Properties of non-IPR fullerene films versus size of the building blocks.

Abstract

This perspective focuses on the cage size dependent properties of novel solid fullerene nanofilms grown by soft-landing of mass-selected C(n)(+) (48, 50, 52, 54, 56, 58, 62, 64, 66 and 68) onto room temperature graphite surfaces under ultra-high vacuum conditions. Such non-isolated-pentagon-ring (non-IPR) fullerene materials are not accessible to standard fullerene preparation methods. The component molecular building blocks of non-IPR films were generated by electron impact induced ionization/fragmentation of sublimed IPR-C(70)(D(5h)) (-->C(n) (n = 68, 66, 64, 62)) or IPR-C(60)(I(h)) (-->C(n) (n = 58, 56, 54, 52, 50)). Non-IPR fullerene films on graphite grow via formation of dendritic C(n) aggregates, whereas deposition of IPR fullerenes under analogous conditions (via deposition of unfragmented C(60)(+) and C(70)(+)) leads to compact islands. The latter are governed by weak van der Waals cage-cage interactions. In contrast, the former are stabilized by covalent intercage bonds as mediated by the non-IPR sites (primarily adjacent pentagon pairs, AP). A significant fraction of the deposited non-IPR C(n) cages can be intactly (re)sublimed by heating. The corresponding mean desorption activation energies, E(des), increase from 2.1 eV for C(68) up to 2.6 eV for C(50). The densities of states in the valence band regions (DOS), surface ionization potentials (sIP) and HOMO-LUMO gaps (Delta) of semiconducting non-IPR films were measured and found to vary strongly with cage size. Overall, the n-dependencies of these properties can be interpreted in terms of covalently interconnected oligomeric structures comprising the most stable (neutral) C(n) isomers-as determined from density functional theory (DFT) calculations. Non-IPR fullerene films are the first known examples of elemental cluster materials in which the cluster building blocks are covalently but reversibly interconnected.

Full text

Page 1

This journal is c the Owner Societies 2010Phys. Chem. Chem. Phys., 2010, 12, 10671–1068410671
Properties of non-IPR fullerene films versus size of the building blocks
Daniel Lo ¨ ffler,waSeyithan Ulas,aStefan-Sven Jester,zaPatrick Weis,a
Artur Bo ¨ ttcher*aand Manfred M. Kappes*ab
Received 6th April 2010, Accepted 8th July 2010
DOI: 10.1039/c0cp00137f
This perspective focuses on the cage size dependent properties of novel solid fullerene nanofilms
grown by soft-landing of mass-selected Cn+(48, 50, 52, 54, 56, 58, 62, 64, 66 and 68) onto room
temperature graphite surfaces under ultra-high vacuum conditions. Such non-isolated-pentagon-ring
(non-IPR) fullerene materials are not accessible to standard fullerene preparation methods.
The component molecular building blocks of non-IPR films were generated by electron impact
induced ionization/fragmentation of sublimed IPR-C70(D5h) (-Cn(n = 68, 66, 64, 62)) or
IPR-C60(Ih) (-Cn(n = 58, 56, 54, 52, 50)). Non-IPR fullerene films on graphite grow via
formation of dendritic Cnaggregates, whereas deposition of IPR fullerenes under analogous
conditions (via deposition of unfragmented C60+and C70+) leads to compact islands. The
latter are governed by weak van der Waals cage–cage interactions. In contrast, the former
are stabilized by covalent intercage bonds as mediated by the non-IPR sites (primarily adjacent
pentagon pairs, AP). A significant fraction of the deposited non-IPR Cncages can be intactly
(re)sublimed by heating. The corresponding mean desorption activation energies, Edes, increase
from 2.1 eV for C68up to 2.6 eV for C50. The densities of states in the valence band regions
(DOS), surface ionization potentials (sIP) and HOMO–LUMO gaps (D) of semiconducting
non-IPR films were measured and found to vary strongly with cage size. Overall, the
n-dependencies of these properties can be interpreted in terms of covalently interconnected
oligomeric structures comprising the most stable (neutral) Cnisomers—as determined from
density functional theory (DFT) calculations. Non-IPR fullerene films are the first known
examples of elemental cluster materials in which the cluster building blocks are covalently
but reversibly interconnected.
1.Introduction
Classical fullerenes are hollow-cage, even numbered, all-
carbon clusters comprising pentagon and hexagon rings with
each constituent carbon atom connected to three neighbours
by strong covalent bonds. The structures of the well known
classical fullerene cages, C60(Ih) and C70(D5h) are governed by
the isolated pentagon rule, IPR,1,2which requires that all 12
constituting pentagons be surrounded by hexagonal rings and
aInstitut fu ¨r Physikalische Chemie, and DFG-Center for Functional
Nanostructures (CFN), Karlsruhe Institute of Technology (KIT),
D-76128 Karlsruhe, Germany.
E-mail: artur.boettcher@chemie.uni-karlsruhe.de
bInstitut fu¨r Nanotechnologie, Karlsruhe Institute of Technology (KIT),
D-76128 Karlsruhe, Germany.
E-mail: manfred.kappes@chemie.uni-karlsruhe.de
w Present address: Fritz-Haber-Institut der Max-Planck-Gesellschaft,
Berlin, Germany.
z Present address: Institut fu ¨ r Organische Chemie, Universita ¨ t Bonn,
Bonn, Germany.
Daniel Lo ¨ fflerSeyithan Ulas Stefan-Sven JesterPatrick Weis
PERSPECTIVEwww.rsc.org/pccp | Physical Chemistry Chemical Physics
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 2

10672Phys. Chem. Chem. Phys., 2010, 12, 10671–10684This journal is c the Owner Societies 2010
consequently separated from each other. Correspondingly,
DFT-based calculations (presently the most accurate quantum
chemical methods applied to this overall molecule class)
predict significantly lower stability of strained non-IPR carbon
cages in comparison to IPR cages of a given size.3It has
furthermore been shown, that all classical fullerene cage
structures possible for C20–C58and C62–C68violate the IPR
rule.4For such Cncage sizes, adjacent pentagon motifs (AP)
are unavoidable. DFT calculations have also established
that the energetic sequence of AP cage isomers of a given
nuclearity typically follows the pentagon adjacency penalty
rule (PAPR5), which relates the most stable carbon cages
to the Cn structures with the minimum number of APs
(except for C506).
Whereas non-IPR Cnhollow cages (n o 60; 62–68) can be
readily generated as stable molecules in gas phase, they remain
practically uncharted in condensed phase.7,8In fact, during the
last decade there have been numerous unsuccessful attempts to
isolate non-IPR Cn cages generated in various fullerene
soots.9,10In contrast, a number of chemical derivatives of
non-IPR cages have recently been prepared in bulk scale and
structurally characterized.11These structures typically indicate
classical fullerene cages containing AP sites. Due to their
increased pyramidal angles, such AP sites act as highly reactive
centres12
andthe corresponding
(exohedral) addition products in which C–C bonds shared
between pairs of adjacent pentagons are saturated.13,14The
addition products C36H6,15C50Cl10,16C54Cl8,13C56Cl10,16
C64H417and C66X4(X = H, F, Cl)18are cases in point. Bare,
small non-IPR cages may also be stabilized by ‘‘endohedral
derivatization’’, i.e. by filling with one or more metal atoms to
generate metallofullerenes (e.g. Sc2@C66,19see also the
review20). Usually, the endohedral metal atoms (ions) are
coordinated to annelated-pentagon pairs thus facilitating
electron transfer to the enclosing carbon cages which then
become negatively charged.20,21So far there are only three
well-documented cases of ‘‘non-classical’’ non-IPR fullerene
derivatives: C58F17CF3, C58F1822and (4-Me–C6H4)2C62.23The
fluorinated derivatives comprise cages with heptagon-rings
(HP) and 13 pentagons each whereas the non-classical C62
cage exhibits a four-membered ring.
Small non-IPR Cn cages may be considered as carbon
spheres functionalized by localized (AP or HP) reaction
derivatives areoften
centres. Thus, one might expect the growth of solid materials
comprising such (bare) building blocks to involve self-
stabilization by bond formation between reaction centres to
yield polymeric chains and 3D structures,24as has been
predicted by DFT calculations for solids consisting of C20,25
C36,26,27and C50.28In all of these calculations, quasi covalent
inter-cage bonds between AP sites were indicated.
Such interlinked non-IPR fullerene films were first prepared
by low-energy neutral cluster beam deposition onto a diamond
target with a broad incident cluster size distribution centred
between C20 and C32.29–32We have more recently become
interested in this issue and have applied mass selected ion
beam soft landing (with incident kinetic energies, Ekino 6 eV)
to generate multilayer films of various monodispersed non-
IPR fullerenes on graphite targets.33In these experiments, the
cation beams of non-IPR cages were generated by 40–100 eV
electron-impact induced ionization and fragmentation of
sublimed C60(Ih). This leads to ionization, sequential C2loss
and (following electrostatic collimation) to the corresponding
generation of intense ion beams containing non-IPR cages
with n = 50, 52, 54, 56 and 58.34The films generated by soft
landing, exhibited thermal and electronic properties, which
were clearly dependent on the size of the constituting non-IPR
Cncages.33,35In particular, we found that a good fraction of
the incident non-IPR cages were intactly desorbable from the
films by heating. However, significantly higher sublimation
temperatures were required than necessary for the corres-
ponding C60(Ih) reference films. Preliminary AFM imaging
revealed that the non-IPR Cnfilms grow according to the
Volmer–Weber mechanism, reflecting aggregation of Cncages
as mediated by AP or HP reactive sites.
In the present study, we revisit thin films of non-IPR full-
erene cages,35extend our previous work by performing further
characterization and significantly expanding upon the size
range of the building blocks. In particular, we have additionally
deposited C68, C66, C64, and C62(generated by electron-impact
induced ionization/heating of C70(D5h)) onto HOPG surfaces
under soft-landing conditions. We compare the properties of
the resulting thin films to those of smaller non-IPR fullerenes.
Electronic properties were characterized by ultraviolet and
X-ray photoelectron spectroscopies (UPS and XPS). Thermal
stabilities were measured by thermal desorption spectroscopy,
TDS. Topography of the resulting films was studied by means
of atomic force microscopy, AFM. Our results demonstrate,
how the electronic and thermodynamic properties of mono-
dispersed Cn films (valence band, work function, surface
ionization potential, HOMO–LUMO gap, intercage binding
energy) depend on the nuclearity of the building blocks over a
wide range of cage sizes (n = 48–70).
2.Experimental methods
Non-IPR Cnfilms were grown on freshly cleaved pyrolitic
graphite (HOPG, SPI Incorporated, SPI-II quality) by low
energy ion beam deposition at a base pressure of better than
5 ? 10?10
by electron impact ionization/fragmentation of thermally
sublimed IPR C60or C70emanating from a Knudsen cell.
C60(Ih) and C70(D5h) were obtained as powders from Alfa
mbar. The Cn+
projectiles were generated
Artur Bo ¨ ttcherManfred M. Kappes
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 3

