Tsukuba Symposium on Carbon Nanotube, Tsukuba Japan, 2001
Contribution to the Proceedings of the Tsukuba Symposium on Carbon Nanotube, October 3-5, 2001, Tsukuba, Japan. Physica B 323 (1-4), 90-96 (2002)
Carbon and metals: a path to single-wall carbon nanotubes
Donald S. Bethune
IBM Almaden Research Center, San Jose, CA 95120-6099, USA
Received 5 October 2001
_____________________________________________________________________________________________
Abstract
Adding metals to carbon vapor leads
to the formation of a rich variety of molecular structures.
Examples include endohedral fullerenes, metallocarbohedrenes,
metal-carbide nanocrystallites and perhaps most spectacularly,
single-wall carbon nanotubes (SWNT). We discovered that pure
carbon SWNT catalytically self-assemble when cobalt is co-vaporized
with carbon in a fullerene generator. Third elements added to the
metal and carbon can have important effects. Sulfur and bismuth
can greatly increase the production efficiency of SWNT, leading to
aggregation of SWNT into nanotube bundles or ropes, and can
significantly extend the tube diameter distribution. Tubes larger
than 5 nm were produced and S increased the mean tube diameter by
~60%. A different type of ‘third atom’ effect was recently found
at Virginia Tech: adding nitrogen to a scandium/carbon vapor
leads to the formation of Sc3N@C80,
a new type of endohedral fullerene. Erbium can substitute for any
of the scandium atoms and we have measured the emission spectra and
lifetimes of ErmSc3-mN@C80
(m=1-3). These were the first emission lifetime measurements
reported for light coming from species trapped inside fullerenes.
PACS: 61.46.+w; 61.48.+c; 81.07.De; 78.67.Bf; 33.15.-e; 33.50.-j
Keywords: Single-wall carbon nanotubes; Metallofullerenes; Er emission spectroscopy
_____________________________________________________________________________________________
1. Introduction
Mixing carbon and metals under the
extreme conditions that prevail in a carbon-arc or laser-plasma based
fullerene generator leads to the formation of a remarkably diverse set
of nanostructures, including fullerene cages with metal atoms inside or
outside, metal-carbon covalently bonded clusters, ionic superconducting
crystals of fullerenes and alkali metals, and pure carbon nanotubes
with single-atomic-layer walls and diameters ~1 nm. Some of these
species have played important roles in the development of the fullerene
field from the start.
At the beginning of the fullerene
era, Richard Smalley and his colleagues at Rice tried vaporizing La
with carbon in hopes of finding species with metal atoms trapped inside
the closed carbon cages they had proposed in their celebrated paper on C60 [1]. The experiment succeeded in finding a LaC60
species that survived conditions deliberately designed to expose any
accessible La to chemical attack, consistent with the proposed
endohedral structure, La@C60 [2]. A variation of the La experiment in 1991, again carried out at Rice, found that LaC82,
produced by laser-vaporizing La loaded carbon, survived solvent
extraction from fullerene soot and Houston’s August weather, thus
providing further evidence for metal encapsulation [3].
La, which forms a marginally stable, water reactive, flammable carbide (LaC2),
turned out to have been an astute choice. Subsequent experiments
that mixed carbon with metals that are either better or worse carbide
formers have had dramatically different outcomes. Ti, for
example, forms a hard, highly refractory carbide (MP ~3100 C).
When laser-vaporized Ti and C condense, covalently bonded clusters
result: met-cars such as Ti8C12 are favored at low metal/carbon ratio [4], while at higher metal concentrations FCC-nanocrystals with TinCn-1 stoichiometry dominate [5].
Recently, a laboratory study of the infrared spectra of TiC
clusters led to the identification of FCC-TiC crystallites as carriers
of the prominent 20.1 mm emission band from carbon-rich red-giant stars at the end of their lives [6].
