IMM Report Number 11

New Components

New components advance nanotechnology by extending the options we have for fabricating structures with desirable capabilities. Particularly valuable are stiff, atomically precise, 3D structures, which can be used to control the positions of functional groups without excessive movement from thermal vibrations.

Y. Yao and J.M.Tour, writing in [J.Org.Chem. 64:1968-1971 19Mar99] describe synthesis of a “caltrop”, a tetrahedral compound designed to sit upright on a surface. The authors’ caltrops are composed of four poly-phenyleneethynylene -(CC-C₆H₄)_n– segments projecting from a central silicon atom. The authors specifically designed their caltrops to serve as sharp asperities on SPM probe tips. Three of the legs of their caltrops bear “thiol-tipped feet for adhesion to a metallic surface such as Au, Pt-Ir, or [semiconducting] Ga-As.”

The authors’ synthesis is a six step process with yields of ~50% on each step. They prepared their final compounds on a ~300 milligram scale. They introduce the differences between their compounds’ tips early in the synthesis. Tetraethyl orthosilicate is reacted with para-iodophenyl lithium to replace 3 of the 4 ethyl groups, yielding ethoxytri(p-iodophenyl)silane. The ethoxy group was then replaced by a phenyleneethynylene trimethylsilane group. This group has been used as a base for building molecular wires. As the authors write, they “have [already] prepared numerous oligo(phenylene-ethynylene)s ranging from 10 to 140 nm [in length] via iterative doubling strategies.” These oligomers can be coupled to this caltrop terminus to serve as the final asperity.

The other 3 terminii of the caltrop are coupled to 1-ethynyl-4-(thioacetyl)benzene, displacing the terminal iodines, lengthening these legs, and introducing the (protected) thiol groups.

This strategy can yield an SPM tip which is atomically precise, and which can be functionalized with well controlled chemistry. For example, the authors prepared caltrops both with trimethylsilane tips and with hydrogen tips. They note that competing approaches using nanotubes also aim toward this capability, but are as yet much less mature than the solution chemistry the authors used.

A. Müller, P. Kögerler, and C. Kuhlmann, writing in [Chem. Commun. :1347-1358 7Aug99], summarize a great deal of recent work on the synthesis of polyoxometalate clusters. They have been able to synthesize species containing as many as 248 molybdenum atoms. The metal atoms are encased in octahedra of oxygen atoms. These clusters are typically bound together by edge-sharing octahedra, so the cluster as a whole is a heavily fused polycyclic structure.

These species are synthesized in aqueous solution, formed by a sequence of acidifications and reductions of simple molybdate species. The clusters grow when surface Mo^VI species are reduced to Mo^V, after which “the intermediate cluster acts both as nucleophile and template in directing the formation of two further electrophilic molybdenum-oxide based intermediates [a yet larger cluster].” Perhaps tunneling currents from an STM could selectively reduce atoms on one of these clusters. The authors write that “…these types of integrated constituents (here situated on the surface of a cluster) are more easily reduced than the discrete ones.” so there is a built-in bias towards enlarging a workpiece cluster over reducing a free-floating Mo^VIO₄^2- feedstock ion.

The {Mo₂₄₈} cluster consists of a {Mo₁₇₆} ring covered with two {Mo₃₆} end caps. The ring has a diameter of 3.5 nm, comparable to the diameter of nanotubes or proteins. The authors were able to grow crystals of these compounds and analyze their structures via x-ray diffraction.

One disadvantage of these clusters is that some of them are too symmetrical. For example, the {Mo₁₃₂} cluster with which the paper opens has full icosahedral symmetry, so there are really only 3 inequivalent types of molybdenum atoms in the structure. In the case of the ring cluster, “it is possible to generate deliberately discrete structural defects…” which could be used as handles to incorporate the ring into a larger structure with a selected orientation.

D.G. Kurth, F. Caruso, and C. Schuumller writing in [Chem. Commun. :1579-1580 21Aug99], describe a new technique for depositing sequences of metal-containing layers on nanometer-scale particles in aqueous solution.

The authors’ technique relies on electrostatic forces to deposit an alternating sequence of positively charged and negatively charged linear polymers on their particles. The negatively charged polymer was sodium polystyrene-sulfonate.

The positively charged polymer was the metal-bearing layer. It consisted of Fe^II ions alternating with the ligand 1,4 bis(2,2′:6′,2″-terpyridin-4′-yl)benzene. This ligand has two binding sites for metal ions, one at each end of the molecule, and two such binding sites can wrap around each ion.

The authors started with positively charged polystyrene spheres, depositing the negative linear polymer on it, washing and centrifuging to remove excess polymer, then repeating with the positive metal-bearing polymer, and so on. The authors deposited 14 discrete layers with this technique. They write that “…different active components can be assembled in each layer, thus opening avenues to construct complex functional materials and devices.”

