In conjunction with Foresight Update 39
Recent Progress: Steps Toward Nanotechnology
By Jeffrey Soreff
The ribosome is in some respects the closest thing to a general assembler that exists today. It translates programs in the form of messenger RNA molecules into proteins, which then fold into 3D structures. Nucleic acid polymerases also perform programmed assembly, but with a much smaller range of local shapes. The initial peptide bond formation reaction in the ribosome tolerates substantial differences in the size, shape and chemical properties of the amino acids’ side chains, which is somewhat analogous to the variations in local workpiece structure that a general assembler will need to tolerate. We can construct peptides synthetically via solid phase synthesis today, but can’t achieve the same chain lengths as ribosomal synthesis achieves.
There has recently been a flood of papers on the structure of the ribosome. J.H. Cate et al., writing in [Science 285:2095-2104 24Sep99] and G.M. Culver et al., writing in ibid. pp 2133-2135, report crystal structures of the full particle at 0.78 nm resolution. The ribosome consists of two subunits, a smaller one labeled 30S, whose structure is reported at 0.55 nm resolution by W.M. Clemons Jr. et al., writing in [Nature 400:833-840 26Aug99] a larger one labeled 50S, whose structure is reported at 0.5 nm resolution by N. Ban et al. in ibid. pp 841-847. Both periodicals also contain some additional commentary on these advances.
While these articles are recent, the ribosome structure has been investigated for 40 years, including x-ray diffraction studies of ribosome subunit crystals for 20 years. None of the current structures is quite at atomic resolution yet, though (Nature, p 846) “As data extending to 3 Å resolution can be obtained from these crystals, computation of a map at significantly higher resolution and its fitting by a complete atomic model of the 50S ribosomal subunit may soon be possible.”
Ribosomes are exceedingly complex structures. They weigh about 2.5 x 106 amus, implying about 105 non-hydrogen atoms. They contain “a tangle of 54 proteins and three RNA strands” (Science, p 2048), extending roughly 21 nm in each direction.
The techniques used to find the structures have been largely x-ray diffraction techniques, though there have been quite a few hybrid methods used in the projects. For example, one problem in interpreting x-ray diffraction data is that the diffraction pattern shows the intensities of the diffracted x-rays, but to convert these intensities back into maps of the diffracting molecule, one also needs the phases of the diffracted x-rays. In the J.H. Cate et al. work, this was started by deriving initial phases from previous “cryo-electron microscopy” data, from a lower resolution model of a ribosome derived from many electron microscope images of individual ribosomes at cryogenic temperatures. N. Ban said that “we had to stretch every available method to its limits” (Science, p 2050). Another technique used to establish phases is to add heavy atoms to the structure being analyzed. One such atom per unit cell adds a high intensity reflection which serves as a phase reference for the rest of the structure. “But because the ribosome is so big and electron-filled, Yonath needed larger concentrations of electrons than single atoms could provide and so decided to use clusters of heavy atoms instead.” (Science, p 2049)
These techniques may also be needed in order to confirm the correct fabrication of large atomically precise artificial structures. In one respect, it is unfortunate that x ray diffraction techniques have dominated ribosome structure determination, because this requires macroscopic amounts of the structures in order to grow the diffracting crystals. Scanning probe methods would have allowed structure determination from a handful of molecules. We are still some way from analytical disassemblers.
Even if we don’t directly use any of their working mechanisms, ribosomes can be useful to nanotechnology as large, asymmetrical, self-assembling platforms for mounting other components in known 3D positions. Several of the techniques for structure determination either added components such as the heavy atom clusters or placed point mutations in ribosomal proteins to bind diagnostic functional groups, and these modifications did not unduly perturb the ribosomes’ structures. The same technique could be used to, for instance, stretch out a molecular electronic circuit across the surface of a ribosome by attaching it to several (possibly mutated) binding sites.
Two groups have reported building molecular motors recently, T.R. Kelley et al., writing in [Nature 401:150-152 9Sep99] and N. Koumura et al., writing in ibid. pp 152-155.
The molecular motor built by Kelley et al. currently rotates through 120 degrees. It consists of two major parts, a triptycene (three benzene ring blades fused to two carbons on a central axis) and a helicene (four fused benzene rings which curve back from an attachment to the triptycene axial carbon to block free rotation of the triptycene). There is an amine group on one blade of the triptycene and a hydroxyl tethered to the end of the helicene furthest from the attachment to the triptycene. This molecule can exist in three rotamers (which take several days to interconvert) depending on where the blade with the amine is relative to the helicene. The authors were able to drag one of the rotamers with the amine pointed away from the hydroxyl into one where they are pointed in the same direction by
The net effect is to apply the chemical energy of hydrolysis of the COCl2 towards crossing the rotational barrier between rotamers in a predetermined direction. This is analogous to the rotation of ATP synthase by hydrolysis of ATP in biological systems, but this system is not yet able to make a full rotation.
