A publication of the Foresight Institute
[Editor's Note: This page has been optimized
for Netscape 2 and later. If you are using a browser, such as
Netscape 1.1, that does not support the html tag for
superscripts, please be aware that an number like
"2x109" is meant to be scientific notation for "2
times ten raised to the 9th power," and that "e2"
means "e squared," etc.]
There were many important papers presented at the Fourth Annual Foresight Nanotechnology Conference. Since, however, the conference was summarized in the Foresight Update 23, this column will focus on information from other sources.
Protein synthesis is one technology currently available for building nanometer scale systems. Its primary advantage is that it allows us to build macroscopic quantities of (potentially) atomically perfect nanometer scale structures. The primary disadvantage of protein synthesis is that we want to construct 3D structures, but we can only specify the sequence of amino acids directly. The next three papers advance the state of the art in allowing design of protein structures that fold into target 3D structures.
The current state of protein design is described by J. W.
Bryson, S. F. Betz, H. S. Lu, D. J. Suich, H. X. Zhou, K. T
O'Neil, and W. F. DeGrado, writing in [Science 270:
935-941 10Nov95]. They focus on the information that stability
studies of synthetic proteins can yield about protein energetics.
"Designed, helical peptides provide model systems for
dissecting and quantifying the multiple interactions that
stabilize secondary structure formation. De novo design is also
useful for exploring the features that specify the stoichiometry
and stability of alpha-helical coiled coils, and for defining the
requirements for folding into structures that resemble native,
The authors describe predictions of structures as rather well established in helical domains, with fairly good agreement (±0.3 kcal/mole out of a range of roughly 0.0-1.0 kcal/mole) on the helical propensities of the various amino acid residues, good understanding of "Specific hydrogen-bonded interactions that 'cap' the ends of helices," and good information on the pairwise "hydrogen bonding and electrostatic interactions between amino acid side chains separated by a single alpha-helical turn."
The associations between these helices can be harder to predict, with the authors citing one case where varying one pair of residues produced every state of aggregation from dimers to hexamers. In a number of four-bundle designs "the association of hydrophobic side chains provides a powerful driving force for the formation and association of helices...However, both lattice models as well as early design attempts lack the diversity of stabilizing interactions and specificity found in natural proteins, which we believe are essential for stabilizing native-like folds and function."
They describe a variant of the four-helix bundle ROP, where the hydrophobic residues were packed "in layers consisting of two small and two large side chains per stack" with the resulting protein behaving "in all respects examined like a native protein." In contrast to the helical peptides, beta sheets have had more experimental difficulties: "...the exposed amides at the edges of beta sheets can hydrogen bond to other sheets, leading to insoluble aggregates." One of the first de novo designed proteins, Betabellin (which contains beta sheets), has had its solubility increased by introducing a special type of turn using D-amino acids. The authors write that "Recent progress in designing structural proteins has set the stage for the engineering of functional proteins." They describe an example of a "four-helix bundle protein with four bound hemes," with spectroscopic properties consistent with the design.
In summary, the helical bundle proteins look like they may be ready for use as structural elements in nanotechnology, with careful attention to residue packing, while other structural motifs are being understood, but more slowly.
A more successful use of beta sheets has been taking place in
M. R. Ghadiri's group at the Scripps Research Institute, as
described by P. S. Zurer in [C&EN 18-20 15Jan96].
Ghadiri's group has been synthesizing peptide rings with
alternating D- and L-amino acids. The "rings adopt flat
conformations with the amide carbonyls and NH groups pointing up
and down, perpendicular to the plane of the ring. These rings
self-assemble into nanotubes by stacking one on top of the other,
linked by intermolecular hydrogen bonds in a beta-pleated sheet
The group has sufficiently fine control over the structures formed that they can add functions to their tubes. "For example, the chemists have engineered a system in which evenly spaced carboxylic acid side chains on the outside surface of the nanotube bind copper ions." Ghadiri's group has also been successful in freezing their self-assembled structures in place with covalent chemistry. "They incorporated two side chains bearing terminal olefins into a ring composed of eight amino acids. In nonpolar organic solvents, a Grubbs ruthenium catalyst initiates a ring closing reaction that couples two cyclic peptides together. The products are two 38-membered ring structures formed through a double metathesis [olefin exchange] reaction, with none of the smaller bridged rings that would result from intramolecular reaction."