This method provides one-dimensional design control on a nanometer scale, much like molecular beam epitaxy or Langmuir-Blodgett film deposition. Perhaps, with an atomically precise substrate and fixed length oligomers, this method might build up a structure that retained atomic precision.

G. Schmid et al., writing in [Chem. Commun. :1303-1304 21Jul99], report the first crystalization of Au₅₅(PPh₃)₁₂Cl₆. This cluster compound is important because it is an atomically precise conducting cluster. It has also been used as a component in DNA-based nanostructures. The authors were able to grow ~1 micron crystals of this compound and confirm intercluster distances of 2.3 nm with some preliminary x-ray diffraction measurements.

Fullerene Nanotubes

Fullerene nanotubes, graphite wrapped into seamless tubes with atomic precision, are useful components mechanically and electrically. They have high strength, high stiffness, and (for certain wrapping patterns) high conductivity. The papers in this section describe advances in the application, and manipulation of these materials.

R.H. Baughman et al., writing in [Science 284:1340-1344 21May99] describe the use of single-walled nanotubes (SWNT) as electromechanical actuators.

The authors prepared mats of SWNT bundles by filtering them from suspension on to a teflon filter under vacuum. Their first actuator was simply two sheets of this “nanotube paper” on the two sides of double sticky tape. The two sides were separately contacted, and the three layer sandwich immersed in 1 M NaCl. Application of a few volts between the sides made a 2 cm long sandwich bend through about a cm, demonstrating differential expansion due to the potential difference.

More quantitatively, the authors measured the absolute strain in one sheet of nanotubes by using it to deflect a cantilever bearing an optical sensor. “The maximum observed actuator strains are large: >0.2%” while the maximum from ferroelectric ceramics is ~0.1%. The authors found that an actuator could endure 140,000 cycles with only a 33% decrease in amplitude.

The authors’ “nanotube paper” contains bundles with diameters of ~10nm of nanotubes which themselves have diameters of ~1.2-1.4 nm. Only roughly 20% of the nanotubes’ surface area is exposed, so the authors expect that comparable actuators built from fully exposed nanotubes would have maximum strains of ~1%. The authors calculate that close-packed nanotubes should have “…a Young’s modulus (Y) of ~0.64 TPa…”, yielding a work density per cycle of ~3 x 10⁷J/m³ “~29 times higher than has been tabulated for for the best known ferroelectric, electrostrictive, and magnetostrictive materials.” It is roughly an order of magnitude less than the energy density (44 x 10⁷J/m³) for a 1 volt/nm field, arguably the strongest one can use even in an atomically precise electrostatic actuator.

This work shows that SWNTs, in addition to their potential as stiff structural members, conductors, and electrically active devices, also show great promise as high performance electromechanical actuators.

L. Roscheir et al., writing in [Appl. Phys. Lett. 75:728-730 2Aug99] describe the fabrication of a single electron transistor (SET) from a multi-walled nanotube using an atomic force microscope (AFM). The authors’ SET consisted of a 410 nm long nanotube bridging a 250nm gap between two gold (source and drain) electrodes and roughly 500 nm away from a third gold electrode which served as the gate. The electrodes were built using e-beam lithography on an oxidized silicon substrate.

The authors’ nanotube was not originally deposited on top of any of their electrodes. They were able to use the forces from their AFM to push it from its original location to bridge their electrodes. The AFM tip was scanned over an end of the tube at smaller and smaller distances till the tube was either rotated (2/3 of the attempts) or dragged (1/3 of the attempts). It took roughly 100 moves over several days to transport the tube from its original location to its position for electrical measurements. The authors “note that it was possible to lift the nanotube on top of the electrode which has a thickness larger than the tube radius.”

Electrical measurements on the SET showed periodic dependence of the current on the gate voltage and on the bias voltage across the tube as the tube gains individual electrons. This system is still a bit too large for room temperature operation. Total capacitance on the tube was ~38 aF, so the charging energy E_c=e²/2C_total=2.1 meV, and the single electron effects wash out between 22K and 77K.

It would be interesting to see if the ability to push nanotubes up significant inclines can be extended to build 3D structures of nanotubes such as cross-overs.

H.T. Soh et al., writing in [Appl. Phys. Lett. 75:627-629 2Aug99] describe a new method for constructing circuits of fullerene nanotubes. The authors patterned a catalyst for nanotube synthesis, a mixture of alumina, ferric nitrate, and Mo₂(acac)₂ then exposed the catalyst to methane at 900 °C. This creates mostly individual single-walled nanotubes (SWNT) with roughly 5% nanotube ropes. The patterning was done by depositing a conventional e-beam resist on an oxide-covered silicon substrate, exposing and developing it, creating 5 micron x 5 micron “petri dishes” with SiO₂ at their bottoms, adding a suspension of the catalyst in methanol, then removing the resist and excess catalyst from everywhere except where the “petri dishes” were.

