IMM Report Number 24
In conjunction with Foresight Update 44
By Jeffrey Soreff
F. Moresco, G. Meyer, K.-H. Rieder, H. Tang, A. Gourdon, and C. Joachim, writing in [Phys. Rev. Lett. 86:672-675 22Jan2001] describe using an STM tip to reversibly switch the conformation of a medium-sized organic molecule at low temperatures. The molecule the authors worked with was Cu-tetra-3,5 di-tert-butyl-phenyl-porphyrin (Cu-TBPP), the same molecule used by Gimzewski et al. to demonstrate controlled positioning of molecules at room temperature on a copper surface. The current authors performed their experiments at low temperatures, 15 K. They also used a copper surface, but used the (211) surface, which “consists of (111) nanofacets and (100) intrinsic steps.”
The 3,5 di-tert-butyl-phenyl (TBP) groups on the Cu-TBPP molecule form hydrocarbon “legs” on the porphyrin plane of the molecule. When the molecule is free, in the gas phase, these legs are rotated perpendicular to the porphyrin. Similarly, on Cu(100), the legs remain vertical, and each provides a conducting path from the STM tip to the copper substrate, appearing on the STM images as one of four bright lobes roughly 5 angstroms high.
On Cu(211), two conformations of a leg are stable, one where it makes an angle of 10 degrees to the surface, lying essentially flat, and the other where it makes an angle of 55 degrees to the surface. The authors used the “elastic scattering quantum chemistry (ESQC)” approximation to calculate the theoretically expected STM images for these conformations, with a good match to experiment. The contrast between the images is sharp, with two lobes seen for the nearly flat 10 degree leg and just one for the 55 degree leg, and with an enhancement of over an order of magnitude for the tunnel current through the more tilted leg (also matching calculated values).
The authors used the STM tip in two different modes in order to switch a leg of a Cu-TBPP molecule between the two states. In order to switch it from the more highly tilted “ON” state to the nearly flat, low current “OFF” state, they brought the tip directly down on the peak of the leg, pushing it vertically down to the surface. In order to switch the “OFF” state to “ON”, they brought the tip into near contact (with a “tunneling resistance of 6 X 104 ohms”) with the Cu(211) surface, then moved it horizontally across the leg, pushing the leg from the side.
There are two implications of this work for nanotechnology. The reversible conformation change under experimental control exhibits another option for densely storing information using atomically precise systems. The agreement between detailed quantum mechanical calculations and experimental observations, including heavy atoms and a moderately large molecule, is encouraging for the use of these computational tools.
DNA is a valuable molecule for the construction of nanoscale structures. The predictable complementary pairing in DNA double helices permits better design of structures and of selective adhesion than in most other chemical systems. The papers in this section describe advances in constructing structures with DNA.
N.C. Seeman, writing in [Nano Letters 1:22-26 28Nov2000], describes a systematic view of the topology of the motifs used in DNA nanotechnology. He conceptually builds up a wide variety of structures from a “reciprocal exchange” operation that joins two DNA strands by nicking them, swapping two of the free ends, and rejoining them.
Seeman starts with the simplest case of reciprocal exchange, joining two DNA hairpin turns to yield a single conventional double helix, then shows how multi-arm junctions, and double and triple crossover molecules fit within this taxonomy. The double crossover molecules consist of two DNA helices with parallel axes, joined at two places by crossovers. The triple crossover molecules consists of three double helices with crossover joints between the center helix and its neighbors to the left and right.
Two double helices can also be bound together by a reciprocal exchange at every half-turn of the helix, to yield “a paranemic joining of two backbone structures [which is] very stable.” Seeman also reports in this article that he and co-workers “have constructed 6-helix bundles of DNA, wherein six helices are held together by means of reciprocally exchanged strands”
The 6-helix should be a remarkably stiff structure. Double-stranded DNA is already sufficiently stiff to behave as a rigid rod over short distances. B-DNA has a persistence length of 50 nm, the distance at which thermal vibrations are sufficient to bend it through about a right angle. The bending stiffness of a cylinder is sharply dependent on its diameter, scaling up as its fourth power. A hexagonal 6-helix, with a diameter triple that of individual double helices, should have 80 times their bending stiffness (and, proportionally, persistence length). This brings it well past the ~100nm scale needed for structural parts in an initial assembler.
