Electron diffraction patterns are consistent with a single wall structure. Van der Waals interactions cause these structures to form various aggregates, which Dr. Iijima has given names like "dahlia flower," "bud," and "seed." The details of bucky horn structures have not yet been worked out; the expectation is that pentagon rings of carbon atoms would be necessary to achieve the conical shape, but so far examination by AFM has shown only the basic graphene hexagon. One unique property of these structures is that they are stable at 1200 °C in a vacuum.
Dr. Iijima finished by describing another class of structures produced by laser ablation, which he and his colleagues have recently published in Science [281, 973-975, 14Aug98]. These structures resemble nanoscale coaxial cable, with a core of beta-phase silicon carbide, a semiconductor material, surrounded by amorphous silicon oxide, an insulator, surrounded by outer layers of graphitic boron nitride or carbon, which could be metallic or semiconducting. These structures are a few tens of nm in diameter, with lengths of up to 50 µm. Such structures suggest applications in nanoscale electronic devices.
Dr. Ching-Hwa Kiang (of UCLA) described the history and current status of carbon nanotube research.
In 1993 Dr. Kiang and Dr. Iijima independently discovered single-walled carbon nanotubes. In Dr. Kiang's experiment, cobalt was added to a C60 generation apparatus in an attempt to encapsulate it in the C60. Instead the result was to change the soot-like C60 deposit into a rubbery nanotube mat. Similar effects are seen with nickel. TEM imaging showed ropes of nanotubes. The ropes proved strong. Pulling a mat apart left aligned tubes stretched across a 2-micron gap. The strain in SWNTs is substantially lower than in spherical fullerenes, 0.02-0.03 eV in SWNTs compared to 0.45 eV in C60 and 0.15 eV in C240.
The original cobalt catalyst gave diameters distributed from 1-1.5 nanometers. Adding "promoters" such as lead and sulfur added a long tail to the distribution, with diameters up to 5 nanometers.
Dr. Kiang proposes that the role of the catalyst is to stitch together carbyne rings (-CC-)n. In this model, the (m,n) indices of the tube may be set when the initial ring is formed. The transition metal catalyst may expedite cis-trans isomerization in double bonds in the initial nanotube seed.
Dr. Kiang sees the primary challenges in current SWNT research as being:
Professor Cees Dekker (from Delft) described experimental work on quantum conduction in single wall carbon nanotubes.
Normally, plausible 1-D wires undergo a Peierls distortion, which opens up a gap in the possible electron energies at the Fermi level and turns the wires into semiconductors. In carbon nanotubes the stiffness from all of the other bonds is sufficient to stabilize the tube against this distortion and retain the conduction.
In ordinary graphite, the valence band and conduction band just barely touch at six points, vertices of a hexagon in reciprocal space. The effect of wrapping the graphite up into a tube is to quantize the angular momentum the electron can have around the axis of the tube. This effectively draws lines of allowed momenta on the band diagram of graphite, and the angle that the tube axis makes with the graphite lattice sets the angle that these lines make with the hexagon of zero-band-gap points in graphite's band diagram. In armchair tubes, one allowed line goes directly through two of the zero-band-gap points, and this creates the two bands that give metallic conduction in these tubes. For more complex reasons, one third of other possible nanotubes are also metallic, while the rest have a finite band gap and are semiconductors.
Dekker's group has used atomic resolution STM images of nanotubes on a gold substrate to directly see the angle between the lattice and the tube axis. In this experiment they can also measure I(V) curves of tunnel current. These curves can distinguish semiconducting tubes from metallic ones. The I(V) curves have kinks where new bands become accessible, so gaps between these kinks correspond to gaps between bands. The semiconducting tubes show ~0.6 volt gaps, while the metallic tubes show ~1.8 volt gaps (displaying the gap between the bands just above and below the bands which cause conduction). The diameter of the semiconducting tubes could also be shown to correctly predict the size of the bandgap seen in these tubes, connecting individual tubes' STM images with their electronic properties. By moving the STM probe while examining the I(V) curves, one could even see the shape of the individual electronic wavefunctions, testing the band theory of nanotubes on a very fine grain.
