Full text of DOI:10.1051/epjap:2001180

Eur. Phys. J. AP 15, 177–180 (2001)
THE EUROPEAN
PHYSICAL JOURNAL
APPLIED PHYSICS
ꢀ
c EDP Sciences 2001
STM induced luminescence in carbon nanotubes
1
,a
2
1
1
1
3
3
R. Coratger , J.-P. Salvetat , A. Carladous , F. Ajustron , J. Beauvillain , J.-M. Bonard , and L. Forr´o
1
CEMES/CNRS, 29 rue J. Marvig, BP 4347, 31055 Toulouse, France
CRMD/CNRS, 1b rue de la F ´e rollerie, 45071 Orl ´e ans Cedex 2, France
´
D ´e partement de Physique, Ecole Polytechnique F ´e d ´e rale de Lausanne, 1015 Lausanne, France
2
3
Received: 12 October 2000 / Revised: 15 February 2001 / Accepted: 17 May 2001
Abstract. Light emission excited by scanning tunnelling microscopy (STM) has been observed on multi-
wall carbon nanotubes. The photon yield depends on the scanned tube but it is constant all along the
tube. Simultaneous photon spectroscopy evidences emitted wavelengths in the 600–1000 nm range in good
agreement with previous field emission data. Emission characteristics are independent of substrate and
of STM tip. We discuss how these new results are consistent with radiative transitions involving localized
states in carbon nanotubes.
PACS. 61.46.+w Nanoscale materials: clusters, nanoparticles, nanotubes, and nanocrystals – 68.37.Ef
Scanning tunneling microscopy (including chemistry induced with STM) – 73.22.-f Electronic structure of
nanoscale materials: clusters, nanoparticles, nanotubes, and nanocrystals
1
Introduction
considered. We report here on an intrinsic STM excited
photon emission from a semimetallic mesoscopic solid that
is independent of substrate and STM tip. We discuss the
role of localized states at the cap of nanotubes and the
influence of the large electron mean free path on the non-
local excitation by the STM tip.
It has been shown recently that visible light emission
occurs simultaneously with electron field emission in
multiwall carbon nanotubes (MWNTs) [1,2]. This lumi-
nescence, which spectrum is characterized by a broad
Gaussian distribution located around 700 nm, is at-
tributed to radiative transitions between discrete levels
that are localized at the tip of the emitting nanotubes [2].
These localized states may be quasi-molecular levels lo-
cated on pentagons or other topological defects of the
curved graphene lattice [3,5]. Here we address the ques-
tion with an STM tip excited MWNTs, whether light is
emitted only when the electronic levels of the nanotube ex-
tremities are involved. A first result is that STM- induced
light is indeed detected on some MWNTs. We found, how-
ever, that the intensity and spectral repartition of the
emitted light does not depend on the position of the STM
tip. Moreover, some nanotubes did not emit light at all
2
Experimental
A customized pocket-size STM operating in the constant
current mode, adapted for photon detection and working
in ambient atmosphere has been used. Optical fibres 1 mm
in diameter were placed at about 1 mm of the tunnelling
junction (Ω ≈ π/4 srd per fibre) and conveyed the emit-
ted light to a photomultiplier (PM) or a spectrometer.
The PM (Hamamatsu R2949) was operated in the pho-
ton counting mode (dark count rate of 1 count per sec-
ond at 255 K). The complete wavelength range (185 to
9
00 nm) was used to acquire luminescence maps. The ac-
quisition time for a complete photon map is about 6 mn.
This photon map is derived from a non local measure-
ment of the emission yield for a given tip location and
can therefore be considered as an excitation measurement.
Spectroscopy was performed with an optical multichannel
analyser (EGG, OMA 1245) connected to a charge coupled
device detector cooled at 173 K and operating in the 300–
(
at least in the visible range and with a bias of about
2
V). The emission light characteristics are independent of
substrate and STM tip, in contrast with STM-induced
light emission from noble metal surfaces or nanopartic-
ules [6,8] and from individual C60 molecules deposited
on Au (110) surfaces [9]. In the former case, light emis-
sion has been attributed to the radiative decay of in-
terface plasmons localized in the tip-sample junction
and excited by inelastic tunnelling [10,11]. In the latter
case, the molecules interact strongly with the electromag-
netic modes of the tip-sample cavity and the whole sys-
tem comprising sample, molecule and cavity has to be
1
050 nm range. The typical detector sensitivity is 10 pho-
tons per count at 650 nm. Using this experimental set-up,
it was possible to measure simultaneously the STM topog-
raphy and the integral emitted light intensity. Emission
spectra were recorded with integration times of typically
5
min, while the STM tip was scanning small areas (about
a
2
e-mail: coratger@cemes.fr
10 × 10 nm ) at very low frequency (typically 0.25 Hz).

