The temperature dependence of Fe-catalysed growth of carbon nanotubes on silicon substrates

Article abstract of DOI:10.1016/S0921-4526(02)00965-1

The catalytic particle size (Fe) and temperature dependence of multi-walled nanotubes growth from C₂H₂ and C₆₀ precursor molecules is studied. The structure and density of the carbon nanotubes produced is critically dependent on the particle size, the growth temperature and the carbon flux rate. Under certain conditions, bundles of single-walled nanotubes, where the bundles appear to consist of nanotubes with the same diameters, can be produced. The nanotubes are characterised by electron microscopy and Raman spectroscopy. Field emission properties of aligned films are studied and the electron emission is correlated with light emission measurements.

Full text of DOI:10.1016/S0921-4526(02)00965-1

Physica B 323 (2002) 51–59
The temperature dependence of Fe-catalysed growth of carbon
nanotubes on silicon substrates
1
O.A. Nerushev , R.-E. Morjan, D.I. Ostrovskii, M. Sveningsson, M. J o. nsson,
F. Rohmund, E.E.B. Campbell*
Department of Experimental Physics, School of Physics and Engineering Physics, G o. teborg University and Chalmers University
of Technology, SE-41296 G o. teborg, Sweden
Abstract
2 2
The catalytic particle size (Fe) and temperature dependence of multi-walled nanotubes growth from C H and C60
precursor molecules is studied. The structure and density of the carbon nanotubes produced is critically dependent on
the particle size, the growth temperature and the carbon flux rate. Under certain conditions, bundles of single-walled
nanotubes, where the bundles appear to consist of nanotubes with the same diameters, can be produced. The nanotubes
are characterised by electron microscopy and Raman spectroscopy. Field emission properties of aligned films are
studied and the electron emission is correlated with light emission measurements. r 2002 Elsevier Science B.V. All
rights reserved.
PACS: 81.07.De; 61.46.+w; 79.70.+q
Keywords: Carbon nanotubes; Fe-catalysed growth; MWNT; SWNT; Raman spectroscopy; Field emission
1
. Introduction
been illustrated in a few cases [1]. In spite of the
widespread use of CVD there are relatively few
systematic studies of the dependence of growth on
important parameters such as the size of the
catalytic particles, the temperature of growth, etc.
[2–6]. Those studies that do exist in the literature
are sometimes apparently contradictory although
it is rarely the case that growth conditions can be
directly compared since small variations in para-
meters can lead to surprisingly different results. In
our group we have been interested in the Fe-
catalysed growth of both single- [7] and multi-
walled [7–10] nanotubes (SWNT and MWNT,
respectively). We have recently reported the
strikingly different structural forms of nanotubes
produced from C H and C gaseous precursors
The growth of carbon nanotubes using chemical
vapour deposition (CVD) is of considerable
interest because of the flexibility of the method
and the possibility of up-scalingproduction to
amounts that could be interestingfor industrial
applications. The method also has considerable
potential for the control of the direct growth of
carbon nanotubes on patterned substrates as has
*Correspondingauthor. Tel.: +46-31-772-3272; fax: +46-
31-772-3496.
E-mail address: eleanor.campbell@fy.chalmers.se
E.E.B. Campbell).
(
1
Permanent address: Institute of Thermophysics, 1 Acad.
Lavrentyev Ave., Novosibirsk 630090, Russia.
2
2
60
0921-4526/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 1 - 4 5 2 6 ( 0 2 ) 0 0 9 6 5 - 1

