Bulk electronic structure of K3C60 as revealed by soft x-rays

Abstract

We present C 1s x-ray absorption, x-ray emission, and resonant inelastic x-ray scattering (RIXS) spectra of single-phase crystalline K3C60. The comparison to valence-band photoelectron spectra from the same sample facilitates identification of the contribution from surface and bulk electronic states in the latter. Bulk-sensitive techniques show that the valence bands of K3C60 and pure C60 are characterized by spectral features of similar width, in agreement with the predictions of band-structure calculations. Symmetry selectivity in the RIXS process allows us to assign peaks in the C 1s absorption spectrum, demonstrating a close correspondence with pure C60 also in the conduction band. The symmetry selectivity is as pronounced in K3C60 as in pure C60, indicating that the local C60 symmetry is not appreciably affected by the K doping, either in the ground state or intermediate state, on the time scale of 6 fs.

Full text

Page 1

Bulk electronic structure of K3C60as revealed by soft x-rays
Tanel Käämbre,1,* Joachim Schiessling,1,†Lisbeth Kjeldgaard,1Limin Qian,1Ingrid Marenne,2James N. O’Shea,1,‡
Joachim Schnadt,1,§Dennis Nordlund,1Chris J. Glover,3,?Jan-Erik Rubensson,1Petra Rudolf,2,¶Nils Mårtensson,1,3
Joseph Nordgren,1and Paul A. Brühwiler1,4,**
1Department of Physics, Uppsala University, P.O. Box 530, S-75121 Uppsala, Sweden
2LISE, Facultés Universitaires Notre Dame de la Paix, Rue de Bruxelles 61, B-5000 Namur, Belgium
3MAX-lab, Lund University, P.O. Box 118, S-22100 Lund, Sweden
4Empa, Swiss Federal Laboratories for Materials Testing and Research, Lerchenfeldstrasse 5, CH-9014 St. Gallen, Switzerland
?Received 23 December 2006; revised manuscript received 3 March 2007; published 22 May 2007?
We present C 1s x-ray absorption, x-ray emission, and resonant inelastic x-ray scattering ?RIXS? spectra of
single-phase crystalline K3C60. The comparison to valence-band photoelectron spectra from the same sample
facilitates identification of the contribution from surface and bulk electronic states in the latter. Bulk-sensitive
techniques show that the valence bands of K3C60and pure C60are characterized by spectral features of similar
width, in agreement with the predictions of band-structure calculations. Symmetry selectivity in the RIXS
process allows us to assign peaks in the C 1s absorption spectrum, demonstrating a close correspondence with
pure C60also in the conduction band. The symmetry selectivity is as pronounced in K3C60as in pure C60,
indicating that the local C60symmetry is not appreciably affected by the K doping, either in the ground state
or intermediate state, on the time scale of 6 fs.
DOI: 10.1103/PhysRevB.75.195432PACS number?s?: 78.40.Ri, 73.20.At, 78.70.En, 79.60.?i
I. INTRODUCTION
The fulleride salts of composition A3C60?where A is an
alkali metal? as well as other fullerides have attracted wide
interest due to their transport properties, not least because
they represent a class of high-temperature superconductors.1
In a one-electron picture with rigid band filling, the AxC60
compounds should be metallic for 0?x?6, which corre-
sponds to partial filling of the threefold degenerate t1ulowest
unoccupied molecular orbital ?LUMO? of C60. However, of
the observed stable phases with partial filling ?x=1, 2, 3, and
4?, only a few stoichiometries show metallic properties. In
particular, metallic A3C60has been investigated with the first
sample K3C60as the primary testing ground.1
Band-structure calculations2of K3C60predict the energy
bands to be narrow, similar to bulk ?undoped? crystalline C60.
However, numerous photoemission spectroscopy ?PES? data
show broad structures for the K3C60stoichiometry.3–13A
number of theoretical studies have been carried out in order
to elucidate the role of effects such as electron-electron cor-
relation or electron-phonon coupling and the Jahn-Teller
effect1,14,15on the electronic structure.
The most recent work concludes that both of these effects
have to be included to describe the K3C60 electronic
structure.16–18Notably, transport-related and optical mea-
surements have supplied the necessary parameters for these
studies, whereas PES has remained an enigma to the field.
PES is normally a natural choice to probe the electronic
structure. The earliest PES studies, however, suffered from
uncertainties in sample stoichiometry and the presence of
multiple phases. Samples were typically produced by evapo-
rating potassium on a film of C60, attempting to enhance the
uniformity of intercalation by heating the sample.3,19The
development of a reliable sample preparation method20has
led to a convergence of the experimental PES results.8,10–12
The interpretations of the spectra, however, diverge. Since
the probing depth of PES is limited to the first few molecular
layers of the sample,13,19,21the important question of whether
the bulk and surface have identical stoichiometry and elec-
tronic structure has been strongly debated.
We have previously presented a different set of PES mea-
surements which are consistently explained by a model of
the surface layer as a correlated two-dimensional insulator
and the bulk as metallic, resolving many contradictory
claims in the literature and offering an explanation of why
transport measurements obtained a higher density of states at
EF, for example.12,13There have been no comparable mea-
surements that are truly bulk sensitive to test the hypothesis
of a surface effect. A similar issue has emerged in the inter-
pretation of the x-ray absorption spectroscopy ?XAS? data,
which was attributed to a bulklike surface layer with contri-
butions from symmetry-split core levels.9
The techniques of soft x-ray emission spectroscopy
?XES?, which measures the same final states as PES, and
resonant inelastic x-ray scattering ?RIXS?, which can be used
to monitor electronic symmetry, offer the means to accom-
plish truly bulk-sensitive measurements; their typical prob-
ing depth is several thousand angstroms, enabling one to
address the issues above from a new perspective. Here we
present XES and RIXS spectra of K3C60in the vicinity of the
C K edge. We find that the spectral features are almost as
narrow as in the spectra of pure C60, in agreement with band-
structure calculations of bulk K3C60and our recent analysis
of PES data.12,13We also use the symmetry information in
RIXS ?Ref. 22? to assign the peaks in the XAS spectrum, and
we find that the orbital symmetries are virtually uninfluenced
by K intercalation, so that the spectrum for the compound
has a direct correspondence to that of solid C60. The fact that
the excitation energy dependence of RIXS is similarly pro-
nounced in both C60and K3C60indicates that the local sym-
metry is preserved to a similar extent for the RIXS process of
K3C60as for solid C60. Overall, the data are consistent with a
largely molecular description of K3C60.
PHYSICAL REVIEW B 75, 195432 ?2007?
1098-0121/2007/75?19?/195432?7?
©2007 The American Physical Society 195432-1