It would be interesting to see if the group could synthesize pairs of cyclic peptides which would assemble into nanotubes, since this could permit a Merrifield-style synthesis of distinct oligomers, as well as the nanotubes that they can currently build, which are impressively long (200-300 µm), but have uncontrolled length.
J. R. Desjarlais and T. M. Handel, writing in [Protein
Science 4: 2006-2018 1995] describe a novel
computational and experimental approach to redesigning the
hydrophobic cores of proteins. They "have designed and
engineered several variants of the 434 cro protein,
containing five, seven, or eight sequence changes to the
hydrophobic core." Prior to this work, "designed
proteins generally lack a well-defined and uniquely structured
folded state. These proteins usually display weak cooperativity
in their unfolding transitions and poorly dispersed NMR spectra.
Because of the poorly structured nature of these proteins,
determination of high-resolution structures for these molecules
has also been hampered." The lack of high-resolution
structural information is a particularly important hurdle to
cross, because this information is crucial in providing detailed
diagnostics for tuning the structures. We have to solidify the
protein structures well enough to get enough information to
really debug them.
The authors' design methods start with the native protein. They build a "custom" rotamer library for the hydrophobic core by removing the core side chains from the protein, but retaining the backbone and the non-core side chains. They examined rotamers at 5° increments of torsion angles, retaining only 18 low energy configurations for each of the hydrophobic residues examined. The selection of well-packed structures using this rotamer library is a substantial computational task. "...for a small protein with 10 core positions, more than 1018 structural solutions exist with roughly 1010 sequence combinations."
The authors dealt with this combinational explosion by optimizing their designs with a genetic algorithm. This step takes roughly 1 to 3 hours on a 150-MHz processor, examining 50,000 candidate structures. The primary result is that the authors were able to design two variations on 434 cro which "are of comparable thermal stability to the C-1 native control." Not all of the designs were this stable. Another variant, designed by the same methods, was significantly destabilized relative to the native protein. In addition to the thermal stability evidence, 1D proton NMR spectra were examined for three of the proteins. "Designed proteins typically have very poorly dispersed NMR spectra due to a combination of exchange broadening and chemical shift averaging caused by a dynamic folded state." In the spectra examined, the spectra are about as well dispersed as in the native protein. "This implies that for these representative variants, the folded state is well ordered."
One class of useful nanometer structures that can be constructed with protein technology is that of catalytic antibodies. These proteins help extend synthetic capabilities by controlling the orientation with which reactants encounter each other, reducing the number of side reactions that take place, increasing the purity of the products formed, and hence improving the synthetic utility of the reactions. This is both an application area for nanotechnology and potentially a mechanism for extending the variety of stiff, polycyclic building blocks available to the nanotechnologist. The next two papers extend the state of the art in this area.
Writing in [Science 270: 1775-1782 15Dec95], P.
Wirsching, J. A. Ashley, C.-H. L. Lo, K. D. Janda, and R. A.
Lerner describe an extension to the technology of antibody
catalysis. Previous work has generated catalytic antibodies by
inducing an immune response to an inert antigen that
models the transition state of a desired reaction. The antibody
produced then binds to the the transition state of the reaction,
stabilizing this transition state, reducing the activation energy
of the reaction, accelerating the reaction, and reducing the
fraction of the substrates that undergo undesired side reactions.
The current work uses "reactive compounds as immunogens designed to promote specific chemistry in the antibody binding site, both in vivo during antibody induction and then later in catalysis. Thus, at the time of antibody-antigen encounter, a component of the binding energy results from complex chemical reactivity as well as from simpler forms of complementarity dependent on electrostatic and hydrophobic forces." There is a trade-off required in using reactive immunogens, because "there must be sufficient reactivity to undergo chemical reactions in the binding site of the antibody, but not too [sic] much lability to be completely degraded by the many chemical entities encountered in vivo during immunization."