178
The European Physical Journal Applied Physics
(
a)
(b)
(a)
STM image
A
(b)
STM image
Photon image
Photon image
A’
C
Tips
A
B
2
3
00 nm
200 nm
35 c/s
3
1
00 nm
0 nm
300 nm
15 c/s
0.5 nm
t
Fig. 1. (a) STM topography (Vsample = 2.3 V, I = 10 nA)
Fig. 2. (a) STM topography (Vsample = 2.1 V, I
t
= 40 nA) and
and (b) photon mapping of a MWNT bundle. The photon map
shows that the emission yield depends on each nanotube but
does not depend on the STM tip position along the tube up to
its extremity (for this, two nanotube tips are indicated by ar-
rows on the STM topography). The maximum rate of emitted
photons is 35 counts per second.
(
b) photon mapping of a MWNT bundle. Only three nanotubes
show a significant light emission. The maximum rate of emitted
photons is 15 counts per second.
presents the same characteristics than Figure 1, i.e. pho-
ton intensity depends on the scanned tube. An interesting
observation is the absence of emission in the left part of the
photon map even if the nanotube number is almost iden-
tical. For example, two nanotubes noted A and A’ seem to
be connected to almost the same number of MWNT. How-
ever, only one of these tubes (A) has a significant emission
yield (one cannot exclude, however, that light is emitted
at a wavelength out of the PM spectral response). In ad-
dition, the emitted photon number is not related to the
position of the tube in the bundle and nanotubes almost
buried in the bundle may give large emission intensities.
STM tips were prepared by cutting 0.25 mm-diameter Pt-
Ir or Au wires. The first type of substrates we used con-
sisted of freshly cleaved highly oriented pyrolitic graphite
(
HOPG). The second type was gold films 80 nm in thick-
ness deposited on MoS2 surfaces. MWNTs were deposited
on the sample surface from an ethanol dispersion.
3
Results
Figure 1 shows a typical example of an STM topography This shows that the emission process is not related to the
(
left image) and corresponding photon mapping (right im- number of nanotubes connected to the scanned one but
age) obtained on a bundle of multiwalled nanotubes de- derives from this single tube. It should be noted that the
2
posited on HOPG. The scanned area was 200 × 200 nm . maximum emission rate is less than 100 counts per sec-
The photon image shows that the emission yield depends ond even for tunnelling currents as high as 50 nA and
on the scanned tube but not on the STM tip position bias voltages of 2.3 V.
along the tube. For example, some nanotubes, labeled A
Three different emission spectra are presented in
in the topography, yield the most important photon emis- Figure 3. The first spectrum, labeled A, has been obtained
sion while two others tubes, labeled B and C, present a on a single nanotube using a bias voltage of 2 V and a tun-
weaker emission. On the other hand, the emission yield is nelling current of 50 nA. This spectrum is characterized by
constant up to the nanotube tip (two of them are indicated a large peak centered at about 800 nm (1.55 eV). Most of
by arrows). Other images also show that tubes located the spectra have this characteristic shape but slight vari-
close to emitting MWNT or crossing some of them in a ations of the peak position have been observed from one
bundle may have emission rates below our detection limit nanotube to another (typically between 780 and 820 nm).
and give rise to dark bands in the photon mapping. Such The two other spectra, B and C, have been obtained
an example is shown in Figure 2. A bundle of MWNT has during another experiment using similar tunnelling con-
2
been scanned (area 300 × 300 nm ) at 2.1 V and with a ditions, i.e. 2.05 V and 35 nA. Spectrum B presents a
tunnelling current of 40 nA. The STM image shows well peak at a wavelength of 900 nm (1.38 eV) and a rather
resolved single tubes whose apparent diameter is in good broad tail towards high energy. The third spectrum, C,
agreement with the value of 20 nm deduced from previous has been obtained on another tube with the same Pt/Ir
transmission electron microscopy experiments. The reso- tip and presents the same peak. In addition, a second peak
lution of about 1 nm estimated from cross sections of high centered at a wavelength of 710 nm (1.75 eV) is revealed.
resolution STM images, is therefore sufficient to scan sin- Note that this low energy peak also appears in the spec-
gle nanotubes for photon spectroscopy. The photon map trum A as a broad shoulder. These experiments also show

R. Coratger et al.: STM induced luminescence in carbon nanotubes
179
1
1
1�
�
�
�
0
�
�
1�
9� �
4
0
1�
�
�
�
�
� �
�
�
�
�
�
�
1�
�
0
0
0
�
8
�
1,9
�
,�
�
,1
�
,�
�
,�
�
,�
Bias voltage (V)
�
�
�
�
�
�
� � � � � � � � � �
0
C
�
�
�
4
00
� 00
� 00
� 000
� � 00
1,8
1,�
� ,�
� ,1
� ,�
� ,�
� ,�
�
� � � � � � � � � � � � � �
Bias voltage (V)
Fig. 3. Emission spectra of three different MWNTs acquired
under similar experimental conditions. Dotted lines highlight Fig. 4. Peak intensity in terms of bias voltage for an A type
position of the background. Spectrum A is one of the most spectrum in the 1.9−2.4 V range. The inset shows the peak
currently observed and presents a peak at about 800 nm position as a function of bias voltage and the error bars give
(V
sample = 2 V, I = 50 nA). The spectra B and C have been the full width at half maximum of the Gaussian function used
obtained on two different nanotubes with the same Pt-Ir tip to fit the spectra.