This journal is c the Owner Societies 2010Phys. Chem. Chem. Phys., 2010, 12, 10671–1068410673
Aesar Company at 99.5% purity. Cations resulting from
ionization were electrostatically collimated into an ion beam,
which was passed through a 901 bender (to eliminate neutrals
and anions) and a quadrupole mass selector (Extrel) before
being directed onto a perpendicularly oriented highly oriented
pyrolyticgraphitetarget (HOPG).
resolution was chosen to allow baseline separation of
neighboring Cn?2+and Cn+2+peaks from the selected Cn+
(Dm/m E 200). Hence, the films grown were monodispersed
without any measurable contaminants. The mean kinetic
energy of Cn+ions passing the quadrupole mass selector
was set to ca. 34 eV in order to maximize ion flux. Soft-landing
conditions were realized by applying an appropriate positive
retarding potential to the HOPG target. This allowed reduc-
tion of the mean incident kinetic energy to B6 eV without
significant loss of ion flux or broadening of the incident kinetic
energy distribution (which had a width of less than 3 eV). All
experiments described below were performed at constant mean
kinetic energies of 6 eV using HOPG substrates held at room
temperature, Ts= 300 K.
The efficiency of electron impact induced fragmentation
of C60 and C70 depends strongly on the electron kinetic
energy, Eel.34When raising Eel, both the integral flux of all
Cn+as well as the degree of fragmentation increases.
Consequently there is an optimal Eelvalue for each selected
Cn+, e.g.: EelE 36 eV for C68and EelE 85 eV for C50.
All deposition experiments reported here were performed at
such optimized Eel values (which also limit the extent of
projectile vibrational excitation). Before deposition, the HOPG
sample was flashed several times up to 1100 K under high
vacuum in order to remove –OH and –C–H terminations of
step edges. After cooling to room temperature and prior to
deposition, the cleanliness of the substrate was established by
taking UP and XP spectra.
Fullerene layer thickness was determined by measuring the
neutralization current during deposition (picoamperemeter
Keithley). The film thickness is given here in monolayer
equivalents, MLE, whereby 1MLE = 1014cm?2(this corres-
ponds roughly to the molecular packing density in crystalline
solid C60). Attainable ion fluxes for film growth were n
dependent, e.g. B2 nA for C68+and B0.1 nA for C50+.
Consequently, the growth of a 6 MLE-thick film required B1 h
and B20 h deposition times, for C68and C50respectively.
Thermal stability of Cnfilms was characterized by thermal
desorption spectroscopy, TDS, using a second quadrupole
mass spectrometer (Extrel; electron impact ionization) to
monitor the flux of the Cncages escaping from the solid film.
All experiments were performed at the same heating rate of
5 K s?1. The sample temperature during the TD scan was
measured by means of a K-type thermocouple spot welded to
the back side of the sample holder. The temperature in the
central part of the sample surface was measured by means of a
calibrated radiation pyrometer (Keller PZ20AF).
Valence-band electronic structure of deposited Cnfilms was
studied by means of ultraviolet photoelectron spectroscopy,
UPS (21.2 eV), using a hemispherical electron energy analyzer
(ESI 125, Omicron, energy resolution of B0.1 eV) and a
HeI-discharge lamp (Omicron). The relative alignment of the
sample, lamp and electron analyzer was chosen to allow
Themass selector
for optimal detection of the normal component of the photo-
emission. The UP spectra mirror the density of states in the
valence region, VB-DOS, and provide an easy method for
measurement of the work function, f, and the surface ioniza-
tion potential, sIP.35The XP spectra were taken by using a
twin anode X-ray source (MgKaline 1253.6 eV, DAR 400
Omicron) and the UPS hemispherical analyzer for electron
energy determination.
Consistent with our foregoing work35all non-IPR fullerene
films studied were found to be semiconducting as indicated by
a significant gap between the highest occupied and the lowest
unoccupied states. HOMO–LUMO gaps were determined by
taking UP spectra after exposing the Cnfilms to a weak flux of
Cs atoms (SAES Getters SpA, B1014Cs cm?2s?1). This leads
to measurable occupation of the LUMO-derived band, which
manifests itself by an extra feature appearing in the UP spectra
above the HOMO-derived band. The energy separation
between the HOMO- and LUMO-derived bands is then a
simple measure of the band gap.
The topography of deposited films was measured ex situ by
atomic force microscopy, AFM, operating in the non-contact
mode (AFM, Veeco Instruments CP-II). A 5 ? 5 micron
scanner and NSC18 cantilevers with a nominal spring constant
of 4.5 N m?1were typically used. Systematic measurements of
the lateral distribution of deposits revealed a rather flat
macroscopic density profile with well recognizable edges—
corresponding to the ion beam spot. The mean area covered
by Cncage deposits in a typical experiment was B0.07 cm2.
3.Results and discussion
3.1AFM topography measurements: cage size dependence
The topography of films created by deposition of neutral
C60(Ih) or C70(D5h) from effusive beams onto room tempera-
ture HOPG has been extensively studied by AFM.36Film
growth is primarily determined by the relative strengths of van
der Waals cage–cage interactions, E(Cn–Cn), and slightly
stronger cage-substrate interactions E(Cn–S).37In the initial
growth stages, some C60(C70) cages are pinned at step edges
and act as nucleation centres for cages migrating on flat
terraces. The activation barrier for lateral diffusion on the
graphite basal plane is very low (13 meV35) and correspondingly
the mobility of cages is high at room temperature.38At
increasing coverages, this situation, Ediff { E(Cn–S) E
E(Cn–Cn), gives rise to the formation of 2D islands with
smooth rims.39A diffusing C60(C70) cage can only stick to
an island rim if it finds a highly coordinating site. This
stabilization criterion leads to lateral island growth, while at
the same time the island peripheries become gradually smoother
due to rim diffusion processes.
Fig. 1 (upper panel) shows two AFM images, which illus-
trate the formation of such compact 2D C60(Ih) islands with
smooth rims. Further deposition leads to the formation of
small islands on top of the original 1 ML (= 2D) deposits—
constituting the onset of second layer growth. These islands
are however no longer smooth-rimmed because the underlying
fullerene monolayer exhibits high corrugation and corres-
pondingly higher activation energies for cage diffusion,
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 4

