Original

Paper

phys. stat. sol. (b) 244, No. 11, 3911–3915 (2007) / DOI 10.1002/pssb.200776136

Catalyst size dependencies for carbon nanotube synthesis

M. H. Rümmeli*, 1, 2, C. Kramberger1, F. Schäffel1, E. Borowiak-Palen2, T. Gemming1,
B. Rellinghaus1, O. Jost3, M. Löffler1, P. Ayala1, 4, T. Pichler1, and R. J. Kalenczuk2
1
IFW Dresden, P.O. Box 270116, 01171 Dresden, Germany
2
Szczecin University of Technology, Centre of Knowledge Based Technologies and Nanomaterials,
Institute of Chemical and Environment Engineering, Szczecin, Poland
3
Fraunhofer IWS, Winterbergstr. 28, 01277 Dresden, Germany
4
Departamento de Fisica, PUC-Rio, CP38071, 22452-970 Rio de Janeiro, Brazil

Received 30 April 2007, revised 13 June 2007, accepted 13 June 2007

Published online 26 September 2007

PACS 61.46.Fg, 61.48.+c, 68.37.Lp, 68.55.Ac, 81.07.De, 81.15.–z

We have systematically explored the synthesis parameters in laser evaporation and CVD for the synthesis
of carbon nanotubes. Our findings point to a dependence of the resultant tubes on the volume to surface
area ratio of the catalyst particle. The data furthers our understanding on the nucleation of carbon nano-
tubes.

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1 Introduction
In terms of their electronic structure, single walled carbon nanotubes (SWCNT), where a single graphene
layer is rolled up to form a tube, are of great interest because they can be metallic or semi-conducting
depending on the SWCNT diameter and chirality (the angle at which the graphene layer is rolled up).
However, to take advantage of these exceptional electronic properties a key prerequisite is that they be
manufactured with sufficient control to accurately select or form the SWCNT with a chosen diameter
and chirality. Such control has thus far eluded scientists and engineers, although significant progress has
and continues to be made. Generally, carbon nanotubes (CNT) require growth to stem from a catalyst
particle, although there are instances in which multi-walled carbon nanotubes (MWCNT) have been
synthesized without catalyst particles. Successful catalyst particles generally are Ni, Co, or Fe based,
however, more recent studies show other metals such as In, Cu, Ag and Au can also be used for SWCNT
formation [1–3]. Since SWCNT nucleation requires a catalyst particle, a great deal of attention has been
paid to the catalysts role in forming the embryonic stage of SWCNT. However, their full role has yet to
be determined and this is in part due to conflicting results, which may indicate that several mechanisms
exist. This is highlighted, for example, by different studies in chemical vapour deposition (CVD) based
synthesis routes in which the catalyst resides on a substrate, henceforth referred to as substrate-based
CVD. Some substrate-based CVD studies suggest that the catalyst is not in a liquid state at nucleation
and it is argued that the initial embryonic cap of a SWCNT forms via surface diffusion of C [4]. In addi-
tion, in situ TEM studies suggest that the catalyst geometry does not template this cap [5]. However,
other substrate-based CVD studies point to the diameter of the catalyst particle actually templating the
diameter of the obtained SWCNT [6, 7], and in addition it is argued that the catalyst particles are in a
liquid state allowing bulk diffusion of C. Our substrate based CVD study, as discussed below, also points

*
Corresponding author: e-mail: m.ruemmeli@ifw-dresden.de, Phone: +49 (0)351 4659 510

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

3912 M. H. Rümmeli et al.: Catalyst size dependencies for carbon nanotube synthesis

to the catalyst diameter templating that of the resultant CNT. Other SWCNT synthesis routes do not
require the catalyst particle to reside on a substrate, but instead the catalyst floats in free space and is
termed floating catalyst. Floating catalyst synthesis routes include CVD, arc-discharge and laser evapora-
tion techniques. In our studies on floating catalysts we use the laser evaporation technique since the ease
with which many of the process parameters can be controlled in laser evaporation makes it an excellent
platform to investigate the growth mechanisms of SWCNT [8]. Our detailed laser evaporation studies
point to a cap formation templated by the catalyst particle (circumference growth). Further, these studies
in conjunction with previous works enable us to develop a nucleation model, which describes not only
our observations but also previously reported observations in the synthesis of SWCNT via laser evapora-
tion.

