Google Custom Search

A Debate About Assemblers — Whitesides Rebuttal


Many Future Nanomachines: A Rebuttal to Whiteside’s Assertion That Mechanical Molecular Assemblers Are Not Workable and Not A Concern.

K. Eric Drexler, Ph.D.; David Forrest, Sc.D.; Robert A. Freitas Jr., J.D.; J. Storrs Hall, Ph.D.; Neil Jacobstein, M.S.; Tom McKendree Ph.D.; Ralph Merkle Ph.D.; Christine Peterson

©Institute for Molecular Manufacturing, 2001


Self replicating molecular machine systems will revolutionize manufacturing at some point in the not too distant future. While there are many architectures for such systems, a frequently mentioned approach is the molecular assembler [2, 413]. Assemblers as proposed would be sized on the order of a few hundred nanometers, would be constrained to replicate only in a carefully controlled environment, and could be programmed to build a wide range of useful and highly desirable products.

In a recent article in Scientific American [1] George Whitesides said “Fabrication based on the assembler is not, in my opinion, a workable strategy and thus not a concern.” As the key technical issues have been published for some time (certainly since the publication of Nanosystems [2] in 1992), and as yet no fundamental barriers have withstood review in any technical forum, the claim by Whitesides that such fundamental barriers exist deserves thoughtful scrutiny.

Here, we review Whitesides’ stated objections to determine if any of them constitute fundamental issues that might make the eventual design and construction of a molecular assembler “unworkable”, or render the paths to a nonbiological assembler unproductive. Whitesides’ claims are interspersed with appropriate references, analysis and commentary.

Whitesides expresses concern about moving parts, friction and stiction:

Many interesting problems plague the fabrication of nanodevices with moving parts. A crucial one is friction and sticking (sometimes combined in talking about small devices in the term “stiction”). Because small devices have very large ratios of surface to volume, surface effects–both good and bad–become much more important for them than for large devices. Some of these types of problems will eventually be resolved if it is worthwhile to do so, but they provide difficult technical challenges now. We will undoubtedly progress toward more complex micromachines and nanomachines modeled on human-scale machines, but we have a long path to travel before we can produce nanomechanical devices in quantity for any practical purpose. Nor is there any reason to assume that nanomachines must resemble human-scale machines.

Analysis of the “stiction” issue dates back to at least 1959 when Feynman [14] suggested running bearings dry. It was subsequently analyzed in more detail by Drexler [15] in 1987 when he examined the symmetry considerations involved in making bearings that exhibit low static friction. The issue was again analyzed in Nanosystems [2], including the more general case of two surfaces sliding over each other. Merkle [16] analyzed bearings in greater depth, again concluding that bearings with very low static friction should be feasible. Experimental evidence that molecules can rotate freely in an appropriate environment is overwhelming, including, for instance, the work of Gimzewski [38] showing that the rotation of individual molecules on a surface can be stopped or started by changes in the local molecular environment, the work of Cumings and Zettl [39] demonstrating near-frictionless sliding of nested carbon nanotubes, and the common observation that molecules can rotate freely around a single bond and that even fairly large molecules often move freely on a surface.

While bearings and sliding surfaces are often of very low quality when made using existing lithographic methods, these involve surfaces that are imperfectly characterized and which are, at the molecular scale, very rough. Properly designed molecularly precise surfaces will be able to slide over each with little friction or wear for long periods of time (years or longer). Whitesides is correct that “we have a long path to travel before we can produce nanomechanical devices in quantity”, but this is an assessment of the current primitive state of molecular systems engineering, not a description of a fundamental barrier presented by physical laws.

Could these systems self-replicate? At present, we do not know how to build self-replicating machines of any size or type.

This statement is factually incorrect. Numerous proof-of-concept mechanical devices physically capable of self-replication from simple substrates have been known since the 1950s, including most notably the early working models of Jacobson [40] and Penrose [41], and the simple demonstration design of Morowitz [42].

We know, from recent biological studies, something about the minimum level of complexity in a living cell that will sustain self-replication: a system of some 300 genes is sufficient for self-replication. We have little sense for how to translate this number into mechanical machines of the types more familiar to us, and no sense of how to design a self-sustaining, self-replicating system of machines. We have barely taken the first steps toward self-replication in nonbiological systems [see “Go Forth and Replicate,” by Moshe Sipper and James A. Reggia; Scientific American, August].

We agree that there is much work to do, and much that can be learned from nature. In the biotechnology arena Arcady Mushegian of Akkadix Corp. [43] has looked at the genes present in the genomes of fully sequenced microbes to see which ones are always conserved in nature. He concludes that as few as 300 genes are all that may be required for life, constituting the minimum possible genome for a functional microbe. An organism containing this minimal gene set would be able to perform the dozen or so functions required for life — manufacturing cellular biomolecules, generating energy, repairing damage, transporting salts and other molecules, responding to environmental chemical cues, and replicating. Thus the minimal microbe — a basic cellular chassis — could be specified by an artificial genome about only 150,000 nucleotides bases in length. Glen Evans, now at Egea BioSciences, can currently produce made-to-order DNA strands that are 10,000 nucleotide bases in length [44] and is striving to increase this length by at least a factor of ten. The engineered full-genome DNA, once synthesized, could then be placed inside an empty cell membrane — most likely a living cell from which the nuclear material had been removed. Used in medicine, these artificial biobots could be designed to produce useful vitamins, hormones, enzymes or cytokines in which a patient’s body was deficient, or to selectively absorb and metabolize into harmless endproducts harmful substances such as poisons, toxins, or indigestible intracellular detritis, or even to perform useful mechanical tasks. At least one private company has already been formed to pursue the construction of these artificial biological devices [45].

Mushegian’s results [43] suggest that the minimal autonomous artificial biological replicator consists of at least 300 different (protein) “nanoparts”, multiple copies of each being assembled into a working biological robot. However, nonbiological engineered replicators: (1) need not be designed for operation on diverse organic substrates, (2) need not be capable of self-assembly but may employ positional assembly techniques, (3) need not be capable of surviving mutation or evolution, and (4) need not contain (and by necessity maintain to high fidelity) their own description within their own onboard structure. Hence, it is likely that a positionally-assembled mechanical replicator can be as simple, and designed using no more different kinds of parts than are found in the biological proof-of-concept.

The study and design of mechanical kinematic replicators [25] is yet in its infancy, but it is incorrect to assert that we have “no sense of how to design a self-sustaining, self-replicating system of machines.” A 1980 NASA Summer Study examined this very question at the insistence of Robert Frosch, the NASA Administrator at the time. This technical study [26] produced two distinct conceptual system designs for a macroscale self-replicating system of machines (both functionally scalable to smaller sizes). The final report [26], published in 1982, offered the following technical conclusions: “The theoretical concept of machine duplication is well developed. There are several alternative strategies by which machine self-replication can be carried out in a practical engineering setting. There is also available a body of theoretical automation concepts in the realm of machine construction by machine, in machine inspection of machines, and machine repair of machines, which can be drawn upon to engineer practical machine systems capable of replication. An engineering demonstration project can be initiated immediately, to begin with simple replication of robot assembler by robot assembler from supplied parts, and proceeding in phased steps to full reproduction of a complete machine processing or factory system by another machine processing system, supplied, ultimately, only with raw materials.” If there were a vital economic need, there is little doubt that a fully teleoperated self-replicating macroscale machine shop could be built today.

Compelling examples of artificial replicators exist in the robotics world. For example, the Japanese manufacturing company Fujitsu Fanuc Ltd. briefly operated the first “unmanned” robot factory in the early 1980s [26], then reopened an improved automated robot-building factory in April 1998 [46] that uses larger two-armed robots to manufacture smaller robots with a minimum of human intervention, starting from inputs of robot parts, at the rate of 1000 daughter copies (of individual robots) per year; apparently a different part of the factory uses a distributive warehouse system for automatically assembling the larger robots [47]. Other robotic manufacturers such as Yasukawa Electric [48] also use robots to make robot parts [49]. The existence of extensive theoretical work on self-replication [2527] and nanoscale manufacturing [2, 424] strongly support the claim — as yet unproven experimentally — that submicron mechanical self-replicating systems are feasible, amenable to human design, and will eventually be developed to support molecular nanotechnology

And other problems cast long shadows. Where is the power to come from for an autonomous nanomachine? There are no electric sockets at the nanoscale. The cell uses chemical reactions of specific compounds to enable it to go about its business; a corresponding strategy for nanoscale machines remains to be developed.