during another experiment (Vsample = 2.05 V, I = 35 nA).
On spectrum B, a peak at 900 nm is evidenced while on spec-
trum C, another peak centred at about 710 nm is also revealed. emitting molecules. The authors suggested that the pres-
For the three spectra, the exposition time is 5 minutes. Back- ence of the molecules considerably changes the tip-surface
ground of spectra A and B has been shifted for clarity.
electromagnetic coupling which is responsible from the
light emission process on metal surfaces. In the present
experiments, MWNTs play the major role in the pho-
ton emission process since light emission is independent
of substrate and tip. Moreover, the light spectra is very
similar to that obtained in field- emission experiments (at
variance with experiments on noble-metal surfaces [6]).
It is worth discussing the different light emission mech-
anisms in relation with previous studies of metal and
semiconductor surfaces. STM-induced luminescence re-
sulting from band-to-band transitions has been observed
in semiconductors with an “optical” gap (i.e., a direct
that the peak intensity increases in terms of bias volt-
age (Fig. 4) while the peak position is independent of this
parameter (see inset of Fig. 4). It should be noted that
degradation of the nanotubes occurred when the bias was
increased beyond 2.5 V.
4
Discussion
Under identical experimental conditions, differences in the gap of more than 1 eV) [12]. Only nanotubes with a
emitted spectrum from one tube to the next are observed. very small diameter (and the right helicity) can have
Conversely, no systematic variations are detected when a a gap higher than 1 eV [13]. For shells of larger di-
different tip or substrate is used. For example, only a slight ameter the energy gap is too small (about 0.1 eV) to
increase in the photon yield is observed when gold tips are produce luminescence in the visible spectrum. Let us
used but the peak wavelengths are identical. The essen- examine if radiative plasmon recombination can be re-
tial spectrum characteristics are therefore independent of sponsible for light emission in MWNTs, as in noble
substrate and tip, in contrast with similar experiments metal surfaces. Surface excitations of graphite are in fact
carried out on noble metal surfaces or nanoparticles [6,8]. very different from metals or semiconductors. This was
Finally, no clear relation between emitted wavelength and demonstrated experimentally by electron energy-loss spec-
tube diameter has been observed. Maybe the most im- troscopy (EELS) [14]. EELS shows two different features:
portant result is that the photon yield is independent on a high intensity background coming from electron-hole ex-
∗
the tip position along the scanned tube. This point will citations between π/π bands, and low energy plasmon
be discussed later. It is also worth keeping in mind that resonances from the free electrons. The free electron (in-
the graphite substrate (which can be considered as an “in- traband) plasmon energy is lower than 0.1 eV due to
finite” graphite surface) does not emit light under STM the low density of states at the Fermi level and thus can
excitation.
not account for visible light emission. Only interband π-
All these results appear at first glance very similar to surface plasmons (peaking at 6 eV) may have an “optical”
those reported by Berndt et al. on C60 molecules [9]. In- energy in the low energy tail [15]. Intense electron-hole
deed, they also obtained photon mappings in which some excitations, which are forbidden in noble metals, are ex-
molecules present no emission although their appearance pected to decrease the plasmon lifetime in graphite and to
in the STM topography did not differ from the one of reduce the radiative recombination rate. This explains

180
The European Physical Journal Applied Physics
characteristics probably reflect the large variety in the tip
structure and could offer an original probe for the local-
ized tip states in combination with scanning tunnelling
spectroscopy.
5
Summary
Light emission excited by scanning tunnelling microscopy
STM) on multiwall carbon nanotubes has been observed.
(
The photon yield is constant all along the scanned tube,
with a fine structure in the intensity versus emitted wave-
length. We suggest that this light emission process is asso-
ciated to radiative transitions from localized states of the
carbon nanotube. These observations could give a better
understanding of STM induced light emission in differ-
ent nanostructures, and help to unravel the extraordinary
electronic properties of carbon nanotubes.
Fig. 5. Schematics of the model of STM induced light emission
via quasi-localized states of carbon nanotubes.
why no STM light emission is observed on graphite. The
same arguments hold for MWNTs, so that it is hardly
probable that light emission could come from radiative
plasmon recombination.
