Table of Contents

Rolled graphic sheets form lattices of tubular "chicken coops" for lithium ions and hydrogen molecules.

Carbon nanotubes are scientifically fascinating objects. The electrical, thermal, mechanical, and chemical behavior of individual tubes forces us to rethink the fundamental properties of quasi- one-dimensional nanoscale objects. Applications already in advanced development include cold-cathode flat-panel displays, microwave sources, and overvoltage protection, all based on their excellent field-emission characteristics.

Preliminary performance results in the areas of hydrogen storage and solid-state electrochemistry show promise for innovations in energy technology, but they also reveal many problems remaining to be solved. Hydrogen storage devices and rechargeable lithium-ion batteries are two energy technologies in which nanostructured bulk carbon single-wall nanotube (SWNT) materials can potentially outperform other forms of carbon. However, in contrast to nanoelectronic devices, the showstopper for these macroscopic applications will be materials cost. We need to discover and develop scalable, continuous, low-energy growth chemistries from inexpensive feedstocks.

From chicken wire to nanotubes

Carbon nanotubes were discovered in 1991 as a minor byproduct of fullerene synthesis (1). Remarkable progress has been made in the ensuing nine years, including the discovery of two basic types of nanotubes (single-wall and multiwall), great strides in synthesis and purification, elucidation of the fundamental physical properties, and important steps toward realistic, practical applications.

A SWNT can be envisioned as a narrow strip of nanoscale graphene “chicken wire”, with a carbon atom at each apex of a hexagonal array and 0.14 nm between neighboring carbons, rolled up in a seamless cylinder 1–10 nm in diameter and as long as several micrometers. (“Graphene” refers to a single sheet of linked carbon atoms having the structure found in graphite.) Multiwall nanotubes (MWNTs) consist of many such nested cylinders, like Russian matroyska dolls, with 0.34 nm between individual tubes. Because the length and width of the chicken-wire strip are arbitrary, so are the lengths and diameters of the tubes. Furthermore, the strip need not be cut along a high-symmetry direction, so chiral and achiral tubes are possible (Figure 1).

One of the major goals of nanotube research is to be able to separate tubes of specific diameters, lengths, and axial symmetries, a prerequisite for the kind of chemistry that has been so fruitful with the fullerenes. In the meantime, there are many opportunities for chemical innovation using bulk materials containing a mixture of tube types.

The underlying science and technological potential of SWNTs and MWNTs can be studied profitably at the scale of individual tubes and macroscopic assemblies. Single-tube electronic devices, including diodes, single-electron transistors, and field-effect transistors, have been demonstrated and may play a role in molecular electronics architectures (2). The large aspect ratio of individual tubes has also been exploited in scanning-probe microscopy, whereby the details of deep trenches in silicon wafers are readily observed (3). Chemically specific tips have been achieved by attaching ligands to the ends of the nanotube probes (4). These single-tube technologies rely on the unique electronic, mechanical, and chemical properties intrinsic to the rolled graphene sheet and are not likely to be severely limited by material cost and scale-up issues. However, the large-scale directed assembly of nanotube devices into molecular “integrated circuits” remains a formidable challenge (2).

On the macroscopic end of the spectrum, if bulk nanotube materials are to perform competitively with respect to other materials, it is necessary to design and control structure and morphology at the nanoscale level. Figure 2 illustrates the range of morphologies that exist in as-grown SWNT “soot”. The as-grown material contains chiral tubes (one of these is shown in the top micrograph) and high-symmetry tubes in which some of the C–C bonds are parallel to or perpendicular to the tube axis. Zooming out, the middle panel shows a semicrystalline bundle, or “rope”, consisting of several SWNTs of similar diameter, packed into a two-dimensional triangular lattice like a box of drinking straws (5). These occur naturally during synthesis, and their number and size can be increased by annealing (6). Zooming farther out, the bottom panel shows typical bulk material consisting of a highly porous, tangled network of ropes with sparse interconnections and lots of empty volume.

