Introduction
Combining the insights of a pioneering team of researchers with the 24/7 work capability of robotically operated furnaces provides a powerful advantage when looking for better ways to make carbon nanotubes. Indeed, this combination has now resulted in a major breakthrough in nanotube synthesis at the National Institute of Advanced Industrial Science and Technology (AIST) in Tsukuba, Japan. On page 131 of this issue Kenji Hata, Sumio Iijima and co-workers1 describe a new process for the size-selective growth of ultrahigh forests of double-walled carbon nanotubes containing negligible amounts of catalyst.
The carbon nanotubes are synthesized by chemical vapour deposition (CVD). In this technique, iron nanoparticles on an oxide-coated silicon substrate catalyse the decomposition of hydrocarbon vapours to produce nanotubes on the surface. The technique depends critically — in largely unpredictable ways — on catalyst composition and structure, gas pressure and flow rates, and the compositions of precursor reactants and carrier gases, not to mention the temperatures for any catalyst pre-treatment and the CVD growth itself. The fully automated system at AIST allows Hata, Iijima and co-workers to methodically tweak each of these parameters to optimize the final carbon nanotube growth. They can watch the robots work tirelessly and monitor results while they rest at home, intervening only when the pre-programmed growth conditions fail to yield the desired results. This strategy increases sample output tenfold compared with humans doing the work.
The AIST team, with the help of their robotic assistants, are growing 'forests' of carbon nanotubes (Fig. 1). Roughly parallel carbon nanotubes grow on the silicon substrate like a forest of bamboo trees. The carbon nanotubes themselves come in different forms: the single-walled nanotubes (SWNTs) are seamless cylinders made of rolled-up graphite sheets, whereas the double-walled (DWNTs) and multiwalled nanotubes (MWNTs) contain concentric layers of their single-walled cousins. All are of practical and fundamental interest, although which type is the most useful will depend on whether the application benefits from high surface area (sensors and supercapacitors) or high density (strong mechanical structures or good thermal and electrical conductors).
Figure 1: An example of the nanotube forests grown in Kenji Hata's laboratory.
The scale is shown by mm markings along the bottom.
Full size image (11 KB)Recent work from this pioneering team has already produced remarkable results2. In 2004 they discovered that just the right trace amount of water dramatically increases catalyst lifetime because the water vapour oxidizes the harmful carbon coating that forms on the catalyst during the growth process. A longer catalyst lifetime means that there is more time for the nanotubes to grow, which permits the synthesis of 2.5-mm-high forests of SWNTs that contain less than 0.02% (by weight) of catalyst. For comparison, SWNTs produced by high-pressure CVD of carbon monoxide are over a thousand times shorter and can contain up to 40% catalyst.
The same trick of using just enough water vapour to protect the catalyst, but not enough to damage the nanotubes, is used in the present advance. The new aspect is a thorough investigation of how the thickness of the iron films, used here as a catalyst precursor, determines the type (SWNT, DWNT or MWNT) and diameter of the nanotubes. When heated, the iron film (which is typically 1–2 nm thick) breaks up into an array of iron particles, each forming a possible catalytic base for growing one of the nanotubes in the forest. Thicker films tend to break up into larger particles, which catalyse nanotubes with larger diameters.
The authors grow forests that are several millimetres high and comprise principally SWNTs, DWNTs, or MWNTs. It is worth noting that if the trees in the aforementioned bamboo forest had diameters of 5 cm and the same ratio of diameter to forest height as the DWNTs, the trees would be 30 km tall!
