IMM Report Number 26

Structural Components

Stiff, strong, atomically precise parts to bear loads, transmit forces, and hold active parts in position are essential parts of any nanomachine design. The papers summarized in this section describe advances in the synthesis of these components.

Writing in [scienceexpress 5Apr2001 (101126/science.1057823)], R.R. Schlittler et al. describe the synthesis of single-walled carbon nanotubes (SWCNTs) from C₆₀. Surprisingly, they find that they form single crystals, at least one of which appears to be of pure (10,10) armchair nanotubes, a major advance over the uncontrolled mix of nanotube chiralities synthesized in other methods.

The authors’ synthesis method has several special features. They evaporate alternating layers of C₆₀ and a nickel catalyst. “Some 6 or 7 Ni and C₆₀ layers with individual thicknesses of 10 to 20 nm were deposited.” The layers are deposited through a mask with holes 300 nm in diameter, though “no mask is necessary if the material is deposited on a rough facetted surface, but fewer tubes are produced.” Substrates of SiO₂ and molybdenum were used, the latter more successfully. The layers are heated to 950^o C in vacuum. The nanotubes grew perpendicularly to the surface when the material was “immersed in a ~1.5 Tesla magnetic field” (itself oriented normal to the surface) during growth.

The authors found that this process produced “almost perfect rod-like crystals of SWCNTs”, with diameters from 40-900 nm and lengths up to 2 µm. They find that, “the wall [tube?] diameters in a given rod are remarkably uniform”, though they vary from 1.4 to 2.3 nm across rods.

The authors selected a rod of 1.66 nm diameter nanotubes and studied it with electron diffraction. Under their conditions, if the crystals were composed of tubes with “a range of chiral angles, the diffraction spots associated with the chirality would form arcs.” In fact, they see only one such diffraction spot, hence “one chiral angle” and “monodispersed crystals of SWCNTs”.

This is the key evidence and the key finding from the viewpoint of bottom-up nanotechnology. SWCNTs have many valuable properties: stiffness, strength, and (for the metallic ones) conductivity. In order to use them as atomically precise components, however, there must be a method to either synthesize or isolate nanotubes of uniform chirality, because the chirality defines the bonding topology within the tube. It would be helpful if the authors analyzed the intensities in their diffraction patterns to set upper bounds on how much contamination from tubes of other diameters or chiralities can be present in their rods. Real-space STM atomic imaging to confirm tube indices would also be helpful.

Further work on clarifying and controlling the growth mechanism of these rods, and hopefully scaling it up to yield macroscopic quantities of uniform nanotubes is highly desirable.

Writing in [Science 292:706-709 27Apr2001], P.G. Collins, M.S. Arnold, and P. Avouris describe using selective removal of nanotubes via electrical breakdown in air to tailor multi-walled nanotubes (MWNTs) and ropes of SWCNTs.

In both cases, nanotube shells are removed by running high current densities, ~10⁹ A/cm² through the shells to be removed. The authors believe that the oxidation of the shells is due to direct “current-induced defect formation and that self-heating plays only a secondary role.”

In MWNTs, the authors say that only the outermost shell is removed because it is in direct contact with the electrodes, while “the innermost shells [are left] carrying little or no current, which protects them during current-induced oxidation.” The authors observe the thinning of the MWNT both electrically and microscopically (via AFM and SEM). During thinning, a stepwise drop in current is seen, with nearly constant current steps (of ~19 µA). This is consistent with the decreasing circumferences of successive cylindrical shells.

In SWCNT nanotube ropes, all of the SWCNTs are in contact with external electrodes, but the authors were able to selectively modulate current through the nanotubes, depending on whether the nanotubes were metallic or semiconducting. A gate voltage applied to the SWCNT rope was used to deplete the charge carriers in the semiconducting tubes, while the metallic tubes continued to conduct. When several volts were applied between the ends of the rope, the metallic tubes were oxidized away, leaving the semiconducting ones in place. For very thick ropes, some of the metallic tubes were buried deeply enough that some of the semiconducting tubes had to be sacrificed in order to removed the buried metallic tubes. This was done by increasing the end-to-end voltage above that needed for thinner ropes.

