A publication of the Foresight Institute
The interplay between theory and experiment is important to the advance of nanotechnology. Nanotech design requires accurate predictions of how components on this scale will behave. The papers in this section describe experimental work that helps refine and correct the theory of the electronic behavior of nanotubes, for the first two papers, and the theory of the detailed dynamics of chemisorption (which is relevant to feedstock capture), for the second two papers.
Two recent papers, [Nature 392:59-62 1Jan98] by J.W.G. Wildöer et. at., and [Nature 392:62-64 1Jan98] by T.W. Odom et. al., describe experimental confirmation of the predicted differences between metallic and semiconducting nanotubes. Nanotubes are essentially cylinders of graphite. They can be made in a variety of diameters, and with a variety of angles between the tube axis and graphite lattice. They are named by the distances (along two lattice directions) that one needs to travel in order to go around the circumference of the tube. For example, (10,0) tubes have a circumference that is 10 hexagons long in one direction, (10,10) tubes have a circumference that corresponds to going 10 hexagons in one direction, then 10 hexagons in a different direction. Theory has predicted that (n,m) tubes where n-m is divisible by 3 will be metallic (including all of the (n,n) "armchair" tubes, notably the (10,10) tubes which appear to dominate some samples) and all other tubes will be semiconducting.
Neither experiment directly probed conductivity along the nanotubes. Both of them examined the density of electron states as a function of energy. In metallic materials, there are a substantial number of states near the highest filled level (the Fermi level). These states allow the electrons to easily respond to an electric field by, roughly speaking, shifting to a nearby state, thus conducting current. In semiconductors, there is a gap between the highest filled level and the lowest empty level, so the electrons can't respond to a field by shifting to a nearby state. These experiments looked for gaps in the density of states.
In the Wildöer paper the nanotubes were probed at 4 K. An STM was used both to image the tubes and to measure their density of states. The images of the tubes showed few of the (n,n) "armchair" tubes, but the authors attributed this to a difference between how they and previous workers sampled the tubes. They looked at isolated tubes, while previous studies (which saw more (n,n) tubes) looked at tubes which clung to each other, forming ropes.
The authors plotted dI/dV vs. V for their tubes. This measurement depends on the density of states. They saw two types of curves, "one with gap values around 0.5-0.6 eV, the other with significantly larger gap values. The first category of tube is identified as semiconducting type, the second as metallic tubes." The larger gap is from the next set of states above and below the conducting band. The absence of the 0.5 eV gap in the dI/dV vs. V plot is an indication of metallic behavior. Another mark in the dI/dV vs. V curves distinguished semiconducting and metallic tubes. The tubes were examined on an Au(111) substrate, which has a higher work function than graphite, so it sucks electrons out of the tubes, making the semiconducting ones behave like p-type semiconductors. "In the semiconducting tubes, the Fermi energy seems to have shifted from the centre of the gap to the valence band edge." In metallic tubes it shifts much less, around 0.3 volts. The authors saw 6 metallic tubes and 12 semiconducting tubes. They concluded that "The central theoretical prediction, that chiral tubes are either semiconducting or metallic depending on minor details of wrapping angle or diameter, has been verified."
The Odom paper examined their nanotubes at higher temperatures, 77 K rather than 4 K, but saw much the same behaviour. They used measurements of diameter and angle between the tube axis and the lattice to assign (n,m) for their tubes. They examined nanotubes both in isolation and on the surface of nanotube ropes but still found that "more than half of the SWNTs observed either as isolated tubes or in ropes were found to be moderate gap semiconductors."