In the direction of non-carbide
forming metals, low concentrations of iron-group metals and carbon tend
to phase separate into metal crystallites and graphite. While
this might lead one to expect that nothing as interesting as the
formation of endohedral fullerenes, met-cars or nanocrystals occurs
when transition metals and carbon are co-vaporized, in fact the
interaction of the condensing carbon with only a few at% of transition
metal modifies the condensation process in a way that is quite
astonishing and beautiful: the metal catalyzes the formation of
single-wall carbon nanotubes.
2. Metallofullerenes
At the IBM Almaden Research Center we were inspired by the Rice La@C82
experiment to begin experimenting with metal- carbon
combinations. We added La to graphite in our arc fullerene
generator in an attempt to generate endohedral species in greater
quantity. The odd electron count of La suggested EPR might provide a
sensitive assay for metallofullerene production, since fullerenes
themselves always have an even electron count. The EPR spectra of
the raw soot and a toluene extract from that soot (Fig. 1) showed coupling of a single unpaired electron with the spin 7/2 La nucleus [7].
This was evidence that the La had transferred 3 electrons to the cage,
leaving it in its preferred +3, S=0 state, while the cage acquired net
spin S=½. These results led many groups to pursue making
metallofullerenes and within a short time a wide variety were produced,
including fullerenes encapsulating most of the lanthanides, alkaline
earth atoms, U, alkali atoms, and Ti, and methods to purify
metallofullerenes were being developed [8], [9].
Fig. 1. EPR spectra of La@C82: a) solid powder; b) in toluene solution [7].
Fig. 2. EPR spectra of Sc3@C82 at 193 K in decalins: upper – liquid; lower – frozen [13].
Species with multiple Sc atoms encapsulated turned out to be particularly interesting. Purified Sc2@C84 was crystallized and imaged using TEM, which allowed the metal atoms to be located inside the cages [10]. The simultaneous observation of Sc3@C82 by our group [11] and Shinohara’s group [12]
was something of a watershed. The remarkable 22-line hyperfine
spectra showed the coupling of a single unpaired electron to three
equivalent Sc nuclei. Further studies of the temperature
dependence of Sc3@C82 spectra, such as those shown in Fig. 2, revealed that the metal triangle can rapidly reorient inside the cage [13].
The fact that three atoms could be encapsulated naturally raised
interesting questions: How many metals can be encapsulated in a
fullerene? Would it be possible to encapsulate enough transition
metal atoms to make an endohedral ferromagnetic particle? Such
questions led us to begin trying to make endohedral transition metal
fullerenes, starting with Co.
3. Single-Wall Carbon Nanotubes
When we first vaporized Co and graphite in
our arc generator, it was immediately apparent that something unusual
was happening. As the arc ran, we could see the chamber filling
with what appeared to be long spider-web like threads. When the
chamber was opened, the soot on the walls was rubbery rather than
crumbly, like normal ‘fullerene’ soot. SEM showed the web
material was a mix of ‘threads’ and nodules of carbon, but TEM revealed
the core secret of the threads: they were made up of bundles of
carbon nanotubes with single atomic layer walls (Fig. 3) [14].
At the same time, Iijima and Ichihashi independently reported the
observation of SWNT grown using Fe catalyst in the presence of CH4 [15].
The diameters of the Co-catalyzed single-wall carbon nanotubes appeared
to be remarkably uniform, and were narrowly distributed around 1.3 nm (Fig. 4a).
Although the first samples contained only
a few percent of SWNT, this was enough to allow the group of Michael
Heben at the National Renewable Energy Laboratory (NREL) to carry out
the first measurements [16] to assess whether H2 can be stored in SWNT, as first suggested by Pederson and Broughton in 1992 [17]. The NREL results provided the first-ever demonstration of stable physisorption of H2 on a carbon above 285 K. This potentially important application remains promising, but is still controversial [18].