On a finer voltage scale, coulomb blockades were also observed, with nanotubes playing the role of the conducting island, with conductance peaks ~2 mV apart. For this experiment, the nanotube was deposited over two electrodes on top of an SiO2/Si substrate. Unfortunately, the insulating substrate requires use of AFM, rather than STM, imaging here, so the individual chiral angles of the tubes aren't available. In these experiments, the voltage of the silicon substrate could also be used to modulate the coulomb blockades. At much higher bias voltages of ~1.5 volts the substrate voltage could vary the conduction of a semiconducting nanotube by a factor of 106, in an analog to a BARITT transistor.
At very fine voltage scales, energy separations due to quantization of momentum along the tube can be seen. A 3 micron tube showed energy gaps of ~0.4 millivolts, implying coherent conduction across essentially the entire length of the tube.
Experiments have now been done which stretch a nanotube across a grid of 7 electrodes. This has permitted 4-point probes of the resistance of the nanotube, separating the contact resistance between electrodes and nanotubes from the internal resistance of the nanotubes themselves. It also permits measurements of the coulomb blockades between several pairs of electrodes. These proved similar, but not identical, which is currently attributed to barriers in the nanotube due to bending as it passes over the raised surfaces of the electrodes above the substrate.
Prof. Donald Brenner of North Carolina State University spoke on the implications for nanoscale device design of the predicted electronic properties of strained fullerene nanotubules.
Prof. Brenner began with the basic observations that
All of the deformations were found to influence the band gaps and conductivity of the nanotubes. One result that emerged was that the density of electronic states was greatly changed by buckling of the tube. One measure of this increased reactivity is chemisorption at the site of the kink as a result of the electronic state of the carbon atom being changed from sp2 to sp3 with an unpaired electron. A consequence is that covalent bonding to the site of kink converts a carbon atom from sp2 to sp3. Pulling the electron from the p electron system divides the p system into two halves, which could be metallic on one side and semiconducting on the other side. The metallic/semiconductor junction would attenuate over 4.5 Å.
Professor Boris Yakobson (from NCSU) described modeling irreversible deformation of carbon nanotubes in terms of the creation and movement of dislocations.
At high strain rates, simulations show nanotubes withstanding strains up to 20-30% before breaking. In principle, even low strains make a nanotube metastable. An armchair tube can tolerate ~100GPa for ~10 years. Molecular dynamics simulations of nanotubes under tension show formation of a "Stone-Wales" defect, a rotated pair of carbon atoms, initiating irreversible deformation. The creation of these defects depends strongly on the indices of the tube. In armchair (n,n) nanotubes, a third of the bonds are perpendicular to the tube axis. A Stone-Wales defect rotates a bonded pair by 90 degrees, (turning the 4 hexagons adjacent to the pair into 2 pentagons and 2 heptagons), so rotation of one of these perpendicular bonds to lie parallel to the axis accommodates the strain. This is energetically favored for strains > ~5%.
The energy of these bond rotations has a sinusoidal dependence on the chiral angle of the tube, being most favored for armchair tubes and least favored for zig-zag tubes. The analytical form of the energy can be regarded as an approximation to the results from atomistic analysis of the whole tube. It matches the more detailed results to within a few percent.
The Stone-Wales defect can be viewed as a pair of dislocations, each containing one pentagon and one heptagon. Under mild strain, at high temperatures (as under nanotube synthesis conditions) these dislocations can glide apart from each other. The lines along which the dislocations can glide are along zigzag directions in the hexagonal lattice. In a nanotube glide lines wrap into helices. When a dislocation glides past a section of tube, it stretches it, necks it down, and changes its indices. The change in indices can produce metal/semiconductor and semiconductor/semiconductor junctions.
At strain rates too high to permit the gliding of dislocations or for large armchair tubes where formation of Stone-Wales defects is energetically unfavorable, nanotubes are predicted to fail by brittle fracture. In atomistic simulations, fracture is shown by the formation of progressively larger rings from the initial defect. Effectively, the defect acts like the start of a widening crack rather than like a pair of migrating dislocations.