These papers help validate the theoretical band calculations that gave predictions of metallic behavior. They place very low limits on any residual band gaps, since the energy resolution of these experiments is much finer than kT at room temperature. Less desirably, they show that preparation methods which had been thought to yield fairly pure (10,10) nanotubes seems to produce a broader mixture, though the Odom authors suggest "that STM characterization combined with growth and purification studies will provide a rational pathway for producing structurally homogeneous samples of nanotubes in the future."
|"...experimental work that helps refine and correct the theory of the electronic behavior of nanotubes..."|
In [Science 279:545-548 23Jan98], M. McEllistrem, M. Allgeier, and J.J. Boland described experiments that examined the diffusion of dangling bonds (DB) on a silicon (100) surface passivated with deuterium. The observations were made with an STM. They found that the dangling bonds attracted each other. At low temperature they were present largely as pairs on underlying silicon dimers on the surface. As the temperature was raised, the dangling bond dimers dissociated, though "below 620 kelvin, the unpaired configuration most often observed corresponded to two DBs on adjacent silicon dimers." so the attraction between DBs extended beyond the confines of a dimer. Above 660 K, the DBs dissociated and diffused independently on the surface.
These experiments are important to nanotechnology because many proposals to construct extended structures start with a passivated surface, then remove selected atoms from it at a work site, then add moieties to the surface at these selected locations. Diffusion of the dangling bonds would interfere with this process. Experimental bounds on the rate of diffusion of dangling bonds sets bounds on this problem. The authors of this article found that "in contrast to previous results reported for Cl atoms on Si(100) that showed intradimer hopping at room temperature, all DB motion in the present study was suppressed below 500K." Thus, the temperature limits for using dangling bonds on a deuterium covered surface are less restrictive than those for a chlorine covered surface.
The paired dangling bonds are also important because "dissociative chemisorption of CVD gas-phase precursors requires at least two DBs at nearby surface sites." Dissociative chemisorption is directly analogous to the capture of a butadiyne feedstock molecule with two radicals in Merkle's http://nano.xerox.com/nanotech/hydroCarbonMetabolism.html so the properties of paired DBs may be important in feedstock capture systems.
Another article that is relevant to the capture of feedstocks in nanotech systems is [Science 279:542-544 23Jan98], written by D.E. Brown, D.J. Moffatt, and R.A. Wolkow.
They studied benzene on a Si(111) surface. Their "... focus is on the intrinsic precursor, a physisorbed intermediate state that does not arise from simple blocking of adsorption sites. ... but is inhibited from forming the stronger bond by a barrier related to the structural distortion or realignment of the molecule, the surface, or both." This is a very pure probe of the potential energy surface for the last stage in binding to a surface, since the physisorbed intermediate state places the benzene in almost the right place to bind to the surface.
The authors imaged both physisorbed and chemisorbed benzene with an STM. They were also able to convert chemisorbed benzene back to the physisorbed state with a voltage pulse. They measured the rate of chemisorption of physisorbed molecules, and used it to calculate a barrier of ~0.3 eV for the process. This will help calibrate models for the chemisorption process. Perhaps an analogous study could be done on a surface which is tailored to produce chemisorption events that correspond exactly to feedstock capture in a nanotech system.
|Foresight Update 32 - Table of Contents|
One of the most versatile materials we have today for building atomically precise 3D structures are proteins. The papers in this section describe advances in protein structural analysis, the design of protein/protein interfaces, and the design of enzymes.
Writing in [Science 278:1111-1114 7Nov97], N. Tjandra and A. Bax describe a novel NMR technique for structural analysis of proteins and certain other large molecules.
Two main techniques are used to find the 3D structure of proteins. One method, x-ray crystallography, measures diffraction of x-rays from crystals of proteins. It determines the positions of the atoms in the protein from the intensities of the diffracted x-rays. This method can be very accurate, but, unfortunately, it can't be applied to proteins that cannot be crystallized, and many proteins fall into this group.
The second method, NMR, is a type of spectroscopy that can be applied to proteins in solution. NMR stands for Nuclear Magnetic Resonance. In this technique the protein solution is placed in a magnetic field, which makes certain common magnetic nuclei in the protein (notably 1H, 13C, and 15N) absorb RF energy. The exact frequency at which they absorb depends on their environment, so the spectrum of absorption reveals some information about the protein. One detail of this process, the NOE (Nuclear Overhauser Effect) tells us "qualitative internuclear distances, which constitute the basis for macromolecular structure determination by NMR." Unfortunately, this effect tells us only local distances, so in building up a 3D model of a large molecule the errors build up and give us poor knowledge of the global structure.