52
O.A. Nerushev et al. / Physica B 323 (2002) 51–59
[
11] and explained this in terms of a growth model
3. Catalytic particle size dependence and the growth
model
that emphasises the role of the carbon feedstock
supply rate [12]. This model can also explain a
number of seemingly contradictory studies in the
literature concerningthe dependence of gr owth on
catalytic particle size [13,14]. In this paper we
report some new experimental data supportingthis
growth model. We also report the results of
systematic experiments to study the dependence
of nanotube structure on growth temperature for
tubes grown on thin Fe films deposited on Si. Very
significant changes in the structure and density of
the carbon nanotube films are observed as the
growth temperature is changed. The field emission
properties of films of aligned MWNT grown in
this way are discussed and preliminary results of
the correlation between electron emission and light
emission are presented.
We have shown recently that, under the above
conditions, nanotube growth is only observed if
catalyst particles with diameters of a few tens of
nanometres are abundant in the particle films. The
average particle size could be adjusted by control-
lingthe position of the substrate inside the CVD
furnace duringthe in situ deposition process or by
controllingthe thickness of the electron-sputtered
Fe film in the range of 0.5–20 nm before annealing
[11]. The presence of catalytic particles that are too
large inhibits nanotube growth and produces
mainly non-nanotube deposits [11]. The maximum
size of catalytic particles for nanotube growth was
shown to depend on the growth conditions. In
contrast to other reports [14,15] where CVD
nanotube growth was observed from C H for
2
2
particles up to some hundreds of nm in size, we do
not observe nanotube formation from C H for Fe
2
. Experimental method
2
2
film thickness’ greater than ca. 10 nm (average
particle diameters of 200 nm). Beyond this, we
produce mainly thick carbon fibres with diameters
in the range of 500 nm. The main difference in the
experiments was in the C H pressure used. This
The production apparatus has been described
previously [8]. Synthesis is carried out by thermal
CVD in a horizontal furnace. For the work
reported here, iron films were either deposited in
2
2
situ from Fe(CO) at 3001C or films of a controlled
was much higher in the experiments from other
laboratories [14,15] compared to our values. Very
low C H pressure MBE-type experiments also
5
thickness (between 0.5 and 20 nm) were deposited
on clean Si substrates by electron beam sputtering.
The films were heated in the furnace in flowingAr
2
2
showed growth below critical catalyst particle
sizes, similar to our results [13,16]. For the C60
precursor, evaporated at 6001C (correspondingto
a vapour pressure of 10 mTorr [17]), no nanotubes
at all could be observed on particles with sizes
larger than the nanotube diameter (mean value
22 nm) and no nanotube or carbon fibre deposits
were obtained for Fe-film thickness’ greater than
10 nm [11,12]. We also observed a striking
difference in the structure of tubes produced by
C H and C . This is illustrated in Fig. 1 where
(
600 sccm) and H (100 sccm) to 7501C, where they
2
were held for 10 min. Duringthis annealingstep,
nano-sized iron particles were formed on the
substrate. Subsequently, nanotube growth was
carried out at the atmospheric pressure by adding
the feedstock gas to the gas flow and adjusting the
temperature to the desired growth temperature.
C H (8 sccm) from the gas tank was passed
2
2
through a glass tube at dry ice temperature to
remove acetone. C60 was evaporated from a
stainless steel boat held by a thermocouple rod
close to the hot zone of the furnace [11]. The iron
particle size distributions were determined by
AFM imaging of annealed iron films. The
nanotube films were imaged by scanning and
transmission electron microscopy (SEM/TEM)
and investigated by micro-Raman spectroscopy
2
2
60
TEM images of the tubes produced on a pre-
deposited Fe-film of 1 nm thickness (producingan
average catalytic particle size of 20–30 nm [12]) at
a growth temperature of 7501C (for 30 min) are
shown. Both sets of nanotubes are clean, with no
amorphous carbon deposits; however, the acety-
lene tubes are much thinner and straighter than
the C60 ones. The nanotube diameters do not
(
514 nm excitation).

O.A. Nerushev et al. / Physica B 323 (2002) 51–59
53
2 2
Fig. 1. TEM images of MWNT obtained by CVD from C H (left) and C60 (right). A i nm thick Fe film was deposited on Si by e-beam
sputtering, followed by CVD at 7501C for 30 min as described in the text.
change significantly with increasing average cata-
lytic particle size. Acetylene produces nanotubes
within a narrow diameter range about 14 nm
whereas the C60 tubes have average diameters of
2
on C H -produced films when the average catalyst
2 nm. A second kind of nanotube deposit is found
2
2
particle size reaches a value of about 90 nm (5 nm
Fe film thickness). These tubes show a narrow
diameter distribution about 33 nm and occur
simultaneously with the narrow 14 nm tubes. The
distribution is clearly bimodal [12]. The situation is
summarised in Fig. 2. These observations have
been explained in terms of a simple growth model
that emphasises the role of the carbon feedstock
supply rate [12]. In our experiments the carbon
supply rate from the C60 precursor was approxi-
mately a factor of 500 lower than that in the C H
Fig. 2. Iron particle diameter and heights (thick lines repre-
sentingspline fits to the data points) and nanotube diameters as
a function of evaporated film thickness. (’, &): nanotube
2
2
diameters for C
0 nm film thicknesses a bimodal size distribution is observed.
ꢀ): diameters of nanotubes grown from C60 with a low carbon
flux (evaporation at 6001C). (%): diameters of nanotubes
grown from C60 with a carbon flux rate comparable to that of
2 2
H grown material. For (’) 5 nm and (&)
case. The model assumes that the carbon feedstock
molecules are adsorbed on the catalyst particle and
are decomposed into carbon atoms that diffuse
through the metal particle. The diffusion process is
most probably driven by a carbon concentration
1
(
2 2
the C H experiments.