The net effect of pruning metallic tubes out of SWCNT ropes is to leave a purely semiconducting rope. These ropes act as efficient Field Effect Transistorss, with G_on/G_off of 10³ or better. They were able to use these techniques to build small arrays of nanotube FETs.

This technique allows present day mixed nanotube ropes to be used for electronic applications. The authors also suggest that similar techniques could be used to exploit other molecular devices where our fabrication capability produces mixtures of molecules rather than a pure sample of the desired species.

Writing in [Science 292:479-481 20Apr 2001] M. Remskar et al. describe the synthesis of MoS₂ nanotubes. The nanotubes proved to be of uniform diameter (.96 nm between tube centers) and chirality, which “corresponds to a (3,3) armchair nanotube.”

The authors grew their nanotubes by vapor transport at low
(10^-3Pa) pressure. They used iodine as a transport agent and C₆₀ as a catalyst. The nanotubes are up to hundreds of nm long. The tubes form as needle-like bundles. Individual tubes can be dispersed by ultrasonification in ethanol.

Unlike carbon nanotubes, these tubes have the strong advantage of uniform size and chirality. Unfortunately, the bonds in these tubes are substantially weaker than carbon-carbon bonds, and the small diameter will also reduce the bending stiffness, reducing the effect of these advantages as an atomically precise mechanical component.

Writing in [C&EN 79:41-42 14May2001], S. Borman summarizes some recent work on peptides containing fluorinated side groups. Two groups modified a 33-residue peptide called GCN4. One group replaced 4 leucine (a hydrophobic amino acid) residues with trifluoroleucine. The other group replaced 4 leucines and 3 valines (another hydrophobic amino acid) with fluorinated versions. In both cases the groups were able to increase the stability of the peptides above that of the natural versions. These peptides form dimers. The first group increased the temperature to which these dimers are stable by 13^o C, the second group by 15^o C.

In natural peptides and proteins, a major contributor to the stability of the folded structure is essentially a phase separation between a core of hydrophobic amino acid residues and the water in which the molecule is dissolved. In these new structures “immiscibility of fluorous phases in water and many organic solvents… provide[s the] driving force for fluorinated amino acid side chains to collapse into a fluorous core.” Because this adds a third phase to the usual aqueous/hydrocarbon separation, it adds considerable design freedom to protein design. This is important to nanotechnology because proteins can fold into atomically precise 3D structures, and protein synthesis permits a great range of design freedom, so improving our ability to predict and control folded protein structures improves the range of precise 3D structures that we can build.

Perhaps one can proceed further along these lines. There is a classic demonstration of seven mutually immiscible liquids: heptane, aniline, water, perfluoro-kerosene, phosphorous, gallium, and mercury. Natural proteins use two (perhaps three, if one counts aniline as analogous to tryptophan) of these phases. This work adds a third phase. Perhaps all of these phases can eventually be used to control the tertiary structure of synthetic proteins.

Moving Parts

In order to construct products, one must be able to move atoms into place. Any nanotech molecular manufacturing system must therefore include some type of motor and other moving parts. The papers summarized in this section describe advances in the synthesis and characterization of these types of components.

Writing in [Science 291:2124-2128 16Mar2001] A.M. Brouwer
et al. describe the operation of a fast, photochemically powered molecular motor. The motor moves in “~1 µs; previous light-driven shuttles generally function on the time scale of minutes to hours.”

The authors’ motor consists of a shuttle that moves between two binding sites. The shuttle is a macrocycle, a 26-atom ring, with 4 hydrogen bond donor sites (two paired sites from each of two isophthalamide groups). The shuttle is threaded on to a linear molecule with two binding sites, each capable of accepting 4 hydrogen bonds, but under different conditions. The ends of the linear molecule are plugged with bulky groups which prevent the macrocycle from slipping off.

