MM product design
From Wise Nano
Contents |
Advanced Product Design and Performance
To benefit from the high performance of nanoscale machines and materials, large products will have to combine large numbers of nanoscale components in efficient structures. Preliminary architectural work indicates that nanoscale performance advantages can be preserved in large-scale integrated systems. Large aggregations of high-powered machinery would need to be cooled, but this would limit performance less than it does in today's products due to more efficient machinery and more effective cooling.
Optimal use of high-performance materials
Strong, stiff solids in tension are often greatly weakened by minor flaws. Strain concentrates around any crack or other flaw, causing local failure which rapidly propagates. Solid carbon lattice (diamond) is no exception. To take advantage of the theoretical strength of carbon-carbon bonds, it will be necessary to prevent crack propagation. This can be done by building the material in long thin fibers attached in a way that does not propagate failure. Similar approaches are used today in advanced polymers and fibers, but molecular manufacturing construction would give more control over structure. This would allow fibers to be perfectly aligned, and crosslinked to each other or attached to other structures with minimal strain. Most important is that most of the fibers would be defect-free.
Strong fibers can also form the basis of energy-absorbing materials. Interlaced fibers with high-friction molecular surfaces could be designed to slip past each other under stress slightly less than that needed to break bonds in the material. The limit to the energy absorbed in such an “unbreakable†material is the heat capacity of the material; when its temperature rises too high, its bonds will be weakened. In theory, this much energy could be thermalized (absorbed) in just a few nanometers of motion; longer fibers would allow the material to absorb repeated impacts after it had cooled.
A solid block, slab, or beam of material typically is not efficient at resisting compressive stress. With no cost penalty for manufacturing more complex objects, it will be possible to make a variety of efficient structures such as honeycombs and fractal trusses. A thin pressurized tank will resist compression at any point, transferring compressive stress to the contents and imposing tensile stress on the walls.
A diamond shaft rotating at high speed can carry power at 10 watts per square micron.25 This may be the most compact way to transmit power within a product. Stretching a spring made of tough diamond structure can store energy equal to a significant fraction of the bond energy of spring's component atoms. Such a storage system could be charged and discharged quite rapidly, and store energy without leakage.
Performance of advanced products
The performance of a product will depend on its mass, power, and heat budgets. To a large extent, mass can be traded for efficiency, by adding more systems and running them more slowly to obtain reduced drag. Given the extremely high power densities of electrostatic motors, and the smaller but still quite high power density of electrochemical (fuel cell) or mechanochemical processors (as much as 1 GW/m3), it is safe to assume that power transformation will not be a significant part of the volume of most meter-scale products. If the product expends its power externally, for example on propulsion through a viscous medium, then only a small fraction (probably well under 1%) of the total power handled need be dissipated as internal frictional heating.
Because signaling over even modest distances slows a computer system significantly, and because erasing bits has an irreducible thermodynamic cost, massively parallel computer systems are likely to be concentrated sources of heat. Reversible logic can be used to reduce the number of bits that need to be erased, but this comes at the cost of increasing the total amount of logic and thus the frictional losses. Increasing the size of the computer also increases the signal path length, requiring slower operation. In general, a computer using reversible logic will be most efficient when it spends about half its energy on erasing bits and half on friction,26 which means that Drexler's reference calculation with its modest use of reversible logic was actually close to ideal.
By today's standards, computers will draw very little power: a billion 32-bit 1-GHz processors would use about 10 watts. For many applications, new algorithms would be required to make use of such a highly parallel system. Cooling a cubic-centimeter volume of computers (which would produce about 105 watts of heat) can be accomplished via branched tubing and a low-viscosity coolant fluid using suspended encapsulated ice particles.27
Using the full strength of diamond, handling compressive stress efficiently, and using much more compact motors, computers, and sensors (with less mass required to mount them), products could be built with a small fraction—usually less than 1%—of the structural mass required today. This would often be inconveniently light, but water ballast could be added after manufacture. This shows that nanofactories will be competitive for more than just high-tech products. Even if diamond-like material cost $100 per kg to produce, nanofactory-built products would generally be competitive with current products.
Design of advanced products
To date, product complexity has been limited by the need to manufacture it via relatively simple and crude processes, and minimize the number of operations to reduce manufacturing cost. A nanofactory will impose essentially zero penalty for each additional feature, and will provide several design choices per cubic nanometer. Although a simple product such as a cube filled with inert matter would be easy to specify, designers will want to use nanoscale features in heterogeneous structures without being overwhelmed by complexity.
Perhaps the simplest design task will be to re-implement existing products in the new technology. In today's products, a volume of the product usually implements a single function: a motor, a computer, a structural beam. Many of these components will be able to be replaced by a higher-performance component without changing the product's functionality. In some cases, inert mass will have to be added to preserve kinematic characteristics and avoid excessive redesign; for example, replacement of electromagnetic motors by nanoscale electrostatic motors may require a flywheel to be added to replace the missing rotor mass.
For drop-in replacement of today's components, the key ingredients are well-characterized nanomachines and design libraries that combine them into larger structures. Product designers should not need to worry about nanoscale phenomena in nano-motors, nor about designing a multiply branching distribution structure or converging gear train, nor about implementing fault tolerance.
It seems likely that design of nanofactory-built products will follow a method similar to software engineering: build high-level designs on top of many levels of predictable, useful capabilities encapsulated in simple interfaces that allow the low-level capabilities to be parameterized. One simple but useful design technique will be to design a structure that can be repeated to fill a volume, and then specify the desired volume. (The structure will likely be part of a pre-supplied library.) That volume, full of whatever functionality was designed into its structure, can then be used as a component of a larger, higher-level structure.
The lowest level of structure that designers of large products will usually be concerned with is the individual microblocks that the product is assembled out of. A microblock will be a small fraction of the size of a human cell. This is a convenient size to contain a basic functional unit such as a motor or CPU. A library of such microblock designs will be available for combining into larger functional units.
At the highest level, designers who are merely trying to recreate today's level of product performance will find it easy to fit the required functionality into the product volume.