The fact that cells can use chemical reactions to power internal nanomechanical operations such as kinesin locomotion along microtubules and positional assembly of amino acids into linear peptide chains by ribosomes constitutes an existence proof that chemically powered mechanical operations by autonomous biological nanomachines are feasible. Corresponding strategies for powering autonomous nonbiological nanomachines have been proposed, studied, and in a few cases even demonstrated experimentally, casting some light into the “long shadows” that may have existed in the past.

Chemical power is most accessible to current experimental techniques. For example, Mao et al. [50] have constructed a mechanical device using DNA as structural scaffolding that might serve as the basis for a nanoscale robotic actuator. The mechanism has two rigid double-stranded arms a few nanometers long that can be made to rotate between fixed positions by introducing a positively charged cobalt compound into the solution surrounding the molecules, causing the bridge region to be converted from the normal B-DNA structure to the unusual Z-DNA structure. The free ends of the arms shift position by 2-6 nanometers during this fully reversible structural conversion, like a hinge opening and closing. The actuator can be powered indefinitely by this means.

Yurke and Turberfield [51] have synthesized another actuator using three single strands of artificial DNA which, when placed together, find their complementary partners and self-assemble to form a V-shaped structure. The open mouth of this nanotweezer can be made to close by adding a special “fuel” strand which binds to the single-stranded DNA dangling from the ends of the arms of the tweezers and zips them closed. A special “removal” strand, when added, binds to the fuel strand and pulls it away, opening the nanotweezers again. The cycle may then be repeated. These are examples that follow Whitesides’ suggestion of using biological parts and mechanisms to achieve new nanotechnology mechanisms.

Numerous similar examples of chemical power delivered to working artificial nanomotors have been reported in the experimental literature. For example, a 78-atom chemically-powered inorganic rotating nanomotor was synthesized in 1999 as a proof of principle by chemist T. Ross Kelly at Boston College [52]. Ben Feringa at the University of Groningen in the Netherlands has built an artificial 58-atom motor molecule that spins when illuminated by solar energy [53]. Montemagno’s group [54] has created an artificial nanomotor comprised of organic and inorganic components that is powered by ATP. The group’s first integrated molecular motor ran for 40 minutes at 3-4 revolutions per second. Subsequent motors have been operated for hours continuously by feeding them plenty of ATP. Such motors potentially may be very useful for powering autonomous nanorobots operating in the intracellular regime, as they could borrow their fuel source directly from the living cell. Montemagno reportedly is also trying to build a solar-powered, biomolecular motor-driven autonomous nanodevice, wherein light energy is converted into ATP which then serves as a fuel source for the motor, and has been quoted as saying: “We think we’ll be able to make autonomous devices that are powered by light on a scale of 1 micron or less, smaller than bacteria.”

Other modes of chemical power for future autonomous nanomachines have been proposed and discussed at length in the theoretical literature, but have not yet been subjected to rigorous experimental scrutiny and verification. For instance, hydrogen-oxygen, methane-oxygen, and ethanol-oxygen fuel cells are well-known and in wide commercial use; a medical nanodevice might prefer a glucose-oxygen fuel cell because of the greater availability of glucose in a typical biological system, as considered by Freitas [55]. Fuel cells appear attractive because the fundamental principles readily scale to the molecular size. To date, the closest laboratory demonstrations are microbial fuel cells containing captive bacteria or immobilized enzymes [56] which, when fed organic material, convert chemical energy into electricity that could be used to power tiny motors. Other possible strategies for chemically powered nanoscale machines have been proposed by Freitas [57] by analogy to demonstrated macroscale techniques, which theory suggests should be scalable to submicron dimensions.

Powering nanoscale machines using acoustic energy has been proposed and analyzed by Drexler [58], Merkle [12], and Freitas [59]. The analysis and design is straightforward. One approach is to use pistons (which have sliding surfaces). Maximum energy output can be readily determined by computing the change in Gibbs free energy between the extended and contracted piston configurations, given by the excess pressure times the change in volume. Small devices (a few thousand cubic nanometers) with modest pressure changes (one atmosphere) can readily produce sufficient energy (~300 zJ/cycle, or ~100 kT at room temperature) to power ratchets able to drive step motions in which individual steps are about a nanometer. The low-friction character of carefully chosen sliding surfaces in a pistonlike configuration of nested carbon nanotubes has been verified experimentally by Cumings and Zettl [39].

Drexler [60] has described nanoscale electrostatic motors that could be externally driven by electrical energy. Other strategies for powering autonomous nanoscale machines have been reviewed or proposed by Freitas [61], including heat engines, Brownian motors, mechanoelectric transducers, piezoelectric actuators, electrochemical transducers, light-driven proton pumps and photoelectric cells, rf antennas, radionuclide batteries, and various alternate strategies for transmitting power to tethered nanoscale machines from macroscale electrical, electromagnetic, hydraulic, acoustic, mechanical, and chemical energy sources. In fact, there appear to be many theoretically feasible organic and inorganic methods of delivering power to nanoscale assemblers or to nanorobots. The problem is not the theoretic feasibility of powering assemblers, but rather the fact that the experimental engineering is in its infancy. This should be a reason to move forward with the engineering work of building assemblers, rather than interpreted as a fundamental barrier.

How would a self-replicating nanomachine store and use information? Biology has demonstrated a strategy based on DNA, so it can be done, but if one wanted a different strategy, it is not clear where to start.

If onboard storage is deemed necessary, the biological strategy of storing information in polymers can readily be adopted — there is no fundamental need to adopt a different approach. It would also be possible to employ any one of numerous different classes of molecular memory. Excellent proposals for molecular data storage and, more recently, experimental results can be found in [110112].

However, for the “foreseeable future” it is likely that onboard storage of information will not be required by nanomechanical replicators. One example of an inherently safer and more flexible approach is the broadcast architecture [7, 62]. In this approach, information is broadcast by any of several means to the replicating component. The replicator can be built with an internal “dead man switch” that is automatically off unless activated by an encrypted broadcast signal [109]. The physical replicator becomes, in essence, a remote-controlled manipulator receiving instructions from the outside that guide it, step by step, in assembling a second remote-controlled manipulator. After some number of repeat cycles, the result is a large number of identical remote-controlled manipulators. These manipulators can then be used to assemble large numbers of useful product objects by altering the stream of instructions sent to the population of replicated manipulator devices. Acoustic broadcast (mentioned earlier) can be used to combine both power and information transmission in one convenient mechanism.

The assembler, with its pick-and-place pincers, eliminates the many difficulties of fabricating nanomachines and of self-replication by ignoring them: positing a machine that can make any composition and any structure by simply placing atoms one at a time dismisses the most vexing aspects of fabrication. The assembler seems, however, from the vantage of a chemist, to be unworkable. Consider just two of the constraints.

First is the pincers, or jaws, of the assembler. If they are to pick up atoms with any dexterity, they should be smaller than the atoms. But the jaws must be built of atoms and are thus larger than the atom they must pick and place. (Imagine trying to build a fine watch with your fingers, unaided by tools.) Second is the nature of atoms. Atoms, especially carbon atoms, bond strongly to their neighbors. Substantial energy would be needed to pull an atom from its place (a problem for the energy supply) and substantial energy released when it is put in place (a problem of cooling). More important, a carbon atom forms bonds with almost everything. It is difficult to imagine how the jaws of the assembler would be built so that, in pulling the atoms away from their starting material, they would not stick. (Imagine trying to build your watch with parts salvaged from another watch in which all the parts were coated with a particularly sticky glue: if you could separate the pieces at all, they would stick to your fingers.) Problems with the assembler are also discussed by Richard E. Smalley in his essay on page 76.

