IMM Report Number 14

Inorganic Lattices

We have a good deal of control over the self-assembly of DNA and proteins into three dimensional, atomically perfect structures. Both of these materials have limited stiffness, roughly 10GPa for proteins and 0.35 GPa for DNA. Stiffness is important in nanotechnology in building structures that can place objects with atomic precision despite thermal vibrations.

Many organisms, however, are able to direct the deposition of inorganic lattices, which can be viewed as fused polycyclic ring systems, with considerably greater stiffnesses than biopolymers. Diatoms and sponges, for example, can deposit silica, which has a stiffness of 70GPa. If we could mimic this process with sufficient precision, perhaps we could use it to condense silica on to DNA or protein structures in order to increase their stiffness. Two recent papers describe advances in our ability to fabricate biomimetic silica structures; another describes an imprinting technique.

J.N. Cha, G.D. Stucky, D.E. Morse, and T.J. Deming, writing in [Nature 403:289-292 20Jan00], demonstrate condensation of silica driven by synthetic peptides. The authors were seeking analogs to silicatein, a protein found in a sponge, which “can hydrolyse and condense the precursor molecule tetraethoxysilane to form silica structures with controlled shapes at ambient conditions.”

The authors’ first step was to synthesize homopolymers (polylysine, polyhistidine, etc.) of a number of amino acids thought to be active in silica condensation. They found that “oligomers of L-cysteine … efficiently produce silica from TEOS in pH 7 buffer…This result was presumably due to the nucleophilic properties of the sulphydryl group, which may enable it to initiate hydrolysis of the silicon alkoxide.” The silica produced was an amorphous powder, with no control over the shape of its particles.

The authors’ second step was to synthesize a series of block copolymers, in order to use the self-assembly of the copolymers into structured aggregates to control the shape of the silica. The copolymers combined the hydrophobic poly-L-cysteine domains with hydrophilic domains such as poly-L-lysine. They found that “poly(L-cysteine₃₀-b-L-lysine₂₀₀)” formed 600 nm aggregates in aqueous solution. These aggregates subsequently precipitated 100 micron silica spheres. Different copolymers gave silica columns and other structures.

While the largest scale structures that the authors produced are too large to be useful for the development of nanotechnology, their silica spheres contained their copolymers intimately embedded within the silica. Roughly 10% of their spheres’ weight was organic, and the spheres were highly porous, with “a surface area of 436 m²/g” implying a length scale of a few nanometers. Perhaps the silica/polycysteine bond is sufficiently well-ordered to permit atomically precise addition of silica to protein structures with it.

N. Kröger, R. Deutzmann, and M. Sumper, writing in [Science 286:1129-1132 5Nov99] describe the discovery of polypeptides that “were shown to generate networks of silica nanospheres within seconds when added to a solution of silicic acid.”

The authors isolated these peptides from diatom (Cylindrotheca fusiformis) walls. The extraction procedure was quite harsh, requiring boiling silica from the diatoms in SDS solutions, then dissolving the silica in HF to extract these peptides. The authors separated three peptides that they call silaffins, which rapidly precipitate silica. The authors sequenced these peptides. Silaffins proved lysine-rich, and also contain “posttranslationally modified lysine residues” The lysine residue are attached to “a repeated N-methyl-propylamine”. At physiological pH this structure is protonated, and is “ideally suited for ionic and hydrogen-bonding interactions with the surface of silica particles.” A peptide lacking this modification precipitates silica, but only at much higher pH than one with the modification.

“At any protein concentration, silaffins completely coprecipitate with the silica as long as silicic acid is present in excess.” This is helpful, since it indicates that the silaffins become bound to all of the silica, rather than catalysing the formation of some unbound activated silica species, which might then condense into networks in an uncontrollable way.

Perhaps some analog to these peptides, possibly just the poly N-methyl propylamine structure, can be used as a component in nanostructures that we wish to reinforce with silica.

A.Katz and M.E. Davis, writing in [Nature 403:286-289 20Jan00], describe synthesizing silica with covalently bound clusters of functional groups.

They started with “imprint” compounds, consisting of benzene ring cores, surrounded by 1-3 side groups. Each side group was a -CH₂O-(C=O)-NH-(CH₂)₃-Si(OCH₂CH₃)₃ group. During the formation of the silica, this is mixed with Si(OCH₂CH₃)₄, and the mixture is hydrolyzed and gelled. At this point the -Si(OCH₂CH₃)₃ groups in the imprint lose their ethoxy groups and become -Si(O-)₃ units anchoring the imprint cores in the silica. The carbamate, -O-(C=O)-NH-, linkage is then cleaved at the C-N bond, leaving aminopropyl, NH₂-(CH₂)₃-, groups bound to the silica in clusters surrounding where the benzene core was.