Professor Reed described three classes of experiments that his group has done. The first class used mechanical controllable break junctions (MCBs), thin wires mounted on beams that could be bent. This bending put the wire in tension, allowing it to be broken. The separation of the two halves is stable to 2 picometers. The beam can be bent either with a piezoelectric actuator or with an electrostatic actuator (which achieves the stability record). The beam bending scales down the movement of the actuator by 103, giving the very fine control of electrode separation. As the wire is first broken, one can see a very clean stepwise reduction of the conductance of the junction, in steps close to the conductance for one quantum channel, G0=2 e2/h = (13 kohm)-1, beyond which the conductance drops by 103-104 in the tunneling regime.
By using gold electrodes and adsorbing organic compounds terminating in thiol groups between them, MCBs could show conduction through individual organic molecules. For example, one could see conduction through benzene-1,4-dithiol molecules. The benzene ring acts like a conducting island in a single electron transistor (SET), with a coulomb blockade gap of ~0.7 volts. Unlike SETs with lithographically defined islands, this system works at room temperature. The conduction is consistent with a model where the SET has a resistance of 22 megaohms and an island capacitance of 1.1 X 10-19 farads.
The second type of experiment used a combination of e-beam lithography and anisotropic etching to build a ~50 nm diameter "nanopore", exposing a gold electrode, which could then be coated with a self-assembled layer of a thiol-terminated organic compound, and covered with an evaporated metal top electrode. This structure was built in order to allow measurement of temperature dependent conduction, which helps distinguish between the many possible mechanisms for conduction. The contact is made small in order to avoid defects, which give competing conducting paths. This structure can place molecules in asymmetrical environments, using gold as one electrode and a different metal (typically titanium) as the other. The Au/S bond acts as a Schottky barrier with a height of 0.7 eV, and the Ti/organic as one with height 0.25 eV.
The nanopore experiments have been used to examine a wide variety of molecules. For example, a bi-phenyl thiol showed rectification. There are plans to try to build ohmic contacts by adjusting the work function of the organic molecule, possibly using phenyl diisocyanide. Another molecule that has been used has three phenyl groups separated by acetylenic bridges. At low temperatures (<20K) this molecule packs on a gold surface in a pattern that rotates the rings so that they aren't coplanar. This suppresses conduction through the material. Increasing the temperature permits the rings to rotate, so they can sample coplanar, conjugated configurations. This raises conduction 100-fold. Professor Reed suggested that this might yield a mechanoelectric switch, a new type.
The third type of experiment was described briefly. In it, a nonconducting alkanethiol monolayer self-assembles on a gold surface. An STM pulse is then used to break holes in this layer at selected positions. A solution of a conducting thiol adds molecules into the gap, and their conductance can be observed. This process's resolution appears to be limited by tip shape, and it is planned to try SWNT tips.
Dr. Massimiliano Di Ventra of Vanderbilt University presented a parameter-free first-principles calculation of the I-V characteristics of a molecular device, the benzene-1,4-dithiolate molecule for which the I-V characteristics have been measured by Reed et al. The calculations could be done for molecules of up to 60 atoms, and are limited by available computer memory. For details:
Dr. Eldon Emberly of Simon Fraser University reported on two different methods to model electron transport in molecular wires. In one application of these methods, conductance was calculated for different geometrical configurations of 1,4 benzene-dithiolate (BDT) bonded to two gold nanocontacts. A second application considers what happens when electrons are totally reflected from a molecular wire. For details:
Dr. Toshishige Yamada of MRJ, NASA Ames Research Center reported calculations on the electronic properties of small chains of atoms (adatom chains) upon bonding to a substrate to provide mechanical stability for the atom chain. For details:
Professor Satyam Priyadarshy (from U. of Pittsburgh) gave a talk on conduction in DNA.
Professor Priyadarshy noted that DNA, besides its genetic role, is useful in nanotechnology for scaffolding and recognition. It would be very useful to make electronic devices from it as well. Roughly 50 papers have been published about DNA conductivity in the last 5 years.
A chemist's view of the archetypal conduction experiment is to build a donor-bridge-acceptor structure and measure the rate of electron transfer from the donor to the acceptor.
In the tunneling regime, the rate of electron transfer and, more generally, the conductivity, drop exponentially at a characteristic rate called beta:
G ~ ket ~ e-beta R
For good conductors (such as nanotubes) beta = 0. Bridge materials with beta > ~ 1.4 Å-1 are considered insulators (e.g. decane's beta = 1.6 Å-1). Another criterion for distinguishing metals and insulators is electron mobility. For metals mobility is ~103 cm2/V-sec, while for insulators mobility is ~10-4 cm2/V-sec. Priyadarshy's group calculates betas with large scale (5000 atomic orbitals) SCF Hartree-Fock quantum mechanical calculations.