54
O.A. Nerushev et al. / Physica B 323 (2002) 51–59
gradient and does not involve the formation of
iron carbide [18], in agreement with TEM studies
of the catalytic Fe particles in our experiments.
Precipitation of solid carbon occurs on the
opposite side of the particle when the carbon
concentration exceeds the temperature-dependent
saturation concentration. The carbon feedstock
supply rate determines the decomposition rate,
which, under stationary conditions, is identical to
the precipitation or nanotube growth rate. We
assume that diffusion occurs through a cylindrical
segment of the catalytic particle with diameter a
and, therefore, the diffusion rate will scale as 1=a
(C H ), which is larger than the maximum
2 2
precipitation rate, no oscillations will occur but
stationary growth with constant nanotube dia-
meter is expected. In order to test the model, we
are carryingout additional experiments where the
carbon supply rate to the metal particles is
adjusted. It is extremely difficult to reduce the
acetylene flow rate to match the C60 carbon supply
at 6001C; however, we can increase the tempera-
ture of the C60 source to match the acetylene
supply rate in the above experiments. First results
are plotted on Fig. 2 for a Fe film thickness of
1 nm. In agreement with the model, we see thin
nanotubes from C60 with the same diameter as
those from acetylene produced under the same
conditions. The nanotubes are also straighter than
those produced at the low supply rate. We can also
observe some nanotubes with higher diameters,
comparable to the higher diameter distribution
seen from acetylene for larger Fe film thicknesses.
This may indicate some differences in the decom-
position kinetics. Further work is in progress.
[
19]. For a low carbon feedstock supply rate (the
situation with the C60 experiments) carbon will
diffuse through a cylindrical segment with a larger
diameter a compared to the acetylene case where
the supply rate is considerably higher. Carbon
precipitation occurs at the end of the cylindrical
segment and forms a nanotube with diameter a: If
the catalyst particle is increased in size, the heat
release due to precursor decomposition will lead to
a lower temperature rise and thus a lower
saturation concentration of carbon and the circle
of precipitation will move closer to the region of
decomposition. Under our conditions this leads to
the formation of thicker nanotubes for acetylene
and, for C , it leads to a cessation of growth since
4. Temperature dependence offilm growth
from C H
2
2
We have studied the temperature dependence of
the growth of nanotubes from acetylene (8 sccm)
on 1 nm thick films prepared by electron sputtering
and then annealed at 7501C, as discussed above.
The growth temperature did not significantly
influence the size of the iron nanoparticles after
the annealingstep i.e. the avera ge diameter of the
catalyst particles remained at 20–30 nm. The gas
flow rates were kept constant and only the
temperature of the oven was adjusted. The growth
time in each case was 30 min. Fig. 3 shows SEM
pictures of the films produced as a function of
temperature. An analysis of the SEM data is
summarised in Fig. 4. TEM pictures of typical
structures produced at 5501C, 7501C and 11001C
are shown in Fig. 5. A combination of thin
MWNT (diameters around 14 nm) and thick
non-hollow carbon fibres is seen for the lowest
temperature investigated (in qualitative agreement
with our model). As the temperature is increased
the density and length of the nanotubes increases
6
0
the precipitation circle moves into the hemisphere
where decomposition occurs. The oscillatory
structure observed in our C60 experiments
(
Fig. 1) can also be explained by the model. If we
assume startingconditions where the gr owth circle
lies on the equator of the particle, the precipitation
rate is low and can become smaller than the
carbon supply rate. The carbon concentration in
the metal particle, therefore, increases with the
consequence that the diffusion rate increases and
the precipitation circle moves away from the
equator leadingto a decrease in the nanotube
diameter. A decrease in diameter implies an
increase in precipitation rate that can deplete the
carbon concentration in the particle if the pre-
cipitation rate then exceeds the decomposition
rate. This in turn leads to the precipitation circle
movingback towards the equator with a corre-
spondingincrease in nanotube diameter, and so
on. In the case of a high carbon supply rate