The specific reactive system described in this paper was an organophosphonate diester, RP=O(OR')2 and one of its hydrolysis products, RP=O(OR')O-. The final antibody catalyzed the hydrolysis of an analogous carboxylic ester, as well as the hydrolysis of the organophosphonate diester. The number of turnovers of the phosphonate diester was limited (typically 1-3) because the antibody can itself be phosphonated by the diester, inactivating it. The net result of this work is to add a new option to catalytic antibody technology, broadening the range of immunogens that can be used to trigger formation of potentially useful antibodies.
Writing in [Science 270: 1797-1800 15Dec95], J.
Wagner, R. A. Lerner, and C. F. Barbas III describe catalytic
antibodies that perform the aldol condensation. The aldol
condensation is formally an addition of a ketone (or aldehyde)
with an alpha hydrogen, R(C=O)CHR'R'' (the aldol
"donor") across the carbonyl of another ketone (or
aldehyde) R'''(C=O)R'''' (the aldol "acceptor") to give
R(C=O)CR'R''C(OH)R'''R''''. This reaction "is, arguably, the
most basic C-C bond forming reaction in chemistry and
biology." There are natural enzymes that catalyze this
reaction but "the most limiting aspect of the application of
natural enzymes in synthesis is their rather poor acceptance of a
range of substrates. Although natural enzymes display broad
specificity with respect to the aldol acceptor, the aldol donor
is usually limited to the natural substrate. For example, among
the ketones studied for antibody catalysis only acetone is a
substrate for a natural enzyme. In contrast, antibody aldolases
can use various aldol donors and acceptors. The antibodies accept
acetone, fluoroacetone, chloroacetone, 2-butanone, 3-pentanone,
2-pentanone, and dihydroxyacetone as aldol donor
substrates." This flexibility has advantages and
disadvantages for applications of these antibodies to synthesis.
It is helpful to be able to catalyze a variety of reactions with one antibody, but it is also helpful to have the catalytic antibody be sufficiently selective to catalyze only one reaction amongst those reactions that can potentially occur within the reaction mixture. The potential reactions include possibilities due to various possible aldol condensations of the initial reactants (due to the variety of alpha hydrogens on the aldol donor and to the possible directions of attack of the aldol donor on the aldol acceptor) and due to possible further reactions of the products of the initial reactions.
In cases where the variety of alpha hydrogens on the aldol donor allow for a variety of possible products, for instance with "reactions with 2-butanone and 2-pentanone, the antibodies exhibit some control of the regioselectivity of the aldol addition by preferential formation of the most substituted enamine [the activated form of the aldol donor within the catalytic antibody]." These cases gave product ratios of 94:4 and 73:27, respectively, in favor of the products formed from the most substituted enamines.
In a test of the selectivity of the direction of attack of an aldol donor (acetone) on an unsymmetrical aldehyde acceptor, an 11:1 ratio of the two possible products was formed under antibody catalysis, demonstrating the stereoselectivity of the catalyzed reaction.
In a test of side reactions, monitoring the concentration of reactants and the expected product during antibody catalyzed addition of acetone to a branched 3-phenyl-propionaldehyde acceptor gave results where "the perfect mass balance (top line [from Fig. 6 in the paper]) indicates that no side reactions, such as elimination or polymerization, occurred over that period [a 35 hour reaction run converting 90% of the acceptor to the addition product]. Thus, the antibody-catalyzed aldol reaction is an exceptionally mild method of C-C bond formation."
In summary, antibody catalyzed aldol condensations provide a promising technique for exploiting protein technology to extend synthetic capabilities, supplying both an application area for protein based nanostructures and possibly extending the range of building blocks available for nanotechnology.