Researchers have looked at DNA in conjunction with donors and acceptors which have ranged from externally tethered to the DNA to intercalated between the bases within the helix. Barton et al. studied electron transfers at R > 40 Å and reported a beta ~ 0.2 Å-1. Brun and Harriman looked at R ~ 17 Å and reported beta ~1.1 Å-1. Meade and Kayyan measured ket at R = 21 Å and reported ket = 106 sec-1 (equivalent to a mobility of 10-7 cm2/V-sec)
DNA has both the continuous backbones of sugar/phosphodiester links and stacks of pi-bonded bases (which tend to be somewhat conductive). Priyadarshy's group separated these effects by calculating beta both with the bases present (1.2 Å-1) and with bases absent (2.3 Å-1). By comparison, a stack of benzene rings by itself yields beta ~ 1.4 Å-1. Neither component of DNA appears to act as a good conductor.
In contrast, there has been work depositing silver metal on DNA strands, and these deposits do form conducting wires.
Dr. Keith Runge (from ORNL) gave a talk on semiclassical methods for calculating conductivity of nanowires.
Full ab-initio quantum mechanical calculations are currently limited to roughly 60 atoms. While nanotube conductors are small in lithographic terms, even a short length of a metallic nanotube still contains hundreds to thousands of atoms.
Normally, quantum mechanics is used on nanometer and smaller scales, while classical mechanics is used on larger scales. Runge talked briefly about the usual approximation where nuclei are treated as classical particles in a field set by the quantum mechanical effects of their accompanying electrons.
In Runge's calculation, individual electrons are treated as point particles. The electrons are injected into the wire, where they move in the potential set by the other electrons and by the nuclei. This approach allows ~5000 trajectories/hour to be calculated on a PC. The analog of the transmission probability in a quantum mechanical calculation of the conductance of a wire is the probability of an injected electron reaching the other end of the wire rather than being scattered back to its original electrode. He ran a test where a line of 20 "hollow core" atoms modeled the wire. The semiclassical calculation showed regions of stable and chaotic dynamics. A more sophisticated "charge-screened atom" also showed regions of chaotic dynamics. The "charge-screened" model of an atom puts electrons in known orbitals (but without a full quantum mechanical calculation of the whole extended molecule), then integrates the charge density in these orbitals to get the electrostatic potential in which the injected electron moves.
In both the main conference and in the tutorial Dr. James C. Ellenbogen (of MITRE) gave talks on architectures for electronic digital computation.
Dr. Ellenbogen surveyed a taxonomy of possible nanoscale digital architectures, including mechanical, coherent quantum, biochemical, and electronic. MITRE is concentrating on the electronic path, using designs where logic states are represented conventionally, by voltages at circuit nodes. This builds on the experience of the electronic industry, while the possibly more energy efficient quantum dot approaches require more radical changes. The taxonomy of possible molecular electronic switching devices include biomolecules (but these are too resistive), fullerene based devices (which are very conductive, but currently have limited chemistry) and small conjugated molecules. The choices of switching devices and choices of architectures interact.
Challenges for molecular electronic architectures include:
Within the small molecule electronic paradigm, Dr. Ellenbogen showed a design for a half-adder that:
The Tour/Reed work probed molecules as two-port elements, both with mechanical break junctions and with "nanopore" structures. The half-adder design has 6 contacts, so two-port experiments won't suffice to probe it. Dr. Ellenbogen suggests contacting the structure with multiple converging nanotubes, which would allow sufficient independent contacts to confirm operation of the design. This is projected to occur over the next two years.
Prof. Helen C. Taylor of King's College London reported experiments on the mechanism of the dynein molecular motors. These cause the microtubule skeletons of cilia and flagella to slide, thus causing these cellular appendages to bend in such a way as to propel cells, or to move objects along the surface of the cell.
Dr. Barry Robson of the IBM T.J. Watson Research Center discussed prospects for building molecular machines based upon the design principles used by 4 billion years of biological evolution, but using different chemistry to build protein-like molecules, rather than biological proteins.