We have critiqued Smalley’s essay elsewhere [113], and found the “problems” he raised concerning the infeasibility of the assembler to be without merit. “Pincers” or “jaws” are imprecise and misleading metaphors for an assembler manipulator end-effector, because the assembler designs proposed by Drexler or Merkle have contained no such pincers or jaws. The active tip of a mechanosynthetic manipulator tool need be no larger than a single atom, covalently attached to a progressively larger inverted conical handle structure. Hence the active site of the proposed tool need be no larger than the atom it must pick and place. A wide range of possible mechanosynthetic reactions and mechanisms has been proposed and analyzed theoretically by Drexler [63], Merkle [9, 11], and others [21, 2833]. Significant modeling work has been done on mechanosynthetic reactions that should be useful in the controlled synthesis of diamondoid structures.

For example, the ability to selectively remove (or abstract) a hydrogen from a hydrogenated diamond surface appears to be a fundamental reaction in the controlled synthesis of diamond and related structures. The theoretical analysis of the hydrogen abstraction tool has involved the work of many people, including: Donald W. Brenner [21, 29, 30], Richard J. Colton [21], K. Eric Drexler [64], William A. Goddard III [28], J.A. Harrison [30], Jason K. Perry [28], Ralph C. Merkle [11, 28], Charles B. Musgrave [28], Michael Page [29], O.A. Shenderova [30], Susan B. Sinnott [21, 30], and Carter T. White [21]. The institutions involved include the Materials and Process Simulation Center at Caltech; the Department of Materials Science and Engineering at North Carolina State University; the Institute for Molecular Manufacturing; the Department of Chemicals and Materials Engineering at the University of Kentucky; the Chemistry Department of the United States Naval Academy; the U.S. Naval Research Laboratories, Surface Chemistry Branch; and the Xerox Palo Alto Research Center. Related work on hydrogen abstraction has progressed on the experimental front as well. For instance, Lyding et al. [6567] have demonstrated the ability to abstract an individual hydrogen atom from a specific atomic position in a covalently-bound hydrogen monolayer on a flat Si(100) surface, using an STM tip in ultrahigh vacuum.

Depositional covalent mechanosynthesis with atomic positional precision has also been demonstrated experimentally. For example, Ho and Lee [68] used an STM tip first to locate two carbon monoxide (CO) molecules and one iron (Fe) atom adsorbed on a silver surface in vacuum at 13 K. Next, they lowered the tip over one CO molecule and increased the voltage and current flow of the instrument to pick up the molecule; then they moved the tip-bound molecule over the surface-bound Fe atom and reversed the current flow, causing the CO molecule to covalently bond to the Fe atom, forming an iron carbonyl Fe(CO) molecule on the surface. Finally, the researchers repeated the procedure, returning to the exact site of the first Fe(CO) and adding a second CO molecule to the Fe(CO), forming a molecule of Fe(CO)2, which in subsequent images of the surface appeared as a tiny “rabbit ears” structure, covalently bound to the silver surface. Ho’s group has also demonstrated single-atom hydrogen abstraction experimentally, using an STM [69].

If steric constraints near the tool tip make it unexpectedly difficult to manipulate particular individual atoms or small molecules experimentally with sufficient reliability, a simple alternative is to rely upon conventional solution or gas phase chemistry for the bulk synthesis of nanoparts consisting of 10-100 atoms.

Would a nanosubmarine work if it could be built? A human-scale submarine moves easily in water by a combination of a rotating propeller—which, in spinning, forces the water backward and the submarine forward—and movable planes that guide its direction.

The term “nanosubmarine” implies the necessity for purposeful powered locomotion by molecular nanorobots. Many proposed molecular machines do not require active locomotion or navigation. Such machines would include assemblers replicating in a vat of liquid [4], respirocytes flowing in the blood [34], or the like.

Even for devices requiring powered locomotion, the term “nanosubmarine” is another instance of an imprecise and misleading metaphor for a medical nanorobot that is commonly employed by the popular media to connote a nanomachine swimming through a liquid sea. Nanotechnology researchers recognize that “nanosubmarine” is not a technically precise description, and hence avoid its use in scientific discourse. This is because they realize, for example, that the Reynolds number (the dimensionless ratio of inertial to viscous forces) may be ~108 for a military submarine, ~105 for a human swimmer in water, ~10-2 for a 1-micron nanorobot locomoting at 1 cm/sec through the bloodstream, and ~10-5 for the even slower-moving typical flagellar bacterial swimmer. Since the onset of turbulent flow occurs at a Reynolds number above ~103, this implies that real submarines and human swimmers move through a fluid in which turbulence is commonplace and Newtonian forces are paramount. In contrast, nanorobots (and bacteria) inhabit a remarkably different environment in which all flow is laminar and viscous forces are paramount [70]. Purcell [71] once observed that for a man to be swimming at the same Reynolds number as his own sperm, he would have to be placed in a swimming pool full of molasses and then be forbidden to move any part of his body faster than 1 cm/min, roughly the speed of the minute-hand of a large wall clock. These facts will not prevent the development of medical nanorobots, even though the design and function of such nanorobots must necessarily differ significantly from those of macroscale submarines.

Bacteria that swim actually use structures—flagella—that look more like flexible spirals or whips but serve a function similar to a propeller. They typically do not steer a very purposeful path but rather dash about, with motion that, if all goes well, tends in the general direction of a source of food.

It is well-known that micron-size bacteria “do not steer a very purposeful path but rather dash about” because the purpose of their motions is to overcome the diffusion limit to absorption of essential nutrients from the local environment [71]. In general, outswimming diffusion requires semirandomized burst movements (most usefully along rising concentration gradients) over a characteristic distance Ls ~ D/vswim [71]. For bacteria moving at vswim ~ 30 micron/sec and absorbing small nutrient molecules with diffusion coefficient D ~ 10-11 m2/sec, then Ls ~ 30 microns, roughly the sprint distance exhibited by flagellar microbes such as E. coli. Interestingly, there seems to be a sharp minimum size limit of ~0.6 microns for free-swimming foraging microbes, below which size locomotion confers no apparent benefit to microbes [72].

In contrast, the theoretical power and materials import requirements for medical nanorobots are usually held to 10% or less of available diffusion currents, as a conservative design practice [34, 36], so that materials diffusion currents are usually not a significant operational consideration. Nanorobots which locomote through fluids will therefore have markedly different performance objectives (e.g. purposeful directional travel) than outswimming diffusion (as exemplified by bacteria), and so nanorobot design, performance, and limitations will likewise be markedly different from bacteria as well. Freitas [73] has reviewed or proposed numerous alternative methods for purposeful directional fluidic locomotion, including the flexible oar, the invaginating torus, the rolling bispheroids, the viscous helicopter, the use of surface traveling waves, propulsive cilia, the counterrotating dual screw drive, volume displacement, and viscous anchoring. The dual screw drive [114] appears to provide the best combination of speed, safety, directional control, and energy efficiency of the proposed designs.

For nanoscale objects, even if one could fabricate a propeller, a new and serious problem would emerge: random battering by water molecules. These water molecules would be smaller than a nanosubmarine but not much smaller, and their thermal motion is rapid on the nanoscale. Collisions with them make a nanoscale object bounce about rapidly (a process called Brownian motion) but in random directions: any effort to steer a purposeful course would be frustrated by the relentless collisions with rapidly moving water molecules. Navigators on the nanoscale would have to accommodate to the Brownian storms that would crash against their hulls. For ships of approximately 100 nanometers in scale, the destination of most voyages would be left to chance, because the tiny craft would probably be impossible to steer, at least in a sense familiar to a submariner. Cells in the bloodstream—objects 10 or 100 times more massive than a nanosubmarine—do not guide themselves in it: they simply tumble along with it. At best, a nanosubmarine might hope to select a general direction but not a specific destination. Regardless of whether one could make or steer devices at the nanoscale, they would not work for the sophisticated tasks required to detect disease if one could make them.