The technologies described in the papers above are useful, but
they all rely on diffusion of chemical species to a molecular
machine such as a catalytic antibody in order for useful work to
be initiated. The paper below describes a mechanism for thermal
diffusion to trigger useful changes in a molecular machine, which
is a much faster mechanism, and which brings us closer to being
able to build molecular machines controlled by broadcast signals
(as in Merkle's replicator architecture) at a reasonable rate.
An advance in control of a protein's ligand binding ability comes from P. S. Stayton, T. Shimoboji, C. Long, A. Chilkoti, G. Chen, J. M. Harris, and A. S. Hoffmann, writing in [Nature 378: 472-474 30Nov95]. They bound a temperature sensitive polymer (poly(N-isopropylacrylamide)) to a mutant streptavidin, a protein that normally binds biotin. "Normal binding of biotin to the modified protein occurs below 32°C, whereas above this temperature the polymer collapses and blocks binding. The collapse of the polymer, and thus the enabling and disabling of binding, is reversible." Below the transition temperature the polymer "adopts a hydrated coil conformation" which spreads it out and keeps it from blocking the biotin binding pocket, while above the transition temperature the polymer "is a collapsed globule" which does block binding. Considered dynamically, the collapsing polymer acts as a thermomechanical actuator to shove the biotin molecule out of its binding pocket.
The authors suggest that this control of binding "could also be used to remove inhibitors, toxins, or fouling agents from the recognition sites of immobilized or free enzymes and affinity molecules, such as those used in biosensors, diagnostic assays or affinity separations. This could be used to 'regenerate' such recognition proteins for extended process use." From a molecular manufacturing point of view, thermal control of affinity is notable since it permits an engineered protein to grasp or discard a feedstock molecule in response to a broadcast thermal signal. Unlike actuation via binding of some "signalling" molecule, this mechanism does not require movement of additional chemical species for each actuation cycle.
The current state of the art of protein design and fabrication in particular - and the fabrication of nanometer scale structures in general - is still far from the point where it would be feasible to design structures in the absence of feedback on how successfully the target structure was actually fabricated. The papers below describe advances in diagnostic techniques which help provide this feedback.
Protein crystal growth is an important and difficult step in
obtaining structural information on proteins from x-ray
diffraction measurements performed on these crystals. A number of
advances in protein crystal growth are described in an article by
D. Normile in [Science 270: 1921-1922 22Dec95]. The
article describes several different advances presented at the
"Sixth International Conference on Crystallization of
The first advance was a quantitative analysis of improved quality of protein crystals grown in space. The analysis found that the "mosaicity" of crystals grown in microgravity was reduced. Macroscopic protein crystals are really composed of many small blocks of crystal which are somewhat misaligned. Mosaicity is a measure of how severe this effect is. "Reduced mosaicity can improve the signal-to-noise ratio and should result in improved precision in determining crystal structures." In experiments where lysozyme crystals were grown on Earth and in space under equivalent conditions, analysis of diffraction data showed "that the Earth-grown crystals had a mosaicity three times greater than the space-grown crystals."
A second advance was made in microbatch techniques, "in which crystals are grown in 1 to 2 microliter drops of a mixture of a protein and a crystallizing agent...Using even smaller droplets of solution isn't practical, because the drops dry up before the protein crystallizes. By covering each droplet with a layer of oil, Chayen and her colleagues found, they could prevent evaporation of the tiny microliter droplets and also protect the sample from contaminants in the air." In addition, the remaining evaporation appears to go through the oil, allowing tuning of the rate by selecting the type of oil and the thickness of the oil layer.