Fig. 3. TEM image showing single-wall carbon nanotubes produced using 2 at% Co [14]. (Enlarged version)
A point I would like to stress is that adding a third element to the metal and carbon can have important effects. Sulfur [19] and Bi [20], [21]
were found to greatly enhance the production efficiency of SWNT, under
some conditions by an order of magnitude. High densities of Co/Bi
catalyzed SWNT aggregated into bundles or ropes with relatively little
amorphous carbon coating, as seen in Fig. 5.
It is clear that the extremely inhomogeneous environment of an arc,
with huge temperature and density gradients and short dwell times,
precludes careful optimization of SWNT production efficiency. But
the important thing that arc production showed was that the catalytic
process is highly effective in building SWNT, so that despite the
relatively unfavorable conditions, significant yields can be
obtained. More recently in tube furnaces, where temperature and
pressure can be precisely controlled, yields of 70% or better have been
obtained [22].
Fig. 4. SWNT diameter distributions obtained using Co alone, or with Co and sulfur. (Adapted from [19])
A second consequence of the addition of co-catalysts like S, Bi, or Pb [21] to Co is that the diameter distribution of the tubes produced is considerably altered. Fig. 4b
shows the diameter distribution obtained when 4 at% each Co and S were
co-vaporized with graphite. With S added, the mean tube diameter
increased from ~1.4 to 2.2 nm and nanotubes with diameters out to
nearly 60 nm were produced. These are very large
structures. For example, a 4.1 nm diameter, 30-30 nanotube has ~
320 carbons per nm of length. The shift in mean diameter has
implications for models of SWNT formation. Consider the simple
model described by Thess et al. in 1996 [22],
in which the diameter of a tube is governed at its birth by competition
between dangling bond energy, which increases as the diameter, and
strain energy due to bending graphene into a cylinder, which is inverse
to the diameter. The energy due to the end cap, with its six
isolated pentagons, is considered to be independent of diameter.
A consequence of this model is that the optimum diameter is
proportional to a-1/2, where a is the energy/edge-length due to dangling bonds. In this model, to increase the mean tube diameter from 1.4 to 2.2 nm, a
must decrease by ~60%. This suggests that sulfur increases the
mean tube diameter by tying up dangling bonds until it is displaced by
carbons adding to the growing tube. Earlier work on
graphitization of carbon found that S promotes graphitization by
interacting with the edges of the basal planes [23].
4. Emission spectroscopy of ErmSc3-mN@C80
I now turn to a different type of ‘third
element’ effect in the realm of metallofullerenes. Several years
ago Dorn and the Virginia Tech group discovered that by introducing a
small amount of N2 into the arc chamber, they could produce Sc3N@C80, which has an icosahedral cage containing a triangular group with Sc at the corners and N at its center [24]. It is possible to replace any or all of the Sc atoms with Er, to give ErmSc3-mN@C80 [25]. A model of the Er3 version is shown in Fig. 6. Such species are especially interesting because Er3+ emits light near 1.5 mm
with spectral characteristics that are sensitive to the Er
environment. Here the isolation and homogeneity of the internal
environment is unique, and opens up the possibility of
spectroscopically studying spin-spin interactions between trapped
species within the cage or within different cages. Additionally,
we can investigate the pathways of energy relaxation.
The spectra obtained at liquid-He temperature for the Er1, Er2 and Er3 species are shown in Fig. 7 [26]. The overall position and number of lines in the spectra confirm that the emitting species are Er3+
ions. The spectral widths of the lines are roughly an order of
magnitude narrower than lines observed in typical glassy environments,
showing that the Er ions indeed see a homogeneous environment and are
isolated from the disordered matrix. Er3N@C80
Fig. 6. Model of Er3N@C80. Cage symmetry is Ih.
shows a clean, simple spectrum, indicating
the predominance of a single molecular structure. In contrast,
the fine structure observed for the Er1 and Er2
species indicates that the internal group has two possible arrangements
in each of those cases, with equivalent Er sites in the latter.