Page 2

II. EXPERIMENT
The experiments were carried out at the surface branch of
beamline I511 at MAX-lab, Lund.23The beamline provides
linearly polarized light from an undulator source, monochro-
matized using a Zeiss SX700 plane grating monochromator
?Au coated, line density of 1221 lines/mm?. The RIXS spec-
tra were recorded using the grazing incidence grating spec-
trometer with a 1200 lines/mm spherical grating and a
10 ?m entrance slit. With these parameters, the resolution of
the spectrometer was 0.15 eV. The resolution of the mono-
chromator was set to 0.1 eV. XAS spectra were acquired
using a high pass yield detector, set to accept C 1s Auger
electrons. The recorded electron yield was normalized to the
incoming photon flux using a clean gold mesh placed after
the optical elements. The energy resolution in the XAS spec-
tra was better than 20 meV. The base pressure was
2?10−10mbar.
Single-crystal K3C60was obtained by sublimation of al-
ternating C60and potassium layers onto a clean Cu?111? sur-
face. In every cycle, the stoichiometry was always kept be-
low x=3 as checked by PES. The deposition rate was
calibrated via the time to obtain a monolayer prior to sample
preparation. After ten deposition cycles, a sample approxi-
mately 300 Å thick was obtained. Excess C60was then sub-
limed under controlled temperature conditions similar to the
vacuum distillation approach.20The PES data of the K3C60
sample12,13were in good agreement with those of earlier
reports.9–11A ?1?1? ?111? surface pattern was confirmed by
low-energy electron diffraction, consistent with previous
work.10
III. RESULTS AND DISCUSSION
A. C 1s excitations
The electronic transitions studied in the present work are
illustrated in Fig. 1. The occupied valence levels are repre-
sented by the highest occupied molecular orbital ?HOMO?
and the level below ?HOMO-1?, and the unoccupied valence
levels by the LUMO. XES and RIXS are illustrated in a
two-step picture, which often is appliciable to XES, but ig-
nores the coherent effects often important in RIXS.24In core
level and valence level PES, electrons are ejected and the
molecule is left in an ionized state. The first step for both
techniques consists of the absorption of a photon. The inter-
mediate state in XES resembles a core hole PES final state,
whereas in RIXS, the system can be said to have a core
exciton. The excited intermediate state decays with the emis-
sion of an x-ray photon. The final state in XES is identical to
the final state in valence PES, whereas that in RIXS is a
valence exciton.
To better understand the RIXS technique, we want to con-
sider the excitation step in more detail. The C 1s XAS spec-
tra of pristine C60and K3C60are shown Fig. 2. For now, it is
useful to note that the energies chosen as excitation steps for
RIXS in Fig. 3 are indicated with arrows and labeled by
capital and small letters for C60and K3C60, respectively. Re-
call that the spectrum of solid C60is quite similar to that of
molecular C60, with an intensity distribution related to the
empty density of states, but strongly modified by core hole
effects.25–27The absorption spectrum of K3C60shows similar
resonances that are, however, somewhat broader than the
corresponding excitation in C60, and the absorption edge is
slightly shifted to lower energies. It is generally accepted that
the reduced intensity in the LUMO-derived resonance of
K3C60is approximately proportional to the number of states
FIG. 1. A schematic of the techniques employed. In each case,
the final state of the indicated process is shown, along with the
associated excitation and/or deexcitation transitions and associated
particles ?photon and/or electron? for both photon-in–photon-out
techniques. Core-level and valence-level PES provide a starting
point, creating hole states. For XES, core photoionization first cre-
ates the same state as in PES as shown, followed by a transfer of the
hole to the valence states and simultaneous emission of a photon
that is detected, which in a single-particle picture creates the same
final state as valence PES. In RIXS, which is analogous to XES, the
intermediate state corresponds to an XAS excitation, which then
decays similarly to XES, but with an extra electron in the initially
unoccupied valence levels. RIXS leaves the molecule in a valence-
excited state, which leads to different emitted photon energies than
in XES, as discussed in detail in the text. The two steps are usually
measurably coherent in RIXS, which further differentiates it from
XES, as discussed in the text.
FIG. 2. C 1s XAS of C60and K3C60measured via Auger elec-
tron yield, with the first three resonances of solid C60labeled ac-
cording to the unoccupied molecular orbitals from which they are
derived. The arrows indicate the excitation energies for the spectra
in Fig. 3.
KÄÄMBRE et al.
PHYSICAL REVIEW B 75, 195432 ?2007?
195432-2