In a second paper, D. Zanchet, C.M. Micheel, W.J. Parak, D. Gerion, and P. Alivisatos, writing in [Nano Letters 1:32-35 16Dec2000] describe the separation of mixtures of gold particle/DNA complexes into fractions bearing known numbers of DNA strands via electrophoresis.
The authors bound their single-stranded DNA to their gold nanocrystals via a terminal thiol. The binding is a statistical process, so that choosing the DNA:Au ratio sets the average number of strands attached to each nanocrystal, but the actual number bound to any individual crystal is sometimes above the average and sometimes below this ratio.
The authors ran several sets of experiments, using 5 nm and 10 nm crystals and DNA from 18 to 100 bases long. They used “2% or 3% agarose gels at 100 V, 6.7 V/cm, 1 h.” for electrophoresis. For the 5 nm crystals and 100 base DNA the 0, 1, 2, and 3 strand conjugates are clearly separated. With larger crystals or shorter DNA the resolution is less clear. The authors were able to recover the conjugates from their bands. They found that 50 base strands did not produce well resolved bands, but they suggest that they “could still use electrophoresis for short ssDNA conjugates, for example, by hybridizing the short strand with the long one, then isolating the conjugates and finally releasing the long ssDNA.”
The authors performed a number of experiments to rule out alternative explanantions for their bands, including non-specific DNA interactions with the nanocrystals and interactions of clusters of nanocrystals.
Nanocrystal/DNA composites are an important alternative approach to building DNA-directed nanostructures, potentially gaining stiffness from the lattices of the nanocrystals rather than from DNA junctions. In addition, the nanocrystals can play functional roles, such as conducting islands.
In a third paper, J. Richter, M. Mertig, W. Pompe, I. Mönch, and H.K. Schackert, writing in [Appl.Phys.Let. 78:536-538 22Jan2001], describe the contruction of metallic nanowires by the deposition of a palladium film on individual DNA molecules.
The authors deposited “lambda-DNA molecules, 48,502 base pairs in length” on interdigitated gold electrodes with micron-scale spacing. They aligned their DNA molecules across the inter-electrode gap by drying the droplet of DNA solution so that “the receding front of the evaporating drop…align the latter [the DNA] perpendicularly to the direction of the drying front.”
The DNA molecules were reacted first with a Pd2+ solution, then with a reducing agent. The Pd metal atom clusters produced by this treatment catalyze further Pd reduction, so initial clusters could be grown by repeating these steps. Initial stages of reduction produced isolated clusters, which further steps grew “through to wires with a continuous metal coating.”
Wires with diameters above about 50 nm showed ohmic resistance. This is an improvement over previous work with silver wires, which showed a gap in conductivity at low bias voltages.
Initial work with these Pd wires showed high contact resistance, ~5 kohms, even for wires as thick as 200 nm. The authors reduced this to <<1 kohm by “deposition of electron-beam-induced carbon lines written over the very ends of the [Pd] wires” improving their contact to the gold electrodes. After this step, resistance of one 50 nm wire dropped to 743 ohms.
This technique appears to be potentially useful in building nanometer-scale wires, but needs significant refinement to be incorporated into atomically precise systems. The authors’ films grow from initially separate clusters. Ideally, an atomically precise wire would grow from the DNA strand as a uniform film, preferably with a thickness comparable to the DNA strand rather than 10-20 times its diameter. Perhaps investigation of the base sequence in the vicinity of the initial clusters would reveal particularly effective sequences for nucleating Pd clusters. Dense, uniform nucleation might allow continuous coatings at smaller thicknesses.
