Kinematic Self-Replicating Machines

© 2004 Robert A. Freitas Jr. and Ralph C. Merkle. All Rights Reserved.

Robert A. Freitas Jr., Ralph C. Merkle, Kinematic Self-Replicating Machines, Landes Bioscience, Georgetown, TX, 2004.


6.3.2 Molecular Assemblers are "Impossible"

The second argument against designing a molecular assembler is that it is simply impossible [14, 15, 202-206]. According to this argument, artificial programmable self-replicating manufacturing systems cannot be built by human engineers, and therefore any attempt to develop them constitutes wasted effort. This claim conveniently eliminates any concerns that might otherwise be raised by the potential development of nanoweapons systems. In this view, the risks of nanotechnology are intrinsically low, or limited to toxicology risks such as the inhalation of nanoparticles [235].

Initial arguments against feasibility were sweeping, and based on supposedly universal principles. Molecular assemblers were deemed impossible because thermal noise or quantum effects made molecular machines in general impossible [202-207]. These arguments lost ground when it was pointed out that biological molecular machines exist, and are able to function despite any limitations that might be imposed by thermal noise or quantum uncertainty. It may be recalled that in 1959, biologist Garrett Hardin [3039] noted that some geneticists called genetic engineering “impossible” as well. Today such criticisms of molecular assemblers survive only among ill-informed authors [3040] who are obviously unfamiliar with the voluminous technical literature on this subject.

More recent arguments concede the feasibility of biological molecular machines but attempt to argue that there are fundamental differences between biological systems and molecular assemblers able to synthesize a wide variety of non-biological materials. For example, in 2001 Vogel [13] wrote: “Future man-made nanosystems will certainly be able to perform a variety of functions, but a robot that is proficient in all three functions – movement in space, recognition of a chemically complex environment, and self-replication – will remain the fabric of dreams.” Most notably, in the same year Nobel chemist Richard Smalley wrote in a now-classic quotation [14]: “Self-replicating, mechanical nanobots are simply not possible in our world.”* In support of this bold claim, Smalley advanced two objections – the “fat fingers” and “sticky fingers” problems – which are, in reality, objections to mechanosynthesis and not to self-replication.

* The utility of organic chemists in commenting on nanosystems engineering may be similar to that of traditional explosives experts commenting on the Manhattan project. Most notably, Fleet Admiral William D. Leahy, Chief of Staff to U.S. President Harry Truman during World War II and formerly Chief of the Naval Bureau of Ordinance during 1927-31 (and later Chief of Naval Operations during 1937-39), maintained until Hiroshima that the atomic bomb being developed under the Manhattan Project would never work. “This is the biggest fool thing we have ever done,” he told Truman [2994] in 1945 after Vannevar Bush had explained to the President how the bomb worked. “The bomb will never go off, and I speak as an expert in explosives.” Five years later in his memoirs [2995], he frankly admitted his error.

The “fat fingers” objection is the assertion [14, 3041] that “there just isn’t enough room in the nanometer-size reaction region to accommodate all the fingers of all the manipulators necessary to have complete control of the chemistry.” How many fingers are necessary? The original claim [3042] that “chemistry is the concerted motion of at least 10 atoms” excludes a large part of well-known chemistry and was subsequently expanded [14] to the claim that: “In an ordinary chemical reaction five to 15 atoms near the reaction site engage in an intricate three-dimensional waltz that is carried out in a cramped region of space measuring no more than a nanometer on each side.” The supposed need for 15 “fingers” in “ordinary” chemical reactions, and the apparent impossibility of placing 15 probe tips in the small volume of a reaction site was then deemed “fundamental.”

But chemical reactions often involve less than five reactants, and frequently involve only two. These two reactants can be brought together with one reactant bound to a substrate and the second reactant positioned and moved by a single “finger” – as has already been demonstrated experimentally, for example, by Ho and Lee [3054], using an STM. Even if steric constraints near the tool tip made it unexpectedly difficult to manipulate particular individual atoms or small molecules with sufficient reliability, a simple alternative would be to rely upon conventional solution or gas phase chemistry for the bulk synthesis of nanoparts consisting of 10-100 atoms. These much larger nanoparts could then be bound to a positional device and assembled into larger (molecularly precise) structures without further significant steric constraints. This is the approach taken by the ribosome [3043] in the synthesis of proteins: individual amino acids are sequentially assembled into a specific atomically precise polypeptide without the need to manipulate individual atoms (Section 4.2). “Atomically precise” is a description of the precision of the final product, not a description of the manufacturing method. The “fat fingers” problem disappears.

The “sticky fingers” objection is the assertion [14] that “...the atoms of the manipulator hands will adhere to the atom that is being moved. So it will often be impossible to release this minuscule building block in precisely the right spot....these problems are fundamental....” The existence of some unworkable reactions does not preclude the possibility of a great number of workable reactions. If the “sticky fingers” problem* is truly fundamental, then no set of reactions could exist which allows the synthesis of a useful range of precise molecular structures. But we know this is untrue. The ribosome, a ubiquitous biological molecular assembler that suffers from neither the “fat finger” nor the “sticky finger” problem, readily synthesizes the class of atomically precise molecular structures known as proteins using positional techniques (Section 4.2). It is unclear why we should expect there to be no other classes of atomically precise molecular structures that can be synthesized using positional techniques. The experimental observation that ribosomes can synthesize polymers such as proteins under programmatic control appears to contradict the hypotheses that the programmatic synthesis of stiffer polycyclic structures such as diamond is “fundamentally” impossible, and that mechanical assemblers will never be built.

