primary electron energy and were equal to 25.0 eV for. C6o and 24.8 eV for C70. The conclusion on the space localization for plasma occilations in fullerenes ...
Z. Phys. B 93, 327-331 (1994)
ZEITSCHRIFT FUR PHYSIKB 9 Springer-Verlag 1994
Reflection electron energy-loss spectra of the fullerenes C6o and C7o Yu.M. Shul'ga, V.I. Rubtsov, A.S. Lobach Institute of Chemical Physics, 142432 Chernogolovka,Moscow Region, Russia Received: 28 April 1993/Revised version: 24 June 1993
Abstract. High purity polycrystalline samples of C60 and C7o were obtained and studied by the electron energy-loss spectroscopy in the reflection mode. The spectra were used for determination of the loss functions of fullerenes. Loss functions of the fullerenes were compared with those of graphite. It was established that the relative intensities of the peaks corresponding to (o-+n)- and 7cplasmons depended on the primary electron energy, while the (a + rr)-plasmon energies did not depend on the primary electron energy and were equal to 25.0 eV for C6o and 24.8 eV for C70. The conclusion on the space localization for plasma occilations in fullerenes was made on the base of the study of the energy dependent loss functions.
cussed in literature [2, 9, 10] but the reasons causing it are still obscure. The goals of the present study have been registration of the electron energy-loss spectra for solid individual C60 and C70 fullerenes in reflection geometry at the different energies of the primary electron beam (250-2000 eV), subtraction of the background brought about by the multiple losses from the experimental spectra, isolation of the loss-function I r a ( - l / e ) for fullerenes and comparison of the data obtained with those for graphite. Besides we have made an effort to consider possible reasons for the disagreement between the experimental and calculated values of hoop (o" + a:) for fullerenes.
PACS: 79.20; 36.40+d
Experimental
Introduction Electron energy-loss spectroscopy (EELS) is widely used in the study of the fullerene electronic structure. The analysis of the EELS data shows such effects as multiple inelastic scattering and losses within the atomic sphere with the hole in the C 1s level created under photoionization ("intrinsic" losses) to be very important, since they affect dramatically the shape of the measured spectra, which differ from Im ( - l/e) (e is a dielectric function). Hence the values of the a + n - p l a s m o n energy (hoop (o-+ ~)) obtained in the experiment at the primary electrons (Ep = 50 - 200 keV) transmission via thin solid films of C60 [1-3] differ from those obtained in the study of the fine structure near the Cls photoelectron peak [4-6] or in the reflection spectra of the low-energy (75-200 eV) electrons [7, 8]. Furthermore the values of hOOp(o- + rr) obtained in a transmission mode exceed considerably those calculated using the carbon atoms density in solid C60 within the model of free electrons. This disagreement is often dis-
Buckminsterfullerene C6o (purity is better than 99.9% according to data obtained by electron spectroscopy and high effective liquid chromatography) and fullerence C70 (99.7%) as black polycrystal powders were prepared by the method analogous to a previously described one [ 11 ]. The soot-like material produced in electric-arc evaporation of graphite rods (4-6 mm in diameter) in a helium atmosphere (100 Tort) was then subjected to an extraction process involving toluene, in a Soxhlet apparatus. The extract containing mainly a mixture of C6o and C70, was divided into individual components by multiple chromatography on neutral alumina (activity I) columns (4 • 60 cm) using a mixture of hexane and toluene as an eluent. Bright shining fine-crystal powders prodused in crystallization from the magenta C6o and red C7o solutions, and further vacuum drying at 170 ~ were used in the present work. Test samples were prepared by fullerene deposition on an aluminum substrate which was preliminary purified and oxidized in air to form a 20-40A thick oxide layer. Such an oxide layer was expected to prevent a possible chemical fullerenes-substrate interaction. The thickness of the fullerene covering in the analysis zone was 0.2-0.7 gin.
328 The electron spectra were measured on a spectrometer PHI-551 equipped with a double-pass cylindrical mirror analyzer with an attachment for angle measurements and an electron gun coaxial to the analyzer. The analyzer was used in the retarding mode at the pass-energy equal to 25 eV and the absolute resolution of 0.7 eV. The primary electron energy was within 250-2000 eV and a F W H M (full width at half maximum) of the energy distribution was 0.5 eV. The sample positions were chosen so that the incidence angle for the primary electrons (#) was equal to the registration angle for the reflection electrons @). At # = y , two sample positions were used: f l = 7 0 ~ (a "normal" angle) and fl = 10 ~ (a "sliding" angle relative to the sample surface). The angle resolution was 8 ~. The base pressure in the spectrometer test chamber did not exceed 3 • 10-10 Torr. The spectra measured were corrected on the dependence of the transmission function of the analyzer on the kinetic energy. The loss functions y (E) were determined from the integral equation [12]:
N in(E) -- ky (E) 9 N in (E) = ky (E),
(1)
where 9 denotes convolution; N ~n(E) is an inelastic part of the measured EEL spectrum, normalized on the square of the elastic peak; k is the factor taking into account the experiment geometry. The y (E) function is proportional to the loss function averaged over the scattering angles [121 y (E) ~,ln (1 + 02,~/02) Im ( - 1/e (E)),
(2)
where 0m~• is the maximal angle of scattering on the valence electrons, 0 e = E/(2Ep). In what follows y (E) will be referred to as the loss-function. Figure 1 illustrates the operating sequence for extraction of the loss-function
C6o EELS
a 1
z
c6o
cr+~
from the experimental spectrum. One of the advantages of the approach proposed is the innecessity of special assumptions on the differential cross-section for the plasmon and other channels of inelastic scattering.
Results and discussion
Figure 2 shows the loss functions for the fullerenes and graphite. All curves have two main peaks whose origin may be attributed to the plasma oscillation excitation of all valence electrons (a + n-plasmon) and re-electrons (rc-plasmon) by analogy with graphite. The peak positions for loss-functions are given in Table 1. The positions of a + rc-plasmon peaks for C6o and C7o are close to each other and have no systematical shifts at going from fl = 70 ~ to fl = 10 ~ On the contrary, the shape of the graphite spectrum has a strong angle dependence. It should be also noted that the subtraction of the background of the multiple electron scattering causes the decrease in the measured value of hcop(O-q-Tr) by about 1 - 2 eV. The frequency of plasma oscillations (cop) of all valence electrons in solid differs from that within the free electron model (cop = (4 7tneZ/m) 1/2) when we take into account the interaction with the electrons of the core levels. The equation for plasmon frequency corrected on the interaction is [13] co; = (4 ~neZ/mK .... )1/2
(3)
The numerical value of ~r.... may be calculated according to [13]
(~Coo,.e-1)/(~oore+ 2)=(4 ~/3) ( ~ Nic~i) ,
(4)
where N i is the density of ions with the polarizability cq. In this case, the ions denote atoms without valence electrons taking part in plasmon excitation. According to Pauling [14], the polarizability for C 4+ is equal to 1.3 • 10 - 2 7 cm 3. Hence,/~?ls = 1.0018 for o- + rcplasmon in graphite. It is obvious that for other carbon materials the K ls values are very close to a unity. The case of plasma oscillations of x-electrons is different due to significant the polarizability of the o--electrons, which are inner in respect to these oscillations. Supposing that plasma oscillations of all valence electrons
b
Table 1. The peak positions (in eV) of the loss-functions at # = 70~ and at # = 10~ (in brackets)
.-
