|
|
|
|
|
|
|
|
We have produced an endohedrally doped fullerene that shows a metal-like density of states at the Fermi level. Individual La@C60 clusters deposited onto graphite exhibit a zero band gap as observed by scanning tunneling spectroscopy on single clusters at room temperature. Moreover, we find that an isolated La@C60 cluster on graphite shows a reversible opening of a band gap at a transition temperature of —28 K. The transition is associated with a freezing of the vibrational motion of the La atom inside the fullerene cage. The metallic behavior of La@C60 is attributed to the presence of a dynamical dipole in the single cluster. © 2001 American Institute of Physics. [DOI: 10.1063/1.1406500] /
Endohedral fullerenes are of great interest due to their diversity and plans for numerous applications.1 Because of the robust carbon cage and its large hollow interior, endohe¬dral fullerenes represent a new class of technologically rel¬evant composites as they incorporate possible metallic and fullerenelike properties. One of the most intriguing aspects of endohedral fullerenes is the expected metallic or even su¬perconducting character.12 Charge transfer from the metal (M) into the unoccupied carbon cage orbitals, which is fos¬tered by the high electron affinity of the fullerenes,3 raises expectations forM@C60 and other endohedral fullerenes to exhibit metallic properties. In spite of tremendous efforts spent on the isolation of endohedral fullerenes from carbon soot, only a few endohedral fullerenes have been isolated in macroscopic amounts. The chromatographic extraction of endohedrally doped M@C60 seems particularly difficult, as only Eu@C60 has been isolated in pure form.4 However, it turns out that all these endohedral fullerites reveal a zero density of states at the Fermi level in contrast to the intuitive metal-to-carbon charge transfer picture. Metallic conductiv¬ity has only been observed for C60 when doped within a very narrow stoichiometry range and some of these metallic phases exhibit remarkable high superconducting transition temperatures.125"13 None of the other fullerenes are found to be metallic regardless whether dopants are introduced exter¬nally or encapsulated endohedrally into the cage. The failure to produce metallic endohedral fullerides might be attributed to the fact that charge transfer in endohedrally doped fullerenes is more complex than that in the exohedral alkali-metal-doped systems. This is most likely due to the hybrid¬ization of metal d orbitals with the carbon orbitals causing a donor-acceptor type of bonding. This hybridization has been proven experimentally for a thin film of La@C82 by resonant photoemission experiments.14
Endohedral M3 + @C3m clusters containing trivalent metal atoms are of special interest as these are isoelectronic to the high-rc superconducting fullerides A3C60 (A = alkali metal). As proposed by Weaver etal, crystals of pure La@C60 are expected to be metallic due to a charge transferbest described as 3 + .15 This has not been experimentally demonstrated so far, because La@C60 has not been isolated in pure form. However, by using a laser vaporization cluster source, a wide range of endohedral lanthanide fullerenes from 30 to about 150 carbon atoms, including endohedrally doped M@C60, are available for deposition from a mass-selected metal-fullerene cluster beam.1617 Combining this deposition technique with scanning tunneling spectroscopy (STS), we have probed the local density of electronic states near the Fermi level of single La@C60 and Ce@C60 fullerenes on highly oriented pyrolytic graphite (HOPG). As STS gives information on both the occupied and unoccupied orbitals the band gap of the individual clusters on HOPG is immediately revealed by these measurements.
The clusters were synthesized by pulsed laser vaporiza¬tion of a graphite-metal rod (100:1) and condensation in He atmosphere.1617 After supersonic expansion, the cluster cat¬ions were mass-selected by a magnetic sector field (m/Am «=* 250) and deposited onto a sample area of ~ 1X1 mm2. In order to avoid damage upon surface impact, the clusters were deposited softly at an ion kinetic energy of <200 meV/atom. Following deposition, the cluster-covered samples were transferred into a variable-temperature scanning tunneling microscope (STM) under ultra high vacuum conditions.18 The STS spectra have been taken with a chemically etched tungsten tip on top of individual clusters.