Tjandra and Bax's new technique adds some information on the orientation of the bonds between certain atoms (and of similar vectors between nearby unbonded atoms). In their technique, the protein molecules are dissolved in a dilute aqueous solution of a liquid crystal. The liquid crystal forms an ordered phase (aligned with the magnetic field), and collisions with it weakly align the protein molecules.
|"...advances in protein structural analysis, the design of protein/protein interfaces, and the design of enzymes."|
The alignment of the protein molecules biases an effect that normally averages to zero in solution and makes it visible in the NMR spectrum. This effect, dipolar coupling, is just the effect of the magnetic field from one nucleus in the protein on the magnetic field seen by nearby nuclei. It depends on the angle between the line between the two nuclei and the direction of the magnetic field. Their liquid crystal forms pancake-shaped structures in water, so, for instance, pancake-shaped proteins bouncing off the structures tend to line up with them, and therefore line up with the magnetic field. The shape of the protein picks out the direction in which it aligns with the liquid crystal and therefore with the magnetic field. The dipolar coupling for a bond in the protein then tells us the angle between the bond and this special, shape-dependent, direction. Because this direction is the same from one end of the protein to the other, these measurements give us global information, relative orientation over long distances. It is nicely complementary to the local distance information that NOEs give us. It is also particularly useful for diagnosing the large structures that we wish to build for nanotechnological applications. The authors write that their "approach also holds the potential of extending NMR structure determination to proteins beyond [a weight of] 30 kD. ... It is not restricted to biological macromolecules and should be immediately applicable to numerous other types of molecules, including carbohydrates, peptides, and natural products."
In order to fully exploit the ability to build 3D structures from proteins, it will be necessary to build structures larger than single proteins, and therefore containing interfaces between proteins. Writing in [Science 278:1125-1128 7Nov97], S. Atwell, M. Ultsch, A.M. De Vos, and J.A. Wells describe how they remodeled an interface between human growth hormone (hGH) and its receptor protein. The binding between natural hGH and its receptor is strong enough to give a low dissociation constant, Kd, of 0.3 nM. The crystal structure of the complex between them is known. The authors use a mutation to change a critical tryptophan in the receptor to alanine, leaving a 150 cubic angstrom hole in the complex, and reducing "the binding affinity ... by a factor of >2500." The first mutation spoils the "lock and key" fit between the hormone and the receptor, essentially carving a hole in the lock.
The authors then looked for complementary changes in the hormone to restore binding to the new receptor. They used some structural information to do this, picking the five residues in hGH that pack against the original tryptophan. They used a selection process, rather than a design process to pick new choices for these hGH residues. They "randomly mutated hGH at five residues ... that pack against [the mutated receptor residue]" Approximately 107 mutant hGHs were examined. The mutants were selected for binding against the mutated receptor, using seven rounds of selection.
The authors found a mutant hGH which bound to the mutant receptor with a dissociation constant, Kd, of 14 nM, effectively retrieving about half of the binding energy that was lost on going to the mutant receptor. The mutant hGH, as expected, bound badly to the natural form of the receptor, with a Kd > 1000nM.
|"...we can indeed redesign protein/ protein contacts now."|
The authors did an x-ray crystal structure of the mutant hGH/mutant receptor complex. They found that two holes remained in this complex, one with a volume of 20 cubic angstroms and another with a volume of 40. They also found quite a bit of structural readjustment, writing: "The global changes that radiate outward from the pivot point necessitate substantial rearrangement in both main-chain and side-chain contacts throughout the interface. Some of the main-chain and side-chain van der Waals and hydrogen bonding groups that are 15 Å from the epicenter of these mutations move by up to 3 Å relative to their neighbors in wild-type hGHbc...As a result, four hydrogen bonds are lost and three new ones are gained between unchanged atoms in the mutant complex"
The implications for nanotechnology are, first, that we can indeed redesign protein/protein contacts now. On the other hand, this technique is clearly much harder than altering a DNA sequence and altering a complementary one to retain Watson-Crick pairing. The global structural readjustments also suggest that combining several redesigned interfaces in one protein could be very difficult, since the changes from one interface change could propagate to the other and interfere with it.