Writing in [Science 270: 1967-1970 22Dec95], D.L. Olson, T. L. Peck, A. G. Webb, R. L. Magin, and J. V. Sweedler describe a new NMR probe that obtains spectra from samples which are a factor of 130 less massive than those usable in a conventional NMR probe. "The microcoil is 1 mm long and encloses a sample of 5 nl within the [fused silica] capillary." A sustantial improvement in sensitivity came from immersing the microcoil in Fluorinert FC-43, which provides a match to the magnetic susceptibility of the copper coil, thereby improving the uniformity of the magnetic field, reducing the line widths of the signals, and improving sensitivity. "The ability to acquire high-resolution spectra on 5-nl samples with improved mass sensitivity enables a variety of uses for microscale NMR. Biological applications will greatly benefit from the ability to structurally identify molecules with submicrogram LODs [limits of detection]. As an example, a microcoil NMR spectrum of a seven-amino acid peptide is shown..." From the point of view of nanotechnology, this advance will assist in using NMR to confirm synthetic protein geometries (for sufficiently rigid proteins), or to screen synthetic proteins for rigidity based on the global dispersion of their NMR spectra, even when the synthetic difficulties limit sample sizes.
In addition to using individual protein molecules for nanotechnological applications, there are some larger scale components from biological systems that may also be useful components for nanotechnology.
One phase of research on microtubules has been capped as Y. Zheng, M. L Wong, B. Alberts, and T. Mitchison, writing in [Nature 378: 578-583 7Dec95], identify a gamma-tubulin complex which "acts as an active microtubule-nucleating unit which can cap the minus ends of microtubules in vitro." The complex appears (via electron microscopy) to be a ring 25-28 nm in diameter ("similar to the outer diameter of a microtubule (25nm)") with a thickness of about 10 nm. This is potentially useful since microtubules are important structural elements in cells. More precise knowledge of how to control their formation and orientation might allow us to exploit them as structural members in early nanomachinery. The control exerted by this complex may be quite precise. Microtubules which spontaneously assemble typically have 14 protofilaments (lines of protein molecules stretched out along the tubule) while those assembled in vivo (nucleated from the centrosome) have 13. Essentially the control of the nucleation acts like an initial circle of bricks in starting a spiral tower. It very precisely sets the pattern for the tower as a whole, even though the bricks themselves may permit several patterns.
Writing in [Science 269: 496-512 28Jul95] J. C. Venter et. al. (40 authors in total) describe the full sequencing of Haemophilus Influenzae Rd. This genome of 1,830,137 base pairs is the first complete genome sequence for a free-living organism. The strategy followed by this group "eliminated the need for initial mapping efforts and is therefore applicable to the vast array of microbial species for which genome maps are unavailable." Rather than preorganize the genome with a mapping approach, in this group's strategy "a single random DNA fragment library may be prepared, and the ends of a sufficient number of randomly selected fragments may be sequenced and assembled to produce the complete genome." A large part of the problem solved was computational, there had been a "lack of sufficient computational approaches that would enable the efficient assembly of a large number (tens of thousands) of independent, random sequences into a single assembly." From the point of view of nanotechnologists, the main effect of this advance is to fully specify an existing system that can replicate itself using simple feedstocks. The analysis of the genome sequence has thus far yielded 1743 regions which appear to code for proteins. These regions were matched against "a database of nonredundant bacterial proteins (NRBP) created specifically for the annotation... NRBP is composed of 21,445 sequences extracted from 23,751 GenBank sequences and 11,183 Swiss-Prot sequences from 1099 different species." Of the 1743 possible proteins, 1007 were identified with sufficient accuracy to allow assignment of their biological role. An additional 347 matched "hypothetical proteins" in the database, and 389 are unidentified. Ideally, it would be useful to know what function each of these proteins plays and to have tertiary structures for all of them, but this is clearly going to take some time. Other regions that have been identified in the genome include ribosomal RNA and transfer RNA.
In the analysis of proposed nanometer scale structures, a variety of analysis methods are useful in evaluating the structures before attempts are made to fabricate them. The most expensive of these methods, but the ones which make the fewest approximations and are therefore potentially the most trustworthy, are the quantum mechanical ab initio methods. The papers described below extend these methods.