2 Experimental
The laser evaporation setup and is described elsewhere [9]. Two sets of SWCNT using the following
catalyst mixes were explored, Pt/Rh/Re (18.98:3.22:0.8 wt%) and Ni/Co (10:10 wt%). In all cases the
purities of the materials, including the target graphite were >99.9%.
For the CVD studies Fe nanoparticles were deposited on thermally oxidised Si with an additional
10 nm Al2O3 layer using an inert gas condensation method based on a direct current sputtering process
[10]. Witness plates are obtained during the procedure for pre-CVD analysis. The CVD reactions were
conducted at 800 °C with Cyclohexane (50 mbar) as the feedstock. More details of the CVD reaction are
given in Ref. [10]. The obtained products from the laser evaporation and CVD routes were analysed
using transmission electron microscopy (TEM) – (FEI Tecnai F30), Fourier transform Raman spectro-
scopy (Bruker IFS100) and Fourier transform optical absorption spectroscopy (OAS) – (Bruker
IFS113V/88) in the energy range 0.35 eV to 2.35 eV. Details on the preparation of the TEM grids and
OAS films can be found elsewhere [9].

3 Results and discussion

Our systematic laser evaporation studies explored various parameters, namely, gas type (He, Ne, Ar, N2,
Kr and Xe), catalyst concentration and type and temperature (800 °C to 1600 °C). The effect of these
parameters on the mean diameter and diameter distribution as well as yield were determined primarily
using OAS which has been shown to be a powerful tool for such studies via the SWCNT absorption
peaks arising from transitions between van Hove singularities.
The studies on the role of gas type showed in all cases that as the oven temperature rose, the mean
diameter also increased in a linear manner until at the higher temperatures the mean diameter increase
suddenly levelled off to a constant value [9, 11], which we term the upper saturation diameter. With
respect to the linear increase in mean diameter of the SWCNT with oven temperature, the rate of change
varied between the gases such that the gradient became less steep as the gas thermal conductivity de-
creased. As for the upper saturation diameter, this remained the same for all gases with the exception of
He, which has a significantly higher thermal conductivity relative to the other investigated gases. In this
case the saturation diameter was reduced. With lower oven temperatures the temperature dependence
also saturates to a constant level. This lower mean diameter limit we term the lower saturation diameter.
Both the upper and lower saturation diameters can be seen in Fig. 1a.
In terms of the yield (SWCNT to amorphous species), N2 provided the highest yields and He led to the
lowest yields. The yield is also dependent on the catalyst content in the ablation target, such that as the
catalyst content increases so does the yield up to a catalyst content of ca. 10 at%. Thereafter the yield
rapidly drops. However, despite the changes in yield, no changes in the diameter dependence are ob-
served with catalyst concentration. Slight differences in the upper and lower saturation diameters are
observed for different catalyst types though.
We also explored the variation of the full width at half maximum (FWHM) of the diameter distribu-
tion with respect to oven temperature. This dependence is illustrated in Fig. 1b which shows the FWHM

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.pss-b.com

 Original
Paper

phys. stat. sol. (b) 244, No. 11 (2007) 3913

1,60 0.70

1,55 Gaussian Fit

Upper saturation diameter
Mean Diamater (nm)

0.65
1,50
0.60

FWHM (nm)
1,45
0.55
1,40
0.50
1,35

1,30 0.45

1,25 Lower saturation diameter 0.40

a) b)
900 1000 1100 1200 1300 1400 1500 1600 900 1000 1100 1200 1300 1400 1500 1600
o o
Oven Temperature ( C) Oven Temperature ( C)

Fig. 1 (online colour at: www.pss-b.com) a) Mean diameters vs. oven temperature in N2 (Target A). b) Diameter
distribution full width at half maximum (FWHM) vs. oven temperature in N2 (Target A). Curves are to provide a
guide to the eye.

increasing to a maximum after which it declines. The above-discussed behaviours can be explained in a
model where there exists a nucleation window, which depends on the catalyst volume to surface area
(Vc/Ac), and a second window in which the carbon saturated catalyst particles precipitate their carbon.
This second precipitation window will be dependent on various factors such as the catalyst-carbon eutec-
tic temperature and the carbon absorption coefficient of the catalyst. We argue that SWCNT are only
formed when these two windows overlap.
SWCNT formation
window