Carbon nanotubes are essentially cylindrical sheets of graphite. They have extraordinary mechanical and electronic properties, with very high stiffness and, for ideal nanotubes with the right lattice indices, conduction as perfect quantum wires. These properties are modified in important ways by irregularities in or on the tubes, including by defects, adsorbed atoms, and the ends of the tubes, sometimes in useful ways and sometimes in undesirable ways. The papers in this section describe advances in the analysis (and, in one case, the manipulation) of these irregularities in single-walled carbon nanotubes (SWNTs).
In the first paper, A. Hassanien, M. Tokumoto, P.Umek, D. Mihailovic, and A. Mrzel, writing in [Appl.Phys.Let. 79:808-810 5Feb2001], describe the observation of standing wave patterns of charge density near the tip of a SWNT.
The authors observed atomically resolved images of their SWNTs at ambient conditions. A Pt-Ir tip was used for imaging. Their tubes were all around 1.6 nm in diameter. At low (50mV) bias in constant current (200 pA) mode they observed corrugations of 0.25 nm when far from the SWNT tip, “which compares nicely with the lattice constant of the armchair nanotube lattice” but a more complicated pattern including “peak pairing features … with period 0.74 nm” when close to the tip. This long period pattern extends roughly 6 nm from the tip, giving was to the lattice corrugations beyond this distance.
The authors interpret the STM image near the tip to reflect charge density variations from the reflections of the electron waves from the topological defects forming the tip. They “conclude that topology related defects such as pentagons or pentagon heptagon [pairs?] can be revealed from low bias STM topographic images.” At higher voltage biases (>60mV) “a wider energy window” contributes to STM current and partially washes out this pattern, giving a tip effect which “tends to decay within 1.5 nm away from the nanotube edge.
It would be very useful to know the exact topology of a SWNT tip, particularly if this SWNT is to be used for an STM probe or to manipulate a workpiece. Ideally, I would like to see this work extended to do explicit model fitting to try and deduce the exact bond topology of the tip from these observations. It would also be very useful to see if similar techniques can characterize derivatized tips, determining, for instance, exactly how many carbonyl groups are on the tip and where they lie.
In a second paper, M. Ouyang, J.-L. Huang, C.L. Cheung, and C.M. Lieber, writing in [Science 291:97-100 5Jan2001], describe STM imaging and spectroscopy of metal-metal and metal-semiconductor junctions within individual SWNTs. They call these intramolecular junctions, IMJs.
The authors imaged roughly 100 SWNTs in UHV at 5K. They found that , “about 10% of these were found to exhibit stable defect features under extended scanning.” The authors saw atomically resolved images at sites several nm from the junctions themselves. They were able to establish the indices of their SWNTs from their images. The authors model the interface in one of their tubes as due to a number of 5-7 defects in the hexagonal SWNT lattice. Unfortunately, the defect itself is not visible in the STM images, which the authors attribute to “(i) the defect may not be located directly at the upper surface of the SWNT circumference and (ii) the local density of states, which are measured in the STM experiment, do not necessarily reflect the atom positions.”
The authors were able to partially confirm their assignment of SWNT indices by spectroscopy. They measured the density of states from I(V) curves at various locations along their tubes. Far from the junctions, they saw a large (1.29 eV) gap between state density peaks in the metallic tube section and a small (0.45 eV) gap in the semiconducting section, as expected.
The authors measured state densities at a series of points along their tubes, crossing the IMJ. They saw a smooth transition from the metallic state density to the semiconducting state density, with no additional peaks in the junction region which would have indicated localized states, “suggesting that the M-S junction may behave as an ideal Schottky diode.” In contrast, they did detect localized states in their metal-metal IMJ.
The authors suggest “that STM characterization of IMJs can provide critical information and a feedback mechanism for growth studies designed to establish rational pathways for controllably producing IMJs in the future.”
In a third paper, M. Bockrath, W. Liang, D. Bozovic, J.H. Hafner, C.M. Lieber, M. Tinkham, and H. Park writing in [Science 291:283-285 12Jan2001] describe detecting scattering by individual defects in metallic single-walled carbon nanotubes.