One of the binding sites is a succinamide. It contains two carboxy oxygens separated by 4 carbon atoms. These oxygens are the hydrogen bond acceptors in the ground state of the motor, when the macrocycle is bound to this station.

The second binding site is a naphathalimide. It contains a
-(C=O)-NR-(C=O)- moiety. In the ground state, this hydrogen bond acceptor is a poorer one than the succinamide. The naphathalimide group absorbs 355 nm photons to give an excited state. This excited state extracts an electron from an external donor to give a radical anion. The radical anion derived from the naphathalimide group is a better hydrogen bond acceptor than the succinamide, so the macrocycle shifts binding sites at this point. Eventually, after ~100 µs, the donor group retrieves its electron from the naphathalimide group. The macrocycle then shifts back to the succinamide, completing the cycle.

The authors also prepared the radical anion form of their compound electrochemically, allowing them to characterize the spectra of both forms of their compound, confirming the translocation of the macrocycle. They measured the timescale of the photochemically induced translocation by monitoring the absorption maximum of the naphathalimide radical anion.

This peak shifts from 416 nm to 411 nm, as a result of the change from the bare anion to one encapsulated in the polar environment of the macrocycle’s hydrogen bonds, over a ~1 µs timescale.

The radical anion’s absorption is also used to monitor the oxidation back to the starting state by the donor cations. The absorption diminishes on a 100 µs time scale, after which the system can again be illuminated with a 355 nm pulse, repeating the whole cycle.

The authors’ calculate that each molecule of their motor can convert ~10^-15 W of optical power to mechanical power, limited by the time in the recovery step. They compare this to kinesin, capable of 5 x 10^-18 W. They explicitly suggest that this system, or “more powerful and efficient analogs” may be useful for tasks such as the “rearrangement of the structure of surfaces or the fetching-and-carrying of molecules or clusters of atoms between specific locations.”

Writing in [Science 292:733-737 27Apr2001] J. Liphardt et al. describe measurements on the unfolding of three RNA molecules. Each of the authors’ three molecules is a hairpin or modified hairpin, a strand of RNA which is self-complementary and folds back on itself to bring its ends close together. The authors pulled on their hairpins with optical tweezers, measuring force vs. distance curves and the dynamics at varied pulling rates.

All three hairpins were roughly 60 bases long in their unfolded states. The simplest, P5ab is just a hairpin loop, with no additional structure. The second, P5abcDA, is a T-shaped structure, a hairpin interrupted by a second hairpin sticking out of its side. In other words, it has a three-helix junction. The third, P5abc, is, like P5abcDA, a T-shaped structure, but it also has some regions which bind to Mg²⁺ ions.

Under constant force conditions, all three molecules show a bistable region, the extension of P5ab jumping up and down by 18 nm at a force of 14.5 pN, P5abcDA by 22 nm at 12.7 pN and P5abc by 26 nm at 22 pN (with an intermediate state as well).

The authors interpret the two-state behavior as evidence of cooperative folding. In this experiment, however, even quite a modest amount of cooperation would appear as bistable behavior. A constant force spring (as expected for successive unpairing of bases in a hairpin in the absence of cooperation) subjected to a constant force is neutrally stable. A few kT of cooperative folding (only 10% of the ~260 x 10^-21 J of folding energy here) would suffice to disfavor intermediate amounts of folding and give the observed two-state behavior.

For the simplest hairpin, P5ab, transitions were faster than the resolution time of the experiment, 10 msec. The more complex molecules showed slower dynamics, with hysteresis visible in the loading/unloading curves for P5abcDA at loading rates above 1 pN/sec. P5abc shows yet slower dynamics, “these molecules do not hop when held at constant force; rather, they unfold suddenly and do not refold for the duration of the experiment (5 min).”