O.A. Nerushev et al. / Physica B 323 (2002) 51–59
55
produced at 7501C have a matt black appearance
but the films produced at 9001C look like the
untreated substrate (i.e. before the nanotube
growth). Closer inspection (Fig. 3) reveals that
there are indeed nanotubes on the film but they are
predominantly bundles of single-walled nanotubes
(difficult to see on the SEM picture) mixed with
some larger MWNT (diameterE30 nm). The
situation changes again at 11001C where there is
a lot of thick fibre growth on the substrate but no
nanotubes can be found.
Micro-Raman spectroscopy provides more de-
tails of the quality and structure of the materials
produced. Raman spectra obtained from films
produced at four different temperatures (excita-
tion wavelength of 514 nm) are shown in Fig. 6.
Unfortunately the spectrum measured on the
À1
9001C film only extends to 2000 cm . The spectra
at 6001C and 7501C are typical for MWNT.
Signals from the underlying Si substrate can be
seen in the bottom spectrum since the films are of
low density. This disappears for the high-density
film produced at 7501C. The two main peaks
observed are the so-called D- and G-lines [20]. The
À1
(2)
G-line, at 1576 cm , corresponds to the E2g
mode of highly oriented pyrolytic graphite and
demonstrates the presence of crystalline graphitic
À1
carbon. The D-line, at 1348 cm , originates from
2
disorder in the sp -hybridised carbon and can be
due to impurities and/or lattice distortions in the
carbon nanotubes. The relative intensity of the D-
line to the G-line decreases with increasing gr owth
temperature indicatingan increasingde rge e of
crystallinity in the material. An increase in
MWNT film quality with growth temperature
Fig. 3. SEM images of carbon nanotube material grown on Si
(8 sccm) for 30 min as a function of oven
substrates from C
temperature.
2
H
2
has been noted before Ref. [3]. The peaks in the
À1
range 2000–3000 cm
are due to second-order
scattering. The spectrum obtained from films
grown at 9001C again shows Si from the exposed
substrate. Evidence for the presence of SWNT in
this sample is obtained from the low wave number
range of the spectrum. This is shown in more detail
in Fig. 7 where micro-Raman spectra obtained
from three different regions of the film are shown
(laser spot diameter of ca. 5 mm). From the density
of bundles of SWNT seen with SEM one can
estimate that one or two bundles will be sampled
by the laser spot. The peaks at 145, 180 and
(
(
7
Fig. 3) although the diameter remains constant
Fig. 4). There is a maximum at a temperature of
501C where films of aligned MWNT are formed
growing perpendicular to the substrate. The left-
hand picture in Fig. 3 shows the top of the film
while the right-hand picture shows the alignment.
The situation changes dramatically at a tempera-
ture between 8501C and 9001C. Also, the macro-
scopic appearance of the films changes. Films

56
O.A. Nerushev et al. / Physica B 323 (2002) 51–59
2 2
Fig. 4. TEM images of material grown from C H (8 sccm) for 30 min at 5501C (left), 7501C (centre) and 11001C (right).
2 2
Fig. 5. Raman spectra from films grown from C H (8 sccm)
for 30 min at different temperatures. The quality of the
nanotubes increases with increasingtemperature up to 900 1C
as seen by the ratio of the Raman D- and G-lines. Only very
little carbon is observed from the material grown at 11001C
with the spectrum dominated by Si. The spectrum obtained
from films grown at 9001C shows evidence of SWNT
formation.
Fig. 6. Micro-Raman spectra measured on three different spots
of the film grown at 9001C. On average only one or two bundles
will be present in the laser focus. The spectra clearly show
different positions for the radial breathingmodes indicating
that the nanotubes within a given bundle have very similar
diameters (indicated on the plot) but the diameters of tubes
within different bundles can differ. (*): laser lines.
1
90 cmÀ1 are due to the radial breathingmode of
SWNT with diameters of 1.6, 1.3, and 1.2 nm,
respectively [21]. The other major peaks in this
part of the spectrum are due to laser lines (marked
with an asterisk). These results are strongly
indicatingthat all the SWNT in a gi ven bundle
have the same diameter. This appears to be similar
to the results reported by Schlittler et al. where
crystalline bundles of SWNT were grown at 9501C
from nano-patterned sandwich layers of C60 and
Ni on a Mo substrate [22].
We can only observe a very small signal from
carbon in the Raman spectrum measured from the