The authors prepared SWNTs by CVD, and examined tubes with diameters <= 1 nm. They connected the nanotubes to leads defined by e-beam lithography and examined their conductivity on an oxide-covered silicon substrate that served as a gate electrode. Under these conditions metallic nanotubes show significant conductivity over the full range of gate voltages accessible under the experimental conditions. The authors found “surprisingly, however, ~60% of >50 metallic nanotube devices showed resistance that changed substantially with Vg.”
The authors investigated these changes by scanning a charged tip over the SWNTs, measuring the resistance between the ends of the nanotube as a function of tip voltage and position. They found that there were defects in the nanotubes at isolated positions which added a maximum increment to the resistance when the positions reached characteristic electrostatic potentials. They attribute this scattering to isolated structural defects within the SWNTs.
The authors found that “the reflection coefficients of defects range from 0.5 to 0.7 at room temperature” with “resistance peaks ~10 kilohms in height” for the defects in the example tube that they analyzed at length in the paper.
These scattering centers are both useful and a hindrance, depending on whether their formation can be controlled. They add a new type of voltage-controlled device to the repertoire of SWNT electronics, but they also interfere with the use of SWNTs as simple wires.
In a fourth paper, C. Zhou, J. Kong, E. Yenilmez, and H. Dai, writing in [Science 290:1552-1555 24Nov2000] describe constructing a diode within a carbon nanotube by adsorbing potassium along half its length.
The authors’ SWNT is 3.5 microns long, with a diameter of around 2 nm. Before adsorbing the potassium on to the tube, it behaves as a p-type semiconductor. The tube lies on an oxidized silicon substrate, which acts as a gate electrode, allowing the authors to establish the sign of the carriers in the tube by varying the gate voltage.
The authors covered half of the tube with a resist, poly-methylmethacrylate. They evaporated potassium on to the uncovered half of the nanotube, changing it from p-type to n-type. More precisely, they were able to see several regime of conduction as a function of gate voltage, “p+n, p+n+, pn+, and nn+ respectively” as gate voltage was raised. This sequence was visible at room temperature.
At 10K, the p+n+ regime displayed negative differential resistance, “consistent with the Esaki diode mechanism operating in heavily doped p-n junctions.” The peak-to-valley current ratio observed was 1.5. The authors estimate that the width of the depletion region between their p+ and n+ regions should be “on the order of several nanometers.” Coulomb blockade behavior is also seen at lower (~4K) temperatures.
The ability to add electrically active devices to SWNTs by modifying their adsorbates after the tubes themselves have been grown adds to the range of options for applying nanoscale systems to electronic applications.
Like SWNTs, nanowires have diameters starting at about a nanometer, though both types of structure can have considerably larger diameters. Unlike SWNTs, nanowires are solid structures, not hollow. They generally grow with a 3-D crystal lattice vector along the axis of the wire, while SWNTs usually have lattice vectors that wrap around the tube axis in a helix.
Y. Huang, X. Duan, Q. Wei, and C.M. Lieber, writing in [Science 291:630-633 26Jan2001], describe the use of fluidic alignment and surface patterning to arrange nanotubes and nanowires into functional structures. These techniques do not directly advance atomically precise fabrication, but they do “demonstrate ordering of NW [nanowire] structure on multiple length scales – organization of nanometer diameter wires with 100-nm to micrometer-scale separations over millimeter-scale areas.”
Flow of a solution of nanotubes or nanowires at a few mm/sec is sufficient to align the nanotubes or nanowires to within a few degrees. At ~10 mm/sec, “more than 80% of the NWs are aligned within ± 5 degrees of the flow direction.”
Under the authors’ conditions, once nanotubes or nanowires bind to a surface, they remained there. Coverage gradually increased over time. They show, for instance, deposition of InP NWs increasing (albeit somewhat nonlinearly) from zero to 250 NW/100 µm over a 40 minute deposition period. After one deposition period the deposited NWs are stable enough that a second layer can be deposited perpendicular to the first simply by changing the flow direction. The authors even display “an equilateral triangle of GaP NWs obtained in a three-step assembly process, with 60° angles between flow directions.”