All extant technical designs for medical nanorobots [3436] describe them as ~1 micron in size and constructed of thousands to millions of nanoscale “parts.” This is approximately the same size as a typical bacterium or mitochondrion and ~1000 times the volume of the 0.1-micron “nanosubmarines” that Whitesides imagines. Medical nanorobots with complex nanoscale parts will be ~1-3 microns in size or larger, but may require mobility in fluid. These devices are distinct from molecular assemblers, which may be as small as 100 nm in size, but do not require mobility in fluid. The statement that “water molecules would be smaller than a nanosubmarine but not much smaller” is grossly in error, since a water molecule is only ~0.00015 microns in diameter, or ~10,000 times smaller than proposed medical nanorobots. This is an important distinction in the “bombardment” argument. According to Einstein’s approximation for Brownian motion [74], after 1 second has elapsed at room temperature a fluidic water molecule has, on average, diffused a distance of ~50 microns (~400,000 molecular diameters) whereas a 1-micron nanorobot immersed in that same fluid has displaced by only ~0.7 microns (only ~0.7 device diameter) during the same time period. Thus Brownian motion is at most a minor source of navigational error [75] for motile medical nanorobots.

Freitas [76] notes that motile nanorobots should be able to safely traverse in vivo fluid volumes at speeds up to ~1 cm/sec, up to 100 times faster than the 100-200 micron/sec velocity achieved during the purposeful locomotion of swimming sperm [77] and more than fast enough to negotiate capillaries and small blood vessels with ease. Vascular wall ambulation has also been proposed and described [78].

The fact that biological cells in the bloodstream do not locomote their way through human fluid volumes at high speed reflects the rather obvious truth that doing so would confer no significant evolutionary advantage — not that doing so is impossible for any fundamental physical reason, as Whitesides implies. Methods for steering fluid-borne motile medical nanorobots have been described by Freitas [73].

Parts of the “little submarine” strategy for detecting and destroying diseased cells in the body, such as cancer cells, would have to focus on finding their prey. In doing so, they would probably have to mimic aspects of the immune system now functioning in us.

While it is true that immunochemical molecular recognition is often the quickest and most reliable technique for detecting specific cell types, and one could do worse than mimic biological immune systems, many other techniques are potentially available to medical nanorobots to detect cancer cells and malignant neoplasms. For example, specific thermographic dermal patterns may signal the presence of osteoid osteomas, breast cancers, or melanomas [79]. The exact spatial coordinates of these patterns could be recorded in the nanorobot computer memory prior to injection, or could be transmitted to in vivo medical nanorobots during a therapeutic procedure [80].

In Nanomedicine, Freitas [80] points out that in vivo thermal mapping nanorobots or “thermographicytes” could detect subtle internal thermal anomalies reflecting the presence of interior hematomas, lipomas, myomas, edemas and hydromas, and slowly growing tumors could be found by searching for localized, shallow, but monotonic thermal gradients in the ~microkelvin/sec range in the historical data. Mobile nanorobots passing near any thermographicyte would interrogate the device’s thermographic data library and thus may examine all or part of the current whole-body coarse thermal map. Similarly, historical and current data can be transmitted to medical personnel who interrogate the in vivo nanorobotic communications network. In principle, an expanding (warm) tumor could be detected after growing to a volume encompassing a maximum pathological cell count of ~100. For continuous surveillance of an entire organ to microkelvin accuracy, Freitas [80] estimates that up to ~16 million monitoring units are required, dissipating at most ~0.3 watts which must also be taken into account. (The liver, for example, typically generates ~10 watts metabolically.) Of course, tumors which are metabolically less active are harder to detect by this means. Also, measuring heat from tumors located in organs having high metabolic activity (e.g. liver, kidney) is even more difficult, especially given the large caloric variations after consuming food or drink, or during normal hormonal responses, and given the relatively high normal rate of cell division in some organs (e.g. liver, gut) that must be distinguished from abnormal tumor growth.

Another approach to cancer detection by medical nanorobots proposed by Freitas [81] is the deployment of chemical demarcation markers to direct the nanorobots to particular locations. Such markers may include injected chemical plumes, time release implants, remote-triggered releasers (e.g. pressure-release multi-chemical-bearing labels), and tissue-coded markers (analogous to radiolabeled iodine concentrating in the thyroid gland, alkali metal ions seeking out bone, or cancer-cell-targeted light-sensitive drug molecules in photodynamic therapy) — all of which may establish useful localized chemical gradients. The use of such chemical beacons for chemonavigational triangulation is not efficient over large distances, since the required volume of signal chemical rises as the cube of the range. However, chemical localization may be moderately useful in close quarters. For instance, a micron-size nanorobot constructed of nanoscale parts, could be equipped with two antipodal chemical sensors that angularly resolve two periodic sources of dissimilar signal molecules to the limit of nanodevice rotation during a Brownian motion-limited measurement time. This means that a pair of 10-probe sensors taking measurements every ~1 millisec (concentration differential Dc/c ~ 20% over the full nanomedically-relevant range of Kd = 10-4 to 10-13 molecules/nm3) could resolve ~ 3° of arc, equivalent to two chemical point-emitters separated by ~1 micron at a range of ~20 microns (~1 cell width) from the nanorobot.

Detailed chemomapping [81] of individual patients may allow customization of standardized human chemographic profiles, e.g. to identify problem areas. Chemographicytes (surveyor-class medical nanorobots) may assist in the mapping process at up to ~KHz sampling frequencies. For example, regions of hypoxia develop in all solid tumors where rapidly dividing cells are supplied by inadequate or poorly developed vasculature. A continuously updated whole-body tissue oxygenation map would allow recognition of growing hypoxic regions, permitting prompt detection of nascent tumors.

The recognition of a cell as “normal” or “pathogen” or “cancer” is an extraordinarily complex process—one that requires the full panoply of our immune system, including the many billions of specialized cells that constitute it. No simple markers on the outside of most cancer cells flag them as dangerous. In many of their characteristics, they are not enormously different from normal cells.

This statement is factually incorrect. The full complexity of our immune system is not necessary to recognize a cell as normal or pathogenic. It is also not true that cancer cells bear no easily detectable extracellular plasma membrane markers. For instance, GM2 ganglioside is a well-known glycolipid present on the surface of ~95% of melanoma cells, with the carbohydrate portion of the molecule conveniently jutting out on the extracellular side of the melanoma cell membrane [8286]. GM2 and another ganglioside, GD2, are expressed in several types of cancer cells including small-cell lung, colon, and gastric cancer, sarcoma, lymphoma, and neuroblastoma [82]. This is not to minimize the complexity and variety of cancer cells, but the point is that while we should learn what we can from our immunological mechanisms, we should also not be limited by them. We have already extended their range in the macroworld by MRI, genetic marker screening, and enzyme concentration testing.

Properly-designed medical nanorobots used for surveillance should find it relatively straightforward and safe to extract a quick cytoplasmic sample from all tissue cells that are even slightly suspicious. For example, a whole-body map of the concentration of intracellularly sampled telomerase (or associated proteins such as TRF1/TRF2 [87], Ku, and tankyrase [88], or even telomerase mRNA levels) may provide a good first-cut toward a whole-body late-stage cancer map, since 85%-90% of all primary tumors display this biochemical marker [89]. (However, ~10% of human cancer cells maintain their telomeres without telomerase [90] and telomerase expression per se is not oncogenic [91].) Cancer cells also display above-normal concentrations of b1 integrins, survivin, sialidase-sensitive cancer mucins and leptin receptors such as galectin-3, and below-normal concentrations of b4 integrins. Numerous specific markers of disease (an emerging field known as molecular epidemiology [92]) are available for detection and mapping by nanorobots.

A little submarine that was to be a hunter-killer for cancer cells would have to carry on board a little diagnostic laboratory, and because that laboratory would require sampling devices and reagents and reaction chambers and analytical devices, it would cease to be little. Operating this device would also require energy. The cells of the immune system use the same nutrients as do other cells; a little submarine would probably have to do the same.