The basic strategy behind macroscopic applications consists of disassembling the bulk material into constituent ropes or tubes, and then reassembling them in a controlled manner to meet a specific design criterion. For example, aligning the axes of a layer of tubes or ropes perpendicular to the substrate should optimize performance for cold-cathode field electron emission (2). Density and porosity can be varied by orders of magnitude, depending on how the solid is reassembled from dispersed tubes and ropes (7). The empty space between tubes in a rope is large (on the scale of atoms and small molecules), which suggests the possibility of creating new guest– host “compounds” analogous to intercalated graphite, doped conjugated polymers, and alkali fullerides

For more information see, A history of carbon-based intercalation compounds.

Hydrogen storage

Hydrogen-based transportation has the potential to reduce carbon monoxide, carbon dioxide, and nitrogen oxide (NO_x) emissions. Hydrogen offers an alternative fuel source in the event of a petroleum shortage. Advances in several areas are necessary before on-board hydrogen storage can be applied to combustion or fuel cell systems. In particular, achieving the 6.5 wt% stored hydrogen of the federal Clean Car Project requires major advances in storage density, energy efficiency, safety, and cost. Most, if not all, of these issues would be resolved by a light-weight material capable of reversibly storing and releasing hydrogen in a modest range of near-ambient temperatures and pressures. Carbon nanotube materials are obvious candidates; the middle panel of Figure 2 shows why. If the ends of the SWNTs constituting a rope could be opened, then hydrogen molecules could, in principle, be physisorbed inside the tubes and in the large open channels between the tubes, as shown in Figure 3 .

Calculations based on this model indicate that the 6.5 wt% goal is realistic; hydrogen can be physisorbed on the interior faces of the nanotubes; and it can form a liquidlike core inside the nanotubes. The experimental situation is at once promising, confusing, and incomplete. At least two measurements of storage density at 300 K confirm that the capacity target is within reach using SWNT materials (9, 10). Other reports on what we assume are similar materials imply that useful storage densities are achieved only at cryogenic temperatures (11). Yet another claim of 20 wt% storage in alkali-metal-doped nano tubes (12) may have been compromised by traces of water (13). Even more dramatic claims concerning “graphitic nanofibers” (14) are not reproducible (15).

None of these experiments address the local structure of the binding sites, so it is unclear how the results connect to theoretical predictions. Furthermore, experiments are performed on real-world material, whereas calculations are based on idealized models. Binding energies of physisorbed hydrogen can be obtained by thermal analysis (temperature-programmed desorption, for example), but these do not provide atomic-scale geometric information about the binding sites. Furthermore, experimental binding energies are generally much larger than values calculated for the idealized structure. What we need is a spectroscopic technique that is sensitive to the local environment in which the hydrogen molecule is trapped.

Neutron cue-ball probes

One such technique is the inelastic scattering of thermal neutron projectiles directed at the trapped hydrogen. A pool player knows instinctively that a cue ball (or neutron) hitting another ball (free hydrogen molecule) will undergo a kinetic energy loss ranging from essentially zero up to its initial energy, depending on the recoil angles of target and projectile. The spectrum of neutron energy losses will thus be continuous and featureless.

On the other hand, when hydrogen is confined to a specific location (as in crystalline solid hydrogen or the interior of a nanotube), the only possibility for energy transfer from neutron to hydrogen must involve the internal vibrations and rotations of the two protons. The quantum physics of the collision process tells us that neutron energy loss will occur in discrete steps by inducing transitions between quantized rotational energy levels of the hydrogen dumbbell, by converting from the para form (antiparallel spins on the two protons) to the ortho form (parallel spins). It also tells us that these levels are shifted and split by the geometry of the confining environment, such that in principle we should be able to determine how much hydrogen is trapped in multiple sites of different symmetry. Furthermore, the temperature dependence of the energy-loss peaks associated with different sites is a direct measure of the binding energy, a quantity that can be directly compared with theory.

So far, this all sounds like ordinary optical spectroscopy, except that the magnetic moment of the neutron is the “driving force” of the ortho-to-para transition. A proof-of-principle experiment was recently performed at the National Institute of Standards and Technology (NIST) Center for Neutron Research in Gaithersburg, MD (16). A 0.6-g sample of as-grown SWNT material was provided by the National Renewable Energy Laboratory (NREL) in Golden, CO (9). Hydrogen gas was introduced directly from a standard cylinder at a pressure of 11 MPa and then cooled to 25 K, an admittedly unrealistic loading condition for practical applications. The “spectros copy” consists of measuring the neutron intensity scattered from the sample as a function of the neutron energy loss. This technique is analogous to Raman spectroscopy, which uses photon projectiles, except that photons cannot induce rotational transitions because they have no spin.