The molecular motor built by Koumura et al. rotates through a full 360 degrees by alternating photochemical and thermal steps. The motor consists of two identical halves, connected by a central double bond. In the absence of light, a carbon-carbon double bond cannot rotate. The four groups on the two carbons maintain their relative positions. Ultraviolet light temporarily permits rotation about the double bond, allowing cis isomers (those with two substituents on opposite carbons, but on the same side of the double bonds) and trans isomers (which have substituents on opposite sides of the double bond) to interconvert. This is cis–trans photoisomerization. Normally, this process would rotate halves of some molecules clockwise and half of some counterclockwise. In Koumura et al.‘s molecule, however, there are cyclohexane rings on both ends of the double bond, and these rings, in turn, have bulky groups (methyl and fused naphthalenes) pointed back towards the double bond. The steric interference between these groups sets the direction of rotation during photoisomerization. The net result is that both photoisomerizations produce strained isomers. The overall cycle is:
The photochemical steps in this process add energy to the molecule, driving the overall motion. The authors demonstrated driving their molecules through 2.5 full cycles of this process.
These results, particularly Koumura et al.‘s, are a major step towards programmable molecular machinery. Placing atoms controllably is central to nanotechnology, and moving them controllably is likewise crucial. Prior to this, we had biological motors, which we had no means of single-stepping, and synthetic actuators which moved a group back and forth between two stations, but which could not perform potentially unlimited unidirectional motion.
A stepping motor could let us build one system that can index through a large number of discrete geometries without additional synthetic work. In particular, it might let us perform some of the same synthetic tasks that now require scanning probe methods for positioning on a macroscopic quantity of workpieces. For instance, a sparse film of these motors, winding up polymer strands connected to catalytic groups, might leave the same sequence of catalytic modifications across macroscopic areas of the surface. The ability to perform multiple revolutions also allows us to apply many low energy steps towards driving high energy processes (e.g. stretching a spring to eventually break a covalent bond).
It would be desirable to have follow up experiments to investigate and optimize the mechanical properties of this motor. For instance, optical tweezer experiments to measure the torque delivered, and the probability of a misstep as a function of the load torque, would be very useful.
J.C. Ellenbogen and J.C. Love, writing in Architectures for molecular electronic computers: 1. Logic structures and an adder built from molecular electronic diodes (MITRE, July 1999), available as an Acrobat PDF file (the “pink book”) describe theoretical work towards Tour wire based molecular electronic logic. This paper is an extension of Ellenbogen’s presentations at the 1998 Foresight conference.
To reiterate the information common to the previous presentation: The authors are pursuing an architecture analogous to conventional electronics. Logic values are represented by voltage levels, not by quantum mechanical phases or positions of atoms. Logical functions are performed by circuits drawing finite currents through diodes, resistors and other circuit elements. The authors consider both nanotubes and Tour wires (poly-(phenylene-acetylene)) as possible conductors. They choose Tour wires due to the experimentally demonstrated ability to synthesize them with atomic precision, which is not yet possible with nanotubes. Tour wires also have roughly 1% of the cross-section of (10,10) nanotubes, so the authors’ designs push closer to the limits of the densest possible electronic circuits than nanotube based designs would. The area of the authors’ designs is roughly a factor of 106 smaller than that of today’s CMOS equivalents.
The authors’ circuit designs rely on components that have been experimentally demonstrated to work at the molecular level:
The authors exhibit classic AND and OR rectifying diode/resistor circuits, showing how they map into their molecular components. They also exhibit an RTD-based exclusion OR design. This completes the set of logical functions needed to compute arbitrary digital functions.
RTDs have a current (voltage) curve which rises to an initial peak (due to tunneling current), then falls to a valley before rising again. The falling section of the curve produces the useful electronic behavior (exploited by the authors’ exclusive OR, for instance). An important figure of merit for the device is the ratio of the peak current to the valley current. The authors write that M. A. Reed and J. Tour, who had experimentally demonstrated an RTD with a peak/valley ratio of 1.3:1 in 1997 have now demonstrated one with a ratio of 1000:1.
The principal new work presented here is a quantum mechanical analysis of a variety of rectifying diode structures. The authors’ structures are based on doped polyphenylene structures, which makes them more compatible with Tour wires than the experimentally tested molecules are. The authors molecules contained donor and acceptor phenyl groups separated by a saturated dimethylene bridge. The dopant groups used were methoxy-, methyl-, trifluoromethyl-, and cyano-, the first two being electron donating and the second two being electron withdrawing. The dopant groups were placed meta to the saturated bridge and the energies and degree of localization of the wavefunctions calculated by the “ab initio Hartree-Fock self consistent field (SCF) molecular orbital method.” The authors treated the energy difference between the lowest unoccupied molecular orbital (LUMO) localized on the acceptor phenyl and the LUMO localized on the donor as a measure of the electrical asymmetry of the diode. They found a roughly 2 eV difference in energy in both the dicyano-dimethyl and the dicyano-dimethoxy diodes.