Figure 1(A) shows a STM overview of the mass-selected endohedral clusters on HOPG. A mean coverage density of — 1/100 nm~2 can be inferred from the STM figure, which corresponds to —1% of a monolayer. Thus the deposited clusters are separated several times their diameter as clearly seen for La@C60 in Fig. 1(B). The intactness of the deposited fullerene cages has been verified by the height of La@C60 and Ce@C60, which is very similar to pristine C60 as dem¬onstrated in Fig. 1(C). Both endohedral clusters appear slightly higher than C60, which is attributed to a different tunneling conductivity owing to the different electronic structures. At room temperature, La@C60 shows an enhanced mobility on the HOPG surface with respect to Ce@C60.
Figures 2 and 3 show the normalized STS curves of single C60, La@C60, and Ce@C60 clusters on HOPG at room temperature, respectively. The most prominent difference be¬tween La@C60 and Ce@C60 is the density of states in the immediate vicinity of the Fermi level. While Ce@C60 shows a gap of —0.3 eV, which is distinctly smaller than that of deposited C60 on HOPG (Fig. 2) and reconstructed Si(100),19 no gap is obvious for La@C60. As can be seen in Fig. 4 the tunneling current of La@C60 at room temperature increases linearly with the bias voltage for a large energy region around the Fermi level. Therefore the density of states at the Fermi level of La@C60 is different from zero. The zero band gap identifies La@C60 to be metal-like. In contrast, Ce@C60 has a semiconductorlike electronic structure as revealed by the sharp drop of the normalized STS signal distinctly below and above the Fermi level. The small peak at -0.1 V in Ce@C60 is attributed to poor statistics emphasized by nor¬malization near zero bias. Up to now, all endohedral com¬pounds, e.g., La@C82, have been found to be semiconduct¬ing when deposited as solid films. La@C60 is the first endohedral fullerene found to have a metal-like density of states. Moreover, the moleculelike peak structure of the STS curve demonstrates that the metal-like character is an in¬tramolecular property of the La@C60 cluster itself.
The metal-like density of states in La@C60 can be inter¬preted in terms of charge donation from the La atom into C60-derived òã orbitals. Figure 2 shows a comparison of STS curves of C60 (top) and La@C60 (bottom) on HOPG. The C60-derived lowest unoccupied molecular orbital (LUMO) is expected to be split in La@C60 by symmetry reduction ow¬ing to both a noncentrosymmetrical position of the La atom2021 and charge donation. Upon charge transfer from the trivalent La atom into the empty LUMO of C60, the ^ „-derived orbital in La@C60 is only partially filled. The Fermi level is thus located within this Jahn-Teller split or¬bital. Consequently, the peaks immediately below (J3) and above (/?') the Fermi energy can be assigned to the ?! „-derived orbitals while the Fermi energy in C60 is located between the hu and tlu orbitals. Peak ó at +0.9 V is inter¬preted to be derived from the tlg orbital (C60 LUMO+1), which is broadened by Jahn-Teller distortion. The same holds true for the C60-derived highest occupied molecular orbital (HOMO) (/?„) around -2 V (a) and the hg/t2u-denved peak at +1.9 V ($). Though the degeneracy of the C60-derived orbitals is lifted upon doping, the overall valence-electron structure of La@C60 is principally similar to that of pristine C60. However, due to the charge transfer from the La dopant into the empty "conduction" band of C60, the LUMO-derived orbital of La@C60 is only partly filled, a nec¬essary condition for a metal-like level density.
The electronic level density at the Fermi level of La@C60 changes upon cooling as demonstrated in Fig. 4. While the electronic structure is not altered very much from room temperature down to —29 K, a distinct gap is opened around the Fermi level at ~28 Ê as deduced from the /(V) curve. A gap of —40 mV is revealed by fitting a flat base line through the data around the Fermi level in Fig. 4. In contrast, the I(V) curves at room temperature and —29 Ê exhibit a nearly constant slope throughout the entire vicinity of the Fermi level. Thus, the I(V) curve of La@C60 reveals a metal-like characteristic above -29 K, while a semiconduc-torlike behavior is obvious below —28 K. The "metal-to-semiconductor" transition in La@C60 is reversible at the transition temperature of —28 K. We note that the gap of 40 mV is too small to be caused by a Coulomb blockade, which amounts to more than 1 eV for a sphere of 10 A diameter. Since no Coulomb blockade is observed at any temperature we conclude that the endohedral fullerene is not locally charged, making a sufficient orbital overlap and Ohmic con¬tact to the HOPG substrate evident. Moreover, no effect re¬lated to the HOPG substrate is known to occur in this tem¬perature region.
