene inside. Let
us
now allow the second layer
to lower its symmetry. This can be done in a simple
way: model the interaction between metal atoms by a
short-range pair potential with an appropriate equi-
librium distance and let the atoms
of
the second layer
move freely within this potential on top of the first
layer. This allows the atoms to move to more highly
coordinated positions. Starting with atoms in the ar-
rangement with Ih-symmetry, the layer will relax
spontaneously by rotating
all
20
triangular faces of at-
oms around their three-fold axes by approximately
19". The resulting structure is shown at the bottom of
Fig.
4.
One of the rotated triangles has been shaded
and the angle of rotation marked. In a projection on
a plane perpendicular to the threefold axis, each pair
of atoms at the edges of the triangle lie on a straight
line with one of the three atoms on the surrounding
icosahedral vertices. The two rows
of
atoms along the
former truncated edges have now shifted by the radius
of one atom relative to each other in direction of the
edge, leading to close packing at the edges. Of course,
the triangles could have been rotated counterclockwise
by the same angle, resulting in the stereoisomer
of
the
structure described above. This structure no longer has
Ih-symmetry. There are no reflection planes and no
inversion symmetry. Only the two-, three-, and five-
fold axes remain. The structure belongs to the point
group
I
(order 60).
I
is the largest subgroup of I,,.
The layer has, thus, undergone the minimum reduc-
tion in symmetry.
Of the three arrangements of atoms in the second
layer shown in Fig.
4,
we find the one on the bottom
(symmetry I) the most probable. It optimizes the co-
ordination of neighboring atoms within the layer and,
as we will see further down, this arrangement can also
be well extended to C,, coated with metal.
Of course, after having observed two complete layers
of
metal around a fullerene, we searched for evidence
for the formation of additional layers. However, be-
fore looking at experimental data, let
us
try to con-
Metal-coated fullerenes
173
C60M32
K=3
C60M104
C60M236
C60M448
Fig.
5.
Proposed arrangements of the atoms in the first four
layers of an alkaline earth metal around
a
Cm molecule: the
atoms at the icosahedral vertices are drawn in black and one
of
the triangular faces of atoms has been shaded in each
layer. Note the spiral
of
atoms (dark grey) in the fourth layer.
struct the third and fourth layers around
c60
in a
manner similar to the second layer with I-symmetry:
place one atom above each of the icosahedral vertices;
for each additional layer, increase the length of the
edges of the triangles between the vertices by one atom
with respect to the underlying layer; rotate the trian-
gles
so
that each edge points toward
a
different icosa-
hedral vertex. For the second layer, this angle of
rotation is
19".
For the third and fourth layer it is ap-
proximately
14"
and
1
lo,
respectively. These angles
are measured relative to the position with full
Ih-
symmetry. The atoms stacked as triangular faces
above the hexagonal rings of
c60
resemble
a
tetra-
hedron with one tip pointing towards the center of the
cluster, having a slight twist due to the difference in
orientation of a few degrees between consecutive lay-
ers. The resulting structures of the first
four
layers are
depicted in Fig.
5.
For clarity, one of the triangular
faces has been shaded. The atoms at the icosahedral
vertices are drawn black. The number of atoms re-
quired to complete the third and fourth layer in this
manner are 236 and
448.
At the bottom of Fig.
5,
the fourth shell is shown
from two directions. Note the spiral of atoms that are
emphasized by a dark grey. This spiral can be wound
around any
of
the five-fold axes from tip to tip. Sim-
ilar spirals exist in the other layers, too. Each layer can
be envisioned to consist of five such spirals of atoms.
For
each layer, there is also the stereoisomer with the
opposite sense of chirality.
To express the number of atoms needed to complete
such layers mathematically, let
us
introduce a layer in-
dex K. Define K as the number of atoms along the
edge of a triangular face without including the atoms
on the vertices above the C60-pentagons. The first
layer then has
K
=
1,
the second
K
=
2. The number
of atoms in the Kth layer can easily be calculated to
10K2
+
10K
+
12.
(1)
The total number of atoms N(K) in a cluster com-
posed of
K
complete layers around
c60
becomes
N(K)
=
i(10K3
+
30K2
+
56K).
(2)
Note that the coefficient of the leading order in
K,
de-
termining the shell spacing on an
N1'3
scale, is equal to
that of an icosahedral cluster of the Mackay type[l7].
Inserting
K
=
1
. .
.4
into eqn
(2),
we find
N(
1)
=
32,
N(2)
=
104,
N(3)
=
236,
N(4)
=
448,
N(5)
=
760, etc.