Page 3

filled in the ground state.3,28This tendency was, however,
implicitly challenged recently,11and the resolution of these
conflicting views is taken up in Sec. III C.
B. Occupied states
The photon-in–photon-out spectra obtained for K3C60are
shown in Fig. 3. The spectra show certain similarities at all
excitation energies, such as the HOMO-derived peak be-
tween 282 and 283 eV, and several other peaks at lower
emission energies which can be associated with molecular
orbitals of the free molecule. A detailed discussion of such
peak assignments was previously given for solid and/or mo-
lecular C60.22For K3C60, a different feature is the half-filled
LUMO-derived band, apparent at an emission energy of
about 284 eV. This band is overlaid by a strong elastic peak
as the excitation energy decreases to the lowest values con-
sidered here, just above 284 eV. Before considering the gen-
eral variations as a function of excitation energy, we want to
compare the XES and RIXS spectra to PES spectra of the
same samples.
Figure 4 provides such a comparison for C60and K3C60.
To plot the data on a common energy scale, we must first
consider the energetics of the different processes in a com-
mon framework. For the sake of simplicity, we focus on the
analysis of the HOMO and start with solid C60. As discussed
elsewhere29,30in some detail, the energetics for PES are rela-
tively straightforward: ?1? The core hole energy is given by
the C 1s ionization potential ?IP? of 289.6 eV.29?2? The
HOMO hole energy is given by the HOMO IP of 6.9 eV.29
?3? This yields an XES energy for the HOMO of ?1?–?2?
=282.7 eV, since the XES energy is to first order merely the
difference between these states, in excellent agreement with
our measurements. The energetics for RIXS are obtained by
a similar series of considerations: ?4? The C 1s→LUMO
resonance has an energy of just under 284.5 eV.25?5? The
HOMO→LUMO exciton can be considered to have an ion-
ization potential of4.8 eV.30
→LUMO transition itself has, for the present purposes, an
energy given by ?2?–?5?=2.1 eV. ?7? This places the
HOMO-derived final state in RIXS at ?4?–?6?=282.4 eV,
i.e., the RIXS data for this transition should lie about 0.3 eV
lower in photon energy than the same feature in XES, which
is consistent with the experimental results ?present work and
Ref. 22?.
For K3C60a similar analysis proceeds as follows: ?1? Be-
cause ionization potentials have not been measured for
K3C60, we take EFas the reference level, for which the C 1s
binding energy is11,13285.0 eV. ?2? The HOMO binding en-
ergy is11,132.3 eV. ?3? This yields an XES energy for the
HOMO of ?1?–?2?=282.7 eV, the same value as for solid
C60, which is several tenths of an eV higher than the value
seen, e.g., for transition “g” in Fig. 3. ?4? The C 1s
→LUMO resonance has an energy of approximately
284.3 eV, as seen in Fig. 2. ?5? The separation of the HOMO
from the unoccupied LUMO is equal to its binding energy in
the present case, giving 2.3 eV. ?6? This places the HOMO-
derived final state in RIXS at ?4?–?5?=282.1 eV, which is
too low compared to the experimental result of 0.2–0.3 eV.
?6?
ThustheHOMO
FIG. 3. Photon-in–photon-out spectra for K3C60recorded at the
indicated excitations of Fig. 2 ?RIXS? and at 306 eV ?g?, which are
taken to correspond to off-resonant excitation, i.e., XES. Peak e in
the spectra is due to elastic ?Rayleigh? scattering at the excitation
energy.
FIG. 4. Valence band PES, XES, and RIXS data of the indicated
samples, measured with the excitation energies 110 eV ?PES?,
306 eV ?XES?, and as indicated by arrows a and A in Fig. 2 for
RIXS. The positions of bulk contributions to the PES spectrum in
?b? are marked by dashed lines.
BULK ELECTRONIC STRUCTURE OF K3C60AS…
PHYSICAL REVIEW B 75, 195432 ?2007?
195432-3