Figure 4 shows the temperature dependence of the neutron energy loss spectra, which tell us three things. First, the discrete loss peak at about 14.5 meV means that the hydrogen is indeed physisorbed at these temperatures, pinned at specific sites and free to rotate but not to translate. The small shift relative to the 14.7-meV value expected for rotations of the free molecule means that the binding forces are not particularly strong, similarly to hydrogen in the pure molecular solid (17) or physisorbed on graphite (18). Second, the peak width exceeds the instrumental resolution, suggesting either a low- symmetry binding site or, more likely, multiple binding sites with similar environments because the sample consists of a mixture of tube types. Finally, the decrease in peak intensity as the sample is warmed from 25 K signals the thermally activated desorption of weakly physisorbed molecules. The details of the temperature dependence suggest that the binding energies in this sample are slightly higher than for hydrogen on flat graphite and considerably weaker than the theoretical predictions pertinent to the idealized model treated in Figure 3.

I consider this proof-of-principle experiment to be a success (even though the test material and loading conditions are of no practical interest), because it demonstrates that rotational spectroscopy using thermal neutrons provides important microscopic information about hydrogen-binding sites in nano-tube materials. The next step is to study nanostructured SWNT materials that have been optimized to store high densities of hydrogen at temperatures closer to ambient. The good news for future studies is that the NIST spectrometer has just been upgraded to afford a 20-fold sensitivity increase, meaning that data comparable to those in Figure 4 can be obtained from much smaller samples of experimental materials, or that many samples obtained with different processing conditions can be screened quickly.

Lithium-ion batteries

Recent developments in rechargeable lithium-ion battery technology have been dominated by efforts to replace metallic lithium electrodes with lithium–carbon guest–host compounds for improved safety and cyclability (17). “Rocking chair” or “shuttlecock” batteries shuttle Li⁺ between two guest–host solids serving as cathode and anode, preventing the reduction of Li⁺ to dangerous metallic lithium at any point in the charge– discharge cycle. Although today’s battery manufacturing is dominated by highly graphitic carbon anodes, alternatives such as “amorphous” carbons, which result from pyrolysis of organic and inorganic precursors, offer some attractive features (18). A major goal is to achieve higher lithium capacities in the carbon anode to partly offset the weight penalty of the carbon.

The maximum Li/C ratio in graphite is limited by the crystal structure of the lithium-saturated intercalation compound LiC₆, in which one layer of lithium is associated with one layer of graphene. If the layers could be separated, then in principle one could “butter the toast” on both sides, doubling the Li/C ratio at the expense of introducing nanoscale porosity to minimize electrostatic repulsion between neighboring Li–C–Li sandwiches. Figures 2 and 3 show why nanotubes are attractive in this context: As with hydrogen storage, if the interstitial channels in the rope lattice and the tube interiors are accessible, we might achieve the same lithium density on the outside as with graphite, while at the same time filling the interior cavity with more than a monolayer. Thus, capacities more than twice that of graphite should be achievable.

Before pursuing that goal, we must realize that a prerequisite for rechargeable battery anodes is reversible insertion and extraction of lithium in the carbon host. This has been demonstrated for SWNT materials by several groups. Our first attempts, made using unpurified material, were unsuccessful; they showed large irreversible capacity and little or no revers ible capacity. Much better results were obtained with purified material (6, 19). So far, we have obtained reversible lithium capacities that are about 25% greater than that of graphite.

A team at the University of North Carolina, Chapel Hill, has achieved better success by studying the effects of ball-milling after a simpler purification procedure (20). The team was able to produce material with about twice the lithium capacity of graphite. These two results are promising, but the details of the charge–discharge behavior tell us that a lot more needs to be done for these materials to be of practical interest. Furthermore, as in the hydrogen storage problem, there is no hard evidence to convince us that we are in fact exploiting the interior and exterior sites implied by the idealized models.