Circuits were formed when “a nanotube bridge forms when a tube growing from one catalyst island falls on and interacts with another island during CVD synthesis.” A last stage of processing deposited metallic contacts over the catalyst islands to serve as leads for electrical measurements. The authors adjusted the spacing between the islands to control the length of the nanotube connections. Below a spacing of about a micron, multiple connections sometimes were made. These were cut with an AFM tip in order to isolate a single connection.

These connections are both mechanically strong and electrically conductive. Resistances down to 20 kiloohms were measured. These “contact resistances [are] roughly 100 times lower than can be obtained by placing contacts either on top of or below randomly located tubes, which previously was the state of the art.”

The authors’ plans for further work “include understanding the precise nature of the interfaces between the nanotube metal or nanotube-catalyst metal and their relations to the low resistance ohmic contact.” Hopefully this will allow the fabrication of these high-quality contacts at arbitrary points in the topology of a nanotube circuit, perhaps by adding them to presynthesized, preplaced nanotubes.

Fabrication Techniques

New techniques for assembling atomically precise components into larger structures are an important component of nanotechnology. The papers summarized in this section advance these techniques.

J. Chen et al., writing in [Appl. Phys. Lett. 75:624-626 2Aug99] describe a technique for placing molecular wires into selected locations in a self-assembled monolayer (SAM).

The authors deposited a dodecanethiol SAM on an annealed gold surface. They used a scanning tunneling microscope (STM) to apply voltage pulses “from 1.8 to 3.6 V and pulse duration (T_p) from 0.5 µs to 0.5 s.” The probability of cutting a pit through the SAM was roughly 50% for a voltage of 2.5 V (at a height giving a tunneling current of 0.5 nA), rising to >90% above 2.7 V. The pits formed had diameters of 10 nm and depths of 1.4 nm, consistent with the known length of dodecanethiol molecules.

The authors placed molecular wires in their pits by immersing their substrate in a solution of 2′-ethyl-4:1′-ethynylphenyl-4′:1″-ethynylphenyl-1,4″thioacetylbenzene and NH₄OH. The latter converts the protected thioacetyl groups to thiols, which can then bind to gold. Over the course of a few minutes these molecular wires fill in patterned pits in the dodecanethiol SAM. A later STM image then shows a raised topography, due to 2.2 nm length of the new molecules. At ambient temperature and pressure “several scans over one hour showed no pattern degradation.”

The authors consider this technique to be “a general method for generating intermixed SAMs of arbitrary shapes and compositions.” They “suggest that future work utilizing a carbon nanotube STM tip might implement single molecule replacement.” Alternatively, perhaps atomically precise patterns can be written with this method by making the molecules in the original SAM wider, perhaps by using a lattice of dendrimers or similar materials.

Y. Arai et al., writing in [Nature 399:446-448 3Jun99], describe tying knots in actin filaments and DNA molecules with optical tweezers.

The authors bound actin filaments to pairs of myosin-covered polystyrene beads. The unknotted length of the filament was roughly 25 microns, with 0.9 micron beads at each end. They looped the end of the filament by moving one of the beads with one of the foci of their optical trap. They knotted their filaments over ~ 1 minute. The actin filaments were prevented from following the moving bead too quickly (preventing knotting) by operating in a viscous syrup of 50% sucrose. The filaments were made visible by fluorescence.

The authors measured the stress on their filaments as a function of the diameter of their knots. They found that the flexural rigidity measured down to a radius of curvature of 0.2 microns “is close to the rigidity reported for free actin filaments undergoing brownian motion” at curvatures around tens of microns. Below a knot diameter of ~0.36 microns these knots break. This only requires a force of 0.9 pN, while a straight actin filament can withstand ~600 pN.

The authors also knotted DNA molecules. These knots were tougher, withstanding the maximum 15 pN that their optical trap could apply.

Typically, work on the manipulation of individual molecules which is covered in this column is usually done with some type of scanning probe apparatus. In this experiment, the authors note “that the transparent grip by the optical tweezers allowed continuous manipulation during knotting, without a shift of the grip as would be required in knotting with fingers or mechanical tweezers.”

One extension of this work suggested by the authors is to “study the curvature dependence of the interaction of intracellular filaments with associated proteins or ligands.” From the viewpoint of nanotechnology this experiment would demonstrate the assembly of a structure with a form of mechanochemistry.

The knots themselves might be useful as prestressed mechanical components. The actin rings have a significant operating range between the 14-18 micron persistence range above which they will distort freely due to thermal motion and the 0.36 micron size at which internal stress breaks them.