Writing in [Nature 391: 301-304 15Jan98], E. Quéméneur, M. Moutiez, J.-B. Charbonnier, and A. Ménez describe how they redesigned the catalytic site in cyclophilin to convert it from an isomerization catalyst to a hydrolysis catalyst.
The modifications do not just enhance an existing function. The authors wrote: "Cyclophilins catalyze cis-trans isomerization of X-Pro bonds but have never been reported to hydrolyze X-Pro bonds." The cyclophilins do, however, bind peptides containing proline (and therefore X-Pro bonds). The authors installed new catalytic machinery in this enzyme to exploit the existing binding site but substitute a new catalytic function for the existing one.
Other workers had found that a serine could cut peptide bonds because serine protease enzymes had been examined, and the hydrolysis had been found to depend on a serine approximately 3.2 angstroms from the carbonyl carbon of the peptide bond to be cut.
The authors of this paper knew the structure of "free and peptide-bound cyclophilins" and exploited that information to find four plausible places where adding a serine might cut the peptide bond of the substrate peptide. Of the four locations, two yielded inactive enzymes, two yielded active enzymes, and, of the two, one was 104 times better than the other.
The best serine mutant was refined by adding some neighboring catalytic machinery (a histidine and an aspartic acid) to generate "a catalytic triad-like machinery" like those in known serine proteases. The geometry of the known serine proteases was again used, this time to pick locations for the histidine and aspartic acid.
The authors found that "the kcat of the triple mutant was nearly 100-fold higher than that of EcypP-A91S [the mutant with the single serine]. In addition, its rate enhancement (kcat/kuncat) was about 109, a value comparable with those observed for a number of natural enzymes."
This work suggests that we will be able to exploit catalytic centers and binding sites of enzymes independently. Analogously, it suggests that separate design of substrate holders and tool tips in artificial nanosystems is likely to work, easing design of both components. More directly, the ability to mix and match catalytic machinery and binding sites from separate natural enzymes may help us exploit modified enzymes to widen the set of building blocks available for nanotechnology.
|Foresight Update 32 - Table of Contents|
Nanotechnology is fundamentally about constructing useful structures at the nanometer scale, particularly stiff ones, resistant to distortion from thermal vibration. The papers in this section describe advances in constructing zeolites, in building precise structures from silica, in constructing 3D structures with an STM, and in extending self-assembled structures.
Writing in [Science 278:2080-2085 19Dec97], X. Bu, P. Feng, and G.D. Stucky describe some advances in zeolite templating. Their zeolites have large cages interconnected with large diameter channels (8 angstroms) along several crystallographic directions. These channels are important because they permit reactants or adsorbates to diffuse more readily into the zeolite than they could if there were wide channels in only a single direction.
The authors used organic templates with both hydrophobic and hydrophilic regions. The hydrophobic regions in the template reserve space for the cage interiors and the hydrophyllic regions in the template generate charged walls in the final zeolite structure. The use of organic templates to control zeolite structure is a potentially information-rich technique for imposing patterns (in this case primarily charge patterns) on the final rigid structure.
In addition, the authors were able to synthesize zeolites with a variety of incorporated metals at their divalent sites: cobalt, manganese, magnesium, and zinc. This adds another degree of freedom to the synthesis. The amount of these metals that the authors were able to incorporate was larger than in previous work. They write: "With about 45% of the Al3+ sites replaced by Co2+, UCSB-6Co [one of their zeolites] has the highest framework transition-metal concentration of any large-pore zeolite type structure. ... Such a high transition-metal concentration could lead to novel catalytic and magnetic properties, but it also contributes to the lower thermal stability of UCSB-6 as compared to Al3+-rich molecular sieves."
The rigid, patterned structures of the zeolites themselves might serve as useful components in early nanomachinery. Alternatively, the zeolites might serve as patterned catalysts to help build nanosystem components.