In Figures 5 and 6, I compare the electrochemical performance of powdered graphite and purified SWNT, with Li⁺ as the mobile ion. This is a so-called half-cell experiment; in a real battery, the lithium would be replaced by another lithium guest–host compound, such as LiCoO₂. The schematic diagram on top shows the situation during the half-cycle in which lithium “intercalates” into the carbon because of the large potential difference between lithium metal and pure carbon. For graphite, most of the action occurs near zero potential difference, ensuring a large and fairly constant output voltage during the discharge of a real battery. The first insertion half-cycle consumes slightly more lithium than is required to achieve the saturated LiC₆. Next, we impose a current in the opposite direction to drive the lithium out of the carbon, but a small fraction remains after we recover the initial cell potential of 2 V. This “irreversible capacity” is small, about 10% of the total, and is mostly limited to the first complete cycle. Subsequent cycles are reversible, and no hysteresis occurs between charge and discharge, two important attributes for practical batteries.

The major drawback of graphite (apart from cost) is illustrated in the inset in Figure 5. The sequence of steps and plateaus indicates that the reversible progression between graphite and LiC₆ occurs through a series of transitions between different stoichiometric compounds. Because these involve changes in volume, the graphite crystallites are subject to continuous expansion and contraction as the battery is cycled, which eventually leads to significant degradation of long-term cyclability.

SWNTs behave quite differently (Figure 6). The first lithium insertion is accompanied by a large irreversible capacity, more than twice the reversible part. In a real battery, this would require tripling the amount of LiCoO₂ required to realize the full reversible capacity on repeated cycling, a clearly unacceptable situation. Unlike graphite, significant irreversi bility persists upon further cycling. Worse yet, lithium insertion and removal occur over a wide range of potentials, which means that the working battery voltage would vary continuously as the battery is discharged. Finally, there is a large hysteresis between charge and discharge, which means that the battery could not simply be recharged at constant voltage.

Nevertheless, the prospect of doubling reversible capacity with respect to graphite (using methods such as ball-milling) is reason enough to try to solve the problems of hysteresis and irreversible capacity. Several solutions are suggested by similar research on disordered carbons (21). A possible saving grace is that, unlike graphite, SWNTs maintain high lithium capacity at high current densities, a property that will be important if the lithium-ion battery market is extended to high-power applications such as electric vehicles and load-leveling devices.

It is not obvious whether energy storage applications using carbon nanotubes will ever be commercially attractive. Efforts to optimize these materials for specific purposes are just getting underway. New methods for nanotube synthesis appear frequently in the literature. Many of these methods are directed to achieving economical large-scale production. The first commercially available fullerenes were priced at $1000/g, and the purity was far from reagent grade. Today, one can buy 99.9% pure C₆₀ for $70/g, but it is still made by a very high-energy batch process. Viable new technologies may emerge with the convergence of successful nanoscale “tailoring” and cost reduction.

Acknowledgments

The neutron spectroscopy experiment is a good example of “big-science–small-science” collaboration, for which I gratefully acknowledge the contributions of friends at NIST and NREL, especially Dan Neumann, Craig Brown, Taner Yildirim, and Mike Heben. The lithium-ion battery work at the University of Pennsylvania is taken from the Ph.D. dissertation of Agnes Claye, to whom I wish all the best. Financial support was provided by the U.S. Department of Energy, the National Science Foundation’s Materials Research Science and Engineering Centers Program, and the New Energy Development Organization.