Did we predict the number of atoms required to
complete additional layers around the metal-coated
c60
correctly? Figure
6
shows a spectrum of
c60
cov-
ered with the largest amount of Ca experimentally pos-
sible (note the logarithmic scale). Aside from the edges
of
x
=
32 and
x
=
104
which we have already dis-
cussed, there are additional clear edges at
x
=
236 and
x
=
448
(completion of a third layer was also observed
at C6OSr236). Note that these values are identical to
the ones just predicted above for the completion of the
third and fourth layer of metal atoms. We, therefore,
feel confident that the alkaline earth metals studied
do, in fact, form the distinct layers around a central
c60
molecule with the structures depicted in Fig.
5.
It should be pointed out again that these layers
would,
of
course, contain identical numbers of atoms
if the triangular faces had not been rotated and, thus,
the Ih-symmetry had been preserved[7]. The reason
for preferring the arrangement with I-symmetry
(which can still be called icosahedral) is that it leads
to higher coordination of the atoms at the borders be-
tween the triangular faces.
0
10000
mass
[amu]
20000
Fig.
6.
Mass spectrum of photoionized C&a, clusters with
high metal content: additional edges, interpreted as comple-
tion
of
a third and fourth layer, are observed at
x
=
236
and
x
=
448.
174
U.
ZIMMERMAN
et
al.
Note that the structures depicted in Fig.
5
are not
self-similar because the angle of rotation of the faces
differs for each layer. The layers should, therefore,
not be called 'shells' as they are called in the case
of
pure alkaline earth-metal clusters. With increasing
size, the shape of the cluster will converge asymptot-
ically to that of
a
perfect icosahedron.
With
C70
at the center
of
the cluster, we observed
the completion of layers at
x
=
37,
114, and 251. For
completion of the observed three layers around C70,
each layer requires
5
atoms more than the correspond-
ing layer around
c60.
The arrangement of atoms in
the first layer is again obvious: place one atom above
each of the
37
rings of the fullerene.
Attempting to preserve the D,,,-symrnetry of
C70
molecule and of the first layer when constructing the
second and third layer, results in some ambiguity
of
placing the atoms on the equator around the five-fold
axis. Also, we found no structure that was sufficiently
close packed to be convincing. Lowering the demand
on symmetry by removing the symmetry elements con-
taining
a
reflection (as was done in the case of the
coated c60) leads to the point group
D,.
Similar to
c60,
close-packed layers can be obtained by rotating
the 10 remaining triangular faces around their normal
by 19". The remaining atoms can be placed in
a
close-
packed arrangement on the remaining faces on the
equator. Fig.
7
shows these first three layers. For the
third layer, shown from two different directions, one
spiral of atoms is indicated by a dark grey shading.
Again, the layers can be envisioned to consist of five
spirals of atoms around the five-fold axis.
Very high metal vapor pressures are required to
C70M37
C70M114
c70M251
Fig.
7.
Proposed arrangements
of
the atoms in the first three
layers
of
an alkaline earth metal around a
C70
molecule: the
atoms at the icosahedral vertices are drawn in black. Note
the spiral
of
atoms shaded in the third layer.
produce the multilayered clusters discussed above,
so
high that large quantities of pure metal clusters may
also be formed. The great variety of isotopic compo-
sitions to be found in large clusters makes it impossi-
ble, beyond some size, to distinguish between these
pure metal clusters and clusters containing
a
fullerene
molecule. This complication limits the amount of
metal atoms that can be placed on one fullerene and,
thus, the number of layers observable. This maximum
amount differs for each alkaline earth metal and is
lowest in the case of Ba coating. For this reason, it is
desirable to suppress pure metal-cluster formation.
This is more easily achieved with certain metals, such
as
Ca
and, as we will see below, Cs, making these el-
ements particularly favorable coating materials.
At the end of this section, let
us
return briefly to
the spectra shown in Fig.
3.
Notice the structure in the
mass spectrum of C60Cax between the completion of
the first metal layer at
32
and the second at 104. This
structure is identical in the fragmentation mass spec-
tra of fullerenes covered with Ca and with Sr. It is
reminiscent of the subshell structure
of
pure Ca clus-
ters. The subshells could be correlated with the for-
mation of stable islands during the growth
of
the
individual shells[ 10,111. The 'sublayer' structure we
observe here may also give some clue to the building
process of these layers. However, the data is presently
insufficient to allow stable islands to be identified with
certainty.
4.
COATING
WITH
ALKALI METALS
The structures observed in the mass spectra of ful-
lerene molecules covered with alkaline earth metals,
as described in the previous section, all seem to have
a
geometric origin, resulting in particularly stable clus-
ter configurations every time
a
highly symmetrical
layer of metal atoms around
a
central fullerene mol-
ecule was completed. When replacing the alkaline
earth metals by an alkali metal (i.e., Li, Na,
K,
Rb,
or Cs),
a
quite different situation arises.