Page 4

These simple calculations, which ignore screening effects in
the intermediate state, Jahn-Teller contributions, and general
vibrational broadening, give nevertheless a semiquantitative
description of the energetics, suggesting that such effects are
limited to a scale of several tenths of an eV. These consider-
ations, furthermore, show that placing the XES and RIXS
spectra on the PES binding energy scale is not possible on
the basis of energetics for K3C60, as it clearly is for pure C60.
Thus the comparison in Fig. 4?b? is based on positioning
the LUMO-derived part of the XES and RIXS spectra near
zero binding energy as carefully as possible. What is obvious
for both samples is that the peaks derived from the frontier
valence bands can be clearly identified. For K3C60, most of
these features are preserved in the XES and RIXS spectra,
whereas in PES, they are greatly broadened and cannot be
easily distinguished.
The spectral features of the two samples have similar
shapes and widths in RIXS and XES. In contrast, the PES
data of K3C60 show structures which are substantially
broader. Both techniques differ significantly in their probing
depth. The present observations suggest that PES of K3C60
displays contributions not limited to the bulk electronic
structure, i.e., that the surface electronic structure has to be
different and make a strong contribution to the PES data. The
observations noted above do not allow us to be conclusive on
this point, because satellites appearing due to K doping may
appear with different intensities in PES and RIXS spectra.
However, the present observations of simpler XES and RIXS
spectra, more in line with the theoretical expectations of nar-
row bands, support our analysis elsewhere12,13that the sur-
face electronic structure is different. To summarize, we find
here that the valence bands of bulk C60and K3C60are simi-
lar, in agreement with band-structure calculations.
Another difference between the two compounds in the
photon-based spectroscopies is an intensity enhancement of
the LUMO-derived band. Starting with pure C60, in the sim-
plest picture the XES spectra reflect the occupied states, and
hence we do not expect any LUMO intensity. The observed
LUMO intensity can be explained by the fact that the emis-
sion does not exclusively come from pure core hole states,
but also from multiply excited states populated via shake-up
processes as also observed in Auger.31–33We therefore as-
sume that recombination of an electron which has been
shaken up to the LUMO in the excitation process gives rise
to the observed “LUMO intensity” in the XES spectrum of
C60. This implies that the electron in the LUMO is localized
at the core hole site during the core hole lifetime, consistent
with work aimed directly at this question.34–36This is in
sharp contrast to the LUMO+1 state which, in spite of a
shake-up probability comparable to that of the LUMO,37
does not give rise to any measurable recombination intensity
in the XES spectrum. While we cannot explain this, we sug-
gest that there are at least two factors to be considered in
explaining this observation: ?1? Using XAS as a guide, the
LUMO-derived state has a cross section which is higher by
approximately at least a factor of 2, and likely much higher,
given the effects of vibronic broadening and the predicted
electronic cross sections25; ?2? the proclivity of electrons ex-
cited to the LUMO+1 to hop to a nearest-neighbor molecule
before the core transition34,36could be important in this re-
spect.
The K3C60features are significantly broader than in C60,
both in the XES and RIXS spectra. In XES, the broadening is
largely due to a superposition of emission from inequivalent
states due to shakeup and different spin configurations38
populated in the excitation process. Note, however, that the
width of the LUMO peak in K3C60is less than 1 eV. The
simplest interpretation is that this represents the upper limit
to the energy separation of all initial states contributing to the
XES spectrum, and hence, indicates that the binding energies
of the C 1s levels corresponding to nonequivalent carbon
atoms differ by less than 1 eV. One could speculate that
different core levels, as proposed by Goldoni et al.,9lead to
excitation to LUMO levels which are qualitatively different,
e.g., differently hybridized with neighbor molecular levels;
this more complicated scenario is, however, not supported by
the angle independence of the XAS ?Ref. 11? and RIXS re-
sults presented here. Thus, we suggest that the narrow XES
LUMO peak is inconsistent with the interpretation of level
splitting9,11in bulk K3C60, implicitly supporting an alterna-
tive interpretation.12,13
On the other hand, the broadening of the XES spectrum is
sufficiently large to smear out any obvious Fermi cutoff,
which is clearly visible in PES. In analogy to the broadening
of the LUMO-derived peak in pure C60, the broadening of
the LUMO-derived peak in K3C60can be attributed to emis-
sion from a continuum of intermediate core-excited states
populated in shake-up processes during the excitation. This
implies that the screening of the core hole is incomplete dur-
ing the lifetime39of the intermediate states, although the sys-
tem is metallic. This behavior is in contrast to what is ob-
served for simple metals, where the screening is fast and the
Fermi cutoff is readily observable in XES.40,41
In RIXS the spectral features are substantially narrower
than in XES. The energy selectivity of the excitation reduces
the distribution of the intermediate states,36and we see that
the RIXS spectrum of C60is in almost perfect agreement
with the PES results. This is in spite of the fact that the RIXS
final states have an additional electron in the LUMO orbital.
The close correspondence indicates that this electron does
not significantly influence the valence electron system, as
previously discussed for resonant Auger.36An important
point for RIXS and XES is that the molecular charge state
does not change in the emission step, further minimizing the
effects of an additional electron on the emitting molecule
with respect to electron spectra.
The K3C60RIXS spectra show an excess broadening as
compared to those of C60. Since the LUMO-derived orbital
in K3C60has an occupancy of 3, there is an unpaired electron
in the ground state. At resonant excitation, the spin of the
excited electron may be parallel or antiparallel to this elec-
tron. Therefore, two exchange-split states for each final state
configuration are expected to be populated. The exchange
splitting for LUMO excitation is estimated to be 0.3 eV,17
which cannot be resolved, but which would account for the
observed additional broadening.
C. Unoccupied bulk electronic states
The XAS of K3C60is typically measured using electron
yield, which implicitly involves bulk and surface contribu-
KÄÄMBRE et al.
PHYSICAL REVIEW B 75, 195432 ?2007?
195432-4