O.A. Nerushev et al. / Physica B 323 (2002) 51–59
57
Fig. 7. Log of the current density versus the applied voltage for an aligned MWNT film produced at a temperature of 7501C. The
behaviour is reproducible as longas the emission current does not exceed 1 mA/cm (stars, triangles). When this current is exceeded
2
(diamonds), subsequent I2V scans (squares and circles) are shifted to higher field strengths.
sample prepared at 11001C in spite of the relatively
high density of nanowires or fibres present. By far
the major contribution in the Raman spectrum is
from Si. This suggests that the structures we
observe in the SEM are mainly due to Si
nanowires; however, this will need to be confirmed
by e.g. EDX measurements and should be
regarded as a preliminary statement.
ported from other groups both for individual
nanotube emitters and for films [23–26]. The I2V
characteristic of a typical film of aligned MWNT is
shown in Fig. 7, plotted on a log-lin scale. As long
as the emission current does not exceed the value
2
of 1 mA/cm the I2V behaviour is reproducible
and can be cycled many times. The current at
which this behaviour breaks down corresponds to
the position in a Fowler–Nordheim plot [27] where
there is a noticeable change of slope indicating that
the local field conditions at the top of the
nanotubes have changed. Once this critical current
has been exceeded, a higher field is required to
obtain comparable emission currents on later I2V
cycles. Changes in the slope of Fowler–Nordheim
plots have been reported previously [10,28,29] and
have been discussed e.g. in terms of space charge
effects [28] or changes in the emitting material due
to the high current [10].
5. The correlation offield emission and light
emission from aligned MWNT films produced
at 7501C from C H
2
2
We have previously reported the field emission
behaviour of films produced from C H at a
2
2
growth temperature of 7501C (very similar to those
shown in Fig. 3) [10]. We were able to show that
such films had field emission properties amongthe
best reported so far and that the presence of an
amorphous carbon coatingon the ali gn ed MWNT
did not significantly affect their field emission
characteristics [10]. Strongli gh t emission accom-
panyingfield emission has previously been re-
We cannot observe any light emission for
emission currents below the critical value corre-
spondingto the re gi on where the Fowler–Nord-
heim slope changes; however, as soon as this
emission current is exceeded we do see strongli gh t

58
O.A. Nerushev et al. / Physica B 323 (2002) 51–59
Fig. 8. Fowler–Nordheim plot from another spot on the film used for Fig. 7. Light emission measured during these meaurements is
also plotted (right-hand scale). The threshold for light emission correlates with the change in the Fowler–Nordheim gradient for the
first I2V scan (marked by dashed line). The fluctuations in the light emission correlate perfectly with fluctuations in the emitted
current. Light emission on the second and third I2V scans on the same spot also begins when the slope of the emission current changes
and shifts continuously to higher applied fields as the number of scans is increased. The light intensity plot marked with 1, 2 and 3,
respectively correlates to the first, second and third I2V scans.
emission. As has been reported for light emission
from individual nanotube tips [24,26], the intensity
of the emitted light closely follows variations in the
emitted current, as can be seen in Fig. 8. The light
emission threshold also moves to higher fields on
subsequent I2V cycles. In contrast to previous
reports [24,25] the light that we observe is mainly
black-body-like indicatinga temperature of the
order of 2000 K. We do not see significant
contributions (although we cannot rule it out
completely on some samples) in the 500–800 nm
range from light emission from localised tip states
as has been reported before [24,25]. We thus
interpret our observations as beingdue to strong
ohmic heatingof the MWNT leadingto black-
body radiation with the temperature increasing
with electron current. For high emission currents
this can lead to emission of carbon particles from
the tips and eventual destruction of the emitting
area, as we have occasionally observed. This
would be in agreement with the conclusions drawn
from the first observations of light emission from
individual MWNT tips [26]. Further work is in
progress.
6. Conclusion
The catalytic particle size and temperature
dependence of iron-catalysed carbon nanotube
growth on silicon substrates has been studied.
The results have suggested a growth model based
on the carbon feedstock supply rate. An interest-
ingtransition is observed at a gr owth temperature
of ca. 9001C for the low carbon flux conditions
that we are usingin our experiments. At this
temperature bundles of SWNT are formed where
all the nanotubes in the bundle appear to have the
same diameter. At higher temperatures the major-
ity of nanostructures observed on the substrate
consist of Si. Aligned films of MWNT are grown
at a temperature of 7501C. The field emission of
these films has been characterised and the electron
emission correlated with light emission. We