Bloodborne phagocytic biological cells such as neutrophils, lymphocytes, and monocytes manage to detect, engulf and digest pathogenic cells without elaborate onboard chemical laboratories, and phagocytes have evolved by trial and error, not through efficient and rational design. As Whitesides points out, human engineers have the advantage of being able to learn from nature. They may extend natural methods, or innovate new ones [36], as they have in the macroworld. Again, this is a medical and engineering problem, not a fundamental problem with physical limits.

Like natural hunter-killer cells, the simplest artificial nanorobotic phagocytes could employ chemical recognition sensors at their outermost surface of sufficient number and diversity to guarantee selectivity for a specific type of cancerous or pathogenic cell. Chemotactic sensor pads can in principle be quite small, on the order of ~100 nm2 each and perhaps 105-107 nm2 for complete cell detection arrays [36, 93]. The use of dedicated molecular binding sites permits the direct detection of target molecular species, largely eliminating the need for the antiquated (in this context) paraphernalia of bulk chemistry such as reagents and reaction chambers.

Chemotactic sensor pads that detect an adhesion force of ~100 pN through a mechanical sensor movement of ~10 nm require an energy input of ~1000 zJ per detection. Hence a 100 pW medical nanorobot that devotes just 1% of its onboard power supply to chemotactic sensing could afford to measure up to 106 adhesion contact events per second and still remain within the stated power budget constraints.

Medical nanorobots may indeed make use of a few of the same nutrients as immune system cells — in particular, as noted earlier, glucose and oxygen. A trillion (1012) therapeutic nanorobots each consuming 100 pW of power would produce a total whole-body waste heat of 100 watts — roughly equivalent to the human basal metabolic rate and representing an insufficient thermal load to trigger any significant response by the human thermoregulatory system [94].

Small machines will eventually be made, but the strategy used to make them, and the purposes they will serve, remain to be devised.

We agree with Whitesides that small machines will eventually be made. Strategies to make them have been discussed in the technical literature [2, 516]. The purposes they will serve, for instance in the medical arena [3, 3437], have also been discussed in the technical literature. As the strategies become more complete and more fully verified by experimental results, the purposes these machines may serve will become clearer and the need to confront their social, financial and ethical implications will become increasingly urgent [4].

Biology provides one brilliantly developed set of examples: in living systems, nanomachines do exist, and they do perform extraordinarily sophisticated functions. What is striking is how different the strategies used in these nanometer-scale machines are from those used in human-scale machines. In thinking about how best to make nanomachines, we come up against two limiting strategies. The first is to take existing nanomachines—those present in the cell—and learn from them. We will undoubtedly be able to extract from these systems concepts and principles that will enable us to make variants of them that will serve our purposes, and others that will have entirely new functions. Genetic engineering is already proceeding down this path, and the development of new types of chemistry may enable us to use biological principles in molecular systems that are not proteins and nucleic acids. The second is to start from scratch and independently to develop fundamental new types of nanosystems. Biology has produced one practical means for fabrication and synthesis of functional nanomachines, and there is no reason to believe that there cannot be others. But this path will be arduous.

It is true that there is much to be learned from biological examples of nanomachines. It is also true that nonbiological nanomachines need to be designed so as to not interfere with biological systems [109]. However, it is not clear why the task of reverse engineering the entire baroque molecular structure and function of living cells involving potentially hundreds of thousands of distinct parts is inherently less difficult than the task of converting inherently scalable macroscale machine systems to nanoscale implementations involving potentially orders of magnitude fewer distinct parts. Our scientific knowledge of biological molecular machines spans mere decades whereas our direct experience with macroscale machines spans millennia.

Looking at the machines that surround us and expecting to be able to build nanoscale versions of them using processes analogous to those employed on a large scale will usually not be practical and in many cases impossible. Machining and welding do not have counterparts at nanometer sizes. Nor do processes such as moving in a straight line through a fluid or generating magnetic fields with electromagnets.

Table 2.1 (page 34) of Nanosystems [2] provides a useful overview of scaling laws. For instance, magnetic force scales adversely with size. Magnetic forces between nanoscale current elements are usually negligible in comparison with typical intermolecular forces [2]. On the other hand, electrostatic forces scale more favorably, leading to the design for an electrostatic motor measuring 100-400 nm in diameter [95].

Some macroscopic machines scale to smaller sizes quite readily. Others scale well in general, but require some modifications. Other approaches do not scale well. Specific proposals need to be examined, and specific conclusions need to be reached based on their merits. In general, macroscopic machines that rely on simple mechanical interactions (gears, belts, pulleys, levers, wheels, bearings, shafts, struts, ratchets, and many others) can be scaled to molecular sizes without undue difficulty. To our knowledge, there have been no proposals for machining or welding parts at the nanoscale. This form of argumentation is simply knocking a strawman.

Techniques devised to manufacture electronic devices will certainly be able to make some simple types of mechanical nanodevices, but they will be limited in what they can do.

The limitations depends upon what those techniques are. If they involve the positional assembly of nanoscale components synthesized via conventional bulk chemistry or other means, then they may have more general applicability.

The dream of the assembler holds seductive charm in that it appears to circumvent these myriad difficulties. This charm is illusory: it is more appealing as metaphor than as reality, and less the solution of a problem than the hope for a miracle. Considering the many constraints on the construction and operation of nanomachines, it seems that new systems for building them might ultimately look much like the ancient systems of biology. It will be a marvelous challenge to see if we can outdesign evolution. It would be a staggering accomplishment to mimic the simplest living cell.

Molecular nanotechnology researchers who accept the proposition that molecular assemblers are physically possible are neither “seduced” nor “charmed.” Rather, they have become convinced of the accuracy of this proposition only after studying the technical literature. These researchers are the least likely people to find “hoping for a miracle”. Rather, they are striving to create the theoretical underpinnings, and to achieve the experimental results that will bring the full potential of molecular assemblers to fruition. There are many constraints on the construction and operation of nanomachines, just as there are many constraints on the construction and operation of biological cells. The mere existence of such constraints does not imply impossibility or even implausibility, as demonstrated by biology.

Mimicking the simplest living cell would indeed be a staggering accomplishment. Fortunately it is not necessary to mimic all of the functions of the living cell in every nanomachine. For example, virtually all medical nanomachines will lack the ability of the biological cell to reproduce and to evolve new structures and functions on their own, making artificial nanomachines inherently safer than biological cells in terms of the potential for a loss of control. Most medical nanomachines will not require the capacity for complex onboard organic synthesis, and many nanorobots will not require the abilities of locomotion, manipulation, or various forms of global navigation or communication as are possessed by different types of biological cells.

Among those biological functions that must be mimicked by the mechanical nanodevice, more compact and efficient implementations can sometimes be found by rational design. A straight line drawn between two points is always a shorter path than a random walk. In some cases, evolution may have produced optimal designs. In other cases, designs survived because they were good enough to do the job, and the primary biological selection pressure was on other parts of the organism. We regularly build macroscale artificial systems that outperform biological systems in some dimensions. Examples include: numerical calculating chips, x-ray imaging systems, and subsonic jets for high-speed locomotion. There is no reason to assume that we will not be able to accomplish unique nonbiological designs at the nanoscale.

Biological structures work in water, and most work only in a narrow range of temperatures and concentrations of salts. They do not, in general, conduct electricity well (although some, such as the chloroplast and the mitochondrion, move electrons around with great sophistication). They do not carry out binary computation and communications. They are not particularly robust mechanically.

This is a useful summary of some of the major limitations of existing biological approaches, and some of the major motivations for considering alternative approaches.

Thus, a great many types of function must be invented if nanomachines are to succeed in nonbiological environments.

This is a useful recognition of the exciting engineering challenges ahead and the likelihood that mechanical nanomachines, once designed and built, may have greater functionality than biological-based systems, when placed in nonbiological environments.

And what have we learned from all this about the doomsday scenario of gray goo? If a hazard were to arise from nanomachines, it would lie in a capability for self-replication. To be self-replicating, a system must contain all the information it needs to make itself …

This assertion is likely to mislead scientists who are not familiar with the extensive literature on self-replication. Replicative architectures exist which do not require a system to explicitly contain all the information needed to make itself. For example, Laing [9698] has described a process of machine replication without self-description, also known as “replication by inspection”. Another example is “exponential assembly” [99] as recently proposed by researchers at Zyvex Corp.