The authors also examined changes in energies resulting from shifts in the positions of the phenyl groups due to rotation of the dimethylene bridge. These were found to be limited to less than 0.1 eV. These geometric sensitivities are important because the topologies of logic designs can force geometrical distortions in gates or wires due to crossovers. The authors designed both a half adder and full adder with their molecular components, requiring 1 and 3 crossovers respectively.
I would like to see both I(V) calculations and experimental measurements on the new diode structures, since they differ somewhat from the experimentally measured molecules. The authors write that software to calculate “the electrical behavior of conductive molecular-scale circuits”, called molSPICE, is being developed at their MITRE site.
The authors describe a number of remaining “design challenges for molecular electronic circuits.” In the version of their paper on the web on 11/20/99, one of their challenges is the need for gain in electronic circuits. That version of the paper described gain as requiring 3-terminal molecular electronic devices. While such devices would make it simpler for circuit designers to incorporate power gain in their systems, there are tunnel diode circuits (such as the Goto pair) which provide power gain, and a revised version of the paper is noting that 3-terminal devices are not strictly necessary.
Scanning probe microscope techniques are important to nanotechnology because they allow experimenters to directly sense and move objects at the atomic scale. While scanning tunneling microscopy (STM) has generally had the highest resolution of these techniques, atomic force microscopy (AFM) has the advantage of not requiring a conducting substrate or electronic conduction through the workpiece. The papers summarized in this section describe advances in AFM techniques.
J.H. Hafner, C.L. Cheung, and C.M. Lieber, writing in [J.Am.Chem.Soc 121:9750-9751 20Oct99] describe a novel method for using a nanotube as an AFM tip by growing it directly on a commercial tip.
Nanotubes have been used as AFM tips for several years. They have excellent resolution because of their tiny diameters. The authors of this paper say that methods of mounting separately synthesized nanotubes onto AFM tips have been time consuming, which has limited the usefulness of nanotubes as tips.
The authors describe electrophoretically depositing a Fe-Mo colloidal catalyst on a commercial AFM cantilever tip with a -1.8 volt pulse. Nanotubes were then grown on the tip at 800 degrees C in an ethylene/hydrogen/argon mix. The authors found that the growing tubes aligned themselves so as to probe directly from the apex of the AFM tip. They find that the attractive forces between an AFM tip and a growing nanotube are sufficient to bend the nanotube into contact with the surface during initial growth. When the growing nanotube encounters an edge of the pyramid, these forces suffice to align its growth along the edge. When the tube encounters the apex of the pyramid, however, it grows off into space.
The authors could reliably grow single-walled nanotubes with diameters of roughly 2.5 nm. They find “that we can obtain effective tip radii of 3 nm or less.”
J. Lefebvre et al., writing in [Appl. Phys. Let. 75:3014-3016 8Nov99], describe constructing electronic circuits from single-wall carbon nanotubes (SWNTs) with an atomic force microscope (AFM).
The authors deposited their SWNTs from a suspension in an organic solvent onto an SiO2-covered silicon wafer. During each move of a tube the sections of the tube were sequentially pushed with the silicon AFM tip. The authors would start near one end of the tube, bring the probe down, move the tip 10 nm in the direction desired, then move the tip further from the tube end and repeat the process till the whole tube had been moved. During the initial move “most of the tube remains in place due to substrate interaction.” A sequence of these 10 nm steps has been used for micron-scale total displacements. A similar sequence of moves can be used to rotate the tubes about an axis perpendicular to the substrate.
Both imaging and displacement are done with the AFM operated in tapping mode. During displacement the scan speed is increased roughly tenfold (from 1-5 µm/s to 20-80 µm/s). This is faster than feedback loop for controlling tip height can respond to, “so the tip interacts strongly with the SWNT.” Nonetheless, forces are smaller than in contact AFM in which “SWNTs are disturbed during imaging and cannot be manipulated.”
The authors were able to push a tube across another tube. In the AFM images, the tube that they manipulated winds up crossing another tube at roughly right angles. During part of the displacement a portion of the tube breaks off, so the forces applied must be comparable to SWNT strength.
The authors used e-beam lithography to contact one of the crossovers in their system, albeit a naturally occurring one. In this case a bundle of SWNTs of 7.5 nm diameter crossed another with 2.5 nm diameter. They measured resistances along the two bundles (“330 and 50 kohm, respectively”), and the resistance of the junction (roughly 10 megaohms, with a nonlinear I(V)). At cryogenic temperatures they also measured quantum dot behavior of the lower bundle, observing periodic conduction as a function of the potential of the upper bundle and the silicon substrate.