Writing in [C&EN 29 14Aug95], S. Borman describes an improved ab initio technique called MEDLA (molecular electron density Lego assembler) developed by P. G. Mezey and P. D. Walker of the University of Saskatchewan, Saskatoon. "To demonstrate the technique, Mezey has published ab initio electron density calculations for bovine insulin, which contains 773 atoms, and for a bacteriophage protein containing more than 1,000 atoms." G. M. Maggiora, of Upjohn Research Laboratories, commented that "People have tried to divide molecules up in various ways...to estimate properties of larger molecules, but it's been largely unsuccessful. At least to my knowledge, this is the first example where something of this accuracy has been accomplished." Maggiora further commented that Mezey and Walker "calculate the smaller fragments that fit into large molecules in a very high level way...So for those molecules, they have information equivalent to the whole quantum mechanical wave function, the electron density, and any other properties (such as the molecular electrostatic potential) that can be derived from it." From the point of view of nanotechnologists, this should permit tip reaction calculations to be extended to much larger neighborhoods of the reaction center than has previously been possible, extending our confidence in these analyses.
Writing in [Scanning Microscopy 9: 381-386 1995]
K. Cho and J. D. Joannopoulos describe ab initio
simulations of interactions between a tungsten tip and a silicon
(100) surface. Their simulations used "a state-of-the-art
density functional pseudopotential conjugate gradient
scheme." The simulation "results predict that the tip
can be used to flip dimers on the surface, from one buckled
configuration to the other, reversibly, and without inducing
damage to either the intrinsic surface or the tip." The top
layer of a silicon (100) surface consists of a layer of silicon
dimers, and authors' simulations show that the dimers can reside
on the surface with a tilt of about 20° in either direction with
respect to the surface. In the absence of a tip, there is a
barrier of about 0.1 eV between the two tilted geometries.
Calculations of the tip-surface system showed that the silicon atom underneath the apex of the tip is stabililized by about 0.2 eV in the configuration which tilts it up, closer to the tip atom. "Consequently, the tip always measures a dimer atom in the up-flip geometry, resulting in a symmetric STM image of the dimer." The calculations indicate that gradually moving the tip up, while keeping it centered on one atom of the dimer, leaves the dimer with that atom elevated. At room temperature, thermal transitions would soon randomize the dimer between its two possible states. At low temperatures the authors suggest that "Since each dimer can be manipulated to exist in one of two equivalent states, it conceivably can be used to write and read one bit of information." This work advances nanotechnology because the authors' calculations, though fully quantum mechanical like Musgraves' hydrogen abstraction tool work, extend those calculations towards a system which is currently experimentally accessible. In addition, the authors' calculations uncovered a mode of operation which may be directly useful in information storage.
Carbon nanotubes, also known as fullerine tubes or graphitic tubes, are potentially useful to nanotechnologists in a variety of ways. Nested nanotubes may prove useful as bearings; nanotubes have been proposed as pores in nanostructures; and nanotubes are strong and stiff enough to be useful in a variety of structural roles. In the nearer term, nanotubes with well-characterized terminations would be attractive probe tips for scanning microscopy. The papers below describe studies of electron emission from nanotubes, which is sensitive to the details of nanotube terminations, and may help to drive control of these terminations.
Writing in [Science 270: 1179-1180 17Nov95], W. A. de Heer, A. Châtelain, and D. Ugarte describe a carbon nanotube field emission electron source, with a current density of 100mA/cm2, that is potentially useful for flat screen display applications. The utility of the nanotubes comes directly from their sharp tips, which concentrate the effective field strength by a factor of as much as 1300 above the uniform field in which the tubes are immersed. By contrast, conventional field emitters typically concentrate the field by a factor of 10. This group has recently been able to align arrays of nanotubes, allowing this technique to be used for large area cathodes. "The large field amplification factors are related to the geometry of the tube terminations. As shown by Iijima et. al. [Nature 356: 776 1992], the terminations have a variety of structures and are often conical with 20° opening angles, with radii of curvature at the tips that may be <1 nm. The density of emitting tips is estimated to be on the order of 105 cm-2. Because this is only a small fraction of the nanotube density (approx. 108 cm-2), only those tubes with particularly sharp tips that are favorably situated on the film emit efficiently." Since nanotubes with sharp, well-controlled terminations would be ideal for proximal probe work, it will be interesting to see if this group's developments become applicable to that area.