Let
us
begin with clusters having
a
low metal content
but containing several fullerene molecules. Figure
8
shows
a
fragmentation mass spectrum of (C60)nRbx
(a weak background has been subtracted). Mass peaks
belonging to groups of singly ionized clusters with the
same number of fullerenes have been joined by
a
con-
nection line to facilitate assigning the various peaks.
This spectrum is clearly dominated by the peaks cor-
responding to (C,Rb6), Rb+. Of the peaks correspond-
ing to doubly ionized clusters, also visible in Fig.
8,
the highest peak of each group (C60Rb6)nRb:+ with
odd
n,
has been labeled
'++'
(note that every other
peak of doubly ionized clusters with an even number
of fullerenes coincides with
a
singly ionized peak).
Writing the chemical formula of these particularly sta-
ble clusters in this way makes the systematics behind
these magic peaks immediately clear: one or two Rb
atoms are needed to provide the electrons for the
charged state of the cluster, the remaining cluster con-
sists of apparently exceptionally stable building blocks
Metal-coated fullerenes
175
300
3
0 2000
4000
'
6000
'
8000
mass [amu]
Fig.
8.
Mass spectrum, with background subtracted, of pho-
toionized (C,),Rb, clusters containing both singly and dou-
bly ionized species: the solid line connects peaks belonging
to groups of singly ionized clusters with a fixed value of
n.
Note the dominant peaks corresponding to (c,&b6),Rb+
and (C60Rb6),Rb$+ (marked
'I++").
C6,Rb6. The corresponding building block can be
found in the mass spectra
of
clusters containing any
alkali metal and
Cm.
Only Na is
a
minor exception to
the extent that the clusters (c60Na6),,Naf do not show
up as especially strong peaks in the fragmentation
mass spectra. They do, however, mark
a
sharp fall-
ing edge and
a
distinct change in the character of the
spectra, as we will see later.
It seems quite obvious that the origin of the stabil-
ity of these building blocks is not geometric. More
likely, the electronic configuration of this unit is re-
sponsible for the stability, the six valence electrons of
the metal transferred to the six-fold degenerate
t,
,
LUMO of the
c60
molecule. Such
a
transfer of six
electrons to the LUMO of
Cm
has also been observed
in the bulk intercalation phases of C60M6 with M
E
(K,
Rb, Cs)[5]. These alkali metal fullerides become in-
sulators due to the complete filling of the
t,,
derived
band (which was found to be only slightly disturbed
by the presence of the alkali ions[5]). The appear-
ance of such
a
building block is not limited to clusters
containing
c60.
Mass spectra of (C70)nMx show ex-
actly the same intensity anomalies at (C70M6)nM+
and (C70M6)nM:+. An explanation similar to the one
given
for
c60
regarding the stability of the building
block observed holds
for
C,,[18].
Adhering to this interpretation, the bonding of the
first six
or
seven alkali metal atoms will be primarily
ionic in nature. How will additional atoms attach to
the
c60
molecule? Will they continue transferring
their valence electrons to the next unoccupied orbital
of
Cmr again showing high stability when this six-fold
degenerate
tl,
orbital becomes filled? Looking for in-
formation supporting this hypothesis, we will begin
with an investigation of clusters having the composi-
tion CbOLix. Based on
ab
initio
calculations, it has
been suggested that the cluster C60Li12 should be sta-
ble with the valence electrons from the Li atoms fill-
ing both the
t,,
and the
t,,
orbitals[l3].
Figure
9
shows fragmentation mass spectra of sin-
2000
v)
Y
FI
2
0
300
v)
Y
0
1
8
0
720
1
LixC,+,
so0
mass [amu]
900
Fig.
9.
Mass spectra
of
singly (top) and doubly (bottom) ion-
ized C,Li, clusters: note the prominent features at
x
=
7
for
singly ionized and
x
=
8
for doubly ionized clusters and at
x
=
12
in both spectra.
gly and doubly ionized CmLiw clusters. Mass peaks
are, again, joined by
a
connecting line. The fine struc-
ture of the peaks is caused by the two natural isotopes
of
Li. Again, we find prominent peaks at
x
=
7
for sin-
gly ionized and
x
=
8
for doubly ionized clusters. Ad-
ditionally, there are prominent peaks at
x
=
12 in both
spectra. Twelve is exactly the number
of
electrons
needed to fill the
t,,
and
t,,
orbitals,
so
it seems, at
first, that we have found what we were looking for.