10674Phys. Chem. Chem. Phys., 2010, 12, 10671–10684This journal is c the Owner Societies 2010
EdiffE E(Cn–S) (various diffusion channels with Ediffranging
from 178 meV to 429 meV40). Correspondingly, Ediff E
E(Cn–S) E E(Cn–Cn), which leads to hindered lateral cage
mobility and consequently dendritic on-top islands.40The
inset in Fig. 1a shows an AFM image illustrating the shape
difference between islands in the first and second layers. Note
that the AFM images shown in the upper panel in Fig. 1 were
obtained after depositing C60+ions onto HOPG under soft-
landing conditions (T = 300 K, Ekin= 1–6 eV). They show
surface topographies essentially identical to those found when
depositing thermal energy (neutral) C60cages.39Consequently,
under our deposition conditions, impact-induced cage neutra-
lization followed by cage thermalization does not appear to
affect the island topography significantly—relative to films
prepared by effusive beam deposition.
Before comparing the topography of IPR and non-IPR
films it is useful to first consider DFT calculations performed
for non-IPR fullerene cages interacting with a variety of
different substrates (Cn(n = 28, ... 48, 60)/Si(001),41C20
oligomers,42C36 2D polymers,26,27C20/Diamond25). These
predict low energy covalently interlinked, three dimensional
Cn
cagearchitectures—with
constituted by non-IPR sites (mostly 2AP-2AP—where 2AP
indicates a cage surface site with two adjacent pentagons).
Thus, the expectation is that non-IPR fullerene cage
deposition should lead to porous/dendritic films exhibiting
local arrangements of adjacent Cn cages, governed mainly
covalentbondstypically
by their respective numbers of non-IPR sites as well as by
the spatial distribution of these non-IPR sites over the cage
surfaces.
Previously, we have focused on the surface topography of
non-IPR C58 films deposited onto HOPG as studied by
AFM.43The island shape was found to be strongly dependent
on Ekin(within the range: 1 eV o Ekino 40 eV) as well as on
the surface temperature, Ts(varied from 200 K to 800 K)
during deposition. Large dendritic 2D C58islands are formed
at low kinetic energies (and room temperature). These become
smaller and more compact with increasing Ekin. The strongly
dendritic islands also become more compact when depositing
at elevated surface temperatures. These trends can be rationa-
lized in terms of a mechanistic picture in which dissipation of
incident kinetic energy must take place prior to cage
attachment to (preformed) island peripheries. Two main loss
mechanisms are involved: (1) inelastic conversion of the
primary kinetic energy E0 of impacting cages into kinetic
energy of parallel motion across terraces, EJ (EJ = eE0
with the conversion coefficient e depending on E0according
to e(E0in eV) = 0.3 exp(?E0/5.95) + 0.088) and (2) decay of
the EJ component, due to molecular friction acting on
cages which move across the terraces with a velocity v,
EJ(t) = EJ(0)exp(?Zvt) (Z = 8400 eVs m?2).43One can
envisage two qualitatively different sticking pathways, depending
on the orientation of the Cncages upon reaching an island:
(i) all non-IPR sites of an incoming cage become involved in
bonds to the island (or are shielded from further reaction)
leaving only the ‘‘IPR’’ cage regions exposed, or (ii) one or
more non-IPR reaction centres of the incoming cage remain
exposed (and accessible for further bonding) after addition to
a growing island. This distinction also applies to the dynamic
case when a hyperthermal Cncage strikes an island periphery
terminated by cages which are oriented according to either
(i) or (ii). One can then formally distinguish between four
different local attachment types, A–A, A–B, B–A and
B–B,differingconsiderably
(Ebind(E(A–A) o E(A–B) = E(B–A) { E(B–B); with A
and B referring to the IPR and non-IPR cage segments
involved in the cage–cage interaction, respectively). In the
A–A case the cage–cage interaction is van der Waals like
(E(A–A) E 0.6 eV44). In the second case, A–B or B–A,
a significantly higher binding energy is expected. For the
strongest B–B interactions, two adjacent non-IPR sites can
form covalent bonds with the corresponding interaction
energy being EbindE1.9 eV.44Note that an on-terrace Cn
cage moving with a lateral kinetic energy, EJ, can only stick at
the periphery of a growing island if EJo Ebind(neglecting
barriers). Consequently, for hyperthermal kinetic energies,
A–A and A–B pathways will have non-binding outcome
unless dissipative processes have slowed the incoming cage
to EJo 1.9 eV.
The C58films grown by soft-landing at room temperature
are metastable in a thermodynamic sense. We have previously
inferred that A–A and A–B contacts can be converted into
more stable B–B bonding states upon thermal annealing.43
This activates the surface mobility of (previously) A–A and
A–B interlinked cages. Upon encountering an exposed
non-IPR reaction center, these can then form new quasi
ininteractionenergies
Fig. 1
surface after exposure to a low C60+dose at two different incident
kinetic energies, 1 eV and 6 eV, (a) and (b) images, respectively (Ts=
300 K). Only smooth rimmed, compact 2D C60islands are observed.
The inset in the left panel shows an example of second layer growth via
formation of compact C60aggregates on top of the first-layer island.
The two deposition experiments also differ in terms of their C60
coverages (0.1 MLE and 0.5 MLE, left and right panel, respectively).
Lower panel shows two AFM images taken after exposing HOPG
surfaces to comparable doses of C68+and C58+, left (c) and right (d)
panels (1.5 and 1 MLE, respectively; Ts= 300 K, Ekin= 6 eV). In all
experiments with non-IPR Cncages, we observed the formation of
dendritic islands (see also ref. 43).
Upper panel shows two AFM images showing an HOPG
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 5

This journal is c the Owner Societies 2010Phys. Chem. Chem. Phys., 2010, 12, 10671–1068410675
covalent bonds, (the B–B state). Correspondingly, tempering
at elevated sample temperatures can raise the overall film
thermodynamic stability due to an increase in the relative
number of B–B interactions. Fig. 2 shows AFM images
obtained for C58deposits, which illustrate such a transforma-
tion as induced by 550 K annealing. Dendritic islands become
converted into a polymeric network. At even higher annealing
temperatures, desorption sets in and begins to compete with
rearrangement. The AFM image taken after heating the
surface to 1100 K shows that most of the deposited C58
material has been desorbed leaving a chain-like residue
pinned at step edges. The molecular nature of this irreversibly
bound material is presently the subject of further studies.
It may conceivably derive from cage coalescence reactions
which become significant within the desorption temperature
range.45
AFM measurements were also performed on deposits of
other non-IPR fullerene cage sizes: C50and C68. Under the
same deposition conditions as used for C58soft-landing, film
growth was observed to proceed analogously. Three successive
growth stages could be distinguished as coverage was
increased: (1) pinning at HOPG step edges, (2) lateral
growth of dendritic 2D islands on basal planes and (3)
formation of the 3D ‘‘pyramidal’’ structures prior to complete
wetting of the first monolayer (Vollmer–Weber growth
mechanism).
The lower panel in Fig. 1 shows AFM images taken after
deposition of comparable amounts of C58 and C68 onto
HOPG (EkinE 6 eV, Ts= 300 K). In contrast to the smooth
rimmed C60(Ih) islands, all AFM images of the non-IPR films
reveal dendritic island structures. Correspondingly, the lateral
growth of an individual non-IPR fullerene island does
notproceedhomogenously.
directions are preferred. These should ultimately be relatable
to the relative alignment of non-IPR sites in individual cages
(and the corresponding molecular structures). One might
therefore expect significant differences between C50, C58
and C68films, due to the fact that the respective cage structures
may have very different numbers (and spatial distributions) of
reactive sites. However, on the mesoscopic length scales
resolvable with our present AFM setup, no striking differences
were found among the island topographies studied. In future,
we plan high-resolution STM studies to further examine
this issue.
Instead,discrete growth
3.2 Thermal stability
3.2.1
shows a series of thermal desorption spectra obtained by
monitoring the flux of C68molecules escaping from linearly
heated HOPG surfaces covered by C68 films of various
thicknesses. The corresponding measurement for a film of
the parent IPR C70(D5h) is also shown (grey curve). Note,
that the maxima of the broad C68-TD spectra are located
at higher temperatures than for C70-TD, indicating consider-
ably higher thermal stability. At comparable coverages,
C68-TD bands are at least twice as broad as those of
C70 (dw(C68) E 115 K and dw(C70) E 55 K). At very
low coverages, the C68-TD desorption maximum is centered
at B760 K. The desorption maximum shifts towards higher
temperatures at higher coverages—levelling off at 815 K
for >1 ML thick films. According to a simple Redhead
analysis, the activation energy for desorption is proportional
to the temperature at which the maximum desorption rate is
observed, Tm.46Thus, the significant upshift with increasing
coverage indicates a corresponding increase in the binding
energy of C68cages. For the thickest films, a C68desorption
activation energy of 2.1 eV per cage can be derived from the
measurements.
As previously mentioned, the initial C68film growth stage is
characterized by step edge decoration (and empty basal
planes). Thus, the peak in the C68-TD spectrum observed for
low coverages r0.5 MLE (TmE 760 K) can be assigned to
cages pinned by reactive sites terminating the step edges,
–C–Cn. The activation energy for desorption of C68 cages
from such sites appears to be slightly lower than for
desorption from 3D islands in thick films.
Fig. 3b shows corresponding C58-TD spectra as a function
of coverage. Compared to C68they show a higher desorption
maximum centred at 860 K (for thick films). In contrast,
comparably thick C60films show a narrower desorption max-
imum (grey curve) located at much lower temperatures-around
580 K. Mean desorption activation energies derived from a
Redhead analysis were 2.35 eV and 1.63 eV for C58and C60,
respectively. The most striking difference between C68- and
C58-TD spectra concerns the width of desorption bands: C58
features are significantly broader (dw(C58) E 180 K versus
dw(C68)
E
115 K).Nevertheless,
much narrower parent IPR fullerene desorption features
Non-IPR fullerene films: C58 versus C68. Fig. 3a
comparedtothe
Fig. 2
deposits to 550 K for 120 s, (c) after desorbing the majority of the deposits by linearly heating the sample up to 1100 K.
AFM images of HOPG surfaces: (a) after deposition of 1.5 MLE of C58+ions (Ekin= 6 eV, T = 300 K), (b) after annealing the C58
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 6

