K.
Eric Drexler
Institute for Molecular Manufacturing,
555 Bryant Street, Suite 253,
Palo Alto, California 94301, USA.
If one omits human hands from the description and recognises molecules and collections of molecules as examples of parts and materials, then the description of conventional manufacturing systems can equally well describe molecular manufacturing systems. The abstract problem, in both cases, is to manipulate pieces of matter so as to build complex structures. Many of the subproblems and required subsystems are entirely parallel.
The differences, of course, are numerous: Molecular manufacturing systems use devices having nanometre-scale rather than millimetre-scale features. Operating frequencies are higher by a factor of ~ 106. The smallest moving parts must be analysed, not in terms of the elastic and surface properties of bulk materials, but in terms of molecular potential energy functions. Typical assembly operations involve the formation of a distinct pattern of chemical bonds via mechanosynthesis, rather than welding, adhesive joining and the like. Unlike conventional systems and their products, molecular manufacturing systems and their products can be eutactic, that is, well-ordered in the molecular sense. Because all products are made with atom-by-atom control, costs are more nearly proportional to mass than to, for example, the number of moving parts. Thermal vibration is a primary source of errors in molecular assembly processes. Radiation damage and photochemical reactions can destroy components. Each of these issues, and many others, must be considered in the design of the products, mechanisms and processes of molecular manufacturing.
The physical principles required for this analysis are described in [1], which analyses nanomechanical components, computational systems and manufacturing systems and describes implementation pathways leading from today's technology base in chemistry, biochemistry and proximal probe mechanisms, e.g. scanning tunnelling and atomic force microscopes.
Nanomechanical systems are easy to analyse and have characteristic frequencies high enough to make them surprisingly attractive as computer devices. In such systems, signal transmission occurs by the mechanical displacement of a slim rod. Switching occurs when the displacement of one rod mechanically obstructs the motion of another. A typical switching time is a respectable ~ 10-10s. An analysis of mechanical nanocomputers indicates that a processor capable of executing 109 instructions per second can fit within a volume of ~ 0.1 m3 and consume a power of ~ 10-7 W. An array of 109 such processors in an air-cooled desktop package could deliver more computational capacity than all the computers in the world today combined. A related technology base can provide information storage densities of1010 bits/m3. This analysis takes account of signal propagation speeds, material strength and stiffness, sliding friction, nonisothermal compression losses, entropy generated by register erasure, error rates caused by thermal vibration, power supply, clocking, clock skew, input, output and other issues [1].
The feasibility of constructing parallel computing systems that execute 1018 instructions per second will enable improved computer-aided design and simulation of complex systems. For example, if the time required to model the structural deformations and aerodynamics of a vehicle is reduced to a second or less, simple and powerful iterative optimisation methods can be applied. In a more direct application to space technology, the feasibility of constructing microscopic computer systems supports the feasibility of constructing small automated spacecraft and of providing computers of high capacity on spacecraft of conventional size.
Scaling laws favour the use of electrostatic mechanisms rather than electromagnet-based mechanisms in small systems and yield high power densities for electrostatic motors of submicron dimensions. A design exercise indicates the feasibility of power conversion densities > 1015 W/m3 with > 99% efficiency for the interconversion of mechanical and electrical energy in direct-current devices; i.e., the same device can serve as a motor or as a generator. Motors providing high power density can facilitate the engineering of a wide range of active structures. As with mechanochemical energy conversion processes, these systems can approach thermodynamic reversibility
These capabilities can work together in energy conversion systems. The ability to convert chemical to mechanical to electrical energy serves the function of a fuel cell. The ability to convert electrical to mechanical to chemical energy, and the reverse, serves the function of a storage battery. The ability to convert electrical to mechanical to electrical energy, but at a different voltage, serves the function of a DC transformer.
For energy collection, molecular manufacturing can be used to make solar photovoltaic cells at least as efficient as those made in the laboratory today. Efficiencies can therefore be > 30% [3]. In space applications, a reflective optical concentrator need consist of little more than a curved aluminium shell < 100 nm thick (photovoltaic cells operate with higher efficiency at high optical power densities). A metal fin with a thickness of 100 nm and a conduction path length of 100 m can radiate thermal energy at a power density as high as 1000 W/m2 with a temperature differential from base to tip of < 1 K.
Accordingly, solar collectors can consist of arrays of photovoltaic cells several microns in thickness and diameter, each at the focus of a mirror of ~ 100 m diameter, the back surface of which serves as a ~ 100 m diameter radiator. If the mean thickness of this system is ~ 1 m, the mass is ~ 10-3 kg/m2 and the power per unit mass, at Earth's distance from the Sun, where the solar constant is 1.4 kW/m2 is > 105 W/kg. This calculation assumes no improvements in basic device physics but only in the ability to fabricate small, precise structures.
Molecular manufacturing lends itself to the mechanosynthesis of carbon-rich materials. One such material is diamond, which offers ~ 15 times the stiffness density ratio and ~ 75 times the strength-to-density ratio of a high-strength aluminium alloy [1]. If coated to prevent oxidation, diamond is stable to ~1800 K; aluminium melts at ~ 930 K. Although bulk diamond is brittle, fibrous structures consisting almost entirely of diamond can be tough. If stresses are applied chiefly along a single axis, a common case, division of the material into fibres need not sacrifice much useful strength. With a further modest sacrifice in the strength-to-density ratio, fibres can be interrupted at intervals by mechanisms containing nanomechanical motors, sensors, controllers and actuators, imposing active control of their lengths. An active material of this sort can simulate perfect rigidity so long as stresses remain below the ultimate strength of the structure and so long as fluctuations in stress are slow compared to the characteristic response frequency of the control system, which can, in the present instance, extend to many kilohertz. Active materials can also provide nearly perfect damping of structural vibration.
Molecular manufacturing systems resemble machining systems in requiring many motions per part, and can likewise build machines. They can be more productive than printing, stamping and lithography, however, because the small size of the moving parts in a molecular manufacturing system permits them to perform motions at high frequencies. If millimetre-scale features in conventional manufacturing systems correspond to nanometre-scale features in molecular manufacturing systems (design exercises indicate that this is a reasonable comparison), then the operating frequency of the latter can be greater by a factor of 106. If the number of motions per part were the same (a rougher approximation), then the productivity would likewise be greater by a factor of 106.
More detailed design exercises [1], not based on scaling arguments, indicate that a macroscopic molecular manufacturing system can produce its own mass in comparable product in about an hour. Applying the scaling argument in the reverse direction would suggest that conventional manufacturing systems can produce their own mass in comparable capital goods in about 100 years. Since this is an underestimate of their true productivity, this comparison suggests that the estimated productivity of molecular systems may also be an underestimate. At conventional interest rates, the contribution of the cost of capital goods to the cost of additional capital goods is negligible for systems with productivity of this order.
This may seem surprising but there is no physical law that demands high costs for manufactured goods. Indeed, agriculture proves that molecular systems can operate on inexpensive feedstock materials to produce material structures of great complexity at costs in the $1/kg range.
Mechanosynthetic operations, like digital switching operations, are subject to thermal and other sources of noise. The probability of an error in a mechanosynthetic operation can be calculated from knowledge of the potential energy surface for the reaction, using transition state theory methods grounded in statistical mechanics. In general, the probability of an error is an exponentially decreasing function of the increment in barrier energy between the trajectory leading to the error state and that of the trajectory leading to the desired state. In a mechanically stiff system, elastic restoring forces with stiffnesses ~ 20 N/m can impose large energy costs (³ 1.5 x 10-19 J) on trajectories that deviate from the desired trajectory by ~ 0.14 nm. At room temperature, an energy difference of this magnitude is usually sufficient to limit the error rate to £ 10-15 per operation [1]. Thus, as in digital logic, errors in molecular manufacturing systems can be made both distinct and extremely rare.
Classical radiation target theory, validated by studies of damage to proteins in vacuum [4], indicates that the probability of a damaging event is proportional to the radiation dose D (rad) and the component mass m (kg), and experimental data indicates that the probability that a component remains operational [1] is
A typical terrestrial dose rate is < 0.5 rad/year. Space environments vary widely with location, time and shielding. In a typical spacecraft interior, outside Earth's magnetosphere but shielded by several millimetres of aluminium, a single solar flare can deliver ~ 100 rad spread over tens of hours. Without shielding, dose rates are higher. A dose of ~ 10-4 rad, which would be received in several thousand seconds of terrestrial exposure, causes a density of defects on the same order as the ~ 10-15 per atom suggested by the previous description of errors in manufacturing.
