scientific correspondence

4. Swenson, K.I., Jordan, J. R., Beyer, E. C. & Paul, D. L. Cell 57, 145155 (1989). 5. Carrasquillo, M. M., Zlotogora, J., Barges, S. & Chakravarti, A. Hum. Mol. Genet. 6, 21632172 (1997). 6. Zelante, L. et al. Hum. Mol. Genet. 6, 16051609 (1997). 7. Denoyelle, F. et al. Hum. Mol. Genet. 6, 21732177 (1997). 8. Estivill, X. et al. Lancet 351, 394398 (1998). 9. Lench, N., Houseman, M., Newton, V., Van Camp, G. & Mueller, R. Lancet 351, 415 (1998). 10. Scott, D. A., Kraft, M. L., Stone, E. M., Sheffield, V. C. & Smith, R. J. Nature 391, 32 (1998). 11. Denoyelle, F. et al. Nature 393, 319320 (1998). 12 White, T. W., Bruzzone, R., Goodenough, D. A. & Paul, D. L. Mol. Biol. Cell 3, 711720 (1992). 13. Goliger, J. A. & Paul, D. L. Dev. Dyn. 200, 113 (1994).

Very long carbon nanotubes

Carbon nanotubes1 can now be produced in large quantities by either arc methods2,3 or thermal decomposition of hydrocarbons4,5. Here we report that pyrolysis of acetylene over iron/silica substrates is an effective method with which to produce very long, multiwalled carbon nanotubes that reach about 2 mm in length, which is an order of magnitude longer than that described in most previous reports15. The method we used is similar to that reported in ref. 5, but we have made some improvements to substrate preparation. We prepared the substrates by a solgel process using the following technique. We mixed tetraethoxysilane (10 ml) with 1.5 M iron nitrate aqueous solution (15 ml) and ethyl

Figure 1 Low-magnification scanning electron microscope images of aligned carbon nanotube arrays. a, Top view of a sample after 5 hours of growth. The triangular substrate is covered with nanotube arrays growing outwards perpendicularly from every surface of the substrate. b, A sample after 48 hours of growth. Only one nanotube array remains after stripping off other arrays from the substrate.
NATURE | VOL 394 | 13 AUGUST 1998

alcohol (10 ml) by magnetic stirring for 20 minutes. We added a few drops of concentrated hydrogen fluoride (0.4 ml), and the mixture was stirred for another 20 minutes. The mixture was then dropped onto a quartz plate to form a film 3050 m thick. After gelation, the film was dried overnight at 80 C, during which time the gel cracked into small pieces of substrate of area 520 mm2. The substrates were calcined at 450 C for 10 hours under vacuum and then reduced at 500 C for 5 hours in a flow of 9% H2 /N2 under 180 torr. At this stage, large quantities of iron/ silica nanoparticles, which acted as catalysts for nanotube growth, formed evenly on all surfaces of the substrates. Subsequently, carbon nanotubes were produced on the substrates in a flow of 9% acetylene in nitrogen at 600 C for 148 hours under 180 torr. We used this method to prepare carbon nanotubes at a very high yield. Every surface of the substrate was covered with a nanotube array composed of large quantities of highly aligned nanotubes (Fig. 1a). The lengths of the nanotube arrays increased with growth time, and reached about 2 mm (Fig. 1b) after 48 hours of growth. We believe the nanotubes could be even longer if the growth time was further increased. High-magnification scanning electron microscope images (Fig. 2a) show that the carbon nanotubes grow outwards separately and perpendicularly from the substrate to form an array. The nanotubes within these arrays are of uniform external diameter ( 2040 nm), with a spacing of about 100 nm between the tubes. Most of the nanotubes in the arrays are highly aligned, although a few appear to be slightly tangled or curved. The nanotube arrays can be easily stripped off from the substrates, and scanning electron microscope observations of the surfaces of the resulting substrates show that the iron/silica nanoparticles that were present on the surfaces of the substrates before nanotube growth disappear after such growth. But subsequent reduction of the substrates after nanotube growth resulted in the transportation of more catalyst particles to the substrate surfaces from within the bulk substrate. Therefore, the substrates can be re-used after reduction to grow new nanotube arrays with characteristics similar to the original ones. High-resolution transmission electron microscope observations (Fig. 2b) show that the multilayered graphite tubes (1030 layers) are coated with a small amount of amorphous carbon on their periphery owing to the low growth temperature (600 C) and long growth time. To determine whether the long nanotubes grow continuously from the bottom
Nature Macmillan Publishers Ltd 1998