The purely fluidic alignment of NWs does not provide perfectly uniform spacing. The authors show that the spacing can be regularized by patterning the substrate to bind to the NWs in lithographically defined stripes. In their example -NH2 terminii on the surface are used to attract GaP NWs. This patterning is not sufficient by itself to control the NW deposition, but it combined synergistically with fluidic control.
In a second paper, X. Duan, Y. Huang, Y. Cui, J. Wang, and C.M.Lieber, writing in [Nature 409:66-69 4Jan2001], describe the construction of light-emitting diodes from InP nanowires.
The authors grew their nanowires via laser-assisted catalytic growth. They used InP targets with “5% Au as the catalyst and 1% of Te or Zn as the doping element.” The wires grew in the <111> direction. TEM photos established the near perfection of the lattice in the core of the wires, albeit surrounded by “1-2 nm amorphous overlayer”, which is attributed to oxidation by air.
The authors established the polarity of the doping in their wires by operating them as FETs, depositing them on oxidized silicon wafers, which served as the gate electrodes, contacting the ends of the wires with evaporated metal electrodes (Ni/In/Au), and measuring conduction through the InP wires as a function of gate voltage. Positive gate voltage suppressed conduction through p-type, Zn-doped, wires and negative voltage through n-type, Te-doped, wires.
The authors were able to make p-n junctions “reproducibly by sequential deposition of dilute solutions of n- and p- type nanowires with intermediate drying”. They were also able to use electric fields between electodes to orient nanowires in the direction of the field. The nanowires are microns long, so the electrodes can be fabricated conventionally. The crossover between p-type and n-type nanowires acts as a diode, passing current from p to n but not the reverse. In the example shown in the paper, about 50 nA flow at about 3 volts of forward bias in a junction between a 29 nm diameter nanowire and a 40 nm diameter nanowire.
Light is emitted from the crossover, albeit at low efficiency (~0.001%). The authors attribute the low efficiency to competing processes from surface states, and hope to suppress these processes by passivating the nanowires.
These devices do have critical dimensions on the order of 30 nm, though they are not atomically precise. Since one of the mechanisms proposed for tuning the emission wavelength from these LEDs is adjusting their diameter, perhaps applications of short-wavelength emitters will prompt refinement towards small diameter nanowires. The light emission from the crossovers may also be directly useful as a light source for near-field optical scanning microscopy.
Organisms contain molecular components, some of which are directly useful, or can be re-engineered to become useful, as components in nanoscale technology. The papers summarized below describe work in employing enzymes and biomolecular motors.
C. Khosla and P.B. Harbury wrote a review article on modular enzymes in [Nature 409:247-252 11Jan2001].
Enzymes are of interest to nanotechnologists, both as guides to what active sites for nanoscale fabrication might resemble and as possible aids in synthesizing novel building blocks.
Normally, an enzyme binds to a substrate with the same active site in which it performs its catalysis, so adjusting one without damaging the other is difficult. Modular enzymes, where these functions are separated, do exist, though they are uncommon. Modularity is valuable because it permits the same reaction to be applied to a variety of substrates, forming varied products.
The most dramatic case in biochemistry is peptide bond formation in a ribosome, forming all the proteins found in living things.
The authors distinguish three types of modularity:
- enzymes for individual substrates where “catalysis and substrate specificity are separable”
- enzymes that bind multiple substrates with separate binding sites, and
- multi-enzyme systems that catalyze entire metabolic pathways
The separation between modular functions requires quite specific geometrical features in the folded enzyme; “it must minimally have a modular architecture.” Generally this requires separate domains, “stable globular fragments of proteins that may refold autonomously and carry out specific functions [here, recognition and catalysis].” This is not, however, a sufficient condition for functional modularity. Chimaeras composed of domains taken from different enzymes may still be non-functional.