* The only actual proposal for using “sticky fingers” that the authors have been able to locate involves the materials-selective glue idea [3044] proposed in connection with a yet-to-be designed macroscale replicator intended to be built from 0.45-cm machined plastic Lucite blocks. In this informal proposal, a polyethylene hand would grasp the individual Lucite blocks and positionally assemble them into larger structures, after dipping in a methylene chloride glue which “will only bond Lucite to Lucite, but does not bond polyethylene.” Presumably the polyethylene hand would later be assembled using a Lucite hand that dips polyethylene building blocks into a glue that does not bond Lucite.

More directly, we can examine both those reactions that have been proposed specifically for use in a mechanical assembler and those reactions that take place using an SPM (scanning probe microscope). Drexler [208] discusses the mechanosynthetic reactions that might be used to synthesize some diamondoid structures of interest. Generally, these involve a single “finger,” i.e., a probe tip with a functionalized end that would cause a site-specific reaction on a growing molecular workpiece. Merkle [216] discusses several reactions which involve two, three, and even four reactants bound to the tips of molecular tools. Merkle and Freitas [1, 2322-2325] have extended this work, and Freitas [2338] in early 2004 filed the first known patent on diamond mechanosynthesis. A large and growing literature on relevant research work with SPMs – both theoretical [2323-2325, 3045-3050] and experimental [2986, 3051-3055] – supports the feasibility of site-specific reactions involving a reactive tip structure interacting with a surface or with a molecule on a surface.

Other than the current lack of working molecular assemblers (Section 6.3.3), which might be compared with the lack of rockets able to go to the moon in 1950, there appears to be little evidence to support the claim of impossibility and much evidence to refute it [16, 17]. The existence of a wide range of self-replicating biological systems, of new developments in biotechnology and programmable microbes (Sections 4.4 and 4.5) or biobotics [1875], and the existence of extensive theoretical work on self-replication and nanoscale manufacturing [197-226, 2323-2325] strongly support the claim that artificial self-replicating molecular manufacturing systems are feasible, amenable to human design, and will eventually be developed.

Others have argued that artificial self replicating systems are, in general, impossible [14, 204]. This claim is contradicted by the fact that compelling examples of artificial replicators exist in the macroscale world (Chapter 3). For example, a number of simple mechanical devices capable of primitive replication from simple substrates have been known since the 1950s [2], and self-replicating computer programs have been known at least since the 1970s [2]. The Japanese manufacturing company Fujitsu Fanuc Ltd. briefly operated the first “unmanned” robot factory in the early 1980s [2], then reopened an improved automated robot-building factory in April 1998 [712] 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 [713]. Other robotic manufacturers such as Yasukawa Electric [714] also use robots to make robot parts [711]. The manufacturing base of most industrialized countries, of many states or provinces, and even of some individual large municipalities can produce most of the material artifacts of which the base itself is composed, constituting yet another existence proof for artificial or technological self-replication. Finally, the world’s first macroscale autonomous machine replicator, made of LEGO® blocks, was built and operated in 2002 (Section 3.23.4). By contrast, the arguments that have been advanced against the feasibility of artificial self replicating systems in general and assemblers in particular [202-205] are of uniformly poor technical quality and display an astonishing ignorance of the relevant literature.

Feasibility can be demonstrated by exhibiting a single feasible design. Impossibility can only be proven by showing that all potential designs are impossible. As there are a vast number of designs that can be imagined, and a potentially even vaster number that have yet to be imagined, the task of proving impossibility is daunting. The great value of impossibility proofs is well understood in computer science, where the impossibility of solving the halting problem is a widely known and very robust result – a result based on rigorous proofs. Knowing that a problem is impossible is very useful, as it means research to solve the problem can be abandoned.

By contrast, attempts to kill research into molecular manufacturing systems by the casual use of the term “impossible,” supported by arguments that collapse under even casual scrutiny, and in the absence of any attempt at rigor, are on par with previous claims that flight to the moon was impossible, or that heavier-than-air flight was impossible.

The related claim of impracticality is sometimes advanced on the basis that because evolution required “billions of years” (actually, ~0.8 billion years; Section 5.10) to produce bacteria, it is impractical to expect to design artificial self-replicating assemblers in any time frame relevant to human effort [3056]. However, the same line of reasoning would suggest that jet airplanes are even more improbable than artificial replicators because an even longer ~4.6 billion years of natural evolution were required to produce a Boeing 747 aircraft. Since aircraft do exist, this argument is proven to be false. Human engineers can act with purpose; nature cannot. It is important to note in this context that if the Wright brothers believed that heavier than air flight was theoretically infeasible, they never would have persisted in engineering an airplane. The claims of theoretical infeasibility and the skepticism surrounding the development of an assembler have inhibited both nanotechnology R&D and education.

A more intellectually honest evaluation might note that the time from the first experimental demonstration of a manually-operated macroscale artificial replicator (Penrose Block Replicators, 1957; Section 3.3) to the first experimental demonstration of a fully automatic macroscale artificial replicator (Suthakorn-Cushing-Chirikjian Autonomous Replicator, 2002; Section 3.23.4) was a mere 45 years. The first “manual” assembly of a microscale replicator was achieved by Jeon et al [1860] in 1970 and by Morowitz [1861] in 1974 when they assembled a viable synthetic amoeba organism starting from three [1860] and later five [1861] separate amoeba parts that had been cannibalized from many different organisms. Applying the same 45-year time increment to achieve the first automated assembly of a microscale replicator might suggest an arrival date of 2015-2019 for the first molecular assembler or nanofactory. The first demonstration of artificial bacterium-building should occur within a decade or sooner (Section 4.4), since the first construction of a polio virus from scratch has already been demonstrated, in 2002 [1949]. As for molecular assemblers or nanofactories, it appears that the schedule is far more sensitive to budgetary support constraints than to technological difficulty.


Last updated on 1 August 2005