Graduate student MinFeng Yu of Prof. Rodney S. Ruoff's group at Washington University presented work done in collaboration with Zyvex L L C on "Manipulation of Carbon nanotubes using Scanning Probe Microscopes." Their four degree of freedom nanomanipulator performs 3-dimensional nanomanipulation studies under vacuum inside a scanning electron microscope, to permit visualization of the results of the nanomanipulation. For details:
Abstract that Mr. Yu submitted to the conference. Full paper, which contains link to movies showing the manipulation of nanotubes.
Dr. Tuan Vo-Dinh of Oak Ridge National Laboratory described the development of an antibody-based nanoprobe capable of monitoring biochemicals within single cells. A fiberoptic sensor is pulled to fine tip and coated with metal except at the tip, to which is bound 102 to 103 molecules of antibody to the molecule to be detected, in this case benzo [a] pyrene tetrol (BPT). The probe is inserted inside a cell, illuminated by a laser, and the fluorescence due to BPT molecules bound to the probe tip measured. Sensitivity was estimated at 40 attomoles of BPT.
Dr. Fumiya Watanabe of Kyushu University described his work on a "poor man's" multitip processor based upon 30 year-old field emitter array technology. The tip arrays produced various patterns etched onto a Si wafer, with 300 µm spacing between each mark, and at best 150 nm line width.
Dr. Pavel Krecmer of the University of Cambridge described how polarized light could cause an AFM cantilever made of a chalcogenide glass to move up and down by about 1 µm in a reversible fashion. He speculated that this effect could be the basis of very small atomic force microscopes, about 1 µm in size. However, the response time seems very slow, about 5 minutes.
Prof. Santosh Devasia of the University of Utah described how applying an inverse function of the known dynamics of piezo-positioners to compensate for vibrations permits optimizing how the probes are tracked, leading to an order of magnitude increase in scanning speed, from 50 Hz to 445 Hz.
Dr. Ralph C. Merkle of Xerox PARC presented a design study on a graphite casing to contain a simple replicating assembler. The simplified assembler would be composed of stiff hydrocarbons and would use highly reactive tools (like radicals and carbenes) in a neon atmosphere. There would be room to build within the assembler two copies of itself so that net replication is achieved even with the destruction of the original assembler during release of the products. A new proposal was presented for using acoustic control to provide both power and information for operation.
Dr. J. Storrs Hall of the Institute for Molecular Manufacturing analyzed the tradeoffs between a system architecture with simple assemblers that self-replicate to grow to macroscopic size, and an architecture composed of many different types of assemblers working together to make complex objects. He concludes that it is desirable to design a replicator to reproduce itself only for a few generations, and then build something else. "Furthermore, it is crucial to design replicators that can cooperate in the construction of objects larger and more complex than themselves."
Dr. Tahir Cagin of the California Institute of Technology reported both simulation and experimental work on friction and wear in diamond as a material for MEMS and NEMS applications. He noted that the bond strength in diamond is 80% greater than in silicon, which translates into 10 times the hardness, 5 times the toughness, and a wear rate 104 times less.
Dr. Stephen P. Walch of ELORET at NASA Ames Research Center presented computational studies of the interaction of hydrogen with diamond and silicon surfaces. These calculations address the issue of using gas phase hydrogen to satisfy dangling bonds on the surfaces of diamondoid nanodevices, rather than having to add individual hydrogen atoms one at a time.
The Tutorial on Critical Enabling Technologies for Nanotechnology was held Nov. 12, 1998. Tutorial Chair, Prof. Jan H. Hoh, Department of Physiology, Johns Hopkins University School of Medicine defined nanotechnology for the purposes of the tutorial as atomically precise manufacturing from the bottom up. The pathway to develop such technology is not yet clear so research in this area is extraordinarily interdisciplinary, ranging from computer science to biology. Sampling the relevant enabling technologies, four topics were presented in the hope of cross-educating researchers on some of the other approaches in use. Dr. James C. Ellenbogen of MITRE discussed molecular electronics both in the tutorial and during the Conference. More detailed summaries are available on his Web site. The discussions of simulation methods, scanning probe microscopy, and self-assembly follow.
At the end of the day, Dr. Ralph Merkle put the talks in perspective, pointing out that each of the four talks covered a specific area in the journey from present-day theoretical studies and experimental techniques to future molecular manufacturing technology. A few comments he made about the path to molecular manufacturing:
Professor Donald W. Brenner (from NCSU) gave a talk on simulating dynamics on the molecular scale.
Professor Brenner began his talk with a description of some application areas for atomistic simulations: Predicting the mechanical limits of nanostructures, the chemical reactivity of nanostructures, the electronic properties of nanostructures. Both small molecules and macroscopic amounts of materials can be analyzed without massive computation, but simulations are critical at the intermediate scales. He used a nanogear analysis as an example, showing how simulation turned local knowledge about bonding into structural level knowledge on what speeds and accelerations would make gear teeth slip or tear themselves apart.
Professor Brenner then described the series of approximations needed to simulate a structure like the nanogear with ~100-200 atoms. Ideally one would solve the time-dependent Schroedinger's equation for electrons and nuclei. In practice, this is too hard, and there are also useful insights from the approximate models.
The first approximation is the Born-Oppenheimer approximation. This separates the electronic and nuclear degrees of freedom. It finds the solution to Schroedinger's equation for the electrons in the presence of fixed nuclei. It then treats the energy from the electron's states as setting a new potential energy for the nuclei. The wavefunctions for the movement of the nuclei are now found from Schroedinger's equation using the new potential. This approximation effectively assumes that the electrons move fast enough that they can always keep up with nuclear motion. This can break down when there are low energy electronic states or when molecules are colliding rapidly in comparison to electronic movement. This approximation drastically reduces the number of particles, hiding the electrons in the potential energy of the nuclei.
The second approximation is to treat the nuclei as classical particles. This loses quantum effects such as nuclear tunneling or zero-point vibrational energy. There is some work on wave packet dynamics which retrieves some of these effects.
The third approximation is to approximate the potential energy from the electrons' motion with some function of the nuclear coordinates which is easier to calculate.
Professor Brenner touched on the complex history of atomistic simulations, extending from 1936 to the present, and involving at least 4 different communities with heterogeneous goals:
Professor Brenner then described three components of an atomistic simulation:
The integrator treats the differential equations of motion for the nuclei as difference equations with a finite step size. Typical step sizes are 10-15 - 10-14 seconds. Important criteria for an integrator are its accuracy and stability.
In an ideal simulation of a large enough system, a heat bath would form a natural part of the system, and the dynamics of the interface between, for instance, a nanogear and the heat bath would hold the nanostructure at nearly constant temperature, with only physically accurate excursions in temperature. Unfortunately, a heat bath would require many atoms, so its effect is simulated by less computationally expensive methods. Typically the velocities of atoms in the model are modified in some way or fictitious forces are added to the model. Many types of thermostats have been used. The Nose' thermostat incorporated many of the innovations of its predecessors and is currently the only one that generates the correct thermal fluctuations in small systems.
Interatomic force calculations have taken a number of approaches. The most CPU intensive approach is to directly calculate them from a quantum mechanical calculation of the electrons' state. Most development effect is currently going to this approach. Historically, they have been calculated with ad-hoc functional forms fitted to known molecular properties. Some functional forms have also been derived from quantum mechanical arguments.
Pairwise interatomic potentials have well known properties. The 1/r repulsion of the nuclei, the exponential attractions and repulsions of the core and valence electrons, and the 1/r6 attractions of Van der Waals forces are well known functions. Three-body (e.g. bond bending) and higher order terms have been used extensively, but they do not have a solid grounding in the electrons' quantum mechanics.
"Molecular Mechanics" generally refers to a popular potential function like Allinger's MM2 that includes bond stretches and angle bends between bonded atoms and pairwise forces between non-bonded atoms. It requires fixed bonding topology during a simulation.
Potential functions that can accommodate dynamic changes in bonding topology calculate effective bond order between nearby atoms on the fly. While they do not do a full quantum mechanical treatment of the electrons, they rely on a theorem that calculates the local density of electronic states from the paths with fixed numbers of "hops" between atoms. In particular, the binding energy of an atom is largely dependent on the width of the electronic orbital energy distribution. This width is largely dependent on the number of paths with two "hops", one away from the atom in question and one back to it.
In general, potential functions of some sort have been done with some success with all elements. What hasn't been done well yet are mixed metallic/covalent systems.
Professor Brenner concluded by touching on the major challenges in atomistic simulations today:
Dr. Jason P. Cleveland (from Digital Instruments) gave a talk on scanning probe microscopes, with an emphasis on force microscopes.
The fundamental physics of force microscopy are very similar to what one would expect from exploring a surface by touch manually, or to the physics of Edison phonographs. The major changes are in the tip radii, now 5-10 nm, and in the size and speed of the cantilevers on which they are mounted, now ~100 µm long, with resonant frequencies of ~1 MHz.
The cantilever deflection (and hence tip position) is typically measured by reflecting light from the cantilever and measuring the light beam's movement. The noise in this measurement is only about 0.1 Å in any 10 kHz bandwidth (above the base 10 kHz bandwidth which contains 1/f noise).
The cantilevers (and their attached tips) are typically moved with piezoelectric actuators. An actuator a few inches in size provides a few µms of travel.
There are three primary modes of AFM operation:
In contact mode a feedback loop maintains constant deflection (constant force) by varying the height of the tip while the tip is scanned. Ideally, one would prefer use tip-sample forces of ~ < 1 nN, but they range from 100 nN - 0.01 nN. In air, capillary and electrostatic forces add to the tip-sample force. Operating AFMs under fluids partially cancel capillary, electrostatic, and Van der Waals forces, allowing tip-sample forces of 0.1 nN - 0.01 nN.
Contact mode imaging can be destructive. Dr. Cleveland showed us an example of a (1 µm)2 image of epitaxial silicon, followed by zooming out to see a (2 µm)2 image of the same area. The second image had a blank square where the first image has been scanned, showing that the tip had leveled the (1 µm)2 area as it measured it.
In non-contact mode the cantilever is vibrated near its resonant frequency, and small shifts in the resonant frequency due to force gradients from the sample are detected. In this mode the tip is typically kept out of any adsorbed liquid layer. It typically senses forces 1 nm - 10 nm from the sample. This mode is nondestructive, but it has limited resolution because of the tip-sample distance. It also is sensitive to gunk and must "scan slowly to avoid contacting and getting stuck in [an] adsorbed layer."
In TappingModeTM the AFM tip is oscillated vertically with a typical amplitude of > 20 nm. A feedback loop maintains constant amplitude as the tip is scanned across the surface. Hitting the surface makes the amplitude drop. Surprisingly, even though the tip-surface interaction is very nonlinear on this scale, the tip oscillation stays close to sinusoidal.
The main advantage to TappingModeTM is that it eliminates the large lateral forces that are generated during contact mode scanning. Dr. Cleveland made an analogy to the macroscopic behavior of sandpaper. One can push on sandpaper with one's finger all day without damage, but sliding across it under pressure does damage. The large oscillation amplitude of this mode lets the tip pull back to where scanning does no damage. It also allows the tip to pull out of adsorbed layers. It "allows imaging of soft, fragile, and adhesive surfaces without risk of sample damage."
Typical oscillation frequencies in air are 50 kHz-500 kHz. At these frequencies, many soft surfaces act stiffer than under DC probing, so more structural details are visible.
Dr. Cleveland displayed some TappingModeTM images made in aqueous solution, including some in DNA. He said that base pairs were not visible, that resolution was limited to helix turns, around 30 Å. The outlines of base pairs are also rather similar. Hansma is working on imaging bases on strands separated by a polymerase, using micron-scale cantilevers to operate at higher frequencies.
Dr. Cleveland touched on a number of specialized techniques:
One notable technique (called LiftModeTM operation), involved scanning a line for topographic information, then using the topographic information to rescan it at constant height above the measured profile of the sample while measuring some other interaction.
In phase imaging, the relative phase of the drive signal on the piezoelectric actuator and of the cantilever deflection is measured. This turns out to be sensitive to materials differences, and is good for seeing material phase boundaries in samples where topography isn't informative. Dr. Cleveland went through the modeling of the cantilever oscillation and showed how amplitude and phase information could be combined to yield power dissipation in the sample, which is a good probe of viscoelastic properties.
Dr. Cleveland described the use of diamond tips mounted on stiff cantilevers to do destructive measurements of materials properties. These tips are currently made by polishing 3 facets by hand. Radii of 10 nm can be reached. Unlike the macroscopic case, where indentation is the better standardized measurement, films are better measured by scratching. Indentation tends to go through films, measuring substrate properties rather than film properties.
Dr. Cleveland went through the various regimes that a tip on a spring experiences as it approaches and retracts from a sample:
In one experiment a calcite sample under water had a complex force curve with multiple minima. It was possible to trace it, not by directly moving the tip, but by letting the tip hop between the minima (and nearby positions) thermally. The statistics of tip positions gave the force curve.
Professor Joseph A. Zasadzinski (from UCSB) gave a talk on Langmuir-Blodgett films.
Langmuir-Blodgett films are monolayers of amphiphiles (soap-like molecules with hydrophobic groups at one end and hydrophilic groups at the other end) which are formed on the surface of a water trough and transferred to a solid substrate. A sequence of layers can be deposited, giving one-dimensional molecular scale control of structure. Langmuir-Blodgett films have been known for seven decades. They have been studied intensively, with over 2700 papers published in the last 4-5 years. There are a long list of potential applications, which have thus far remained potential.
Langmuir-Blodgett films are made by
Zasadzinski described a great deal of work his group had done on the structure of Langmuir-Blodgett films. The films were imaged with AFMs, which was able to see individual methyl groups at the hydrophobic ends of individual molecules. The films were made of divalent metal (Cd2+, Mn2+, Pb2+, Mg2+, Ca2+, Ba2+, Zn2+) salts of arachidic acid (CH3(CH2)18CO2H).
Roughly speaking, small metal ions gave a simple, nearly hexagonal packing (though the unit cell is rectangular), very similar to the packing in polyethylene. Larger metal ions gave much more complex structures. These ions force the hydrocarbon tails to tilt in order to contact their neighbors. The spacing of CH2 groups along the tails favors certain tilt angles. The net effect is to generate complex structures with long repeat distances - even chiral packings.
Langmuir-Blodgett films haven't found applications because they are unstable. Professor Zasadzinski showed a dramatic series of images showing micron-scale cracks appear in a film within 15 minutes of formation. Over 48 hours, the film reconstructs to yield islands with roughly polygonal edges - analogs to facets in a 3D crystal.
The problem appears to be in the vertical structure of the film. A 3D cadmium arachidate crystal has the two arachidate groups bound to the metal going off in opposite directions ("splayed"). A Langmuir-Blodgett film of cadmium arachidate has to have all of the hydrophilic groups (the metal and both CO2H groups) on one side of the film and the hydrocarbon tails on the other (a "hairpin" structure). The hairpin structure has an area of 19.4 Å2 per molecule, while the splayed structure has an area of 18.0 Å2. The transformation "means holes. Lots of holes."
Films have been made stronger by putting in double bonds and crosslinking. The problem with this is that the act of crosslinking shrinks the film and shreds it. A new approach is to add one -OH group to the cadmium arachidate. This ties up one bond to the cadmium, so only one arachidate group is needed. This allows a monolayer structure that is more similar to the stable 3D crystal structure, and is less prone to reorganize.
Foresight staff Elaine Tschorn (left) and Tanya Jones (center) are supported by volunteer Norma Peterson (right) which keeps the conference running smoothly. (Not shown in picture Harriet Hillyer)
The portion of Update 35 that constitutes the IMM Report is on the IMM Web site: http://www.imm.org/.
From Foresight Update 35, originally published 30 January 1999.
Foresight materials on the Web are ©1986–2014 Foresight Institute. All rights reserved. Legal Notices.