Page 5

tions. The surface contribution will be reflected in the ratio
of surface to bulk intensity. For a significantly shifted surface
signal, the weight of different parts of each peak in the spec-
trum will be modified. This issue should be considered when
trying to understand why Goldoni et al.11observed only
small variations in the XAS spectrum as a function of elec-
tron detection energy, i.e., of the relevant mean free path,
showing that the surface and bulk contributions do not differ
enough to have a strong impact on the spectrum.
A general shift of the K3C60spectrum to higher energy by
approximately 0.35 eV would make most of the spectral fea-
tures coincide with those of the C60spectrum. The largest
differences between the two compounds are found close to
the edge, in the region of the LUMO and LUMO+1 reso-
nances. The relative depletion of the LUMO intensity is eas-
ily understood by considering that the LUMO is half-filled in
the ground state of K3C60. The region just above the
LUMO?-derived? absorption is also strongly modified. These
modifications have provoked a proposal that the peak be-
tween the C60LUMO and LUMO+1 absorption energies is
not a LUMO+1-derived level, as commonly assumed, but of
different origin. For instance, split states, which are dis-
cussed as the reason for the broad photoemission lineshape
of the LUMO-derived band,9would be expected to appear as
additional peaks in the XAS.As shown in Fig. 2, the LUMO-
derived resonance shifts more strongly than the other peaks,
which indeed raises important questions about the changes in
the spectrum upon K intercalation.
We show here that this issue can be settled by studying
how the RIXS spectra change as the excitation energy is
varied across the different absorption resonances. Figure 5
shows that changing the excitation energy from the LUMO
to the second absorption resonance has a large impact on the
RIXS spectrum. For pure C60, this variation has been ana-
lyzed in detail elsewhere.22In the dipole approximation, the
RIXS process is governed by the selection rules for second-
order optical processes. For systems with inversion symme-
try, parity must be conserved; i.e., if a gerade ?ungerade?
orbital is excited in the absorption step, only radiative decay
from gerade ?ungerade? orbitals is allowed. In systems with
large vibronic coupling ?like C60?, and systems where the
symmetry is slightly disturbed ?like K3C60?, the selection
rules are relaxed to propensity rules, which nevertheless de-
termine important characteristics of the excitation energy de-
pendence of the RIXS spectra.22
From this discussion and previous work, it can be under-
stood that scattering at the ungerade LUMO-derived ?t1u?
intermediate state in C60emphasizes the E1and E5features
of the emission band, exclusively, corresponding to ungerade
valenceorbitals. Consequently,
attenuated when the excitation energy is tuned to the gerade
LUMO+1 resonance. Raising the excitation energy from the
LUMO-derived band in K3C60to the absorption structure “b”
produces changes in the emission band, which in great detail
correspond to the changes in the C60emission when the en-
ergy is raised from the LUMO to the LUMO+1 resonance.
Hence, it is straightforward to assign gerade parity to the “b”
resonance, and we conclude that the state labeled as “b” in
the XAS is indeed LUMO+1 derived.
To explain the larger shift of the LUMO+1-derived fea-
ture, we postulate a hybridization of the K valence orbital
these features are
with the LUMO+1, which shows a nonrigid shift in the
spectra. This state consistently shows the strongest effects of
such interactions compared to the neighboring frontier
states.34,42–45
The separation of surface and bulk contributions in the
electron yield spectrum cannot be solved quantitatively here.
The LUMO intensity is, as pointed out above, expected to be
about 1/2 that of pure C60for the bulk, but significantly
higher for the surface molecules.12,13To explain the excess
broadening of the spectral features compared to pristine C60,
we suggest that one should also consider the exchange split-
ting, expected in the absorption spectrum due to the unpaired
electron in the ground state. Important to note, however, is
that the energy of the LUMO resonance could be quite simi-
lar for each of the three charge states suggested for mol-
ecules near the K3C60surface, since the excitation is neutral
and local to the probed molecule.
There is still an open question regarding the expected in-
tensity of possible satellites in XES compared to PES for
such systems. Since recent work on bulk K3C60?Refs. 12 and
13? suggests that such satellites are weak, as shown directly
for surface compounds,46employment of Occam’s Razor
would lead one to reject an important role for satellites in the
present data. Nevertheless, calculations of XAS, XES, and
RIXS of such materials would be useful to fully explore the
implications of the present data. The parity selectivity of
RIXS in K3C60is observed even when higher intermediate
states are involved, as clearly observable in Fig. 3: the exci-
tation dependence of the relative intensities of the emission
spectral peaks closely follows the behavior in pristine C60
?see, e.g., Ref. 22?, and we can therefore conclude that the
local symmetry is well preserved upon the K intercalation at
FIG. 5. Resonant inelastic scattering via the two lowest-energy
absorption resonances. Excitation energies as marked in Fig. 2 by a
and b ?K3C60, lower panel?, and A and B ?C60, upper panel?. ?The e
in each spectrum marks the elastic-scattering peak.?
BULK ELECTRONIC STRUCTURE OF K3C60AS…
PHYSICAL REVIEW B 75, 195432 ?2007?
195432-5