to the top within the nanotube arrays, we cleaved some thinner nanotube bundles (Fig. 2c) from the arrays and measured their resistivity ( ) at temperatures from room temperature to 10 K by a four-probe method. The temperature dependence was semiconductor-like for most samples, but metallic behaviour was also seen ( 10 2 to 10 3 cm and d /dT 10 5 to 10 6 cm K 1 at room temperature). This resistivity is an order of magnitude larger than the values reported previously6,7, presumably because of the presence of large numbers of defects in our very long nanotubes. These results indicate that these very long nanotubes do indeed grow continuously without any interruption within the nanotube arrays. Possible uses of the nanotubes are related mainly to their dimensions and electronic8 and mechanical911 properties. The main factors that may limit their practical use are the cost, yield and quality. Very long nanotubes may be useful in, for example, the probe tips of scanning tunnelling microscopes12, field emission materials13 and nanotube-reinforced materials911. We are now

Figure 2 Scanning and transmission electron microscope images of nanotubes after 48 hours of growth. a, High-magnification scanning electron microscope image of an aligned carbon nanotube array. b, High-resolution transmission electron microscope image of a carbon nanotube. c, Scanning electron microscope image of a thin nanotube bundle cleaved from a nanotube array along the growth direction of the nanotubes. The bundle contains about 20 nanotubes. The total length of the

631

scientific correspondence
attempting to measure the tensile strength of very long nanotubes.
Z. W. Pan, S. S. Xie*, B. H. Chang, C. Y. Wang, L. Lu, W. Liu, W. Y. Zhou, W. Z. Li
Institute of Physics, Chinese Academy of Sciences, Beijing 100080, China * e-mail: user412@aphy01.iphy.ac.cn

L. X. Qian
Department of Physics, Central University of Nationalities, Beijing 100081, China
1. 2. 3. 4. 5. 6. 7. 8. Iijima, S. Nature 354, 5658 (1991). Ebbesen, T. W. & Ajayan, P. M. Nature 358, 220222 (1992). Bethune, D. S. et al. Nature 363, 605607 (1993). Endo, M. et al. Carbon 33, 873881 (1995). Li, W. Z. et al. Science 274, 17011703 (1996). Ebbesen, T. W. et al. Nature 382, 5456 (1996). Fisher, J. E. et al. Phys. Rev. B 55, 49214924 (1997). Dai, H. J., Wong, E. W. & Lieber, C. M. Science 272, 523526 (1996). 9. Treacy, M. M. J., Ebbesen, T. W. & Gibson, J. M. Nature 381, 678680 (1996). 10. Wong, E. W., Sheehan, P. E. & Lieber, C. M. Science 277, 19711975 (1997). 11. Falvo, M. R. et al. Nature 389, 582584 (1997). 12. Dai, H. J. et al. Nature 384, 147150 (1996). 13. de Heer, W. A., Chatelain, A. & Ugarte, D. Science 270, 11791181 (1995).

Switch from specialized to generalized pollination

The once prevalent view that the evolution of extreme ecological specialization is accompanied by a loss of potential for adapting to new conditions, and thus is irreversible1-4, has been challenged by several recent examples1,2,5. However, we know of no modern phylogenetic studies showing reversal in pollination relationships from extreme specialization to generalization, although such reversals are theoretically expected6,7. Here we present molecular phylogenetic evidence for an evolutionary shift in Dalechampia (Euphorbiaceae) vines from a highly specialized relationship (pollination by one or a few animal species2,7) with resin-collecting bees to generalized pollination by a variety of pollen-feeding insects. This shift was associated with dispersal from Africa to Madagascar, where the specific resin-collecting pollinators are absent. These results show that plants dispersing beyond the range of their specific pollinators may succeed by evolving more generalized pollination systems. Only a few genera of bees in the families Megachilidae and Apidae use floral resin in nest construction8,9, and only two genera of plants, Dalechampia and Clusia (Clusiaceae), are known to secrete resins that attract pollinators8. Partly because the reward they offer is non-nutritive, Dalechampia blossoms that secrete resin are pollinated by only one or two species of bees at any one location, and so depend on
632