We attribute the opening of the gap to the freezing of the dopant's vibrational motion within the fullerene cage. The period of circular motion at room temperature has been cal¬culated by Andreoni and Curioni to be 1.1 ps.20>21 This La motion is expected to freeze at a temperature that corre¬sponds to an energy smaller than the potential barrier. Our measured freezing temperature of —28 Ê corresponds to an energy of —2.4 meV, which defines the minimum energy for movement. This potential barrier agrees quite well with the calculations of Andreoni and Curioni,2021 who predict vibra¬tional frequencies involving La to be as small as 30 cm"1 (3.7 meV). The motion of La inside the C60 cage is associ¬ated with a dynamical dipole moment of La@C60. In contrast to the dynamical Jahn-Teller distortion above —29 Ê where the dipole averages to zero on a picosecond time scale, the dipole is locked in the frozen state that gives rise to a static Jahn-Teller distortion. Obviously, the La dynamics intro-duces electronic states at the Fermi level that are well sepa¬rated by 40 meV in the static state. This agrees with the calculated splitting of the C60 tlu-derived levels for the ground-state geometry of La@C60.22 Both the freezing of the low-frequency modes involving the La atom and the result¬ing gap opening is consistent with the nonmetallic electronic structure of La@C82 for which a fixed position of La has been predicted at room temperature.2021
The question why Ce@C60 behaves like a semiconductor at room temperature in contrast to La@C60 can be attributed to a formal 4+ charge state due to the extra 4/electron in Ce. In this case the Fermi level is expected to be located some¬where between the split /^„-derived C60 orbitals (/?,/?',/?" in Fig. 3), and the existence of the gap follows from a closed shell structure of the tx„-derived levels. As the Jahn-Teller distortion of the tlu band is much less than the splitting of the hu and tlu frontier orbitals of C60, the reduced HOMO-LUMO gap in Ce@C60 is consistent with this interpretation. Considering an on-site two-electron Coulomb repulsion similar to the Hubbard U, a gap might occur even in the case of a noninteger charge transfer. Three electrons occupying the ?i„-derived levels would then give rise to four peaks in the spectrum. Peaks /?' and /?" below and above the Fermi level, respectively, represent the same single-particle orbital, split by the Coulomb interaction U. The other tl„-derived levels are represented by f3 and /?'"', the latter indicated as a shoulder of peak /?". The Jahn-Teller distortion then amounts to -0.2 eV (/?-/?') and -0.1 eV (/?"-/?'"). Note that for a detailed understanding of the charge transfer theexact hybridization between the Ce 5d/6s and fullerene or-bitals has to be taken into account. Nevertheless, the pres¬ence of the additional 4f electron will definitely have an ef¬fect on the hybridization. Moreover, Ce is located in a more eccentric off-center position which is —10% larger than that in La@C60.20'21'23 This and the larger charge transfer could lead to a permanent dipole moment in Ce@C60 in contrast to the dynamical dipole moment of La@C60.
In conclusion, endohedral doping can lend metallic prop¬erties to a single C60 cluster by intramolecular orbital hybrid¬ization, whereas intermolecular band overlap will cause a line of several adjacent C60 clusters to form a conducting wire.24 The deposition from a mass-selected cluster beam opens the possibility for a systematic exploration of diverse physical properties such as metallicity, superconductivity, and ferromagnetism on a vast number of endohedral fullerenes doped with rare-earth-metal atoms. If a method of large-scale synthetic production can be found, this could open the pathway for the design of technologically novel endohedral fullerene-based materials such as molecular met¬als of extremely low weight with magnetic or superconduct¬ing properties or as novel catalyst materials.
We are grateful to G. Seifert for many valuable discussions. Technical support by H. Pfeifer and J. Lauer is grate¬fully acknowledged. This work has been supported by the Sonderforschungsbereich 341 of the Deutsche Forschungsg-meinschaft.
Carbone Technologies &Materials © 2008
+7 (812) 363 2051
+7 (812) 332 1078
info@fulleren.com