C.P. Collier et al., writing in [Science 285:391-394 16Jul99] describe the fabrication of electrically programmable logic gates which use an organic monolayer as their active component.

The authors deposited a micron-scale array of aluminum wires on an oxidized silicon substrate, deposited a Langmuir-Blodgett monolayer of a organic compound on these wires, then evaporated a titanium electrode (followed by a thicker aluminum layer) “through a contact shadow mask by using electron beam deposition.” Three different types of organic compounds were used, all variants of rotaxanes, all containing aromatic moieties which introduced energy states near resonance with the Fermi level in aluminum.

When operated at a bias near -0.5 V, these sandwiches acted as diodes. The current increases sharply as the biased is changed from -0.5 V to -1.5 V, while it stays near zero from -0.5 V to 0.5 V. The authors used this nonlinearity in their device characteristics to operate parallel devices as AND and OR gates, using classic diode-resistor logic circuit topologies.

The authors could also permanently remove diodes from their circuits by applying positive voltages to them. These oxidized the rotaxanes, removing the energy states in the gap between the electrodes, bringing “…the measured current…[to] a factor of 60 to 80 less than that for a closed switch.”

The authors note that “the configuration [oxidation] voltage can be a factor of 2 greater than the logic levels used when they are operating. This means that it should be easy to design circuits that are safe from accidental reconfiguration under operation conditions.” Unfortunately it also means that an additional set of circuits must be developed to steer the configuration voltages to the diodes that need to be programmed, requiring at least different topology and possibly different materials. In this paper, steering was accomplished by macroscopic apparatus.

The authors have demonstrated nonlinear molecular elements and circuits approximately equivalent to a fuse-programmable PLA, which (together with NOT gates) is a universal logic element. Further development is needed to

• provide a NOT circuit
• provide power gain for fanout and noise immunity
• scale the interconnects down from micron scale to nanometer scale
• develop circuits to steer the programming currents

R.F. Service wrote an article summarizing the state of total synthesis of natural products in [Science 285:184-187 9Jul99]. From the viewpoint of nanotechnology, this subfield of organic chemistry is important in showing how far covalent synthesis can be pushed today.

The article describes the postwar history of the field as having three phases:

1. synthesis to confirm structure
2. synthesis to extend and systematize synthetic reactions
3. synthesis to “make and modify drug candidates that are either rare
   or hard to isolate in abundance”

Service writes that “the synthesis race is intensifying while the number of
targets is shrinking, because of a slowdown in in searches for promising
new natural products.” As a result, synthetic groups are concentrating on
applications of existing reactions and are discovering less new synthetic
chemistry.

Unfortunately, the use of existing reactions is still far from straightforward. Service describes P. Baran’s view as that “…most good ideas fail…So it’s typically those who work harder and try more reactions in the lab who come out ahead.” This implies that we still don’t have as much predictive ability as we’d need to, for instance, automate the synthesis of custom parts for a nanomachine unless they use some very standard chemistry such as peptide synthesis.

Service describes the total synthesis community as being composed of roughly a dozen labs, with 20-40 chemists in each lab, multiple projects underway at any given lab, and synthetic projects requiring 1-10 years.

Computational Elements

A major application area expected for molecular nanotechnology is computation. The paper summarized below suggests new approaches to this application.

Y. Wada, writing in [J. Vac. Sci. Technol. A 17:1399-1405 JulAug99], describes two proposed technologies for atomically precise electronic computation and some theoretical and experimental work directed towards their implementation.

The first technology is an “Atom Relay Transistor” (ART). Its active device consists of a linear wire one atom wide with two perpendicular wires, a “switching gate” and a “reset gate”, almost touching the first wire. There is a central “switching atom” in the first wire. When this atom is present, the wire conducts. When the “switching gate” extracts this atom from the first wire, the gap makes the wire much less conductive. When the “reset gate” pulls the “switching atom” back into place, conductivity is restored. The frequency at which this atom can be switched “was simulated at around 30 THz by using the first principles method.”

Experimental work towards this technology has been an extension of hydrogen extraction from a hydrogenated silicon (100) surface by inelastic collisions with electrons from an STM tip. Wada evaporated gallium atoms on to a silicon surface which had had an atomically perfect line of hydrogens removed, leaving dangling bonds. Yada found that “…the gallium atoms stick to the dangling bonds to form metal wires selectively and spontaneously.”

The second technology Wada describes is “the molecular single electron switching transistor (MOSES).” This is basically a covalently bound version of a single electron transistor (SET). Yada gives examples where the conducting islands are polyacetylene in one case and C₆₀ in another. Yada reported that “a novel micromachine STM (µ-STM) was designed to manipulate a single molecule, which was fabricated using a 0.4µm ULSI fabrication technology.”

Jeffrey Soreff is an IMM research associate.