We also made the first lifetime measurements for excited species inside fullerenes by time-resolving the infrared emission (Fig. 8) [26]. The lifetimes are all ~1 ms,
in contrast to the ~10 ms lifetimes typically observed in inorganic
crystals, corresponding to an emission quantum efficiency of ~10-4.
Raising the excitation energy, for example by going to a species such
as Yb, could decrease the vibrational deactivation efficiency and give
longer lifetimes.
Fig. 7. Spectra of ErmSc3-mN@C80, T=1.6 K. (Adapted from Ref. [26])
Fig. 8. Time-resolved fluorescence at 1.6 K from ErmSc3-mN@C80. (Adapted from Ref. [26])
5. Conclusions
Adding
metals to carbon vapor has provided researchers with a fascinating set
of molecular structures, and led to the discovery of efficient
catalytic routes to the production of SWNT. The characteristics
of the SWNT can be altered in important ways by the addition of a third
element promoter such as S or Bi to the transition metal
catalyst. In addition, the presence of a third element besides
metal and carbon can also give entirely new types of species, such as Er3N@C80.
Isolation of light-emitting Er atoms inside fullerene cages provides a
homogeneous environment and leads to narrow spectral lines.
Lifetimes for the Er excitations in fullerenes were measured for the
first time and were found to be ~1 ms.
Acknowledgements
I am grateful to numerous collaborators
for their contributions to the work described in this paper,
particularly R. D. Johnson, C. S. Yannoni, R. Wendt, H. Hunziker, M. S.
de Vries, P. H. M van Loosdrecht, R. Macfarlane, C-H. Kiang, R. Beyers,
J. R. Salem, H. C. Dorn, S. Stevenson and M. J. Heben.
References
[1] H. W. Kroto et al., Nature 318, 162 (1985).
[2] J. R. Heath et al., J. Am. Chem. Soc. 352, 7779 (1985).
[3] Y. Chai et al., J. Phys. Chem. 95, 7564 (1991).
[4] B. C. Guo, K. P. Kerns and A. W. Castleman, Science 255, 1411 (1992).
[5] J. S. Pilgrim and M. A. Duncan, J. Am. Chem. Soc. 115, 9724 (1993).
[6] G. von Helden et al., Science 288, 313 (2000).
[7] R. D. Johnson et al., Nature 355, 239 (1992).
[8] D. S. Bethune et al., Nature 366, 123 (1993).
[9] H. Shinohara, Rep. Prog. Phys. 63 no. 6 (UK), 843 (2000).
[10] R. Beyers et al., Nature 370, 196 (1994).
[11] C. S. Yannoni et al., Science 256, 1191-2 (1992).
[12] H. Shinohara et al., Nature 357, 52 (1992).
[13] P. H. M. van Loosdrecht et al., Phys. Rev. Lett. 73, 1415 (1994)
[14] D. S. Bethune et al., Nature 363, 605 (1993).
[15] S. Iijima & T. Ichihashi, Nature 363, 603 (1993).
[16] A. C. Dillon et al., Nature 386, 377 (1997).
[17] M. R. Pederson and J. Q. Broughton, Phys. Rev. Lett. 69, 2692 (1992).
[18] C. Zandonella, Nature 410, 734 (2001).
[19] C-H. Kiang et al., J. Phys. Chem. 98, 6612 (1994).
[20] C-H. Kiang et al., Carbon 33, 903 (1995).
[21] C-H. Kiang et al., J. Phys. Chem. Solids 57, 35 (1996).
[22] A. Thess et al., Science 273, 483 (1996).
[23] H. Katsuki et al., Carbon 19 148 (1981).
[24] S. Stevenson et al., Nature 401, 55 (1999).
[25] M. M. Olmstead et al., J. Am. Chem. Soc. 122, 12220 (2000).
[26] R. M. Macfarlane et al., Chem. Phys. Lett. 343, 229 (2001). |