Why is this notably heroic effort being expended to synthesize nanotube forests? One reason is that the nanotube forests can be deployed for applications such as chemical and biological sensors, supercapacitors, batteries, fuel cells, and thermally conducting interconnects for cooling electronic circuits3, 4. Another possibility is to use the forest as a supply of roughly aligned nanotubes for the fabrication of yarns and sheets (Fig. 2). In this process, the nanotube forest serves the same function as the 'roving' of oriented wool fibres used for spinning wool5, 6, 7.
m-diameter nanotube yarn in the vertical direction (weft) and polyester yarn in the horizontal direction (warp). Increasing the nanotube forest height from 300 ![Figure 2 : Left: Scanning electron micrographs of twist-based spinning of a MWNT yarn from a carbon nanotube forest (centre) and derived four-ply (top) and knitted single-ply yarn (bottom) Copyright (2006) AAAS. Right: Optical micrograph of a nanotube textile comprising a 10-ply, 50-|[mu]|m-diameter nanotube yarn in the vertical direction (weft) and polyester yarn in the horizontal direction (warp). Increasing the nanotube forest height from 300 |[mu]|m (used to make the yarn in these images) by the factor of ten made possible by Hata, Iijima and co-workers could considerably increase mass throughput rate.](/nnano/journal/v1/n2/thumbs/nnano.2006.109-f2.jpg)
The demands of this last application on nanotube assembly rate are high: for each kilogram of yarn, over three billion kilometres of nanotubes must be assembled at commercially viable rates. Current nanotube-yarn assembly rates of 30 m per minute now exceed those for commercial wool spinning, but increased mass throughput is required. This could be enabled by use of the ultra-long nanotubes fabricated by the Japanese group. As nanotube forest growth can be quite fast, one can envisage a conveyer belt with nanotubes grown on one end and stripped off at the other.
However, the forest topology needed for solid-state yarn and sheet fabrication is complex. Few forest types currently fit the bill, and the ultra-long forests grown by the AIST team1 are not yet optimized for these processes. The yarn- and sheet-drawing processes require a certain amount of bonding between the nanotubes in the forest, which can be achieved by intermittently bundling the nanotubes together. If either this bundling or the density of nanotubes in the forest is too low (or too high), the fabrication processes fail5, 6.
Now that forests of predominately SWNTs, DWNTs and MWNTs can be grown, we need to ask which type of carbon nanotube is best suited for a given application. Conventional wisdom says that unbundled SWNTs provide the best properties. However, this conclusion largely rests on data for nanotubes that are just micrometres in length. Such short nanotubes cannot provide efficient mechanical stress transfer and electron and phonon transport — either between the walls in MWNTs or DWNTs or between bundled nanotubes.
Relative performance results for different nanotube types will change with the availability of increasingly longer nanotubes, and the performance achieved by all nanotube types is likely to dramatically increase. This is one reason why bulk preparation of ultra-long, easily processed nanotubes is so important. With ever-longer nanotubes becoming available, the highest bulk mechanical strength, stiffness, and electrical and thermal conductivities will likely come from MWNTs that are closely packed. MWNTs provide the highest density of nanotube walls that can carry currents, transport heat, and support stress (as long as the inner walls are similar in diameter to the SWNTs). For electrochemical devices where high gravimetric surface area is needed (such as supercapacitors, fuel cells, and artificial muscles), SWNTs or DWNTs might be best, as long as the degree of nanotube bundling is not too large.
What challenges lie ahead for fabricating nanotube forests? Reduction of growth time, further increase of nanotube length and optimization of the structures grown, both in terms of nanotube perfection and forest morphology, should be high on the agenda. The size-selective formation of DWNTs, as opposed to SWNTs, was achieved for DWNT diameters of about 3.75 nm (ref. 1). Such large diameters and corresponding large void space in the nanotube centre decrease area-normalized properties, unless, like a deflated hose, the tube collapses into a ribbon. Hence, strategies to provide both selectivity of DWNT growth and small DWNT radii are desirable.
The ultimate goal is to synthesize SWNTs of one type and DWNTs and MWNTs with pre-determined wall geometry. This might eventually be possible with a suitable choice of the seeds for forest growth — as in Rick Smalley's proposed cloning method for SWNTs8. Considering the cascade of important results arising from the AIST researchers, their brilliance, and their equipment capabilities, even the most impossible of these tasks seem realizable.