Writing in [Science 269: 1550-1553 15Sep95], A. G. Rinzler, J. H. Hafner, P. Nikolaev, L. Lou, S. G. Kim, D. Tománek, P. Nordlander, D. T. Colbert, and R. E. Smalley describe some electron emission experiments from individual nanotubes. They argue that the active field emitters are "individual linear carbon chains -- Cn atomic wires -- that have been pulled out from the open edges of the graphene sheets of the nanotube as shown in Fig. 4 and are held taut under the influence of the electric field." At room temperature the field emission current was 0.4-0.8 microamp. Oddly enough, the emission goes down sharply on heating the nanotube with a laser to approx. 1500°C, an effect that the authors attribute to "thermally induced evaporation of C3 and other small carbon radicals from the tip of the chain until chain is so short that the electric field is no longer sufficient to produce efficient emission. We expect that there is a steep temperature dependence of the effective resistance of the carbon chain, with nearly ballistic transport when the chain is cool, but frequent scattering and consequent chain heating and further increase in resistance once the vibrations of the chain become excited." The authors suggest that the chains "may turn out to be excellent coherent point sources of monochromatic electron beams and to have wide applications as probes, emitters, and connectors on the nanometer scale."
Fabrication with scanning probes has the advantage of giving
the experimenter direct control over the position where the
modification occurs, at the cost of fabricating structures one at
a time. Most of the atomically precise techniques using scanning
probes, however, have used feedback from an STM current to
determine if a selected atom has been moved. They are therefore
limited to conducting substrates. The new technique described
below relies on a novel feedback mechanism which avoids this
Writing in [Science 270: 1639-1641 8Dec95], E. S. Snow and P. M. Campbell described a novel nanometer scale fabrication method. The authors anodically oxidized a Ti film with an electrically biased (-10 volts with respect to the substrate) silicon AFM tip. The innovation in their technique was to monitor the electrical resistance of their structure during fabrication, automatically switching off the bias when the target resistance was reached. The current flow in the anodic oxidation itself is sufficiently low that it does not interfere with the resistance measurement. Thus far, the narrowest wire that the authors have fabricated "was obtained with a resistance increase that corresponds to a final wire width of 3 nm." It will be interesting to see if the authors are able to extend this technique to produce atomic scale constrictions. Unlike most fabrication techniques, the introduction of feedback into this technique might allow atom-by-atom monitoring of the oxidation process as a target structure is approached.
There have been many articles over the last several years
about nanometer scale particles. Typically, these articles have
described particles with a fairly narrow distribution of
diameters, but with curved surfaces that imply a fairly wide
distribution of surface structures. These particles have
interesting and potentially useful electronic properties, but
consist of too broad a range of isomers to be attractive building
blocks for atomically precise structures. The article below
presents some experimental evidence for better control, which
might make the new particles plausible components for atomically
Writing in [MRS Bulletin 23-32 Aug95], A. P. Alivisatos describes some work on nanometer-scale (1nm-5nm, in various experiments) crystallites, mostly of CdSe, with some Si and some HgS examples. Most of the article was on spectroscopy, but what I found notable was that many of the micrographs of the nanocrystals showed not merely well ordered interiors but also well-defined facets. "The crystallite, at 350 °C, has been made in just the right way in that the temperature is high enough that it will become crystalline inside and that, even during the few seconds during which it is formed, there is enough time for it to arrange itself and to facet." These crystallites may be potentially useful as building blocks for nanotechnology. The bonds in a crystal lattice can be considered to form a rigid, polycyclic molecule. The open question is whether a useful concentration of a single isomer of these molecules can be produced. The rounded crystallites that have appeared in many previous articles looked quite unpromising, considering the range of local structures that must be energetically or kinetically accessible to produce such a surface. Faceted crystallites, on the other hand, must include a much more limited range of surface structures, perhaps a range sufficiently small to make isolating a single species approachable.
Jeffrey Soreff is a researcher at IBM with an interest in nanotechnology.
From Foresight Update 24, originally published 15 April 1996.