…and must be able to collect from its environment all the materials necessary both for energy and for fabrication. It must also be able to manufacture and assemble (or allow to assemble) all the pieces needed to make a copy of itself. Biology has solved all these problems,…

Given that biology has had over 3 billion years to solve all of these problems, whereas a small number of human scientists and engineers have been trying to design and build kinematic replicators for only a few decades, we think progress has been remarkably rapid on this front. As Whitesides points out, it is always instructive to examine the solutions in nature’s 108-fold headstart, but this an opportunity, not a fundamental problem.

…and self-replicating biological systems—from pathogenic bacteria to cancer cells—are a danger to us. In computer systems, self-replicating strings of bits (computer viruses), although not material objects, are also at least a great nuisance, but only indirectly a danger, to us. If a new system—any system—were able to replicate itself using materials present in the environment, it would be cause for concern. So biology and chemistry, not a mechanical engineering textbook, point in the direction we should look for answers—and it is also where our fears about organisms or devices that multiply uncontrollably are most justified. In thinking about self-replication, and about the characteristics of systems that make them “alive,” one should start with biology, which offers a cornucopia of designs and strategies that have been successful at the highest levels of sophistication.

We agree that biological machines such as viruses and bacteria that have been selected to be robust in the natural environment are a greater danger than artificial nanomachines that have been designed with multiple safeguards and specifically engineered to not operate in the natural environment. However, we should all be seriously concerned about the potential for danger from microbial agents engineered from scratch [45], from accidentally engineered microbial agents [100], from accidental or malicious releases of modified or natural (archived) microbial agents [101], from the acquisition of multidrug resistance by existing microbial strains [102106], or from the natural emergence or evolution of new and deadly natural pathogens against which we currently have no defenses [107]. Nonbiological nanomachines could be designed to greatly reduce every one of these risks, and they should be developed as rapidly as possible for this purpose. This includes their potential use in detecting and defending against malicious or accidental misuse of biological or nonbiological systems.

But we now know enough to realize how far we are from reproducing self-replication in a nonbiological system. Fabrication based on the assembler is not, in my opinion, a workable strategy and thus not a concern. For the foreseeable future, we have nothing to fear from gray goo. If robust self-replicating micro (or perhaps nano) structures were ultimately to emerge, they would probably be chemical systems as complex as primitive bacteria. Any such system would be both an incredible accomplishment and a cause for careful assessment. Any threat will not be from assemblers gone amok but from currently unimaginable systems of self-catalyzing reactions.

As we have seen, Whitesides did not demonstrate that there is anything inherently “unworkable” about molecular assemblers or nanorobots. We believe that the medical, environmental, and industrial benefits are well worth the engineering challenges. However, we also believe that all powerful technologies are double-edge swords, and molecular nanotechnology is no exception to this rule [4, 108, 109]. It is not clear what time frame Whitesides regards as the “foreseeable future.” Estimates of the time frame needed to develop assemblers vary, but they are typically measured in decades. Thus, if the “foreseeable future” is the next five to ten years, then perhaps gray goo is not an immediate concern. On the other hand, scientific and engineering knowledge are on a growth curve that is accelerating exponentially, and international progress in molecular nanotechnology may surprise us. If the “foreseeable future” encompasses our lifetimes and the lifetimes of our children, then we certainly need to address these issues forthrightly [4, 108, 109].

References

1. George M. Whitesides, “The Once and Future Nanomachine,” Scientific American 285(September 2001):78-83.

2. K. Eric Drexler, Nanosystems: Molecular Machinery, Manufacturing, and Computation, John Wiley & Sons, New York, 1992; http://www.zyvex.com/nanotech/nanosystems.html; http://www.foresight.org/NanoRev/Bookstore.html#anchor1025139

3. Robert A. Freitas Jr., Nanomedicine, Volume I: Basic Capabilities, Landes Bioscience, Georgetown, TX, 1999; http://www.nanomedicine.com

4. K. Eric Drexler, Engines of Creation: The Coming Era of Nanotechnology, Anchor Press/Doubleday, New York, 1986; http://www.foresight.org/EOC/

5. Ralph C. Merkle, “Self replicating systems and molecular manufacturing,” J. Brit. Interplanet. Soc. 45(1992):407-413; http://www.zyvex.com/nanotech/selfRepJBIS.html

6. K. Eric Drexler, “Molecular nanomachines: Physical principles and implementation strategies,” Annu. Rev. Biophys. Biomol. Struct. 23(1994):377-405.

7. Ralph C. Merkle, “Self-replicating systems and low cost manufacturing,” in M.E. Welland, J.K. Gimzewski, eds., The Ultimate Limits of Fabrication and Measurement, Kluwer, Dordrecht, 1994, pp. 25-32; http://www.zyvex.com/nanotech/selfRepNATO.html

8. Ralph C. Merkle, “Design considerations for an assembler,” Nanotechnology 7(1996):210-215; http://www.zyvex.com/nanotech/nano4/merklePaper.html

9. Ralph C. Merkle, “Binding sites for use in a simple assembler,” Nanotechnology 8(1997):23-28; http://www.zyvex.com/nanotech/bindingSites.html

10. Ralph C. Merkle, “A new family of six degree of freedom positional devices,” Nanotechnology 8(1997):47-52; http://www.zyvex.com/nanotech/6dof.html

11. Ralph C. Merkle, “A proposed ‘metabolism’ for a hydrocarbon assembler,” Nanotechnology 8(1997):149-162; http://www.zyvex.com/nanotech/hydroCarbonMetabolism.html

12. Ralph C. Merkle, “Casing an assembler,” Nanotechnology 10(1999):315-322; http://www.zyvex.com/nanotech/casing

13. Nanomedicine [3], Chapter 2.

14. Richard P. Feynman, “There’s Plenty of Room at the Bottom,” Engineering and Science (California Institute of Technology), February 1960, pp. 22-36. Reprinted in B.C. Crandall, James Lewis, eds., Nanotechnology: Research and Perspectives, MIT Press, 1992. pp. 347-363, and in D.H. Gilbert, ed., Miniaturization, Reinhold, New York, 1961, pp. 282-296; http://www.zyvex.com/nanotech/feynman.html

15. K. Eric Drexler, “Nanomachinery: Atomically precise gears and bearings,” IEEE Micro Robots and Teleoperators Workshop, Hyannis, Cape Cod, November 1987.

16. Ralph C. Merkle, “A Proof About Molecular Bearings,” Nanotechnology 4(1993):86-90; http://www.zyvex.com/nanotech/bearingProof.html

17. K. Eric Drexler, “Molecular engineering: an approach to the development of general capabilities for molecular manipulation,” Proc. Natl. Acad. Sci. (USA) 78(1981):5275-5278.

18. Ralph C. Merkle, “Design-ahead for nanotechnology,” in Markus Krummenacker, James Lewis, eds., Prospects in Nanotechnology: Toward Molecular Manufacturing, John Wiley & Sons, New York, 1995, pp. 23-52.

19. Ralph C. Merkle, “Convergent assembly,” Nanotechnology 8(1997):18-22; http://www.zyvex.com/nanotech/convergent.html

20. Al Globus, Charles W. Bauschlicher Jr., Jie Han, Richard L. Jaffe, Creon Levit, Deepack Srivastava, “Machine phase fullerene nanotechnology,” Nanotechnology 9(1998):192-199.

21. Susan B. Sinnott, Richard J. Colton, Carter T. White, Donald W. Brenner, “Surface patterning by atomically-controlled chemical forces: molecular dynamics simulations,” Surf. Sci. 316(1994):L1055-L1060.