10676Phys. Chem. Chem. Phys., 2010, 12, 10671–10684 This journal is c the Owner Societies 2010
(dw(C60) E 35 K and dw(C70) E 55 K), C58and C68TD bands
are qualitatively similar.
As noted above, theoretical predictions of stable networks
of non-IPR cages, C20, C36 and C50 have demonstrated
enhanced cohesion due to quasi-covalent intercage links as
e.g. mediated by adjacent pentagon (AP) sites.47Similarly, we
interpret the enhanced thermal stabilities of non-IPR C58and
C68films as reflecting covalent –Cn–Cn– oligomer formation.
The differences in sublimation energies and widths of
desorption bands (Edes(C68) o Edes(C58) and dw(C68) o dw(C58))
must then reflect differences between the respective non-IPR
cage molecular structures and associated differences between
the strengths of their intercage bonding. Following the
terminology of the previous section, we equate the average
sublimation energy with the total binding energy of a
given cage (to all surrounding cages (Esubl= kE(A–A) +
lE(A–B) + mE(B–B), where k, l, m are the mean numbers of
type A–A, A–B and B–B bonds in a typical subliming cage,
respectively)). For equivalent bonds, Edes scales with the
mean coordination number of a cage, k + l + m. However,
E(B–B) c E(B–A) > E(A–A) and consequently, Edes E
mE(B–B), where m corresponds to the mean number of non-
IPR sites per cage. We therefore rationalize the higher thermal
stability of the smaller non-IPR fullerene film, Edes(C68) o
Edes(C58), as indicating that m(C68) o m(C58) (see also note 48).
3.2.2
We presently have no unequivocal experimental data on the
molecular structures (or isomer distributions) of deposited
cages and therefore proceed to a discussion based on structure
predictions from DFT theory. Given the comparatively long
experimental timescale between ionization and deposition
(B100 ms), it is plausible, that in each case the lowest (free)
energy cage isomers of the corresponding cations are
soft-landed. However, the size-dependent details of energy
dissipation and cage rearrangements following electron impact
are unclear. Furthermore, (additional) cage rearrangement
may be associated with surface neutralization and impact
heating. Consequently, we must discuss the structures of
deposited cages in terms of predictions for the lowest energy
isomeric forms of both singly positively and neutral cages of a
given nuclearity. DFT calculations find that the lowest energy
cage isomers for both charge states are conventional fullerene
cages with one exception (n = 62 which is predicted to have a
lowest energy HP-containing isomer). Furthermore, these
DFT calculations49,50(see Table 1) suggest that in the size
range of interest here, the lowest energy singly charged cations
and neutrals have the same cage structures—except for n = 50,
66 and 68. As mentioned previously, the most stable Cn(+)
cages differ by the number of reactive C–C bonds m shared
between two adjacent pentagons. Interestingly, the most stable
Cnand Cn+pairs always exhibit the same m values (except for
C50, see Table 1).
C68[C2: 0112] and C58[C3v: 0001] are calculated to have
m = 2 and m = 3 reactive bonds, respectively.51,52Thus, if we
assume that the overall desorption activation energy scales
linearly with the number of such sites, the inequality
Edes(C68) E 2E(B–B) o 3 E(B–B) E Edes(C58) holds.
(note 53). The broad desorption curves obtained for the two
non-IPR fullerene films (Fig. 3) can then be rationalized to
first order as the result of a broad distributions of differently
bonded cages (which would correspondingly escape at
differenttemperatures). Furthermore,
dw(C68) o dw(C58), may be explained by the fact that a
C58[C3v: 0112] cage can be stabilized by forming one, two or
three 2AP–2AP bonds with the surrounding cages whereas a
C68[C2: 0112] cage can only be one- or two-fold coordinated
(as m = 2).
The thermal desorption behaviour discussed so far for C68
and C58 films was found to be qualitatively the same for
all other non-IPR fullerene films studied (from C50to C68).
Fig. 4 illustrates this generality. Shown are TD spectra for
Film compositions: DFT predictions of cage isomers.
thedifference,
Fig. 3
(0.4, 1, 2.5 and 5 MLE; Ekin= 6 eV, T = 300 K). The grey curve
shows a C70-TD spectrum taken for a B2 ML thick C70film. Note the
considerably higher thermal stability of the non-IPR C68 film.
The inset schematically shows the oligomeric structures responsible
for the observed temperature shift, TA- TB. All spectra were taken at
the same constant heating rate of 5 K s?1. (b) C58-TD spectra taken for
several nominal coverages (0.5, 1.1, 1.7, 4.4, 8.5 and 16 MLE,
respectively; deposition conditions as in the previous figure). For
comparison the C60-TD spectrum obtained for a 1.6 MLE thick film
is also shown (grey curve). Note the pronounced difference in the
thermal stability of the IPR and non-IPR films. All spectra were taken
at the same constant heating rate of 5 K s?1.
(a) C68-TD spectra obtained for four different C68coverages
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 7

This journal is c the Owner Societies 2010Phys. Chem. Chem. Phys., 2010, 12, 10671–1068410677
B20 MLE-thick monodispersed Cnfilms (C68, C66, C64and
C62; deposited at Ts= 300 K and EkinE 6 eV). In comparison
to a C70(D5h) film of comparable thickness, all non-IPR films
exhibit much higher sublimation temperatures. Corresponding
Tmvalues lie between 775 K and 825 K and exhibit a clear cage-
size dependence. Again all TD spectra manifest considerably
broader desorption bands than found for C70. Interestingly,
C64-TD spectra are approximately twice as wide as the others
and clearly consist of at least two Gauss-like components
indicating the presence of two distinguishable isomers.54
We have previously reported analogous TD measurements33,35
for films of Cn(n = 50, 52, 54, 56 and 58) resulting from
fragmentation of C60(Ih) cages (Ts= 300 K, EkinE 6 eV). In
analogy to the C62–C68 series, the thermal properties of
C50–C58films were also strongly dependent on cage size, with
Tmvalues ranging from 815 K for C58up to 950 K for C50.
Similarly, the smaller the cage, the wider the TD spectra—
ranging up to dw(C50) B210 K.
Fig. 5 summarizes the size dependencies of the mean
desorption activation energies for all TD measurements. In
general, the smaller the non-IPR cages, the higher the film
desorption activation energies, Edes(Cn). These ranged from
2.1 eV for C68 to 2.6 eV for C50 films—depending rather
smoothly on n. The only significant departures from this trend
were found for the parent C60and C70IPR fullerene films
(1.55 eV and 1.65 eV, respectively, in good agreement with
published data55). For the non-IPR cages, we note that the
m values corresponding to the lowest energy, DFT-based, cage
isomers also increase with decreasing size—from 2 for C68up
to 5 and 6 for the two lowest C50isomers. Consequently, the
general trend apparent in Fig. 5 can be qualitatively rationa-
lized in terms of the mean numbers m of pentagon-adjacencies
on the respective cages and by the associated mean coordination
numbers (see Table 1). Assuming, that these are in fact the
deposited molecular structures, one can calculate the mean
dissociation energy per –2AP–2AP– connection, e(n) =
Edes(Cn)/m.56This results in: e(n) = ?1.47 + 0.038n. Note,
that the dissociation energies of 2AP–2AP connections as
derived from DFT calculations for isolated dimers (2.61 eV
Table 1
calculations46,64 a
Structural parameters of the most stable neutral and singly positively charged fullerenes, Cn and Cn+, according to DFT
Cage isomer Symmetry Structural motifsm
Cn
Cn(0)
Cn(+)
Cn(0)
Cn(+)
Cn(0)
Cn(+)
m
C70
C68
C66
C64
C62
C60
C58
C56
C54
C52
C50
(6000)
(3981)
(1789)
(1998)
(1994)
(1812)
(1205)
(916)
(540)
(422)
(270)
(271)
D5h
C2
Cs
D2
C2
Ih
C3v
D2
C2v
C2
D3
D5h
IPR
2-2AP
1-C3AP
2-2AP
1-C4AP + 1 HP
IPR
3-2AP
4-2AP
2-C3AP
2-C3AP + 1-2AP
6-2AP
5-2AP
0
2
2
2
3 + HP
0
3
4
4
5
6
5
0
2
2
2
3
0
3
4
4
5
5.5
b
(C2)
(C2v)
(2-2AP)
(2-2AP)
(2)
(2)
(D5h)(5-2AP)(5)
aThe cage isomers are labelled according to the Fowler and Manolopoulos scheme.69In addition to molecular symmetries, we have also tabulated
the corresponding structural motifs (C4AP, C3AP, 2AP and HP refer to chain of four adjacent pentagons, chain of three adjacent pentagons, two
adjacent pentagons and heptagon ring, respectively). Also listed are the number of common C–C bonds shared between adjacent pentagons, m.
Where the structures of neutral and cationic Cncages of a given n differ, the entries for the cationic cages are given in italic parentheses
(and bold face). This is only the case for: C50, C66and C68. Excepting n = 50, the same m is found for all Cn/Cn+. Whereas the most stable form of
C50exhibits six 2AP sites (and therefore six common bonds) the corresponding cation C50+has only five. Tabulated m(n) values are the averages of
the m values for Cn0and Cn+.bWe included two C50isomers (D3and D5h) as representing the common lowest isomer because according to DFT
calculations they exhibit nearly the same values of the total energy.
Fig. 4
non-IPR fullerenes obtained by soft-landing the corresponding
[C70]-derived fragment ions: C68, C66, C64and C62(Ts= 300 K and
Ekin= 6 eV, 20 MLE thick films). For comparison the TD spectrum of
the parent C70cages is also shown (grey line curve). All spectra were
taken at the same constant heating rate of 5 K s?1.
Thermal desorption spectra obtained for thick films of
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 8