Writing in [Science 278:1934-1936], J.W. Klaus, O. Sneh, and S.M. George described a way to produce very precise coatings of SiO2. When SiCl4 hydrolyses, it forms SiO2 and HCl, but simply mixing SiCl4 and water gives no control over how the formula units of SiO2 link up. In the authors' technique they dose a surface with SiCl4 and H2O alternately, separating the overall reaction into stepwise half reactions:
Si-OH* + SiCl4 -> SiO-Si-Cl3* + HCl
Si-Cl* + H2O -> Si-OH + HCl
(where the starred species are those bound to the surface). The reactants were applied as gasses, typically with pressures in the 15-200 millitorr range.
The authors note that: "The half-reactions are self-limiting; once a half-reaction goes to completion, additional reactant produces no additional film growth." As a result, each pair of reaction steps adds exactly one atomic layer of silica to the surface. This is analogous to what happens in the Merrifield technique for peptide synthesis. The reagents added in solution are present in excess, so they react with all of the resin-bound peptide. They add precisely one unit at a time because, after the reaction in a given step has occurred, the bound species is left in the wrong form to add extra units.
The authors confirmed the consistent growth with AFM measurements. Starting with a substrate with 2 angstroms of RMS roughness, 50 AB cycles of silica addition (adding 2.1 angstroms per cycle) left the surface with 3 angstroms of RMS roughness. They concluded that "the SiO2 film grows conformally over the Si(100) substrate with negligible roughening"
When only H2O and SiCl4 were used in these reactions, temperatures of >600 K were needed to grow the silica films. The authors were able to reduce the temperature to 300 K by adding pyridine as a catalyst. Analysis of the resulting films showed no detectable carbon, so the catalyst was indeed excluded from the product films. In addition, the measured "Cl/Si atomic ratio was <0.2 to 0.5%", so virtually all of the Si-Cl bonds were successfully broken during the hydrolysis half-reaction.
This technique allows the one-dimensionally atomically precise deposition of a stiff, polycyclic silica film. Perhaps the method can be combined with AFM patterning of the surface to allow 3D precision in depositing silica. The authors also suggest that "this catalytic technique may be general and should facilitate the chemical vapor deposition of other oxide and nitride materials," allowing other stiff materials to be deposited in a well controlled manner.
Writing in [MRS Bulletin 23:28-32 Jan98], G. Meyer and K.H. Rieder describe cryogenic, high vacuum manipulation of atoms and molecules on a copper surface with an STM. They moved Pb and Cu atoms and CO molecules. They operated in several different physical regimes, pulling Pb atoms with attractive forces from an STM tip, dragging atoms smoothly by pulling with the tip close to the substrate, and pushing a CO molecule with a repulsive interaction with the tip. This part of their work is reminiscent of Eigler's work in the early '90s.
A novel feature of their work was their choice of a Cu(211) surface. This surface is stepped. In the words of the authors "the surface consists of (111) terraces separated by (100) single step facets." Because of the inhomogeneity of the steps on this surface, Cu atoms can be extracted out of a step with the STM, eating into the terrace behind it. The ability to extract these atoms "leads to the possibility that extended parts of the substrate can be restructured in an atom-by-atom way. We demonstrate this in Figures 5d-5g, in which more than 30 atoms were laterally manipulated so that a rectangular area on the terrace below is laid open."
By moving atoms back and forth on a number of levels (readily available here because of the many steps), it should be possible to create atomically precise pits of almost arbitrary shape with multilayer depth, without ever needing to drag atoms over step edges (which was a limit in earlier work).
R.P. Sijbesma et. al., writing in [Science 278:1601-1604 28Nov97], describe an extension to the hydrogen bonding between bases in DNA. In their materials "units of 2-ureido-4-pyrimidone that dimerize strongly in a self-complementary array of four cooperative hydrogen bonds," rather than the three bonds in DNA, produce the associations that they use. The authors primarily use these units to build polymers held together by their hydrogen bonding units. They find that "the unidirectional design of the binding sites prevents uncontrolled multidirectional association or gelation." Their applications use the bonds to tune the rheology of their polymers, but not to generate atomically precise structures. It is possible that this new material, together with other quadruple bonding structures, might be useful in a generalization of DNA with equal or better specificity and stability. This could then be used to build self-assembled nanoscale structures.
Jeffrey Soreff is a researcher at IBM with an interest in nanotechnology.
From Foresight Update 32, originally published 15 March 1998.