22. A. Globus, D. Bailey, J. Han, R. Jaffe, C. Levit, R.C. Merkle, D. Srivastava, “NASA applications of molecular nanotechnology,” J. Brit. Interplanet. Soc. 51(1998):145-152; http://www.nas.nasa.gov/Groups/Nanotechnology/publications/1997/applications/

23. J. Storrs Hall, “Architectural considerations for self-replicating manufacturing systems,” Nanotechnology 10(1999):323-330; http://www.foresight.org/Conferences/MNT6/Papers/Hall/index.html

24. Ralph C. Merkle, “Molecular building blocks and development strategies for molecular nanotechnology,” Nanotechnology 11(2000):89-99; http://www.zyvex.com/nanotech/mbb/mbb.html

25. John von Neumann, Theory of Self-Reproducing Automata, A.W. Burks, ed., University of Illinois Press, Urbana IL, 1966.

26. Robert A. Freitas Jr., William P. Gilbreath, eds., Advanced Automation for Space Missions, NASA Conference Publication CP-2255, 1982; http://www.islandone.org/MMSG/aasm/

27. Moshe Sipper, “Fifty Years of Research on Self-Replication: An Overview,” Artificial Life 4(Summer 1998):237-257, in “Special Issue on Self-Replication.” See also: Moshe Sipper, “The Artificial Self-Replication Page,” http://lslwww.epfl.ch/~moshes/selfrep/

28. Charles B. Musgrave, Jason K. Perry, Ralph C. Merkle, William A. Goddard III, “Theoretical studies of a hydrogen abstraction tool for nanotechnology,” Nanotechnology 2(1991):187-195; http://www.zyvex.com/nanotech/Habs/Habs.html

29. Michael Page, Donald W. Brenner, “Hydrogen abstraction from a diamond surface: Ab initio quantum chemical study using constrained isobutane as a model,” J. Am. Chem. Soc. 113(1991):3270-3274.

30. D.W. Brenner, S.B. Sinnott, J.A. Harrison, O.A. Shenderova, “Simulated engineering of nanostructures,” Nanotechnology 7(1996):161-167; http://www.zyvex.com/nanotech/nano4/brennerPaper.pdf

31. S.P. Walch, W.A. Goddard III, R.C. Merkle, “Theoretical studies of reactions on diamond surfaces,” Fifth Foresight Conference on Molecular Nanotechnology, 1997; http://www.foresight.org/Conferences/MNT05/Abstracts/Walcabst.html

32. Stephen P. Walch, Ralph C. Merkle, “Theoretical studies of diamond mechanosynthesis reactions,” Nanotechnology 9(1998):285-296.

33. Fedor N. Dzegilenko, Deepak Srivastava, Subhash Saini, “Simulations of carbon nanotube tip assisted mechano-chemical reactions on a diamond surface,” Nanotechnology 9(December 1998):325-330.

34. Robert A. Freitas Jr., “Exploratory design in medical nanotechnology: A mechanical artificial red cell,” Artificial Cells, Blood Substitutes, and Immobil. Biotech. 26(1998):411-430; http://www.foresight.org/Nanomedicine/Respirocytes.html

35. Robert A. Freitas Jr., “Clottocytes: Artificial Mechanical Platelets,” Foresight Update No. 41, 30 June 2000, pp. 9-11; http://www.imm.org/Reports/Rep018.php

36. Robert A. Freitas Jr., “Microbivores: Artificial Mechanical Phagocytes using Digest and Discharge Protocol,” Zyvex preprint, March 2001; http://www.zyvex.com/Publications/articles/Microbivores.html. See also: Robert A. Freitas Jr., “Microbivores: Artificial Mechanical Phagocytes,” Foresight Update No. 44, 31 March 2001, pp. 11-13; http://www.imm.org/Reports/Rep025.php.

37. Robert A. Freitas Jr., “Nanodentistry,” J. Amer. Dent. Assoc. 131(November 2000):1559-1566.

38. J.K. Gimzewski, C. Joachim, R.R. Schlittler, V. Langlais, H. Tang, I. Johannsen, “Rotation of a Single Molecule Within a Supramolecular Bearing,” Science 281(24 July 1998):531-533.

39. John Cumings, A. Zettl, “Low-friction nanoscale linear bearing realized from multiwall carbon nanotubes,” Science 289(28 July 2000):602-604.

40. Homer Jacobson, “On Models of Reproduction,” American Scientist 46(1958):255-284.

41. L.S. Penrose, “Self-Reproducing Machines,” Scientific American 200(June 1959):105-114.

42. Harold J. Morowitz, “A Model of Reproduction,” American Scientist 47(1959):261-263.

43. A.R. Mushegian, “The minimal genome concept,” Curr. Opin. Genet. Dev. 9(December 1999):709-714.

44. “Researchers build huge DNA chains,” BBC, 27 January 2000; http://www.crystalinks.com/biology5.1.html

45. EngeneOS, Inc.; http://www.engeneos.com/

46. FANUC, “New Robot Factory started the full Operation,” News Track, April 1998; http://www.fanuc.co.jp/en/news/h10/h10_04.htm

47. Robot Factory, http://www.fanuc.co.jp/eannai/servo.htm; citation courtesy of Tihamer Toth-Fejel, 2000.

48. Tohru Watanabe, Watanabefv@aol.com, Department of Robotics, Ritsumeikan University; personal communication to Tihamer Toth-Fejel, 21 March 2000.

49. Tihamer Toth-Fejel, “Modeling kinematic cellular automata: An approach to self-replication,” Environmental Research Institute of Michigan, Ann Arbor, MI, 4 October 2000.

50. C. Mao, W. Sun, Z. Shen, N.C. Seeman, “A nanomechanical device based on the B-Z transition of DNA,” Nature 397(14 January 1999):144-146.

51. B. Yurke, A.J. Turberfield, A.P. Mills, Jr., F.C. Simmel, J.L. Neumann, “A DNA-fuelled molecular machine made of DNA,” Nature 406(10 August 2000):605-608.

52. T.R. Kelly, H. De Silva, R.A. Silva, “Unidirectional rotary motion in a molecular system,” Nature 401(9 September 1999):150-152.

53. N. Koumura, R.W. Zijlstra, R.A. van Delden, N. Harada, B.L. Feringa, “Light-driven monodirectional molecular rotor,” Nature 401(9 September 1999):152-155.

54. C.D. Montemagno, G.D. Bachand, “Constructing nanomechanical devices powered by biomolecular motors,” Nanotechnology 10(1999):225-231; G.D. Bachand, C.D. Montemagno, “Constructing organic/inorganic NEMS devices powered by biomolecular motors,” Biomedical Microdevices 2(2000):179-184.

55. Nanomedicine [3], Section 6.3.4.5.

56. S. Sasaki, I. Karube, “The development of microfabricated biocatalytic fuel cells,” Trends Biotechnol. 17(February 1999):50-52.

57. Nanomedicine [3], Section 6.3.4.

58. Nanosystems [2], Section 16.3.2.

59. Nanomedicine [3], Section 6.3.3.

60. Nanosystems [2], Sections 11.7, 13.3.8, 14.4.8, 14.5.5(e), and 16.3.2.

61. Nanomedicine [3], Chapter 6.

62. Nanosystems [2], Section 16.3.2(a).

63. Nanosystems [2], Chapter 8.

64. Nanosystems [2], Section 8.5.4(c).

65. J.W. Lyding, K. Hess, G.C. Abeln, D.S. Thompson, J.S. Moore, M.C. Hersam, E.T. Foley, J. Lee, Z. Chen, S.T. Hwang, H. Choi, P.H. Avouris, I.C. Kizilyalli, “UHV-STM nanofabrication and hydrogen/deuterium desorption from silicon surfaces: implications for CMOS technology,” Appl. Surf. Sci. 130(1998):221-230.

66. E.T. Foley, A.F. Kam, J.W. Lyding, P.H. Avouris, P. H. (1998), “Cryogenic UHV-STM study of hydrogen and deuterium desorption from Si(100),” Phys. Rev. Lett. 80(1998):1336-1339.

67. M.C. Hersam, G.C. Abeln, J.W. Lyding, “An approach for efficiently locating and electrically contacting nanostructures fabricated via UHV-STM lithography on Si(100),” Microelectronic Engineering 47(1999):235-.