Page 6

this composition. This conclusion is valid both regarding
static symmetry breaking in the ground state and any addi-
tional dynamic symmetry breaking in the RIXS process. Any
departure from C60symmetry would readily appear as a re-
laxation of the parity selection rules associated with the full
symmetry.
IV. SUMMARY
In conclusion, XAS, XES, and RIXS consistently show
strong analogies between single-phase K3C60and pristine
C60. The lines in the photon spectra are narrower than the
corresponding valence-band PES lines from K3C60. As XES
and RIXS are sensitive to the bulk of the material, and PES
to the surface layer, this points to differences in the bulk and
surface electronic structure of K3C60. This strongly supports
the earlier proposition that the surface layer has a stoichiom-
etry and associated electronic structure which differ from the
bulk.12,13
Considering the parity selection rules in the RIXS pro-
cess, we also find that the C60symmetry largely is preserved
in K3C60, both in the ground state and during core excita-
tions. This allows an unambiguous assignment of the absorp-
tion peaks to unoccupied energy levels with molecular C60
character, contradicting the notion of on-ball chemical shifts.
ACKNOWLEDGMENTS
Thiswork wasperformed withinthe EU-TMR
“FULPROP” network, Contract No. ERBFMRX-CT97-
0155, with additional funding from the EU Access to Large
Scale Facilities Program. We would also like to acknowledge
financial support by the Consortium on Clusters and Ul-
trafine Particles and the Caramel Consortium, which in turn
were supported by Stiftelsen för Strategisk Forskning, as
well as Vetenskapsrådet and Göran Gustafsons Stiftelse.
*Present address: Institute of Physics, Tartu University, Riia 142,
EE-51014 Tartu, Estonia. Electronic address: tanel@fi.tartu.ee
†Present address: MAX-lab, Lund University, P.O. Box 118,
S-22100Lund,Sweden.
joachim.schiessling@maxlab.lu.se
‡Present address: School of Physics and Astronomy, University of
Nottingham, Nottingham, NG7 2RD, United Kingdom.
§Present address: Department of Synchrotron Radiation Research,
Lund University, P.O. Box 118, S-22100 Lund, Sweden.
?Present address: Department of Electronic Materials Engineering,
Research School of Physical Sciences and Engineering, Australian
National University Canberra, Australia.
¶Present address: Materials Science Centre, University of Gronin-
gen, Nijenborgh 4, NL-9747 AG Groningen, The Netherlands.
**Electronic address: paul.bruehwiler@empa.ch
1O. Gunnarsson, Rev. Mod. Phys. 69, 575 ?1997?.
2S. Satpathy, V. P. Antropov, O. K. Andersen, O. Jepsen, O. Gun-
narsson, and A. I. Liechtenstein, Phys. Rev. B 46, 1773 ?1992?.
3C. T. Chen, L. H. Tjeng, P. Rudolf, G. Meigs, J. E. Rowe, J. Chen,
J. P. McCauley, A. B. Smith, A. R. McGhie, W. J. Romanow,
and E. W. Plummer, Nature ?London? 352, 603 ?1991?.
