N. Jiang et al. / Solid State Communications 149 (2009) 111–114
113
Fig. 3. Comparison of subtracted spectrum in the 7.5 nm Bi nanoparticle with
calculated volume plasmon loss using Drude approximation, reference spectra of
Bi and SiO2.
Fig. 4. Spectra acquired across a 35 nm bi nanoparticle. The subtracted spectrum
is compared with the calculated volume plasmon loss using Drude approximation.
to interpret the loss spectrum by the volume fractions of Bi particle
and SiO2 matrix, i.e. A + B = 1. In other words, the spectrum in
the particle region is the sum of the contributions from Bi and SiO2
respectively.
dominates along the path of the electron beam, and thus the peak
at 15.1 eV due to volume plasmon excitation can be separated from
other features without the need of any data processing. In the area
near the edge of the particle, on the other hand, only a shoulder
can be seen at around 15 eV in the spectrum, which resembles the
spectrum in the 7.5 nm Bi nanoparticle in Fig. 2(b). We normalize
spectra a and b to their integrated Si L23 edge intensities, and then
subtract spectrum a from b. As shown in Fig. 4, the position of
the maximum of subtracted spectrum is also at 15.1 eV, which is
consistent with the direct measurement from spectrum c.
This work confirms that the plasmon shift in Bi does occur
at very large sizes. However, the shift energy is very small, only
about 0.4 eV and 1.1 eV although R decreases from 90 nm to
35 nm and 7.5 nm, respectively. This indicates that the quantum
confinement effect on bulk plasmon in the Bi nanoparticle may not
be the reason if it obeys rule of R−2. The origin of the blueshift of
volume plasmon in Bi nanoparticles in EELS measurements can,
however, be complicated due to the poor q-resolution of EELS in
TEM/STEM. The integration effect suggested by [2] is definitely an
important source. The lattice distortion due to large surface area
in small particles can be also a determining factor, although the
accurate measurement of lattice distortion in such a small single
nanoparticle is not feasible.
To extract the volume plasmon loss of the Bi nanoparticle
from Fig. 2(b), we can take advantage of the thin specimen
approximation, in which the probabilities of plural scattering
events are negligible. Compared with the SiO2 region outside
the particle, the reduction of the Si L23 intensity in the particle
region is not great, and reaches only 1/3 around the particle center
(Fig. 2(a)). Since the particle size is quite small, its dimension along
the beam direction must also be small. In addition, the estimated
thickness of the SiO2 matrix adjacent to the particle is only around
55 nm. We then normalize all spectra to the intensity of the Si L23
edge, and subtract the reference SiO2 spectrum from those in the
particle region. The average of the subtracted spectra is given in
Fig. 3. As compared with the references of the Bi (90 nm) and SiO2
matrix, the subtracted spectrum is dominated by the contribution
of the Bi nanoparticle, justifying our kinematical assumption. We
note that there is a small shoulder around 10.5 eV in the subtracted
spectrum. This feature can be tentatively interpreted as the surface
plasmon of the Bi particle [17]. A small peak around 25.0 eV is also
seen in the big Bi particles, which is due to excitation of the Bi O45
In conclusion, we have carefully measured the energy at the
maximum energy loss indifferent sized Bi nanoparticles. These val-
ues contradict previously reported results for Bi nanoparticles [8].
It is more reasonable to interpret the blueshift as induced by the
plasmon dispersion effect [2], rather than the quantum confine-
In Fig. 3, the position of the maximum loss is at 15.8 eV in the
7.5 nm Bi nanoparticle, which is only 1.1 eV higher than that in
the 90 nm one. If the quantum confinement effect starts to occur
in the Bi nanoparticle at around 40 nm in diameter, this value is
much smaller than that predicted by the R−2 rule [3], and the result
clearly contradicts the observations in [8].
It should be pointed out that the kinematical approximation
used in this work does not guarantee an error-free determination
of volume plasmon energy in the embedded or partially embedded
nanoparticles, but the error should not be large as long as the
specimen is not too thick. In this case significant error may
be introduced in the integrated Si L23 intensity due to plural
scattering. To confirm the validity of our method, we also carried
out measurements in a 35 nm Bi nanoparticle. Fig. 4 shows three
representative spectra acquired from areas adjacent to the particle
(spectrum a), near the edge (spectrum b) and in the center region
(spectrum c) of the particle, respectively. In the central region, Bi
Acknowledgements
This work is funded by NSF DMR0603993. The use of
facilities within the center for solid state science at ASU is also
acknowledged.
References
[1] E.L. Wolf, Nanophysics and Nanotechnology, Wiley-VCH Verlag GmbH & Co.
KGaA, Weinheim, 2004.
[2] M. Acheche, C. Colliex, P. Trebbia, Scan. Electron Microsc. 1 (1986) 25.
[3] M. Mitome, H. Yamazaki, H. Takagi, T. Nakagiri, J. Appl. Phys. 72 (1992) 812.