The authors confirmed that their aminopropyl groups were indeed present as clusters via several types of evidence.

The final amino-bearing silicas were reacted with a diacid chloride, ClCO(CH₂)₇COCl, which is the right length to bond to both amines in the cluster with two amino groups. They found that this yielded diamide in the two amino group silica, but yielded a monoamide in the silica with the single amino group, since the diacid chloride couldn’t straddle dilute randomly placed sites to react at both ends.

The three modified silicas were soaked in pyrenebutyric acid (PBA), a fluorescent aromatic hydrocarbon bound to a carboxylic acid. In the one amino silica the PBA bound as a monomer, but in the three amino silica the PBA bound as a dimer, which could be observed as a shift of fluorescence to longer wavelengths due to excimer formation between closely spaced PBA molecules.

From the viewpoint of nanotechnology, this might be useful either in:

1. building tips with several converging functional groups held rigidly or
2. in building “molds” to orient subassemblies together

Imprint methods let us take advantage of rigid structures that we cannot control with atomic precision but into which we can nonetheless embed groups (and pores) whose relative positions are controlled with atomic precision (or nearly so).

Nanotubes

Nanotubes are important components in nanotechnology due to their stiffness, strength, and electrical conductivity. Two recent papers describe advances in understanding how nanotubes form.

Writing in [Appl. Phys. Lett. 76:161-163 10Jan00], F. Okuyama, T. Hayashi, M. Kawasaki, and K. Ibe, showed “that atomically thin filaments grow from the edge of an open graphite cage under a glow discharge.”

In their experiments a vaccum diode was operated in naphthelene plus Cr(CO)₆ vapors at ~0.06 Torr. In the absence of a magnetic field, these conditions produce multiwalled carbon nanotubes. In the presence of a magnetic field, slightly different carbon cages are deposited. In either case, the tips of the structures are sometimes open, showing active edges that may yield clues to the growth process.

In the authors’ experiments, filaments with diameters of 1.5-1.8 Å were seen in the field-emission electron microscope at the boundary of open carbon cages. This diameter is consistent with either atomic chain or with graphene ribbon seen edge on. In the image, the filament curls and forms a complex shape, so the authors consider an edge-on ribbon to be improbable.

The filament extends from a “woven” edge of graphite lattice, where a number of filaments come together and the bundle aligns with the lattice. The authors suggest that this structure might reflect the formation of a graphite lattice from self-assembly from these filaments.

While the growth conditions for these filaments are different than for arc-induced nanotube growth, they may still be similar enough to provide hints to structures of growing nanotubes.

Writing in [Appl. Phys. Lett. 76:161-163 10Jan00], A.A. Puretzky, D.B. Geohegan, X. Fan, and S.J. Pennycook describe experiments examining the species present during laser ablation growth of single-walled nanotubes (SWNTs).

The major techniques for nanotube synthesis are chemical vapor deposition, arc ablation, and laser ablation. The authors chose laser ablation for study because after the short (8 nsec in this case) laser pulse, the plume of ablation products cools and reacts without further perturbation. In the authors’ experiments the laser target consists of carbon with 1% each of cobalt and nickel. The target is immersed in flowing argon at 1000 degrees C and 500 Torr.

The laser pulses hit the upstream side of the target. They form “smoke rings” which propagate upstream for ~2 sec. The authors examined optical emission and luminescence induced by a second, probe laser beam. They found that “both imaging and spectroscopy indicate that within the first few milliseconds after laser ablation, atoms and molecules of both carbon and metal catalyst disappear due to condensation into nanoparticles. Unless SWNTs grow very rapidly from atoms and molecules within these first few milliseconds, the majority of growth appears to occur from a feedstock of mixed nanoparticles over seconds of annealing time.” They also found that when the plume was sampled 0.5 seconds after the laser pulse, only short (~100 nm) nanotubes were found, while 10 micron nanotubes were found on a downstream collector after ~>2 sec plume annealing.

These results suggest that nanotubes grow primarily from some intermediate particles that condense rapidly from the initial plume. Since these intermediate species are likely to be less ferociously reactive than isolated atoms, they may be amenable to growth of nanotubes with controlled diameter, chirality, and length, perhaps in some unreactive solvent.