4G. K. Wertheim, J. E. Rowe, D. N. E. Buchanan, E. E. Chaban, A.
F. Hebard, A. R. Kortan, A. V. Makhija, and R. C. Haddon,
Science 252, 1419 ?1991?.
5P. J. Benning, J. L. Martins, J. H. Weaver, L. P. F. Chibante, and
R. E. Smalley, Science 252, 1417 ?1991?.
6M. Knupfer, M. Merkel, M. S. Golden, J. Fink, O. Gunnarsson,
and V. P. Antropov, Phys. Rev. B 47, 13944 ?1993?.
7P. Rudolf, M. S. Golden, and P. A. Brühwiler, J. Electron Spec-
trosc. Relat. Phenom. 100, 409 ?1999?.
8R. Hesper, L. H. Tjeng, A. Heeres, and G. A. Sawatzky, Phys.
Rev. B 62, 16046 ?2000?.
9A. Goldoni, L. Sangaletti, F. Parmigiani, S. L. Friedmann, Z.-X.
Shen, M. Peloi, G. Comelli, and G. Paolucci, Phys. Rev. B 59,
16071 ?1999?.
10A. Goldoni, S. L. Friedmann, Z.-X. Shen, and F. Parmigiani,
Phys. Rev. B 58, 11023 ?1998?.
Electronicaddress:
11A. Goldoni, L. Sangaletti, S. L. Friedmann, Z.-X. Shen, M. Peloi,
F. Parmigiani, G. Comelli, and G. Paolucci, J. Chem. Phys. 113,
8266 ?2000?.
12J. Schiessling, L. Kjeldgaard, T. Käämbre, I. Marenne, L. Qian, J.
N. O’Shea, J. Schnadt, M. G. Garnier, D. Nordlund, M. Nagas-
ono, C. J. Glover, P. Rudolf, J.-E. Rubensson, N. Mårtensson, J.
Nordgren, and P. A. Brühwiler, Eur. Phys. J. B 41, 435 ?2004?.
13J. Schiessling, L. Kjeldgaard, T. Käämbre, I. Marenne, J. N.
O’Shea, J. Schnadt, C. J. Glover, M. Nagasono, D. Nordlund,
M. G. Garnier, L. Qian, J.-E. Rubensson, P. Rudolf, N. Mårtens-
son, J. Nordgren, and P. A. Brühwiler, Phys. Rev. B 71, 165420
?2005?.
14O. Gunnarsson, H. Handschuh, P. S. Bechthold, B. Kessler, G.
Ganteför, and W. Eberhardt, Phys. Rev. Lett. 74, 1875 ?1995?.
15O. Gunnarsson, S. C. Erwin, E. Koch, and R. M. Martin, Phys.
Rev. B 57, 2159 ?1998?.
16M. Fabrizio and E. Tosatti, Phys. Rev. B 55, 13465 ?1998?.
17J. E. Han, E. Koch, and O. Gunnarsson, Phys. Rev. Lett. 84, 1276
?2000?.
18M. Knupfer and J. Fink, Phys. Rev. Lett. 79, 2714 ?1997?.
19G. K. Wertheim and D. N. E. Buchanan, Phys. Rev. B 47, 12912
?1993?.
20D. M. Poirier, Appl. Phys. Lett. 64, 1356 ?1994?.
21A. Goldoni, L. Sangaletti, F. Parmigiani, G. Comelli, and G. Pa-
olucci, Phys. Rev. Lett. 87, 076401 ?2001?.
22Y. Luo, H. Ågren, F. K. Gel’mukhanov, J. Guo, P. Skytt, N. Wass-
dahl, and J. Nordgren, Phys. Rev. B 52, 14479 ?1995?.
23R. Denecke, P. Väterlein, M. Bässler, N. Wassdahl, S. M. Butorin,
A. Nilsson, J.-E. Rubensson, J. Nordgren, N. Mårtensson, and R.
Nyholm, J. Electron Spectrosc. Relat. Phenom. 103, 971 ?1999?.
24J.-E. Rubensson, J. Electron Spectrosc. Relat. Phenom. 110-111,
135 ?2000?.
25B. Wästberg, S. Lunell, C. Enkvist, P. A. Brühwiler, A. J. Max-
well, and N. Mårtensson, Phys. Rev. B 50, 13031 ?1994?.
26M. Nyberg, Y. Luo, L. Triguero, L. G. M. Pettersson, and H.
Ågren, Phys. Rev. B 60, 7956 ?1999?.
27O. Wessely, O. Eriksson, and M. I. Katsnelson, Phys. Rev. B 73,
KÄÄMBRE et al.
PHYSICAL REVIEW B 75, 195432 ?2007?
195432-6