1
No stable cap Stable carbon cap Carbon encapsulation

Precipitating catalyst cluster size distribution

Fig. 2 (online colour at: www.pss-b.com) Proposed nucleation mechanism. The schematic shows the
precipitating particle diameter distributions at three snapshots in time for 3 increased oven temperatures
(1 to 3). Within the SWCNT formation window stable nucleation caps allow SWCNT growth. SWCNT
are only formed in the overlap (shaded region) between the precipitating catalyst (size) distribution and
SWCNT growth window. The degree of the overlap determines the SWCNT diameter distribution and
mean diameter. Particles that are too large are encapsulated by graphite and particles that are too small are
unlikely to form a stable cap.

www.pss-b.com © 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

3914 M. H. Rümmeli et al.: Catalyst size dependencies for carbon nanotube synthesis

Fig. 3 Upper panels: TEM images of the starting Fe catalyst particles. Increasing size from left to right.
Lower panels: Resultant CNT after CVD synthesis on Si/SiO2/Al2O3 (800 °C in cyclohaxane – see Ref. [10]).

Figure 2 illustrates the model and shows the requirement of the SWCNT formation window where
only catalyst particles, which precipitate sufficient carbon to form a hemisphere, nucleate SWCNT. Cata-
lyst particles that are too small cannot form a stable cap for nucleation and particles, which are too big
also do not nucleate SWCNT since they are simply encapsulated by the precipitating carbon. These cata-
lyst volume to surface area restrictions can also be translated to substrate based CVD growth. In this case
the catalysts reside on the substrate and so the catalysts geometry is essentially limiting the Vc/Ac con-
straints by the height of the molten droplets. Further, the interaction between the catalyst particle and the
substrate prevents the particle from being encapsulated, thus as the particle gets bigger, one would expect
the number of nucleation caps to increase. Studies of ours in which we were able to deposit catalyst par-
ticles, as obtained from inert gas condensation, onto substrates in a controlled manner (size and density)
enabled us to directly examine the catalyst particles prior to CVD synthesis and compare them with the
resultant CNT in terms of their diameter and number of walls. The studies show that both the diameter
and number of walls of the resultant CNT increase proportionally to the size of the catalyst particle.
Figure 3 shows typical TEM images for catalyst (Fe) particles with different size distributions and their
resultant CNT after CVD synthesis [7, 10]. One sees clearly the increasing CNT diameter and number of
walls as the catalyst diameter increases. This behaviour is fully consistent with catalyst volume to surface
area constraints.

4 Conclusions
To conclude, we have conducted systematic studies on the formation of CNT in laser evaporation and
CVD. The data points to the resultant CNT having a dependence on the catalyst volume to surface area
ratio. In the case of floating catalysts e.g. laser evaporation this restricts the upper and lower diameter
attainable for SWCNT. In the case of substrate CVD the restrictions can allow one to tailor the CNT with
a given diameter or number of walls.
Acknowledgements M.H.R. acknowledges EU (CARBIO) for support. C.K. thanks the MPI and F.S. thanks the
Cusanuswerk for financial support.

References
[1] M. H. Rümmeli et al., Nano Lett. 5, 1209 (2005).
[2] D. Takagi et al., Nano Lett. 6, 2642 (2006).

© 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.pss-b.com

 Original
Paper

phys. stat. sol. (b) 244, No. 11 (2007) 3915

[3] M. H. Rümmeli et al., phys. stat. sol. (b) 243, 3101 (2006).
[4] S. Hofmann et al., Appl. Phys. Lett. 83, 135 (2003).
[5] S. Hofmann et al., Nano Lett. 7, 602 (2007).
[6] Y. Zhang et al., Appl. Phys. A 74, 325 (2002).
[7] F. Schäffel et al., Chem. Mater., published online (2007), DOI 10.1021/cm070950k.
[8] C. T. Kingston et al., J. Nanosci. Nanotechnol. 6, 1 (2006).
[9] M. H. Rümmeli et al., J. Phys. Chem. C 111, 4094 (2007).
[10] F. Schäffel et al., phys. stat. sol. (a) 204, 1786 (2007).
[11] M. H. Rümmeli et al., J. Phys. Chem. B 111, 8234 (2007).
[12] C. T. Kingston et al., J. Nanosci. Nanotechnol. 6, 1 (2006).

www.pss-b.com © 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim