4.1 Miniaturization of Manufacturing Systems as Driver

This section takes a brief look at the consequences of realizing the Feynman vision of atomically precise nanoscale machines. They are anticipated to have significantly superior performance to conventional machines. The main features are:

Scaling. Assuming that the speed of linear motion remains constant at all scales, a miniature machine can operate at a much higher frequency compared with a macroscopic machine. Hence, throughput expressed as the fraction of the machine's own mass that it can process per unit time (relative throughput) will increase proportionally with diminishing size (assuming that the components handled by the machine are similarly miniaturized). Since mass decreases as the cube of linear dimension, absolute throughput will decrease proportionally to the square of the diminished size. This implies that vast parallelization is required to maintain throughput. The diminished mass also implies negligible inertia and gravitational influence. Stiffness, too, decreases with size, implying that the stiffest possible covalent solids should be used, which is why diamond is a preferred material for the construction of nanoscale machines and most simulations of their operation assume diamond as the material.

Wear. Covalent bonding between atoms is “digital” (or binary)—either there is a bond or there is not. Hence incremental wear, familiar in large material-processing machines, cannot occur. Instead, there is merely a damage probability, which should be much lower than for a large machine, since it is largely dependent on the stress on a part, which sharply diminishes with decreasing size. Furthermore, manufacturing precision can be retained through multiple generations of manufacturing, just as digitally encoded information can be transmitted and retransmitted many times without loss of fidelity (especially with the help of error-correcting program codes).

Precision. It is well known from macroscale machinery that friction and wear diminish with increasing precision of finishing the parts. Atomic precision is the apotheosis of this trend. The familiar microscopic post-processing machining operations such as grinding and polishing are unnecessary for products assembled atom-by-atom. A further corollary of atomic-scale precision is the great intricacy with which products can readily be made.

Parallelization. In order to make humanly useful quantities of products, atomic assemblers first have to make replicas of themselves. The key parameter is the doubling time. A few dozen doublings suffice for a single nano-assembler to produce 1 kg of copies. This strategy also makes the cost of the machines very low, since the development costs for the first machine can be amortized by all the extra copies.

4.2 Facile Fabrication as Driver

Provided performance can be maintained, the smaller a device, the less material is used, leading to cost savings per device. Miniaturization also facilitates massively parallel fabrication procedures—indeed great use has already been made of this possibility in the semiconductor processing industry, as evinced by the continuing validity of Moore's law.

Nanoparticles made from electronically useful materials can be printed at high speed onto a suitable substrates (making printed electronics). This innovation enables single-use (disposable) devices, with obvious advantages in applications such as medicine, avoiding the extra work (and expense and resources) of sterilization and the risks of cross-patient infection.

At present, products with features a few nanometers in size (i.e., not quite atomically precise) are not produced by assemblers, nor will they in the foreseeable future. Devices such as single-electron transistors have to be made by top–down semiconductor processing technology, with which one-off nanoscale artifacts suitable for testing can be made in the laboratory. Low-cost, high-volume manufacture, however, has the following three requirements [6]:

These requirements are fulfilled by the mainstream micro- (becoming nano-)electronics industry. But is there a lower size limit on what can be got out of devices made using conventional semiconductor processing technology (i.e., epitaxy, lithography, and etching)? Suppose we are trying to make an array consisting of 3 nm diameter features on a 6 nm pitch [6]. Each layer of each pillar contains about 80 atoms and is formed by adding or removing atoms one at a time, by deposition or etching, respectively. These processes can be considered to obey Poisson statistics (i.e., the variance equals the mean). Hence the coefficient of variation (the standard deviation, equal to the square root of the variance, divided by the mean) of the area of the pillars is 12%, whereas the reliable performance of the transistors on a VLSI chip requires a coefficient of variation of less than 2%. Fundamental statistics intrinsic to the process prevent manufacturability below feature sizes of around 10 nm. Possibly this statistical barrier could be traversed by fabrication in a eutactic environment (i.e., by assemblers), but as already mentioned they are not currently realizable.

Nevertheless, the computational requirements for some applications, such as presenting visual and audio data (i.e., pictures and music) might be able to tolerate circuits less uniform and reliable than those of current VLSI chips [7].

While devices assembled from preformed nano-objects such as carbon nanotubes might not be subject to Poisson statistics (e.g., if they are assembled by a tip-based manipulator—although this is scarcely a high-throughput manufacturing technology at present), another problem then arises—that of the uniformity, reproducibility, and reliability of the nano-objects. One of the principle ways of making carbon nanotubes is plasma-enhanced chemical vapor deposition, a process that is subject to Poisson statistics.

Nevertheless, it may be possible to evade this intrinsic limitation. Using highly specialized and intricate machinery in the nanoscale, living systems managed to make highly uniform objects in the nanoscale, most notably proteins which serves as a powerful inspiration to achieve the same artificially.

4.3 Performance as Driver

Performance may often be enhanced by reducing the size. If the reason for the size reduction is accessibility or ease of fabrication, the scaling of performance with size must be analyzed to ensure that performance specifications can still be achieved. It is worth noting that the performance of many microsystems (microelectromechanical systems, i.e. MEMS) devices actually degrades with further miniaturization [8], and the currently available sizes reflect a compromise between performance and other desired attributes.

If vast quantities of components can be made in parallel very cheaply, devices can be designed to incorporate a certain degree of redundancy, immunizing the system as a whole against malfunction of some of its components [9]. Low unit cost and low resources required for operation make redundancy feasible, hence high system reliability. A more advanced approach is to design the circuit such that it can itself detect and switch out faulty components. However, for malfunctions that depend on the presence of at least one defect (e.g., an unwanted impurity atom) in the material constituting the component, if the defects are spatially distributed at random, the smaller the component, the smaller the fraction that are defective [10].

4.4 Agile Manufacturing

Outside the nanoworld, manufacturing is being transformed by digitization. Hitherto digitization has been closely tied up with information transmission and processing (based on the general-purpose digital computer). Economists have found that economic growth is correlated with the degree of digitization, and a recent refinement of the analysis of the phenomenon suggests that economic growth is actually caused by digitization [11]. In terms of manufacturing, this has meant the computer control of machine tools in factories, in turn permitting the widespread introduction of robots. These factories were still engaged in the mass production of identical objects. The concept of additive manufacturing (AM), in which objects are built up by microwelding small lumps of metal (or other materials) to each other, while initially developed about 30 years ago for the rapid prototyping of artifacts that would ultimately be manufactured conventionally, has permitted digitization to penetrate more deeply into manufacturing practice: AM (also known as three-dimensional (3D) printing) is now used as a manufacturing technology in its own right. Miniaturized down to atomic dimensions, it becomes Feynman–Drexler bottom-to-bottom assembly. A similar trend has occurred in the printing industry: books can now be printed digitally on demand. The traditional economies of scale associated with assembly lines no longer apply: it is possible to make customized objects for practically the same cost as identical ones (mass customization).

In essence, nanotechnology represents the “digitization” of materials as well as of the information, in which objects are fabricated from preformed nanoblocks. By analogy, we can propose that this further digitization will have a similarly beneficial effect on economic growth. The fundamental reason for these effects of digitization, which do not appear to have been discerned by the economists, must be the universalization enabled by digitization. A digital computer can be programmed to carry out any kind of computation; a digital storage medium can store any kind of (digital) information—text, music or images; and so forth. This in turn enormously increases the adaptability of the system as a whole. Ironically enough, if this is considered as a revolution, it makes further industrial development easier because new challenges can be met with only minimal need for brand-new materials; it will suffice to tweak the specifications of a nanoblock and the assembly procedure to create something with quite different properties, for example—the very antithesis of a revolution based on disruptive technology. It also enables everything to be precisely tailored for optimal use. To give just a simple example, at present hip prostheses are provided in about half a dozen standard sizes and the surgeon selects the best match to the patient needing a replacement. The digital concept means that a prosthesis will be rapidly machined using patient data in order to precisely fit the body. For the same amount of work, or indeed less work, by the surgeon the result will be far superior, with the tangible economic benefit in terms of the activity of the patient post-operation, as well as other intangible benefits of a superior quality of life.

The nearer nanotechnology approaches the ultimate goal of productive nanosystems (see Section 18.1), the more flexible manufacturing should be. One aspect that needs careful consideration is the fact that agile (adaptive) production systems are necessarily rooted in algorithms: hence, agile factories must necessarily be computer-controlled, due to the large volume of information that has to be processed very rapidly (i.e., in real time) during production. The computer must (at least with present technology) run according to a preloaded program, which is necessarily closed and hence cannot but reflect present knowledge. The control center of the factory is therefore intrinsically ill-equipped to adapt to the ever-unfolding events that constitute the course of existence, which is largely constituted by the unknowable part of the future. That the financial turbulence of 2008 turned out to have serious industrial consequences is an all-too-obvious illustration of this truism. Different sectors seem to be intrinsically more or less sensitive to future uncertainty—rational planning demands that this sensitivity be quantified to determine the sectorial appropriateness of agile manufacturing. We need to anticipate that real-world contexts will raise challenges of increasing uncertainty and diversity and so require agility to be achieved by means that are suitably resilient and adaptable to change (“agile agility” or “adaptable adaptability”); that is, agility needs to be explicitly designed for the unknown and unexpected, not merely to cope with well-understood tempos and boundaries. Naturally this will have a cost: presumably the more adaptable an industrial installation, the more expensive, hence it will be appropriate to determine a suitable limit to the necessary adaptability for a particular sector.

4.5 Nano–Info–Bio–Cogno (NIBC)

The quartet of advanced modern technologies—nanotechnology, information technology, biotechnology, and cognitive technology—are often considered to be converging and becoming inseparable and indistinguishable. Practically speaking, it is therefore supposed to make more sense to consider them as a unified whole. While the reasoning behind this idea often seems to be either trivial or obscure, it is nevertheless striking to note that all four have a common feature, namely the discretization of their core elements. In nanotechnology it is the atom—engineering with atomic-scale precision; in information technologies it is the bit (binary digit)—all information must be digitized before it can be processed further; in biotechnology it is the gene—genes can be introduced into, or removed from organisms discretely at will, as required; and in the cognitive technologies it is the neuron—as represented by a discrete cell in a cellular automaton, for example.

In practice the irreducible units are somewhat larger: a nanoblock (e.g., a nanoparticle) containing at least some hundreds of atoms; 8 bits are grouped into a byte, which can represent an 8-bit word (such as any ASCII character); a gene requires a promoter and several genes and may be grouped into an operon; neurons work in groups, although here our knowledge is still very imperfect.

Just as discrete mathematics is different and distinct from continuous mathematics, so is the discrete nature of these advanced technologies different from the continuous nature of traditional, established technologies.

Working at a smaller scale, albeit with nanoblocks rather than atoms, nanotechnology can be viewed as a kind of additive manufacturing process. If so, nanotechnology is qualitatively different from ultraprecision engineering in the nanoscale, despite the fact that it was the first technology to be labeled “nanotechnology” (Section 1.1).

Analog computers are indeed quite different in their architecture from digital computers. The latter do not, however, resemble cellular automata, some classes of which may be capable of universal computation [12].

Modern biotechnology may be predicated on the “one gene, one enzyme” notion of Beadle and Tatum, but this was shown to be wrong by Wright more than 80 years ago [13]. Practical gene engineering does, in fact, require the introduction of many auxiliary genes into an organism to ensure the expression of one desired new gene introduced into its DNA. Much criticism of the use of genetically-modified plants as food for humans and domestic animals rests on the possibly deleterious effects these auxiliary genes may have on human health.

Immense efforts are now being put into understanding the human brain by modeling it. The flagship effort in this direction is currently the “Human Brain Project” led by Henry Markram at the Ecole polytechnique fédérale de Lausanne (EPFL) and funded by the European Union (EU). It is felt that success in this endeavor holds the key to developing artificial intelligence beyond sophisticated expert systems.

Undoubtedly in this sense NIBC represents a clearly identifiable concept of discreteness of matter and process. This aligns it closely with cybernetics, in which machines—information processors—are analyzed by considering their discrete states [14].

Prior to the full nano–info–bio–cogno convergence, we already have nanobiotechnology (cf. Section 1.5)—the application of nanotechnology to biology, including biotechnology and medicine (see Chapter 6 for nanomedicine)—and bionanotechnology, which is the application of biology to nanotechnology, which mainly means biomimetics (Section 10.7) at the nanoscale [15].

4.6 Cost–Benefit Analysis of Nanotechnology

There are two main motivations for undertaking such an analysis: from the viewpoint of an individual company, (i) is it worth adopting a “nano-solution” (that is, a solution involving nanotechnology) to a given problem?—and (ii) from the viewpoint of the government of the country, responsible for creating a favorable environment for its industrial and commercial development, is it worth investing taxpayers' money into nanotechnology infrastructure, such as common facilities for nanofacture and nanometrology centers, in which expensive instruments are acquired for shared use?

Regarding (i), one can differentiate between a manufacturing company considering introducing nano-objects into an existing product, and a group of investors considering whether to launch a nanofacturing company. In the former case, the calculation is straightforward. If a simple substitution is being considered, a fraction $1/x$ of the quantity of existing, non-nano, material being used may be required to achieve the same performance. If the nano:non-nano ratio of costs per unit quantity is less than x, then it may be worth making the substitution, provided that the expense associated with modifying the production process is small. This expense will generally be incurred only once, when the modification is undertaken. Another one-off expense is that of tests carried out on the modified product to demonstrate that it is the same, or better than, the existing product. This testing might, however, be carried out under the R& D program of the company and would include determination of the optimal value of $1/x$ as well as the optimal choice of nano-objects from a range differing in such attributes as size and surface properties. Once that value is established, the cost of making the substitution in production can be evaluated.

If the product is improved by incorporating nano-objects into it, then the demand for it may increase and this also needs to be evaluated. The overall goal is to determine the likely return on the investment needed to adopt the nano-solution. For this determination, investors may avail themselves of the kinds of predictions of nanotechnology market size presented in Chapter 5, although those predictions are of a generic nature and based on extrapolating past trends, and may not be reliable for evaluating the likely market for a specific product, especially if it is new.

Regarding (ii), the required analysis is (or should be) also an evaluation of the return on investment. Will the proposed facilities cause economic growth that will generate increased tax revenues exceeding the expenditure on the facilities? The degree of uncertainty in making such an evaluation can be expected to be so great that it is scarcely worth doing. For example, it is difficult to predict whether the existence of the facilities will encourage new manufacturing companies to be formed. The encouragement could be boosted by appointing indefatigable “nano-ambassadors” who would engage with industrialists to make sure that they were aware of the latest technology, but this might be considered to be too interventionist and run into political difficulties. On the whole, the track record of governments in doing this kind of thing is not good. If there were real demand for a nanometrology center, then it would surely be a worthy object for private investment.