One key step was accomplished several years ago (see PE Dawson et al., Science 266:776-779) with the demonstration that proteins of moderate size could be directly synthesized by chemical ligation of unprotected peptide segments. This process bypasses ribosomal synthesis of proteins so that non-biological amino acids could be incorporated. Next steps would include developing alternatives to a peptide backbone, such as thioether, thioester, oxime, etc., linkages. Other possibilities include cyclic peptides, radial peptides, and branching structures built around lysine residues, which contain a second amino group and thus permit a second peptide bond.
Another approach that Dr. Robson described was the production of proteins using D-amino acids, the mirror images of the L-amino acids used in natural proteins. D-amino acid counterparts of natural enzymes, such as the HIV protease, have been produced that are mirror images of the natural protein, and that are enzymatically active on mirror images of the natural substrate. A major advantage of D-amino acid proteins is that they are much less biodegradable than are L-amino acid proteins.
Dr. Robson described a complex process to generate structures made from D-amino acids that interact with proteins made of L-amino acids. D-amino acid proteins can be made immunogenic via conjugation to natural proteins, the resulting antibodies sequenced to identify the protein sequence(s) that bind to the D-protein, and then this sequence can be made from D-amino acids and will therefore bind to the L-protein counterpart of the original D-protein.
Finally, he described an approach to speed molecular dynamics calculations of protein interactions ("docking"). The proteins involved in such interactions are usually composed of several fairly rigid domains, so calculating an interaction potential for each domain, and considering how objects move in the fields of each of the domains, simplifies the calculations.
Professor M. Reza Ghadiri of the Scripps Research Institute presented a molecular Darwinistic approach to looking at how the complex properties of living systems emerged from molecular ecosystems, which in turn emerged from the information encoded in the covalent bonds of molecules.
Prof. Ghadiri opened with the observation that the emergent properties of living systems, including consciousness, are produced by ~500 chemical reactions catalyzed by the enzymes of intermediate metabolism. His research topic lay between the molecular world (in which the complexity of noncovalent chemistry -- including molecular recognition, catalysis and self-assembly -- emerges from the information inherent in covalent bonds in molecules) and the ecosystem of living organisms. This is the world of molecular ecosystems, in which self-organization and replication are emergent properties resulting from nonlinear chemical systems. A simple example of a nonlinear chemical system is an autocatalytic system, in which the product of a reaction catalyses the reaction forming it. Such a reaction will quickly win out over a competing reaction that is not autocatalytic.
Prof. Ghadiri described the self-replication of a 32 amino acid residue helical peptide R formed from a "leucine zipper" motif, which causes two molecules of the peptide to bind in a coiled-coil interaction, mutually shielding their hydrophobic regions from contact with solvent. R catalyzes its own replication from two fragment peptides of 15 and 17 residues. One of the peptide fragments (E) has an electrophilic end group, and the other (N) a nucleophilic end group, such that when they are both bound to R template molecule, the end groups react to join the two peptides together, forming a new molecule of R. The product is thus the catalyst of its own synthesis.
To turn this simple self-replicating peptide into a molecular ecosystem, Prof. Ghadiri introduced variant E and N fragments, thus producing variants of R. Some variants replicate more efficiently, some less, some are parasites, some are mutualistic, etc. Currently he is studying the test tube evolution of a 256-member ecosystem, with 16 E and 16 N and 32,000 binary interactions.
Prof. Carlo D. Montemagno of Cornell University showed an elegant video in which the 12 nm-diameter F1-ATPase molecular motor rotated a 2 µm-long actin filament. He showed experiments in which the F1-ATPase was altered by genetic engineering to add specific amino acid residues to act as handles. One type of handle added to the base of the motor was used to attach individual motor molecules to metal islands that had been deposited on a substrate using electron beam lithography. Another type of handle was added to the shaft of the motor to attach microspheres to be moved by the motors. He is currently working on a method to turn such molecular motors on and off within the cell.
Prof. Viola Vogel of the University of Washington Center for Nanotechnology reported steps towards using motor proteins to transport objects in non-biological nanoscale environments.