O.A. Nerushev et al. / Physica B 323 (2002) 51–59
59
observe mainly black-body-like emission indicat-
ingan ohmic heatingof the nanotubes rather than
discrete light emission from tip states as has been
reported previously by others.
[10] M. Sveningsson, R.-E. Morjan, O.A. Nerushev, et al.,
Appl. Phys. A 73 (2001) 409.
[
[
[
[
[
11] R.-E. Morjan, O.A. Nerushev, M. Sveningsson, et al.,
001, submitted for publication.
12] O.A. Nerushev, R.-E. Morjan, E.E.B. Campbell, et al.,
001, submitted for publication.
13] Y.Y. Wei, G. Eres, V.I. Merkulov, et al., Appl. Phys. Lett.
8 (2001) 1394.
14] J.I. Sohn, C.-J. Choi, S. Lee, et al., Appl. Phys. Lett. 78
2001) 3130.
15] C.J. Lee, S.C. Lyu, Y.R. Cho, et al., Chem. Phys. Lett. 341
2001) 245.
[16] J. Wan, Y.H. Luo, J.L. Liu, et al., J. Crystal Growth
27–228 (2001) 820.
2
2
7
Acknowledgements
(
This work was supported by Stiftelsen f o. r
Strategisk Forskning (SSF) within the CARA-
MEL consortium and by Vetenskapsr a( det. The
authors would like to thank Lena Falk and
Heinrich Riedl for their help.
(
2
[17] J. Abrefah, D.R. Orlander, M. Balooch, et al., Appl. Phys.
Lett. 60 (1992) 1313.
[
[
18] W.L. Holstein, J. Catal. 152 (1995) 42.
19] R.T.K. Baker, P.S. Harris, R.B. Thomas, et al.,
J. Catalysis 30 (1973) 86.
References
[
[
[
[
[
[
[
[
[
[
20] P.C. Eklund, J.M. Holden, R.A. Jishi, Carbon 33 (1995)
9
59.
21] Y.F. Wang, X.W. Cao, S.F. Hu, et al., Chem. Phys. Lett.
36 (2001) 47.
22] R.R. Schlittler, J.W. Seo, J.K. Gimzewski, et al., Science
92 (2001) 1136.
23] J.-M. Bonard, T. St o. ckli, F. Maier, et al., Phys. Rev. Lett.
1 (1998) 1441.
[
[
[
[
[
1] H. Soh, C. Quate, A. Morpurgo, et al., Appl. Phys. Lett.
5 (1999) 627.
7
3
2] O.Z.C. Bower, W. Zhu, D.J. Werder, S. Jin, Appl. Phys.
Lett. 77 (2000) 2767.
3] J.P.C.J. Lin, Y. Huh, J.Y. Lee, Chem. Phys. Lett. 343
2
(
2001) 33.
4] Y.C. Choi, Y.M. Shin, Y.H. Lee, et al., Appl. Phys. Lett.
6 (2000) 2367.
8
24] J.-M. Bonard, J.-P. Salvetat, T. St o. ckli, et al., Appl. Phys.
A 69 (1999) 245.
25] A.G. Umnov, V.Z. Mordkovich, Appl. Phys. A 73 (2001)
7
5] Z.K.I. Willems, J.F. Colomer, G. Van Tendeloo, N.
Nagaraju, A. Fonseca, J.B. Nagy, Chem. Phys. Lett. 317
3
01.
26] A.G. Rinzler, J.H. Hafner, P. Nikolaev, et al., Science 269
1995) 1550.
27] R.H. Fowler, L. Nordheim, Proc. R. Soc. London Ser. A
19 (1928) 173.
(2000) 71.
[
[
[
[
6] R.A.S.B. Sinnot, D. Qian, A.M. Rao, Z. Mao, E.C.
Dickey, F. Derbyshire, Chem. Phys. Lett. 315 (1999) 25.
7] L.K.L.F.K. Bladh, F. Rohmund, Appl. Phys. A 70 (2000)
(
1
317.
28] J.-M. Bonard, T. St o. ckli, O. Noury, et al., Appl. Phys.
Lett. 78 (2001) 2775.
29] Y. Saito, K. Hamaguchi, K. Hata, et al., Ultramicroscopy
8] O.A. Nerushev, M. Sveningsson, L.K.L. Falk, et al.,
J. Mater. Chem. 11 (2001) 1122.
9] F. Rohmund, L.K.L. Falk, E.E.B. Campbell, Chem. Phys.
Lett. 328 (2000) 369.
7
3 (1998) 1.