10678Phys. Chem. Chem. Phys., 2010, 12, 10671–10684This journal is c the Owner Societies 2010
for a C50= C50connection, and 1.93 eV for C58= C5844) are
considerably higher than the e(n) values determined from
experiment. Moreover, e(n) values were observed to increase
with cage size—in contrast to computed dissociation energies,
which show the opposite trend. We conclude, that the actual
mean number of 2AP–2AP inter-cage interactions present in
non-IPR Cn solids, m*, must be significantly smaller than
m because of steric constraints (dependent on cage size and
spatial distribution of AP sites) which can hinder the forma-
tion of intercage connections. Additionally (and particularly
on the highly corrugated surface of a multi-ML film),
A–A and A–B states might present a kinetic hindrance for
formation of B–B links because of the random nature of the
soft-landing and surface transport events.
As indicated by Fig. 1 and 2 as well as by our previous
work,35,43in the low-coverage range the desorption yield as well
as the temperature of maximum desorption rate depend
significantly on substrate topography (atomic defects, step
edges, grain boundaries, etc.). The above discussion of Cn
desorption/sublimation activation energies is based on data
obtained in the sublimation regime (thick films) in which the
positions of TD maxima are essentially independent of film
thickness. Note, however, that sublimation of individual
non-IPR cages at elevated sample temperatures competes with
irreversible coalescence/fusion to yield stable non-desorbable
conducting networks. The latter reaction is expected to be
facilitated by a precursor arrangement of Cncages with highest
possible m*-values whereas sublimation preferentially occurs
from low m*-value areas. Details will be discussed in a
separate paper (Ulas et al.57).
3.3Electronic structure
3.3.1
structure of non-IPR Cnfilms was characterized using UPS
HOMO-derived valence bands.The electronic
(He 21.2 eV) and XPS (MgKa 1253.6 eV) with particular
emphasis on the valence band (VB) regions and their cage size
dependence.
We first focus on C68and C58films and compare their VB
structure to those of IPR C70and C60films (Fig. 6). The latter
spectra are both in good agreement with the literature.58
Whereas the HOMO-derived band of C70films exhibits five
distinguishable components, the C68-HOMO-derived band
comprises a broad doublet significantly shifted towards the
Fermi level (DE E 0.3 eV) (Fig. 6a). By contrast, a comparison
of C60and C58films reveals qualitatively different behaviour.
Whereas the C60-HOMO-derived band consists of a doublet
with two sharp nearly Gaussian components, the HOMO-
band of the C58film exhibits three distinguishable components
shifted towards EFby B0.5 eV (Fig. 6b). For both non-IPR
films, the observed changes relative to films of the respective
Fig. 5
carbon atoms per cage, n. The data were gained in the sublimation
regime, i.e. for film thickness B20 MLE. Note that Edes values
scale with the average number of non-IPR sites per cage m
(= AP sites)—when assuming that the films consist of most stable
Cnisomers as determined by DFT calculations (the corresponding m
values are indicated at the right margin as well as in Table 1). For C64
films two distinguishable desorption states were found—marked as A
and B. In this context, see also ref. 66 which reports a desorption
activation energy of B1.9 eV for C60(C2v) having m = 2.
Desorption activation energies, Edes, versus the number of
Fig. 6
monitored by ultraviolet photoelectron spectroscopy, UPS (hv =
21.2 eV) for 8 MLE C68and C58films, panels a and b, respectively.
For comparison, the DOS-VB bands obtained for comparably thick
films of the parent C70and C60IPR cages are also shown (grey curves)
in the upper and lower panels, respectively.
Densities of states in the valence bands, DOS-VB, as
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 9

This journal is c the Owner Societies 2010Phys. Chem. Chem. Phys., 2010, 12, 10671–1068410679
precursor IPR cages are likely due to the 2AP structural motifs
of the building blocks as well their covalent interconnections
(m = 2 and 3 for ground state C68 and C58 isomers,
respectively). As we have previously shown, DFT-based
calculations of the densities of electronic states for short C58
oligomers interlinked by –AP–AP– bonds, recover the triplet
HOMO-DOS structure shown in Fig. 6b. Correspondingly, we
have tentatively suggested that this triplet—in particular its
central component—may be a general marker for AP sites and
inter-cage bonds in non-IPR fullerene films. Support for this
view was obtained by analogous features seen in UPS spectra
of C56, C54, C52and C50films. However, the C68-HOMO-
derived broadened doublet band shown in Fig. 6b is in
contradiction with this simple 2AP ‘‘electronic fingerprint’’
view. Fig. 7 illustrates this point further. It shows expanded
scale segments of the UP spectra obtained for comparably
thick Cnfilms, from C70 to C48.59Superimposed on these
HOMO-derived band regions are Gaussian fits to the data.
Whereas the HOMO-derived bands of the [C60] fragment
family could be well fit with only three Gaussian components
(labelled n, Z and m), the [C70]-derived fragment Cn cages
required at least four Gaussian components (a, b, g and d), in
order to obtain comparably good fits (Fig. 7). Apparently, the
electronic signature of interdependent/interconnected AP–AP
sites depends strongly on the size and symmetry of the
component cages. The distinct trends in the HOMO(n)
Fig. 7
ionization of C70and C60, panels a and b, respectively. The DOS-HOMO-derived bands in the [C70] group have been best-fitted with four Gaussian
components, a, b, g and d, centred at 1.8, 2.35, 3.4 and 4 eV, respectively. In contrast, the DOS-HOMO-derived bands of the [C60] family were
satisfactorily fitted using only three Gaussian profiles, n, Z, and m peaked at 1.7, 2.55 and 3.35 eV, respectively. In each case we show the best fits for
the minimum number of components.
DOS-HOMO-derived bands versus cage size for 6–12 MLE thick films of non-IPR fullerenes generated as fragment ions by electron impact
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 10