Page 7

075402 ?2006?.
28T. Pichler, M. Knupfer, M. S. Golden, J. Fink, and T. Cabioch,
Phys. Rev. B 63, 155415 ?2001?.
29A. J. Maxwell, P. A. Brühwiler, D. Arvanitis, J. Hasselström, and
N. Mårtensson, Chem. Phys. Lett. 260, 71 ?1996?.
30J. Schnadt, J. Schiessling, and P. A. Brühwiler, Chem. Phys. 312,
39 ?2005?.
31R. W. Lof, M. A. van Veenendaal, B. Koopmans, H. T. Jonkman,
and G. A. Sawatzky, Phys. Rev. Lett. 68, 3924 ?1992?.
32P. A. Brühwiler, A. J. Maxwell, A. Nilsson, N. Mårtensson, and
O. Gunnarsson, Phys. Rev. B 48, 18296 ?1993?.
33R. W. Lof, M. A. van Veenendaal, H. T. Jonkman, and G. A.
Sawatzky, J. Electron Spectrosc. Relat. Phenom. 72, 83 ?1995?.
34P. A. Brühwiler, A. J. Maxwell, P. Rudolf, C. D. Gutleben, B.
Wästberg, and N. Mårtensson, Phys. Rev. Lett. 71, 3721 ?1993?.
35A. J. Maxwell, P. A. Brühwiler, D. Arvanitis, J. Hasselström, and
N. Mårtensson, Phys. Rev. Lett. 79, 1567 ?1997?.
36P. A. Brühwiler, O. Karis, and N. Mårtensson, Rev. Mod. Phys.
74, 703 ?2002?.
37C. Enkvist, S. Lunell, B. Sjögren, S. Svensson, P. A. Brühwiler,
A. Nilsson, A. J. Maxwell, and N. Mårtensson, Phys. Rev. B 48,
14629 ?1993?.
38W. L. O’Brien, J. Jia, Q.-Y. Dong, T. A. Callcott, K. E. Miyano,
D. L. Ederer, D. R. Mueller, and C.-C. Kao, Phys. Rev. Lett. 70,
238 ?1993?.
39The lifetime of the intermediate state can be estimated from the
widths of the C 1s XPS line shape. For C60, this is in the order
of 6 fs.
40H. Jones, N. F. Mott, and H. W. B. Skinner, Phys. Rev. 45, 379
?1934?.
41T. A. Callcott, E. T. Arakawa, and D. L. Ederer, Phys. Rev. B 18,
6622 ?1978?.
42A. J. Maxwell, P. A. Brühwiler, A. Nilsson, N. Mårtensson, and P.
Rudolf, Phys. Rev. B 49, 10717 ?1994?.
43A. J. Maxwell, P. A. Brühwiler, D. Arvanitis, J. Hasselström, M.
K.-J. Johansson, and N. Mårtensson, Phys. Rev. B 57, 7312
?1998?.
44P. A. Brühwiler, P. Baltzer, S. Andersson, D. Arvanitis, and N.
Mårtensson, in AIP Conference Proceedings, edited by H. Kuz-
many, J. Fink, M. Mehring, and S. Roth ?AIP, Melville, New
York, 2001?, Vol. 591.
45X. Liu, T. Pichler, M. Knupfer, M. S. Golden, J. Fink, H. Kataura,
Y. Achiba, K. Hirahara, and S. Iijima, Phys. Rev. B 65, 045419
?2002?.
46W. L. Yang, V. Brouet, X. J. Zhou, H. J. Choi, S. G. Louie, M. L.
Cohen, S. A. Kellar, P. V. Bogdanov, A. Lanzara, A. Goldoni, F.
Parmigiani, Z. Hussain, and Z.-X. Shen, Science 300, 303
?2003?.
BULK ELECTRONIC STRUCTURE OF K3C60AS…
PHYSICAL REVIEW B 75, 195432 ?2007?
195432-7