From the viewpoint of using these molecules as mechanical components, the simpler hairpin is clearly preferable, with faster dynamics. It would be useful to

• look at force-extension curves in a stiff apparatus, to see the magnitude of the deviation from constant-force behavior,
• examine varying RNA hairpins to see which comes closest to an ideal constant-force spring
• examine the dynamics of folding at higher frequencies, to see the useful speed range of this component

Several pieces of recent work have clarified the structures of programmable biochemical molecular machines.

In [Science 291:2526-2527 30Mar2001], E. Pennisi summarizes some recent work on ribosome structure. The latest structure, at 5.5 Å resolution, is still not quite at atomic resolution. One of the new techniques that improved the recent structures was “adding two or three tRNAs [transfer RNAs] to the crystallizing ribosome.” These filled up docking sites and helped stabilize it, helping to sharpen the image.

In [Science 292:411-414 20Apr2001], J. Marx summarizes work on the structure of RNA polymerase II, an enzyme that copies base sequences from DNA to messenger RNA. This structure was solved to 2.8 Å resolution, sufficient to “see where every amino acid goes.” The enzyme is quite a large one, containing “12 different proteins bound together in a complex that has a molecular weight of about 500,000.”

This is something of a milestone because, like ribosomes, nucleic acid polymerases are programmable enzymes, generating not just one product, but an exponentially large set of them, controlled by their nucleic acid templates. As far as I know, this is the first time that one of these machines has had its structure determined to essentially atomic resolution.

From the perspective of nanotechnology, knowing the structure of these machines may help design programmable machinery of our own. For instance, they help show how to pick the right degree of flexibility of the active sites so that it is flexible enough to accept any of the base pairings that must occur to accommodate programmability, yet is firm enough to hold their substrates in the proper positions to catalyze their respective synthetic reactions.

Scanning Probe Techniques

Writing in [Science 291:2580-2583 30Mar2001], M.A. Lantz et al. describe AFM measurements on an Si (111) 7×7 surface precise enough to measure the force curve for the formation of a single bond between the tip and the surface.

The authors used an Si tip (with a native oxide coating) with nominally 10 nm radius. Initial imaging showed “weak contrast”, but, “with continued scanning, an abrupt improvement in contrast occurred …[which the authors attributed] to the transfer of one or more silicon atoms from the sample onto an oxide-terminated tip.” The authors scanned their samples with an oscillating tip, using frequency changes for feedback. At specific sites, full curves of frequency versus height were measured, and these were converted into force versus distance curves. They ran their experiments at 7.2 K, finding that cryogenic operation reduced problems with reproducible positioning.

One change from previous work was that the AFM tip is now biased (to 1.16 V) to minimize the long range electrostatic force. The bias cancels the surface charge from “the contact potential difference between tip and sample.”

The Van der Waals (vdW) force between the tip and the bulk sample was modeled and subtracted from the measurements over specific lattice atoms. The authors took advantage of the “corner holes” in the Si 7×7 structure, places where an Si hexagon surrounds a vacancy, to model the vdW force. Force-distance curves were taken at this site, and fit to a theoretical -A_HR/6(z+z₀)² force function for a spherical tip over a plane. An excellent fit was found, confirming the use of this model for the vdW force.

The authors were able to subtract out the vdW component of the force curve above the adatoms, leaving short-range forces. They probed the short-ranged forces through their maximum magnitude, measured at -2.1 nN (attractive), though not quite to the classical equilibrium position at zero force. The measured forces are in good agreement with forces calculated from density functional theory. The authors saw reproducible measurements on individual atoms, and can distinguish between inequivalent atoms in the surface structure. They interpret the differences to reflect local surface charges.

The agreement between the calculated and measured force curves indicates that only one tip adatom is contributing to the short range forces, that “no other dangling bonds on the tip contribute to the short-range force measured above the adatoms.”

From the perspective of nanotechnology, this work

• demonstrates the controlled formation and scission of nearly a full covalent bond with an AFM,
• helps calibrate computational chemical models, and
• demonstrates the utility of a relatively simple model of an AFM tip.

Jeffrey Soreff is an IMM Research Associate.

IMM would appreciate learning your thoughts on the above article.