10680Phys. Chem. Chem. Phys., 2010, 12, 10671–10684This journal is c the Owner Societies 2010
dependencies found here for members of the two families [C60]
and [C70] implies that there is no common electronic signature
of AP–AP links for all cages. Instead the HOMO-DOS seems
to mirror merely the origin of the Cncage ([C60] or [C70]
family). In future, it will be of interest to systematically
compare experimental spectra with DOS calculations of model
oligomers—taking into account DFT calculations of the most
stable Cnisomers for each n (and/or experimental data on
actual molecular structures).47–52This will help to understand
the dependence of the DOS: (i) on the respective 2AP
local geometries (e.g. pentagon face sharing C–C, bond
lengths47,51,60and associated pyramidalization angles13) as
well as (ii) on the lateral distributions of AP sites over the
cage ‘‘surfaces’’.
3.3.2
ionization energies of isolated Cncages can be modelled in
terms of a hollow metallic sphere with homogeneous charge
distribution61leading to the relationship: IP = IPN+ e2/2R,
where R is the cage radius and IPNis the ionization potential
for R - N. Corresponding measurements on gas-phase Cn
cages confirm the main trend but also resolve significant cage-
size dependent deviations. It is interesting to consider to what
extent matrix interactions and in particular the formation of
oligomeric chains in non-IPR solid films modifies the ‘‘surface
ionization potentials’’, sIP(R), of these cages relative to the IP
values of individual cages isolated in gas-phase. Surface
ionisation potentials can be determined from film work
functions, f, and the energies of the topmost occupied states,
EB(HOMO), according to sIP = EB(HOMO) + f. We
obtained f values from the UP spectral widths, W, according
to f = hn ? W. This in turn required identification/
assignment of the vacuum and Fermi levels in the spectra,
which together constrain W. In our UP spectra, the HOPG
substrate defines the Fermi level. For all thick Cnfilms, we
measured similar values ranging from 4.7 eV up to 4.8 eV.
Thus, within experimental error of ?0.06 eV, non-IPR films
have a common work-function of 4.75 eV. To within similar
accuracies, we determined the work functions of C60and C70
films to be 4.6 eV and 4.7 eV, respectively. EB(HOMO) values
were obtained by fitting HOMO-derived-bands with Gaussian
functions and assigning EB(HOMO) as the energy of the peak
maximum corresponding to lowest binding energy Gaussian
component. Fig. 8 illustrates the resulting sIP(n) values for
48 r n r 70 (black squares).62For comparison, we also show
IP values as obtained from DFT calculations for the lowest
energy Cncage isomers as found at the B3LYP/6-31 G level
of theory44,63,64(open circles) as well as the experimental data
obtained for gas phase cages65(grey circles). There is an
unexpectedly clear correlation between experimental sIP(n)
and DFT-IP(n) values. We note that our procedure for
determining sIP(n) as applied to a mixture, would yield a
value corresponding to its lowest sIP component. This could
conceivably correspond to the selective ionization of weakly
A–A bound non-IPR cages embedded within a more strongly
(B–B) covalently bound oligomeric matrix.
Ionization potentials. The size dependence of the
3.3.3
HOMO- and LUMO-derived bands of non-IPR Cnfilms can
HOMO–LUMOgaps. Thegaps between the
be determined by UPS if measurable occupation of LUMO-
derived bands is provided for by doping with electropositive
atoms. In such doping, two requirements have to be fulfilled:
(1) the ionisation energy of the dopant atoms should be
lower than the binding energy of the LUMO-derived band,
(Eio E(LUMO) with respect to the vacuum level) and (2) the
relative concentration of dopant atoms should be as low as
possible in order to avoid the formation of stoichiometric
fulleride phases stabilized by periodically arranged cations
(resulting from the associated electron transfer process,
Cn+ A - Cn?+ A+). Consequently, we have deposited
small and comparable amounts of Cs atoms (B2 ? 1015cm?2)
onto thick Cnfilms and have then measured the complete
occupied valence band regions by UPS. Fig. 9 illustrates two
representative examples of the modifications observed upon
Cs-doping C70and C68, panels (a) and (b), respectively. In
order to interpret these spectra, the energetic positions of the
newly occupied LUMO-derived bands were determined by
fitting the valence band regions with (the minimum necessary
number of) Gaussian components. The striped Gaussian
profiles shown in Fig. 9 indicate the corresponding LUMO-
derived bands. We take the energy difference between the peak
maximum of the highest HOMO component and that of the
new LUMO-band as a simple measure of the actual band gap, D.
Fig. 10 summarizes the D data obtained for all Cn films
(48 r n r 70).
For C70 films we measured gap values ranging between
2.05 and 2.15 eV—somewhat higher than previously found
for C60 films (D E 1.95 eV66). In general, non-IPR films
exhibited lower D values than did films of their IPR C60
and C70 precursors. They ranged from 0.8 eV to 1.8 eV.
[C70]-fragment derived non-IPR films generally have signifi-
cantly higher D values than do films of (smaller) [C60] derived
cages (including C48). As previously pointed out, C50
represents an exception in this regard in that its D value
(1.7 eV), is comparable to those of C60and C70films. This
Fig. 8
fullerene films as a function of cage size, black squares. Corresponding,
DFT-derived IP predictions for the most stable isomers are presented
by open circles.64The IP values measured for gas-phase Cn(n o 60)
cages are marked by grey full circles65(see also Table 2 which lists IP
values calculated for the most stable Cnisomers).
Surface ionization potentials, sIP, as measured for 6–12 MLE
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 11

This journal is c the Owner Societies 2010Phys. Chem. Chem. Phys., 2010, 12, 10671–1068410681
implies that the contributing C50cages have comparatively
high molecular symmetry, consistent with the formation and
deposition of the most stable (highly symmetric) isomers as
determined by DFT calculations at the B3LYP/6-31G* level
of theory. We note that the seven most stable cages (with D5h,
D3, Cs, C2symmetries) are predicted to have a mean D value of
1.74 eV.6
D(n) values have not previously been calculated for poly-
meric Cnsolids. Hence, we can only compare measured D
values with the corresponding DFT-based results for isolated
cages. This is illustrated in Fig. 10 which shows the experi-
mental values versus DFT (B3LYP/6-31G*) predictions for
the lowest energy neutral Cnisomers (48 r n r 70).62The
corresponding D(n) data sets appear to be highly correlated
but are systematically offset from each other. The mean offset
is B0.34 eV—likely due to in-matrix stabilization of cage-
localized excess charges relative to the situation for an isolated
gas-phase Cn. As in the case of the ‘‘surface ionization
potentials’’, the high correlation between film measurements
and isolated molecule DFT-D calculations is rather surprising.
The implication is again that the films created here by soft-
landing at room temperature must consist of a considerable
amount of cages with free non-IPR sites. Alternatively
(or additionally), inter-cage bonds do not significantly modify
the band gap relative to a hypothetical non-IPR Cnvan der
Waals solid. We note in closing that, in some cases, the
agreement between experimental and theoretical D(n) values
can be further improved by also including the next most stable
isomers (e.g. for C58, the mean DFT-D value determined for
the two the most stable isomers, C3v: 0001 (0.91 eV) and C3:
hept (1.55 eV)51,52is in better agreement with the experimental
value (1.2 eV)) (see also Table 2).
3.3.4
by means of UPS, non-IPR films exhibit several distinguishing
electronic markers in their valence band regions. We looked
for analogous markers in C1s core states by means of XP
spectroscopy. Fig. 11 shows corresponding measurements for
8 MLE-thick C60and C58films deposited onto SiO2under
standard conditions (Ekin E 6 eV, T = 300 K, surface:
n-doped and oxidized Si wafer annealed under UHV).
X-Ray photoelectron spectroscopy. As already shown
Fig. 9
DOS-VB, showing the effect of Cs doping of 8 MLE-thick films, C70
and C68 films, (a) and (b), respectively. Upper and lower panels
correspond to pristine and Cs-doped phases, respectively. LUMO
derived bands were best fit by five superimposed Gaussian functions.
UPS derived densities of states in the valence band regions,
Fig. 10
D(n), versus the number of carbon atoms per cage n (black squares).
For comparison, the DFT derived D values for the most stable isolated
cages are also shown in49,60(open triangles). See also Table 2 which
lists theoretical D values of the most stable Cnisomers as well as the
related mean D* values obtained for several lowest Cnisomers.
Experimentally determined HOMO–LUMO gap values,
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 12