However, remember that these clusters are charged,
so
the
tl,
orbital obviously cannot be filled com-
pletely. Since the appearance of the magic number 12
is independent of charge, it seems more promising to
try
a
geometric interpretation.
Ab
initio
calculation
shows that the twelve Li atoms have their equilibrium
position above each of the twelve pentagonal faces
and, thus, retain the icosahedral symmetry[l3]. It
seems likely that this highly symmetrical arrangement
of atoms is responsible for the high stability
of
C60LilL, independent of the state of charge, rather
than a complete occupation of vacant molecular
orbitals.
To support this interpretation, we performed semi-
empirical quantum chemical calculations using the
modified-neglect-of-diatomic-overlap (MNDO) meth-
od[19,20]. For
x
=
1
.
.
.
14, we searched for the most
stable ground state geometries
of
C,,Li,. We found
that
for
x
=
1
.
.
.8 for Li atoms preferred to be cen-
tered above the hexagonal faces
of
c60[12]. Exem-
plarily, the geometry of C60Li8 is shown in Fig. 10 on
the left. The eight Li atoms are situated at the corners
176
U.
ZIMMERMAN
et
al.
Fig. 10.
Most stable ground-state geometries found for
Cdi,
and C&i14
by
the
MNDO
calculations:
the
Li atoms
are represented
by
the
filled black circles.
of a cube. The bonds between the Li atoms (black)
and the carbon atoms (white) were drawn merely to
clarify the geometry and are not meant
to
imply any
specific bonds. After
a
transition at
x
=
9,
all Li at-
oms are found to be most stable when centered above
the pentagonal rings for
x
=
10.
.
.
12. For C6,,Li12,
the icosahedral arrangement
of
Li atoms proved to be
significantly lower in energy than all other isomers, in-
dependent of the charge of the cluster, while
for
clus-
ters with
x
around
7,
the number of electrons in the
cluster dominated over the geometry in determining
the total binding energy of the cluster. Interpreting the
magic numbers
x
=
7
and
x
=
8
to be of electronic and
x
=
12
to
be
of
geometric origin thus seems reasonable.
For CsoLi13, the most stable geometry has
12
Li
atoms above the pentagons and one above
a
hexagon.
If
a
fourteenth atom is placed near the Li atom above
a hexagon, the arrangement
of
Li atoms becomes un-
stable. The two Li atoms initially not above a penta-
gon
of
c6(,
will then slide on top
of
a pentagon. The
resulting most stable geometry of C60Li,4 has one
equilateral Li trimer (Li-Li bond length of
2.23
A)
lying
flat
above
a
pentagon and 11 Li atoms centered
above the remaining pentagons of
C,o
as shown in
Fig.
10
on the right. For comparison: MNDO calcu-
lates a bond length of 2.45
A
for the isolated Li:
(equilateral triangle) and
2.19
A
for the two short
bonds
of
neutral Li3.
From the binding energies calculated for the dif-
ferent cluster compositions, we determined abundance
mass spectra for heated C6,LiX clusters from a simple
Monte Carlo simulation. Figure 11 shows the simu-
lated mass spectra resulting from these calculations,
including the Li and
C,
isotope distributions. The
peaks at
x
=
12
and at
x
=
6
+
n
(where
n
is the clus-
ter charge) observed in the experiment (Fig.
9)
are well
reproduced. For more details, see ref. [12].
For values of
x
greater than
14,
a strong even-odd
alternation becomes visible in the spectra shown in
Fig.
9,
peaks corresponding
to
clusters with an even
number of available metal valence electrons being
stronger. We suggest that this even-odd alternation,
similarly observed in pure alkali metal clusters, signals
the onset
of
metal-metal bonding of the metal atoms
1
LiXC60".
'
"'
'""'
12
I
12
j
4
4
x=7 n
I
Li,C6,"
?
J
0
2
4
6
8101214
#
of
Li-atoms
on
c,,
Fig.
11.
Abundance mass spectra of differently charged
hot
C,,Li,
clusters
evaporating atoms calculated
with
a
Monte-
Carlo simulation
(the
Li
and
C,,
isotope distributions are
included).
Energies
required
to remove
Li
atoms were calcu-
lated using the
MNDO
method. The peaks at
x
=
12
and at
x
=
6
+
n
(where
n
is
the
cluster charge) observed in experi-
ment
(Fig.
9)
are
well reproduced.
on the surface of
Cs0
(remember that the MNDO cal-
culations already show the formation
of
a metal
tri-
mer for
x
=
14). The electronic configuration of the
clusters would, then, again determine their relative sta-
bility just as it does for pure alkali metal clusters. Con-
sistent with this
'electronic'interpretation,
the even-odd
alternation displayed by the doubly ionized clusters is
shifted by one atom with respect
to
the singly ionized
clusters, an additional Li ion required
to
supply the
charge
of
the cluster.