References

1. Iijima, S. Nature 1991, 354, 56–57.
2. Physics World 2000, 13 (6).
3. Dai, H.; Hafner, J. H.; Rinzler, A. G.; Colbert D. T.; Smalley, R. E. Nature 1996, 384, 147–149.
4. Wong, S. S.; Woolley, A. T.; Joselevich, E.; Lieber, C. M. Chem. Phys. Lett. 1999, 306, 219–221.
5. Thess, A.; Lee, R.; Nikolaev, P.; Dai, H.; Petit, P.; Robert, J.; Xu, C.; Lee, Y. H.; Kim, S. G.; Rinzler, A. G.; Colbert, D. T.; Scuseria, G. E.; Tomanek, D.; Fischer, J. E.; Smalley, R. E. Science 1996, 273, 483–487.
6. Rinzler, A. G.; Liu, J.; Nikolaev, P.; Huffman, C. B.; Rodriguez-Macias, F. J.; Boul, P. J.; Lu, A. H.; Heymann, D.; Colbert, D. T.; Lee, R. S.; Fischer, J. E.; Rao, A. M.; Eklund P. C.; Smalley, R. E. Appl. Phys. A 1998, 67, 29–38.
7. Smith, B. W.; Benes, Z.; Luzzi, D. E.; Fischer, J. E.; Walters, D. E.; Casavant, M. J.; Schmidt, J.; Smalley, R. E. Appl. Phys. Lett. 2000, 77, 663–665.
8. Fitzgerald, S. A.; Yildirim, T.; Santodonato, L. J.; Neumann, D. A.; Copley, J. R. D.; Rush, J. J.; Trouw, F. Phys. Rev. B 1999, 60, 6439–6453.
9. Dillon, A. C.; Jones, K. M.; Bekkedahl, T. A.; Kiang, C. H.; Bethune, D. S.; Heben, M. J. Nature 1997, 386, 377–379.
10. Liu, C. Y.; Fan, Y.; Lu, M.; Cong, H. T.; Cheng, H. M.; Dresselhaus, M. S. Science 1999, 286, 1127–1130.
11. Ye, Y.; Ahn, C. C.; Witham, C.; Fultz, B.; Liu, J.; Rinzler, A. G.; Colbert, D.; Smith, K. A.; Smalley, R. E. Appl. Phys. Lett. 1999, 74, 2307–2309.
12. Chen, P.; Wu, X.; Lin, J.; Tan, K. L. Science 1999, 285, 91–94.
13. Yang, R. T. Carbon 2000, 38, 623–641.
14. Park, C.; Anderson, P. E.; Chambers, A.; Tan, C. D.; Hidalgo, R.; Rodriguez, N. M. J. Phys. Chem. B 1999, 103, 10572–10578.
15. Ahn, C. C.; Ye, Y.; Ratnakumar, B. V.; Witham, C.; Bowman, R. C.; Fultz, B. Appl. Phys. Lett. 1998, 73, 3378–3380.
16. Brown, C. M.; Yildirim, T.; Neumann, D. A.; Heben, M. J.; Gennett, T. A.; Dillon, C.; Alleman, J. L.; Fischer, J. E., submitted for publication.
17. Dahn, J. R.; Zheng, T.; Liu, Y.; Xue, J. S. Science 1995, 270, 590–593.
18. Claye, A. S.; Fischer, J. E. Electrochim. Acta 1999, 45, 107–120.
19. Claye, A. S.; Fischer, J. E.; Huffman, C. B.; Rinzler, A. G.; Smalley, R. E. J. Electrochem. Soc. 2000, in press.
20. Gao, B.; Kleinhammes, A.; Tang, X. P.; Bower, C.; Fleming, L.; Wu, Y.; Zhou, O. Chem. Phys. Lett. 1999, 307, 153–157.
21. Buiel, E.R.P.; Dahn, J. R. J. Electrochem. Soc. 1997, 145, 1977–1984.
22. www.pa.msu.edu/cmp/csc/nanotube.html, Delft University of Technology: Cees Dekker’s home page (accessed July 18, 2000)
23. http://cnst.rice.edu/images/ropes_1c.jpg (accessed July 18, 2000).
24. http://cnst.rice.edu/tubes/sem.html (accessed July 18, 2000).

John E. Fischer is a professor of materials science and engineering at the University of Pennsylvania (3231 Walnut St., Philadelphia, PA 19104-6272; 215-898-6924; fischer@sol1.lrsm.upenn. edu). He received his B.S. degree in mechanical engineering from the Rensselaer Polytechnic Institute, his M.S. degree in mechanical engineering from the California Institute of Technology, and his Ph.D. in nuclear science and engineering from the Rensselaer Polytechnic Institute. He was a postdoctoral fellow in solid-state physics at École Normale Supérieure, Paris; a physicist and head of semiconductor physics at Michelson Laboratory, China Lake, CA; and associate professor and professor of electrical engineering at the University of Pennsylvania. He was a Fulbright–Science Research Council visitor at the Cavendish Laboratory, Cambridge, U.K., and he has been a visiting professor at several universities. He researches the solid-state physics and chemistry of intercalation compounds, doped conjugated polymers, ful lerenes, fullerides, nanotubes, and disordered carbons as anodes in recharge able lithium batteries.