10682Phys. Chem. Chem. Phys., 2010, 12, 10671–10684This journal is c the Owner Societies 2010
The C60XP-C1s spectrum is fully consistent with the literature.67
Its main band is peaked at 285.1 eV and can be well fit by one
Gaussian component, a, having FWHM E 1.6 eV. Towards
higher binding energy, weak shake-up satellites are also
visible, which together form a rather broad band with two
separately fitable features at B286.4 eV and B289.5 eV. The
latter has been attributed to the decay of the low-energy
plasmon.68In contrast to the C60film, the main band in the
C58XP-C1s spectrum is significantly broader (B3.1 eV) and
shifted towards higher binding energy (by B0.55 eV). It can be
well fitted with only two components, a and b of roughly equal
intensity and separated by B1.2 eV. Component b is peaked at
286.3 eV and does not appear in the C60 spectrum. The
shake-up features recognizable as a separate band in the
C60-XP spectrum are not as clearly present in the C58-XP
spectrum. The C58-C1s band resembles the C1s peak found for
amorphous carbon films.67Both exhibit similar broadening
with respect to the C60-C1s peak. However, while the C58-C1s
band is broadened and shifted towards higher binding
energies, the corresponding broad band of amorphous carbon
has its maximum at the C60-C1s peak position. Apparently,
the b component in the C58-C1s spectrum is a measure of non-
IPR sites in the C58oligomers. In future, we intend to perform
systematic XPS measurements in order to explore the relation
between broadening of the C1s peak and cage size.
4. Summary and outlook
Mass-selective soft landing (Ekin= 6 eV) of fragment cations
generated by electron impact ionization of thermally sublimed
C60(Ih) and C70(D5h) has been used to generate a new class of
multilayer fullerene-nanofilms comprising classical non-IPR
carbon cages (Cn: 48 o n o 60 and 60 o n o 70) on HOPG.
Relative to the IPR reference phases (i.e. C60(Ih), C70(D5h)),
non-IPR films show significantly (and qualitatively) different
properties.
For all of the non-IPR fullerene cage sizes studied here,
resultant film topographies, thermal stabilities as well as
general electronic properties (VB-DOS, sIP and HOMO–LUMO
gaps (D)), can be rationalized in terms of the formation
of covalent inter-cage bonds involving reactive non-IPR-
sites (which typically comprise adjacent pentagon pairs).
Table 2Electronic parameters determined in this study for as-deposited Cnfilmsa
Isomer
(label)
Symmetry
D/eV
Ref. 60
D*/eV
Ref. 49
Dexp/eV IP/eV
Ref. 64
IP*/eVsIPexp/eV
C70(6000)
C68(3981)
C66(1789)
C64(1998)
C62(1994)
C60(1812)
C58(1205)
D5h
C2
Cs
D2
C2
Ih
C3v
2.681
2.337
1.951
2.219
1.296
2.758
0.910
2.68 (1)
1.45 (11)
1.62 (5)
2.13 (3)
1.38 (10)
2.76 (1)
1.13 (5)
2.09
1.92
1.70
1.60
1.40
1.95
1.25
7.02
6.87
6.86
6.78
6.36
7.14
6.41
6.93
6.85
6.65
6.64
6.45
6.95
6.55 6.52 (1)
Ref. 52
C56(916)
C54(540)
C52(422)
D2
C2v
C2
1.650
1.311
1.296
1.70 (4)
1.43 (3)
1.32 (5)
1.30
1.10
0.80
6.65
6.61
6.50
6.64
6.70
6.856.45 (4)
Ref. 63
C50(270)
C48(171)
D3
C2
2.26 (1.37)
1.564
1.85 (5)
1.65 (9)
1.75
1.16
7.26
6.85
6.95
6.70
aListed are HOMO–LUMO gaps, Dexp, and surface ionization potentials, sIPexp, in comparison to the corresponding D and IP values calculated
for the most stable classical non-IPR Cnisomers in gas-phase. D* and IP* indicate the mean values obtained when taking into account all other
(less stable) Cnisomers within 0.78 eV mol?1of the lowest energy forms. The overall number of different Cnisomers included is given in
parentheses. Lowest energy isomers Cnare labelled according to the Fowler and Manolopoulos scheme.69
Fig. 11
and C58films, upper and lower panels, respectively. The b component
is present only in the XP spectra of the C58film and correspondingly
may be regarded as an electronic marker for (quasi covalently inter-
linked) non-IPR sites.
XP spectra in the C1s state regions as obtained for thick C60
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 13

This journal is c the Owner Societies 2010Phys. Chem. Chem. Phys., 2010, 12, 10671–1068410683
Furthermore, our measurements show that the properties of
non-IPR Cnfilms exhibit a systematic dependence on the size
of the building blocks. The corresponding size dependencies
found for Edes(n), D(n), sIP(n), can be rationalized by assuming
that the non-IPR films are composed predominantly of the
most stable (neutral) Cnisomers at that particular size—as
predicted by density functional theory calculations. In
particular, the observed decrease in Edes(n) with increasing
size can be interpreted to first order in terms of the numbers
and spatial distributions of non-IPR reaction centres on the
corresponding cages. However, slight but systematic differ-
ences between the DOS(n) functions of [C70]- and [C60]-derived
fragment families indicate that the number of these non-IPR
reaction centres alone is insufficient to fully explain the
electronic structure trends observed. Additionally, the cage-
specific local structure and relative orientation of pentagon
adjacency sites (i.e. C–C bond lengths and pyramidal angles)
also play an important role.
In future, we will attempt to more directly determine the
corresponding non-IPR isomer compositions and linkages—
using a combination of scanning tunnelling microscopy,
vibrational spectroscopy and diffraction measurements for
molecular structure determination. This should in turn facil-
itate further studies of the characteristic chemical properties of
non-IPR films. So far, the smallest cluster size for which we
have succeeded in generating multilayer non-IPR fullerene
films is n = 48. It will be of interest to develop appropriate
ion sources so as to access molecularly uniform solids
(or mixed/layered films) comprising even smaller all-carbon
cluster building blocks. From the viewpoint of deposition
dynamics, the extent of activation barriers for covalent bond
formation and their influence on island growth remains to be
explored. Finally, non-IPR cages deposited onto HOPG at
room temperature provide a conveniently accessible and
chemically well-defined model system with which to study
the principles of two-dimensional fractal formation and
(thermally- or reaction-induced) dissolution. In combination
with appropriate surface prestructuring on the nanometer scale,
a more quantitative understanding of the associated mass
transport processes should facilitate novel ‘‘self-assembly’’
strategies of potential interest for all-carbon nanotechnology.
Acknowledgements
This research was supported by the Deutsche Forschungsge-
meinschaft (DFG) and the State of Baden-Wu ¨ rttemberg
through the DFG-Center for Functional Nanostructures
(CFN) within subproject C4.6.
References
1 H. W. Kroto, Nature, 1987, 329, 529.
2 Fullerene: Chemistry, Physics and Technology, ed. K. M. Kadish
and R. S. Ruoff, Wiley, New York, 2002.
3 E. Albertazzi, C. Domene, P. W. Fowler, T. G. Heine, G. Seifert,
C. V. Alsenov and F. Zerbetto, Phys. Chem. Chem. Phys., 1999, 1,
2913.
4 P. W. Fowler, Contemp. Phys., 1996, 37, 235.
5 E. E. E. Campbell, P. W. Fowler, D. Mitchell and F. Zerbetto,
Chem. Phys. Lett., 1996, 250, 544.
6 S. Dı´az-Tendero, M. Alcamı´ and F. Martı´n, Chem. Phys. Lett.,
2005, 407, 153.
7 H. Kietzmann, R. Rochov, G. Gantefo ¨ r, W. Eberhardt, K. Vietze,
G. Seifert and P. W. Fowler, Phys. Rev. Lett., 1998, 81, 5378.
8 H. Prinzbach, A. Weiler, P. Landenberger, F. Wahl, J. Wo ¨ rth,
L. T. Scott, M. Gelmont, D. Olevano and B. v. Issendorff, Nature,
2000, 407, 60.
9 C. Piskoti, J. Yarger and A. Zettel, Nature, 1998, 393, 771.
10 W.Kra ¨ tschmer, L.D. Lamb,
D. R. Huffmann, Nature, 1990, 347, 354.
11 A. Koshio, M. Inakuma, T. Sugai and H. Shinohara, J. Am. Chem.
Soc., 2000, 122, 398.
12 X. Lu, Z. Chen, W. Thiel, P. von Rague Schleyer, R. Huang and
L. Zheng, J. Am. Chem. Soc., 2004, 126, 14871.
13 X. Gao and Y. Zhao, J. Comput. Chem., 2007, 28, 795.
14 Q.-Bo. Yan, Q.-R. Zheng and G. Su, Carbon, 2007, 45, 1821.
15 T. Heine, P. W. Fowler, K. M. Rogers and G. Seifert, J. Chem.
Soc., Perkin Trans. 2, 1999, 707.
16 S. Y. Xie, F. Gao, X. Lu, R. B. Huang, C. R. Wang, X. Zhang,
M. L. Liu, S. L. Deng and L. S. Zheng, Science, 2004, 304, 699.
17 C.-R. Wang, Z.-Q. Shi, L.-J. Wan, X. Lu, L. Dunsch, C.-Y. Shu,
Y.-L. Tang and H. Shinohara, J. Am. Chem. Soc., 2006, 128, 6605.
18 Q.-B. Yan, Q.-R. Zheng and G. Su, Carbon, 2007, 45, 1821.
19 C. R. Wang, T. Kai, T. Tomiyama, T. Yoshida, Y. Kobayashi and
E. Nishibori, Nature, 2000, 408, 426.
20 Y.-Z. Tan, S.-Y. Xie, R.-B. Huang and L.-S. Zheng, Nat. Chem.,
2009, 1, 450.
21 Z. Q. Shi, X. Wu, C. R. Wang, X. Lu and H. Shinohara, Angew.
Chem., Int. Ed., 2006, 45, 2107.
22 P. A. Troshin, A. G. Avent, A. D. Darwish, N. Martsinovich,
A. K. Abdul-Sada, J. M. Street and R. Taylor, Science, 2004, 309, 278.
23 W. Qian, S.-C. Chuang, R. B. Amador, T. Jarrosson, M. Sander,
S. Pieniazek, S. I. Khan and Y. Rubin, J. Am. Chem. Soc., 2003,
125, 2066.
24 M. Co ˆ te ´ , J. C. Grossman, M. L. Cohen and S. G. Louie, Phys. Rev.
Lett., 1998, 81, 697.
25 A. J. Du, Z. Y. Pan, Y. K. Ho, Z. Huang and Z. X. Zhang, Phys.
Rev. B: Condens. Matter, 2002, 66, 035405.
26 P. W. Fowler, T. Heine, K. M. Rogers, J. P. B. Sandall, G. Seifert
and F. Zerbetto, Chem. Phys. Lett., 1999, 300, 369.
27 E. G. Gal’pern, I. V. Stankevich, A. L. Chistyakov and
L. A. Chernozatonskii, J. Mol. Graphics Modell., 2001, 19, 189.
28 L. Zhechkov, T. Heine and G. Seifert, J. Phys. Chem. A, 2004, 108,
11733.
29 V. Paillard, P. Melion, V. Dupuis, A. Perez, J. P. Perez,
G. Guiraud, J. Fornazero and G. Panczer, Phys. Rev. B: Condens.
Matter, 1994, 49, 11433.
30 P. Melion, V. Paillard, V. Dupuis, A. Perez, P. Jensen, A. Hoareau,
J. P. Perez, J. Tuaillon, M. Broyer, J. L. Vialle, M. Pellarin,
B. Baguenard and J. Lerme, Int. J. Mod. Phys. B, 1995, 9, 339.
31 P. Milani and S. Iannotta, Cluster Beam Synthesis of Nanostructured
Materials, Springer, Berlin, 1999.
32 B. G. Levi, Phys. Today, 2001, 1, 15.
33 A. Bo ¨ ttcher, P. Weis, S.-S. Jester, D. Lo ¨ ffler, A. Bihlmeier,
W. Klopper and M. M. Kappes, Phys. Chem. Chem. Phys., 2005,
7, 2816.
34 S. Matt, P. Scheier, A. Stamatovic, H. Deutsch, K. Becker and
T. D. Ma ¨ rk, Philos. Trans. R. Soc. London, Ser. A, 1999, 357, 1201.
35 D. Lo ¨ ffler, S.-S. Jester, P. Weis, A. Bo ¨ ttcher and M. M. Kappes,
J. Chem. Phys., 2006, 124, 054705.
36 Z. Y. Li, Surf. Sci., 1999, 441, 366.
37 P. A. Gravil, M. Devel, Ph. Lambin, X. Bouju, Ch. Girard and
A. A. Lucas, Phys. Rev. B: Condens. Matter, 1996, 53, 1622.
38 D. J. Kenny and R. E. Palmer, Surf. Sci., 2000, 447, 126.
39 H. Liu and P. J. Reinke, J. Chem. Phys., 2006, 124, 164707.
40 V. Liu, Z. Lin, L. V. Zhigilei and P. Reinke, J. Phys. Chem. C,
2008, 112, 4687.
41 S. Yao, C. Zhou, B. Han, T. Fan, J. Wu, L. Chen and H. Cheng,
Phys. Rev. B: Condens. Matter, 2009, 79, 155304.
42 Y. Miyamoto and M. Saito, Phys. Rev. B: Condens. Matter, 2001,
63, 161401.
43 S.-S. Jester, D. Lo ¨ ffler, P. Weis, A. Bo ¨ ttcher and M. M. Kappes,
Surf. Sci., 2009, 603, 1863.
44 A. Bihlmeier, C. C. M. Samson and W. Klopper, ChemPhysChem,
2005, 6, 2625.
K. Fostiropoulosand
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online