Potential solutions to this problem include a shift to still smaller modules (the feasibility and performance costs of doing so will require a detailed logic system design), or a shift to components large enough to tolerate radiation damage while remaining functional (this strategy entails substantial performance penalties). Minimising size while ensuring damage tolerance is not a simple problem. To design a system in which all atoms occupy defined sites, only one structure need be considered, but to design a damage-tolerant system, all possible radiation-induced defects must somehow be taken into account. This becomes simple only when all components are large compared to the expected defects.
Manufacturing systems can be designed to tolerate an accumulated dose of 10 rad by exploiting small module size and extensive redundancy in some subsystems and larger damage-resistant structures in other subsystems. Information storage systems can use small modules, moderate redundancy and error correcting codes to provide essentially effort-free records despite extensive radiation damage.
tolerance can be used to design systems suited to the space radiation environment. Special problems arise only when molecular manufacturing has been used to make systems that depend on the operation of numerous nanoscale parts. Typical aerospace systems – power conversion equipment, structures (even active structures), and so forth – can benefit greatly from molecular manufacturing without incorporating such sensitive parts.
A simple calculation indicates the minimum performance to be expected from a single stage to orbit vehicle constructed using a molecular manufacturing technology base. In present aerospace practice, launch vehicles constructed chiefly of aluminium have a dry mass that is a small fraction of their fuel mass. A hypothetical vehicle in which the dry mass equalled the fuel mass would include enough structural material to make wings, landing gear, additional engines and so forth, all with large safety margins in their strength and stiffness, but such a vehicle would be too massive to reach orbit. Yet if one were to modify a conservative design of this sort by substituting diamond-based structures for aluminium structures, the dry mass could be reduced to < 0.02 that of the fuel while retaining both the capabilities and the safety margins. (Reductions in the engine mass, etc., can roughly parallel reductions in the general structural mass). The ideal mass-ratio for a rocket using liquid oxygen and liquid hydrogen is ~ 5 for a launch to low Earth orbit. Since wings can both reduce gravity losses during takeoff and enable transit of the dense lower atmosphere at low speed and, hence, low drag, this ideal mass ratio can be approximated in practice and without using air-breathing engines beyond subsonic speeds. With this mass ratio, ~ 0.2 of the gross lift-off mass can be delivered to orbit. With a dry mass fraction of ~ 0.02, ~ 0.18 of this mass can be payload. To be more concrete: if four passengers with luggage, air, seating, and so forth have a mass of 500 kg, the gross lift-off mass of the vehicle can be ~ 3 tons. The dry mass of the vehicle is only ~ 60 kg (the strength-equivalent of ~ 3 tons of aerospace aluminium).
For a given charge density, fields caused by space-charge effects decline as devices are reduced in size. This enables arrays of small ion thrusters produced by molecular manufacturing to have high thrust-to-mass ratios. If the total mass of a solar-electric vehicle is ~ 10-5 kg/W then a system operating with an exhaust velocity of ~ 250 m/s can accelerate at ~ 0.8 m/s2, or ~ 70 km/s per day. After escape from a planetary orbit, a vehicle capable of this exhaust velocity and acceleration can follow what are essentially two-impulse interplanetary trajectories, travelling at a mean velocity of 100 km/s (the total DV of 200 km/s requires a mass ration of ~ 2.2, implying a somewhat lower initial acceleration). This speed covers one astronomical unit in ~ 17 days. In the inner Solar System, solar-electric propulsion systems in this class can outperform proposed gas-core nuclear rockets [5].
Molecular manufacturing can produce materials and energy conversion
systems of sufficiently high performance that single stage to orbit vehicles
become easy to implement. They can be used to construct superior solar
sails, as well as solar electric propulsion systems capable of crossing
distances in the inner Solar System with travel times on the order of a
month. Molecular manufacturing and its products can make space flight inexpensive,
fast and practical. The challenge is to develop the required tools and
to learn to use them to build the proper products [1].