Information Processing

One of the first areas expected to apply nanoscale fabrication is the information industry, due to the high potential value per mass of atomically dense bits. Two recent papers describe advances in nanoscale logic and memory respectively.

Writing in [Science 286:1550-1552 19Nov99], J. Chen, M.A. Reed, A.M. Rawlett, and J.M. Tour describe the fabrication of a molecular device with “an on-off peak-to-valley [current] ratio in excess of 1000:1.”

A simple linear resistance transmits a current that rises smoothly with the voltage across it. A tunnel diode transmits a current that rises to an initial peak, then falls to a valley, then rises again. The authors’ device shows this structure. In 1997 the authors reported a device which also showed an I(V) curve with this structure, but the old device had a much less dramatic peak-to-valley ratio, only 1.3:1.

The peak-to-valley ratio is important, for instance, in the noise resistance of circuits that can be built with the device. If it is put in series with a sufficient resistance, the circuit is bistable. The maximum difference beween the currents flowing in the two states is the difference between the peak and valley currents, so a large difference makes these states tolerate more noise.

The region of operation between the current peak and the valley has current through the device decreasing as the voltage across it increases. This is called negative differential resistance, NDR. It is the opposite of the behavior for a resistor, and it is responsible for the ability of circuits containing an NDR device to yield power gain and/or bistability.

The authors’ device is 2′-amino-4-ethynlphenyl-4’ethynlphenyl-5′-nitro-1-benzenethiol, a 3 ring Tour wire C₆H₅-(-CC-(C₆H₄)-)₂-SH modified by the addition of nitro and amino groups to the central phenyl ring.

The authors measured the electrical characteristics of their device with a fairly complex microfabricated structure that they have used in previous experiments. The structure contains a silicon nitride membrane (with a complex support structure) with a 30-50 nm hole in it. Gold is evaporated into this hole from one side, then the thiol is applied to it to form a self-assembled layer, then gold is slowly evaporated on to the other side. The net result is avoid “pinhole and other defect mechanisms that hamper through-monolayer electronic transport measurements” when larger contact areas are used. Roughly 1000 molecules’ currents are paralleled in this structure.

The authors’ “candidate mechanism for the NDR is a two-step reduction process”, the first reduction introducing an unpaired free carrier and the second completing the pair and returning the molecule to an insulating state. The authors excluded the possibility of an irreversible electrochemical change by sweeping the voltage bias in both directions, finding that “the I(V) curve was fully reversible upon change in bias sweep direction.”

This device is important because it is an atomically precise electronic element that provides power gain. Together with previously known molecular diodes and resistors, it is sufficient to allow the construction of arbitrary logic functions. In my view, the primary challenges in molecular electronics are now:

• to interconnect several devices into a circuit with atomic precision in
  the whole structure and
• to attach more than two electrodes to an atomically precise structure

We clearly have the pieces to build an atomically precise NOT gate with power gain. In order to demonstrate a NOT gate’s function, we must attach power, ground, input, and output leads to the molecule.

[Editor’s note: these research results generated considerable media attention when first announced last Autumn. See both the “Recent Progress” and “Media Watch” columns in Foresight Update #39 for references.]

Writing in [J.Vac.Sci.Technol.B 17:2467-2470 Nov/Dec99], S.M. Hou et al. describe storing data using an STM to change an organic film from insulating to conducting using 1.4 nm marks.

The material that the authors used in their memory film was N-(3-nitrobenzylidene)-p-phenylenediamine (NBPDA). Previous work had used a two-component organic film, but the authors chose to put both nitro and amino groups into a single compound in order to avoid “the complexity of accurately characterizing the structure of an organic complex thin film.” The NBPDA films were very smooth, with rms irregularity of only 0.16 nm. They deposited their films on HOPG.

The authors created data marks “by applying voltage pulses of 4 V for 2 ms” at ambient conditions. The marks were visible as areas of increased conductivity, showing metallic conduction while the unmarked area showed a semiconducting I(V) curve with a threshold of 1.3 volts. Oddly, the authors report an inter-mark distance of 9 nm, yielding a data density of ~10¹² bits/cm². Since their mark size is nearly an order of magnitude smaller than this spacing, I don’t see why they didn’t report trying closer mark spacings. They report that their marks persisted for at least a 30 minute scanning session.

The authors suggest that their marks were due to an inverse Peierls transition due to the intense electric field under the STM tip. This would create a vertical string of bonds between their conjugated molecules, creating a 1D conductor. They present quantum mechanical calculations supporting this possibility.