Page 14

10684Phys. Chem. Chem. Phys., 2010, 12, 10671–10684This journal is c the Owner Societies 2010
45 (a) I.-H. Lee, et al., Appl. Phys. Lett., 2006, 88, 011913;
(b) S. Bandow,T. Hiraoka,
H. Shinohara and S. Iijima, Chem. Phys. Lett., 2004, 384, 320.
46 P. A. Redhead, Vacuum, 1962, 12, 203.
47 X. Lu and Z. Chen, Chem. Rev., 2005, 105, 3643.
48 Note that in the initial deposition stages only surface defects and
step edges are decorated by Cncages (even for very gentle deposition
conditions, Ekino 1 eV33,35). This indicates that non-IPR cage-
HOPG binding is insignificant by comparison.
49 N. Shao, Y. Gao and X. C. Zeng, J. Phys. Chem. C, 2007, 111,
17671.
50 S. Diaz-Tendero, M. Alcami and F. Martin, J. Chem. Phys., 2003,
119, 5545.
51 D.-L.Chen, W. Q.Tian,
ChemPhysChem, 2008, 9, 454.
52 D.-L.Chen,W. Q. Tian,
ChemPhysChem, 2007, 8, 1029.
53 Note that this remains true even when considering the second most
stable neutral isomers: C68[C2: 0178] and C58[Cs: hept] which
would be present in a hot neutral beam (TBo 1000 K) at relative
concentrations of less than 38% and 9%, respectively51,52.
54 D. Lo ¨ ffler, et al., in preparation.
55 H.Ulbricht, G. Moosand T.Hertel,Phys.Rev. Lett., 2003, 90,095501.
56 The m(n) values were obtained by averaging the theoretical values
predicted for the most stable Cn(0)and Cn(+)cages. In the case of
C62the presence of a heptagonal ring has been ignored by taking
into account the 3-2AP motifs only.
57 S. Ulas, et al., in preparation.
58 (a) P. He, S. Bao, C. Yu and Y. Xu, Surf. Sci., 1995, 328, 287;
(b) S. He, M. Arita, H. Namatame, M. Taniguchi, H.-N. Li and
H.-Y.Li, J.Phys.:Condens.
T.Yumura, K. Hirahara,
J.-K.Feng andC.-C.Sun,
J.-K. Fengand C.-C.Sun,
Matter, 2007,19,026202;
(c) M. Knupfer, D. M. Poirier and J. H. Weaver, Phys. Rev. B:
Condens. Matter, 1994, 49, 2284.
59 Expanded scale segments are shown following background
subtraction. For this, a function Aexp(Ebind/B) has been fitted in
the spectral region between Ebind= 0 and first minimum at binding
energy B4 eV.
60 M. Alcami, G. Sanches, S. Diaz-Tendero, Y. Wang and F. Martin,
J. Nanosci. Nanotechnol., 2007, 7, 1329.
61 O. Boltalina, I. N. Ioffe, L. N. Sidorov, G. Seifert and K. Vietze,
J. Am. Chem. Soc., 2000, 122, 9745.
62 In order to apply the same fit procedure to the HOMO-bands of all
members of both [C60] and [C70] families we repeated the measure-
ments of C58, C56, C54, and C52films. Whereas the IP(Cn) values
well reproduce those presented in ref. 35 the new IP(C52) value
differs significantly. The error in our old IP(C52) value was traced
to an error in pinning the referencing Fermi level.
63 S. Dı´az-Tendero, F. Martin and M. Alcamı´, ChemPhysChem,
2005, 6, 92.
64 S. Dı´az-Tendero, G. Sa ´ nchez, M. Alcamı´ and F. Martin, Int. J.
Mass Spectrom., 2006, 252, 133.
65 J. A. Zimmerman, J. R. Eyler, S. B. H. Bach and S. W. McElvany,
J. Chem. Phys., 1991, 94, 3556.
66 D. Lo ¨ ffler, N. Bajales, M. Cudaj, P. Weis, S. Lebedkin,
A. Bihlmeier, D. P. Tew, W. Klopper, A. Bo ¨ ttcher and
M. M. Kappes, J. Chem. Phys., 2009, 130, 164705.
67 J. A. Leiro, M. H. Heinonen, T. Laihoa and I. G. Batirev,
J. Electron Spectrosc. Relat. Phenom., 2003, 128, 205.
68 M. Ramm, M. Ata, K.-W. Brzezinka, T. Gross and W. Unger,
Thin Solid Films, 1999, 354, 106.
69 P. W. Fowler and D. E. Manolopoulos, An Atlas of Fullerenes,
Clarendon, Oxford, UK, 1995.
Downloaded on 02 September 2010
Published on 20 August 2010 on http://pubs.rsc.org | doi:10.1039/C0CP00137F
View Online