In the case where catalysis and substrate specificity are separable, there appear to be two main modular strategies:
- Enzymes where substrate specificity is due to binding to a portion of the substrate molecule separate from the reaction center, in a separate domain in the enzyme. Examples include the “Fok I restriction endonuclease” which has one domain that binds to a specific DNA sequence and a second domain that cleaves the DNA “nine and thirteen bases away”. The first domain can be altered to change the target sequence.
- Enzymes where there is no substrate specificity other than for the presence of the reaction center. Here, “The substrate can be practically anything as long as the recognition feature is present.” Examples include lipases which act on any straight chain hydrocarbon. Structural studies have found that large portions of these substrates protrude out of the enzyme into the solvent.
From the viewpoint of nanotechnology, the first type of modularity would be ideal for synthesis of custom building blocks. For example, it would be useful to have a binding domain that recognized a carbon nanotube with a particular chemisorbed feature (e.g. an oxygen atom epoxidizing a particular double bond) on a tube of a particular size and chirality with the ability to combine it with a catalytic domain that would add additional groups to the surface at known relative locations.
The review cautions that the specificity of modular binding domains is somewhat limited because “the inter-domain linker must be floppy, so that binding by the recognition domain does not prevent correct positioning of the substrate in the active site.” Hopefully combinatorial exploration of variant linker sequences could reduce the impact of this constraint for specific choices of recognition structure and reaction center, albeit at a cost in true modularity.
The second type of modularity would be ideal in a more mature nanotechnology, where known workpieces are mechanically transported to work sites rather than being moved by diffusion. These reaction-site-specific enzymes could then be applied to specific sites on the workpiece, effectively acting as a tool tip.
In a second paper, R.K. Soong, G.D Bachand, H.P. Neves, A.G. Olkhovets, H.G. Craighead, and C.D. Montemagno, writing in [Science 290:1555-1558 24Nov2000], describe harnessing a biochemical rotary motor in an otherwise inorganic nanoscale system.
The authors picked F1-ATPase as their rotary motor. This molecule rotates while hydrolyzing 3 molecules of ATP to ADP per revolution. In has a diameter of 8 nm, a length of 14 nm, and as stall torque of roughly 90 pN-nm, which the authors note “are compatible with the sizes and force constants of currently producible nanomechanical structures.”
The authors attached the base of their motors to nickel metal posts and attached the gamma subunit, the rotor, to a nickel propeller. Both attachment points were selected by genetically engineering modifications into the ATPase.
The attachment to the base was made by adding a tail of 10 histidine residues to the three beta subunits of the ATPase. The histidine residues contain a nitrogenous heterocycle which binds directly to the nickel metal.
The attachment to the rotor was made by adding a unique cysteine residue to the gamma subunit, then covalently bonding a biotin derivative to it, then binding that through a series of biotin-streptavidin links to the nickel propellers.
Both the posts and the propellers were made with e-beam lithography. The posts were 200 nm high and 80 nm in diameter and were spaced 2.5 microns apart. The propellers were 150 nm in diameter and 0.75 to 1.4 microns long.
The authors measured rotation rates for their propellers of around 8 revolutions per second for the 0.75 micron propellers. Based on a calculation of viscous drag on the propellers, the authors calculate a motor efficiency of around 80%.
Of 400 propellers, only 5 were seen to rotate. I’m somewhat surprised that the cysteine attachment to the gamma subunit ever transmits a torque, because the sulfur on the cysteine only attaches to the subunit through a single bond. It would be interesting to see a detailed model of this region of the interface. Perhaps additional attachment points could transmit torque more reliably.
This motor is compact, with a well-understood structure. It has the advantage of functioning in a homogeneous solution, which could be useful in bootstrapping replicating assemblers, in contrast to, for instance, electrostatic motors, which would require a more complex infrastructure to connect to their power source. It will be directly useful in systems which require continuous motion, though it will require some sort of clutch in order to produce discrete motions in digitally controlled systems.
Jeffrey Soreff is an IMM Research Associate.