Such an even-odd alternation is observed to a dif-
ferent degree
for
all alkali metals covering fullerene
molecules (see also Fig.
8).
It is especially strong for
Na. Fig.
12
shows
a
fragmentation mass spectrum
of
singly charged C&ax.
A
strong even-odd alternation
starts above
x
=
7,
the point at which we suggested the
metal-metal bonding to begin, and extends up to ap-
proximately
x
=
66.
Note that
x
=
12 does not appear
as a magic number in these spectra. In fact, Li is the
only metal for which this magic number is observed.
One possible explanation
as
to
why Li behaves differ-
ently
is
the ability of Li atoms to form covalent bonds
with carbon because the Li
2s
orbital is close enough
in energy to the carbon valence orbitals. Other than
Li, the higher alkali metals form essentially ion pairs
0
20
40
60
No.
of
Na-atoms
on
CG0
Fig.
12.
Mass spectra
of
singly charged clusters composed
of
a single
C,
molecule coated with
a
large
amount
of
Na (background
subtracted).
The even-odd alternation extends
up
to
approximately
x
=
66.
Note
that
x
=
12
does
not
appear as a magic
number
in
these
spectra.
in the gas phase (a Li' ion is exceptionally small and
has, therefore, an exceptionally high charge-radius ra-
tio, comparable to that of Mg2+).
A
neighboring neg-
atively charged fuIlerene would be polarized to such
an
extent that the description as ion pair would not be
justified. The configuration of Li atoms around Cb0
might, therefore, be influenced more strongly by the
structure
of
the fullerene molecule than is the case for
other alkali metals, resulting in the unique configura-
tion and stability of C6,,Li12.
TJnfortunately, in the case of fullerenes covered
with alkali metals, clear evidence
is
lacking regarding
the geometry of the clusters. We can, therefore, only
present speculation that may appear plausible but can-
not be proven presently. The first seven Na ions of the
C,Na: clusters arrange themselves as far from each
other as possible
to
minimize coulomb repulsion while
adhering
to
the
C,
molecule. Additional Na atoms
might successively attach to these
7
ions in pairs
of
two, forming Na: trimers similar to the one calcu-
lated
for
Cs0Lil4. Every time such a stable trimer,
each containing two metal valence electrons, is com-
pleted, a strong peak is observed in the spectrum, re-
sulting in an even-odd alternation. The abrupt change
in the strength of this alternation at
x
=
21
=
3
X
7
Na atoms fits this speculation.
When coating fuIlerenes with larger alkali metal at-
oms, the even-odd alternation is interrupted before
reaching
x
=
21,
so
the structural sequence must be
different for these. Nevertheless, we do suggest that
the first
6
alkali metal
atoms,
having transferred their
valence electron
to
the fullerene molecule, will remain
distributed over the surface of the fullerene, gather-
ing additional metal atoms around them as the clus-
ter increases its metal content. This would result in at
least one metallic layer coating the molecule
(so
speak-
ing
of
metal-coated fullerenes seems justified). How-
ever, we do not have any evidence from the spectra
indicating when this layer will be completed (a rough
estimate shows that a first metal layer, for example
of
Cs,
would require around
30
atoms for completion).
As we have already mentioned, the stability
of
the
alkali-fullerene clusters seems to be primarily deter-
mined by the electronic configuration. Therefore,
it
is not too surprising that completion of a Payer of at-
oms, which would be a geometrically favorable struc-
ture, does not lead to any pronounced features in the
mass spectra. Furthermore, it should be emphasized
that to obtain these fragmentation spectra, the clus-
ters have been heated up to
a
temperature at which
they evaporate atoms on a psec time scale. This cor-
responds
to
a
temperature
at
which bulk alkali metals
are molten. Incidentally,
a
similar behavior is observed
in pure metal clusters: small alkali clusters (less than
1500 atoms) show electronic shells and alkaline earth
clusters show geometric shells[9,10].
When the cluster, containing one fullerene, contin-
ues to grow by adding more metal, it will probably as-
sume the more or less spherical shape observed for
pure alkali metal clusters. It could, then, be viewed as
a
metal cluster with
a
large 'impurity': the fullerene.
Alkali metal clusters containing small impurities, such
as
(SO,),
or
On,
have already been studied[21,22],
showing that the main influence of the impurity is to
shift the number
of
atoms at which electronic shell
closings are observed upwards by 2n,
2
being the num-