68. Wilson Ho, Hyojune Lee, “Single bond formation and characterization with a scanning tunneling microscope.,” Science 286(26 November 1999):1719-1722; http://www.physics.uci.edu/~wilsonho/stm-iets.html

69. L.J. Lauhon, W. Ho, “Inducing and observing the abstraction of a single hydrogen atom in bimolecular reaction with a scanning tunneling microscope,” J. Phys. Chem. 105(2000):3987-3992.

70. Nanomedicine [3], Sections 9.2.5 and 9.4.2.1.

71. E.M. Purcell, “Life at low Reynolds number,” Am. J. Physics 45(January 1977):3-11.

72. D.B. Dusenbery, “Minimum size limit for useful locomotion by free-swimming microbes,” Proc. Natl. Acad. Sci. USA 94(30 September 1997):10949-10954.

73. Nanomedicine [3], Section 9.4.2.5.

74. Albert Einstein, “Investigations on the Theory of Brownian Movement,” with notes by R. Furth, translated into English from German by A.D. Cowper, Methuen, London (1926), Dover Publications (1956); original paper in Ann. Phys. 17(1905):549.

75. Nanomedicine [3], Chapter 8.

76. Nanomedicine [3], Sections 9.4.2.4 and 9.4.2.6.

77. Michael E.J. Holwill, Peter Satir, “II.4 Generation of Propulsive Forces by Cilia and Flagella,” in J. Bereiter-Hahn, O.R. Anderson, W.-E. Reif, eds., Cytomechanics: The Mechanical Basis of Cell Form and Structure, Springer-Verlag, Berlin, 1987, pp. 120-130.

78. Nanomedicine [3], Section 9.4.3.

79. E. Francis J. Ring, Barbara Phillips, eds., Recent Advances in Medical Thermology, Plenum Press, New York, 1984.

80. Nanomedicine [3], Section 8.4.1.3.

81. Nanomedicine [3], Section 8.4.3.

82. Vicki Glaser, “Carbohydrate-Based Drugs Move Closer to Market,” Genetic Engineering News, 15 April 1998, pp. 1, 12, 32, 34.

83. S. Zhang, C. Cordon-Cardo, H.S. Zhang, V.E. Reuter, S. Adluri, W.B. Hamilton, K.O. Lloyd, P.O. Livingston, “Selection of tumor antigens as targets for immune attack using immunohistochemistry: I. Focus on gangliosides,” Int. J. Cancer 73(26 September 1997):42-49.

84. C.T. Kuo, P.J. Bostick, R.F. Irie, D.L. Morton, A.J. Conrad, D.S. Hoon, “Assessment of messenger RNA of beta 1–>4-N-acetylgalactosaminyl-transferase as a molecular marker for metastatic melanoma,” Clin. Cancer Res. 4(February 1998):411-418.

85. P. Livingston, “Ganglioside vaccines with emphasis on GM2,” Semin. Oncol. 25(December 1998):636-645.

86. N. Hanai, K. Nakamura, K. Shitara, “Recombinant antibodies against gangliosides expressed on tumor cells,” Cancer Chemother. Pharmacol. 46(2000):S13-S17.

87. L. Chong, B. van Steensel, D. Broccoli, H. Erdjument-Bromage, J. Hanish, P. Tempst, T. de Lange, “A Human Telomeric Protein,” Science 270(8 December 1995):1663-1667; D. Broccoli, A. Smogorzewska, L. Chong, T. de Lange, “Human telomeres contain two distinct Myb-related proteins, TRF1 and TRF2,” Nature Genet. 17(October 1997):231-235.

88. Susan Smith, Izabela Giriat, Anja Schmitt, Titia de Lange, “Tankyrase, a poly(ADP-ribose) polymerase at human telomeres,” Science 282(20 November 1998):1484-1487.

89. Nam W. Kim, Mieczyslaw A. Piatyszek, et al., “Specific association of human telomerase activity with immortal cells and cancer,” Science 266(23 December 1994):2011-2015; Lisa Seachrist, “Telomeres draw a crowd at Toronto cancer meeting,” Science 268(7 April 1995):29-30.

90. J. Travis, “Tick, tock, enzyme rewinds cellular clock,” Science News 153(17 January 1998):37.

91. Andrea G. Bodnar, Michel Ouellette, et al., “Extension of life-span by introduction of telomerase into normal human cells,” Science 279(16 January 1998):349-352.

92. Paul A. Schulte, Frederica P. Perera, eds., Molecular Epidemiology: Principles and Practices, Academic Press, San Diego, CA, 1993.

93. Nanomedicine [3], Sections 4.2.8 and 8.5.2.

94. Nanomedicine [3], Section 6.5.2.

95. Nanosystems [2], Section 11.7.

96. Richard A. Laing, “Some alternative reproductive strategies in artificial molecular machines,” J. Theoret. Biol. 54(1975):63-84.

97. Richard A. Laing, “Automation introspection,” J. Comp. System Sci. 13(1976):172-183.

98. Richard A. Laing, “Automaton models of reproduction by self-inspection,” J. Theoret. Biol. 66(1977):437-456.

99. George D. Skidmore, Eric Parker, Matthew Ellis, Neil Sarkar, Ralph Merkle, “Exponential assembly,” paper presented at the 8th Foresight Conference, Washington DC, November 2000; http://www.foresight.org/Conferences/MNT8/Papers/Skidmore/index.html

100. R.J. Jackson, A.J. Ramsay, C.D. Christensen, S. Beaton, D.F. Hall, I.A. Ramshaw, “Expression of mouse interleukin-4 by a recombinant ectromelia virus suppresses cytolytic lymphocyte responses and overcomes genetic resistance to mousepox,” J. Virol. 75(February 2001):1205-1210.

101. D.T. Dennis et al, “Tularemia as a biological weapon: medical and public health management,” J. Amer. Med. Assoc. 285(6 June 2001):2763-2773.

102. W.C. Albrich, M. Angstwurm, L. Bader, R. Gartner, “Drug resistance in intensive care units,” Infection 27(1999):S19-S23.

103. R.G. Postier, “Antibiotic-resistant organism infection,” Am. Surg. 66(February 2000):112-116.

104. A.S. Omrani, D. Pillay, “Multi-drug resistant HIV-1,” J. Infect. 41(July 2000):5-11.

105. S.B. Levy, “The future of antibiotics: facing antibiotic resistance,” Clin. Microbiol. Infect. 6(2000):101-106 (Suppl. 3).

106. R.N. Jones, “Resistance patterns among nosocomial pathogens: trends over the past few years,” Chest 119(February 2001):397S-404S.

107. Joshua Lederberg, “Infectious History,” Science 288(14 April 2000):287-293.

108. Robert A. Freitas Jr., “Some Limits to Global Ecophagy by Biovorous Nanoreplicators, with Public Policy Recommendations,” Zyvex preprint, April 2000; http://www.foresight.org/NanoRev/Ecophagy.html

109. “Foresight Guidelines on Molecular Nanotechnology,” Revised Draft Version 3.7: June 4, 2000; http://www.imm.org/guidelines.php

110. Mark A. Reed, James M. Tour, “Computing with molecules,” Scientific American 282(June 2000):86-93.

111. Philip G. Collins, Phaedon Avouris, “Nanotubes for electronics,” Scientific American 283(December 2000):62-69.

112. Philip G. Collins, Michael S. Arnold, Phaedon Avouris, “Engineering carbon nanotubes and nanotube circuits using electrical breakdown,” Science 292(27 April 2001):706-709.

113. K. Eric Drexler, David Forrest, Robert A. Freitas Jr., J. Storrs Hall, Neil Jacobstein, Ralph Merkle, Christine Peterson, “On Physics, Fundamentals, and Nanorobots: A Rebuttal to Smalley’s Assertion that Self-Replicating Mechanical Nanorobots Are Simply Not Possible”; http://www.imm.org/SciAmDebate2/smalley.php.

114. Nanomedicine [3], Section 9.4.2.5.2.

IMM would appreciate learning your thoughts on the above article.