Information storage is one of the classic potential application areas for nanotechnology. If this technique or some similar one becomes a mainstream product, then all of the effort normally applied to tuning production products will be available to increase the density of nanostructure fabrication.

On another note, if this semiconducting to metallic transition can induce lateral conductivity in an NBPDA film on an insulator (perhaps by “drawing” lines with a charged tip starting from a predeposited micron scale electrode), then perhaps it can be used to draw many closely spaced 1-2 nm wires on a surface.

Mechanochemistry

An important feature of nanotechnology is the use of precisely applied mechanical forces to alter the course of chemical reactions. This is important both in creating highly reactive molecular tools needed to build a wide variety of structures and, over the very long term, in allowing nanotechnology to use energy efficiently.

An example of this technique is described in [Science 286:1543-1545 19Nov99]. T. Oya et al. synthesized a gel that changes its affinity to a target when it swells and shrinks.

The authors picked two highly charged target molecules, pyranine-3 and pyranine-4. These are fluorescent polycyclic aromatic hydrocarbons with 3 or 4 sulfonate groups (each bearing a negative charge), respectively, attached. The authors’ gel contains a minority monomer, “methacrylamidopropyltrimethylammonium chloride (MAPTAC), which carries one positive charge.” The minority monomer was mixed with a large excess of N-isopropylacryamide and a cross-linking agent and the mixture polymerized. The resulting polymer is swelled by water at temperatures below 33 degrees C, with a phase transition to a shrunken phase above this temperature.

In the swollen phase the MAPTAC groups are too far apart to bind the targets cooperatively. The affinity of the gel for the pyranines is then proportional to the MAPTAC concentration, indicating independent binding to each MAPTAC. In the shrunken phase the MAPTAC groups are close enough for random clusters of them to cooperatively bind the pyranines. Random clusters of 3 and 4 MAPTAC groups have concentrations proportional to the 3rd and 4th powers of the MAPTAC concentration. In the shrunken phase, the affinity of the polymer for pyrazine-3 and -4 is indeed proportional to these powers of the MAPTAC concentration. Note that these clusters were not templated on their targets. These clusters’ formation is statistically controlled, hence the power law dependence on MAPTAC concentration.

This approach to polymer binding sites is a demonstration of a kind of mechanochemistry. When the majority monomers absorb water and swell the polymer, they pull apart the binding site formed by the minority monomers. This use of a mechanical coupling to modulate binding of a target molecule is essentially the same type of mechanism as Drexler’s sorting rotors use.

On a shorter timescale, this approach may be useful where we wish to connect two subassemblies, and we have a catalytic structure that properly orients them for a reaction to connect them, but the product assembly binds too tightly to the catalytic structure for easy ejection. If the catalytic site can be built of several pieces which are independently connected to a swellable polymer, then they can catalyze the connection of the subassemblies with the polymer in the shrunken state but be forced to release the combined product by swelling the polymer, pulling the pieces of the catalytic site apart.

Fabrication Techniques

Writing in [J.Vac.Sci.Technol.B 17:2452-2456 Nov/Dec99], F. Iwata, T. Matsumoto, R. Ogawa, and A. Sasaki describe a novel way to pattern polystyrene, albeit at a vertical scale of ~15 nm and a lateral scale of a micron.

AFMs have been used to pattern polymers by scratching their surfaces. A problem with conventional scratching is that “the scanning speed of conventional scratching is very slow (~µm/s), which induces an elasto-plastic response of polymeric materials.” The plastic flow produces bumps on the polymer surface. Under the authors’ conditions the bumps are 2nm high on the first pass, increasing to 6nm after three passes.

The authors’ innovation is to move the surface rapidly. They mounted their sample on a 6.5 MHz quartz resonator, moving them laterally by a few nm at this frequency. This is faster than the AFM cantilever can respond, so the diamond tip is effectively fixed by its inertia while the sample moves. Because the tip-sample distance changes faster than the sample can flow, the tip cuts the surface much more cleanly. The mean roughness of the ultrasonically scratched surfaces was less than 1 nm, regardless of the number of scratching passes. With a loading force of 1.4 µN, about 1 nm of material was removed per pass, about the same mean material removal as for conventional scratching.

While this technique has been demonstrated on scales somewhat too large for direct application to atomically precise nanotechnology, perhaps similar non-quasi-static methods will be important on smaller scales as well. Some dynamics of single molecules can be surprisingly slow, so some atomically precise techniques might rely on outracing these modes.

Jeffrey Soreff is an IMM research associate.