Having in mind these considerations, we suggest that
light emission comes from radiative transitions involving
localized states. These states may be intrinsic to MWNTs,
for instance localized on pentagons at the tip, or extrin-
sic, i.e., localized on vacancies, impurities or interstitial
The work in Lausanne is supported by the Swiss National Sci-
ence Foundation program NFP 36.
atoms. The fact that the light intensity remains homoge- References
neous along a particular emitting tube could suggest that
defects in the tube body are responsible for light emis-
sion. But the absence of emission from a large proportion
of MWNTs is then hardly explainable. In fact, although
it was not possible with our experimental set-up to de-
termine precisely the position and the density of localized
defects in MWNTs, the work of Carroll et al. [4] shows
clearly that the localized states are located at the tip in
arc-grown MWNTs. As demonstrated by theoretical stud-
ies and STS experiments [3,5], the energy positions of
these levels are compatible with optical transitions and
depend strongly on the specific tip structure.
1. A.G. Rinzler, J.H. Hafner, P. Nikolaev, L. Lou, S.G. Kim,
D. Tomanek, P. Nordlander, D.T. Colbert, R.E. Smalley,
Science 269, 1550 (1995).
. J.M. Bonard, T. St ¨o ckli, F. Maier, W.A. de Heer, A.
2
Chatelain, J.P. Salvetat, L. Forro, Phys. Rev. Lett. 81,
1441 (1998).
3. R. Tamura, M. Tsukada, Phys. Rev. B 52, 6015 (1995).
4. D.L. Carroll et al., Phys. Rev. Lett. 78, 2811 (1997).
5. P. Kim, T. Odom, J.-L. Huang, C.M. Lieber, Phys. Rev.
Lett. 82, 1225 (1999).
6
. J.K. Gimzewski, B. Reihl, J.H. Coombs, R.R. Schlitter,
Z. Phys. B 72, 497 (1988); R. Berndt, J.K. Gimzewski, P.
Johansson, Phys. Rev. Lett. 67, 3796 (1991).
The origin of the fact that light wavelength and inten-
sity are independent on the STM tip position might be
the following. We propose that the large mean free path
of the injected electrons combined with an efficient reso-
nant coupling of localized tip states to continuum states
are responsible for this behavior according to the follow-
7. V. Sivel, R. Coratger, F. Ajustron, J. Beauvillain, Phys.
Rev. B 45, 8634 (1992).
8. T. Imeno, R. Nishitani, A. Kasuya, Y. Nishina, Phys. Rev.
B 54, 13499 (1996).
9. R. Berndt, R. Gaisch, J.K. Gimzewski, B. Reihl, R.R.
Schittler, W.D. Schneider, M. Tschudy, Science 262, 1425
∗
ing mechanism: electrons are first injected in the π/π
(
1993).
0. P. Johansson, R. Monreal, P. Apell, Phys. Rev. B 42, 9210
1990); P. Johansson, Phys. Rev. B 58, 10823 (1998).
1. J. Aizpurua, S.P. Apell, R. Berndt, Phys. Rev. B 62, 2065
2000).
extended states of the outer shell, large number of them
flows towards the substrate, but a fraction of the electrons
spreads all over the tube in a quasi-ballistic way, like in
a large molecule. Some are trapped during a short time
1
1
(
(
in localized states at the tip which are resonantly cou- 12. D.L. Abraham, A. Veider, Ch. Sch ¨o nenberger, H.P. Meier,
pled with extended states, and occasionally decay in an-
other neighboring localized state situated approximately
D.J. Arent, S.F. Alvarado, Appl. Phys. Lett. 56, 1564
(1990); A. Carladous, R. Coratger, G. Seine, F. Ajustron,
J. Beauvillain, J. Appl. Phys. 84, 1085 (1998).
2
eV below by emitting a photon. The population differ-
1
1
1
1
3. J.W. Mintmire, D.H. Robertson, C.T. White, J. Phys.
Chem. Solids 54, 1835 (1993).
4. P. Laitenberger, R.E. Palmer, Phys. Rev. Lett. 76, 1952
(1996).
5. L.S. Caputi, G. Chiarello, A. Santaniello, E. Colavita, L.
ence between the two quasi-localized states comes from
the V-shape band structure of the graphene layer so that
states higher in energy are more populated than those
situated near the Fermi energy (Fig. 5). Indeed, recent
transport experiments show that the coherence length is
exceptionally large in carbon nanotubes even at room
temperature (it is already the case in graphite in com-
parison with metals) [16]. Differences in the emitted light
Papagno, Phys. Rev. B 34, 6080 (1986).
6. S. Frank, P. Poncharal, Z.L. Wang, W.A. de Heer, Science
280, 1744 (1998).