In general, this technique could be very useful, not only in constructing circuits from nanotubes themselves, but in making multiple contacts to other nanostructures that we need to evaluate or modify. The authors were able to align their e-beam fabricated electrodes “to better than 100 nm with respect to the nanotube circuits.” The nanotubes provide the ability to bring the signals from those lithographed electrodes down to terminals a few nanometers from each other. For example, even the simplest electronic amplifier requires 4 connections: power, ground, input, and output. If we synthesize molecules that act as amplifiers without additional external components, we’ll need these connections to demonstrate this. Extending the molecule to connect to 4 nanotubes with 2 nm diameters terminating in a 4 nm x 4 nm square looks much more feasible than extending the molecule to connect to electrodes with 100 nm spacings.
A cluster of electrically accessible converging nanotubes could also build structures by selective electrochemistry. One could immerse them in a series of solutions, plating out a different material on each of the nanotubes by applying a voltage to just one nanotube at each step.
Finally, S.B. Carlsson et al., writing in [Appl. Phys. Let. 75:1461-1463 6Sep99] describe controlling electron tunneling barriers between Au disks by sliding them around with an AFM tip.
The authors fabricated 3 electrodes and a central island with roughly 100 nm dimensions using e-beam lithography on an SiO2 layer on a silicon substrate. In a second step, they fabricated a grid of Au disks “30-100 nm in diameter and 30 nm thick.” They were able to slide the disks around on the surface with an AFM tip. They adjusted the width of their tunnel junctions by first pushing them into solid contact, then breaking the contact, then monitoring the resistance as they were pushed towards contact again. They could adjust tunnel junction resistances in the range from 100 kohms to 1 Gohm, corresponding to tunnel gaps from 0.3-1.0 nm. The authors have also used this technique to form atomic-scale contacts. The authors found that their tunnel barriers were stable for hours at room temperature and for weeks at liquid helium temperature.
While the Au disks are much too large to be atomically precise, this technique does permit adjustments of the gaps between them to atomic precision. This is typical precision for vertical control in scanning probe microscopy, but is noteworthy in the construction of stable lateral structures.
The authors’ technique is essentially forming and breaking chemical bonds (albeit metallic, not covalent bonds) under experimental control. It is notable that they were able to choose materials and experimental conditions under which, simultaneously:
That these conditions could be satisfied concurrently bodes well for constructing covalently bonded structures with the similar tolerances needed for atomic precision.
From the viewpoint of applications of single electron devices, the authors’ technique provides a kind of breadboard, a way to make “individually tuneable tunnel junctions… important for the investigation of tolerances in critical parameters for future single-electron circuits.”
Writing in [Science 285:2113-2115 24Sep99], K. Kageyama, J. Tamazawa, and T. Aida describe the use of a hexagonal array of nanometer-scale catalytic reactors to directly synthesize linear fibers of polyethylene.
In normal polyethylene, the polymer chains are folded back on themselves, contacting other chains only at the periphery of the resulting crystals. As a result, total forces between the chains are weak, as is the bulk polymer, even though each individual chain is held together by strong covalent bonds. When polyethylene chains are aligned in a uniform direction, each chain contacts adjacent chains at each atom, so the chains are held together much more strongly. Like strands in a rope, they transmit forces efficiently to each other, yielding strengths as much as a factor of 100 greater than that of normal polyethylene.
It has been possible to orient the chains of polyethylene after polymerization, but this is a difficult, expensive process. The authors’ technique grows the polyethylene strands directly out of a honeycombed catalyst with pores too small (2.7 nm) to permit the chains to fold, so instead they bind to each other, forming aligned fibers, as they emerge from the catalytic particles.
The authors’ catalyst was titanocene ((C5H5)2Ti), a standard olefin polymerization catalyst, adsorbed on the pores in a nanostructured silica. The silica is “mesoporous silica fiber”, which has 2.7 nm pores in a hexagonal array aligned along the fiber axis. On exposure of the catalyst to ethylene at 10 atm at room temperature, fibers of polyethylene 30-50 nm in diameter formed. X-ray diffraction and thermal analysis showed the alignment of the chains along the fibers.
This work illustrates an application of nanostructured materials in controlling the fabrication of products with desirable supramolecular structures, even if the molecules of the final product are not themselves novel. Viewing the titanocene molecules as fabrication machines, it shows that controlling the placement of these machines can be valuable. The formation of the aligned fibers also requires the simultaneous operation of many molecular machines, which will eventually be needed for many applications of nanotechnology. It would be interesting to see if a patterned placement of several types of catalyst molecules via scanning probe methods could generate fibers with several types of polymers interleaved on a molecular scale.
Jeffrey Soreff is an IMM research associate.