Figure 1 A phylogenetic hypothesis for Dalechampia sects. Dalechampia and Tiliifoliae based on maximum parsimony (MP) analysis of combined nuclear ribosomal (ITS-1, 5,8S, ITS-2) and chloroplast (trnK intron) DNA sequences. The strict consensus of MP trees is depicted, with bootstrap values (above 50%) along branches. The tree shown is part of a larger tree from an analysis including 16 representatives of the other sections of Dalechampia and two species from candidate sister genera (Plukenetia and Tragia). Evolution of pollination ecology was mapped using MP onto the tree based on the observed pollination of each species. The species are endemic to the regions indicated except D. parvifolia, which occurs in both Africa and Asia.

specific pollinators. In eastern and southern Africa, for example, Dalechampia populations are usually pollinated by only one species of Pachyanthidium or Heriades (Megachilidae)9,10. Recent fieldwork in Madagascar has shown that the species of Dalechampia found there offer only pollen as a reward for pollinators, and that most are pollinated by a variety of pollen-feeding insects, including beetles (Cerambycidae, Scarabidae), muscoid flies (Diptera) and several bees (Halictidae, Anthophoridae, Apidae). Open presentation of a common food reward (pollen) results in interactions with numerous pollinators10, a finding typical of plants with open flowers6. Phylogenetic analysis of combined nuclear ribosomal and chloroplast DNA data sets, and mapping of pollination and morphological traits onto the molecular tree, indicate that Malagasy species of Dalechampia are descended from an ancestor pollinated by resin-collecting bees (Fig. 1). These results also indicate that Dalechampia colonized Madagascar from Africa (Fig. 1). This finding is further supported by morphological data and biogeographical considerations10. But why did the Malagasy colonists abandon efficient, specialized pollination by resin-collecting bees and switch to a generalized pollination system? It seems that resin-collecting megachilid bees, which are the only pollinators of Dalechampia in Africa9, failed to colonize Madagascar10. The ancestral Dalechampia colonists of
Nature Macmillan Publishers Ltd 1998

Madagascar were probably pollinated incidentally by other pollen-feeding insects. They subsequently adapted to the absence of their specific pollinators by losing the gland that secretes the resin reward and by effectively using diverse pollen-feeding insects as pollinators. These changes were sufficiently successful to allow secondary diversification on the isolated island of Madagascar.
W. Scott Armbruster
Institute of Arctic Biology, University of Alaska, Fairbanks, Alaska 99775, USA and Department of Botany, Norwegian University of Science and Technology, N-7034 Trondheim, Norway e-mail: ffwsa@uaf.edu

Bruce G. Baldwin
Jepson Herbarium and Department of Integrative Biology, University of California, Berkeley, California 94720, USA
1. Futuyma, D. J. & Morena, G. Annu. Rev. Ecol. Syst. 19, 207233 (1988). 2. Thompson, J. N. The Coevolutionary Process (Univ. Chicago Press, 1994). 3. Smith, A. B. & Jeffery, C. H. Nature 392, 6971 (1998). 4. Givnish, T. J., Sytsma, K. J., Smith, J. F. & Hahn, W. J. in Hawaiian Biogeography: Evolution on a Hot Spot Archipelago (eds Wagner, W. L. & Funk, V. A.) 288337 (Smithsonian Institution Press, Washington D C, 1995). 5. Lanyon, S. M. Science 255, 7779 (1992). 6. Proctor, M., Yeo, P. & Lack, A. The Natural History of Pollination (HarperCollins, London, 1996). 7. Waser, N. M., Chittka, L., Price, M. V., Williams, N. M. & Ollerton, J. Ecology 77, 10431060 (1996). 8. Armbruster, W. S. Am. J. Bot. 71, 11491160 (1984). 9. Armbruster, W. S. & Steiner, K. E. Am. J. Bot. 79, 306313 (1992). 10. Armbruster, W. S. et al. Natl Geogr. Res. Expl. 9, 430444 (1993).

NATURE | VOL 394 | 13 AUGUST 1998

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.