The biological model for Prof. Vogel's work is protein transport in brain cells, where large numbers of proteins produced in the main body of the cell must be transported, usually as complexes or organelles containing many molecules, along the long axons of the nerve cell to where they are needed. Brain microtubules act as molecular rails along which a variety of motor proteins of the kinesin and dynein superfamilies [reviewed by N. Hirokawa, Science 279:519-526 (23Jan1998)] transport the protein cargoes. The goals of Dr. Vogel's adaptation of these molecular motors to non-biological environments are:
In the conventional non-biological environment, or motility assay, for these shuttles, the kinesin motor molecules are adsorbed to a surface and the microtubules are added, and movement is seen. However, there is no control over the direction of the movement because the microtubules are oriented randomly. Prof. Vogel's work explores using nanometer-scale surface topography to guide molecular motors.
Contoured surfaces are produced by shearing teflon across glass to give a glass surface covered with teflon containing parallel grooves of several to 30 nm in depth. When fluorescently labelled microtubules and kinesin are added to this surface,the microtubules move in a controlled fashion aligned with the grooves, although in both forward and backward directions. At the optimum ratio of kinesin to microtubules, most of the tubules are moving at a single constant velocity.
Detailed studies of the movement rule out some potential mechanisms to account for the guidance, but the details are still unclear. Current observations are consistent with preferential adsorption sites of the kinesin to the grooves such that at moderate concentrations of kinesin, the microtubules must likewise follow the teflon grooves.
Dr. Michael J. Heller of Nanogen described Nanogen's active DNA chip technology in which electric fields are use to address DNA, or µm-scale beads containing DNA molecules, to a specific 80-µm location on the chip (~109 DNA molecules per location). This technology is currently used in genomics research and DNA diagnostics, and Dr. Heller speculated that it could be used to build a hierarchy of structures from self-assembly on the molecular scale through nm- and µm-scale structures.
Dr. Wolfgang Fritzsche of the Institute of Physical High Technology, Jena, Germany, reported on characterizing nanowires made by covering microtubules (25-nm diameter protein assemblies) with a Ni coating to make 60- to 100-nm diameter metallized wires. These were attached to µm-scale electrodes by electron beam deposition of Au contacts and shown to conduct, although more like semiconductors than like metallic wires. The metallization process produces variable thickness of the metal layer and also warps the microtubules so that they are no longer straight. For details:
Dr. Mark Akeson (of UC Santa Cruz) gave a talk on examining individual RNA or DNA strands by passing them a nanometer-scale pore.
The detection mechanism that Akeson and co-workers used was a large change in the ionic conduction through an ion channel when a nucleic acid base was present. They used alpha-hemolysin as their pore. This protein forms a heptameric conducting pore 2.6 nm in diameter when embedded in a lipid bilayer. An open pore conducts ~120 pA in 1M KCl with 120 mV across it. As individual bases thread through the pore, the current drops by 80-90% as each base passes through.
The alpha-hemolysin ion channel is particularly well suited to these experiments. It has a funnel shaped aperture that captures DNA or RNA strands efficiently. It has a good chemical environment at the narrowest point in the pore.
It was shown that strands really has passed through the pore, rather than only transiently blocking it from one side, by amplifying the strands that had threaded through with PCR. It was also shown that the pore could pass single stranded DNA, but not double stranded DNA, so the bases had to go through in single file.
In one experiment an adenine30cytosine70 oligomer was synthesized and put through the pore. The current drops for the two bases were sharply distinguishable. Adenine permitted 20 pA of residual current, while cytosine reduced the current to 5 pA. The numbers of bases of each type could be counted and compared with the number expected from the oligomer synthesized. The differences in residual current are thought to be due to differences in the sizes of the poly-A and poly-C helices. Additional detection mechanisms may be necessary to distinguish all 4 bases. Work is being done on transverse tunneling across a pore.
Up to 180,000 bases an hour can be threaded through the pore. This isn't diffusive transport. The transport rate depends on the voltage across the pore, and goes to zero below a certain voltage. Akeson mentioned the possibility of using a molecular motor to pull the DNA or RNA through the pore. Currently, the time taken for the strand to traverse the pore can be matched to the length of the strand, and the times taken for the poly-A and poly-C sections of a mixed strand can be matched to their lengths.
From Foresight Update 35, originally published 30 January 1999.
Foresight materials on the Web are ©1986–2018 Foresight Institute. All rights reserved. Legal Notices.