The Physical Basis
of High-Throughput
Atomically Precise Manufacturing
by Eric Drexler on 2009/06/12
The section below, adapted from a longer work, discusses the physical basis for understanding atomically precise fabrication systems: first, a very general class of systems, and second, the specific characteristics of high-throughput systems of a kind several technology levels above where we are today. (In my previous post, “A Telescope Aimed at the Future” I said a bit about science, modeling, and as-yet-unimplemented technologies.)
Regarding next-stage objectives for laboratory research and the trajectory of technology development, I’ve previously discussed:
- The path that led from hand tools to automated factories
- Next-stage experimental objectives for nanosystems development
- Self-assembly and directed assembly
- Thermal fluctuations, mechanical stiffness, and error rates
- Next steps in software for macromolecular engineering
Atomically precise manufacturing
Current understanding of potential systems for atomically precise manufacturing (APM) is based on long-established, well-understood physical principles and phenomena. This is not accidental: the methodology of the analysis deliberately excludes speculations regarding new or poorly understood physical phenomena. Design and analysis have shown that a composition of unproblematic parts is sufficient to implement surprising system-level capabilities.
Although there is a temptation to think that surprising results at a system level must stem from something new at a fundamental, physical level. A moment’s thought shows that, while this may often be true, it cannot be a fixed rule; indeed, most innovation in engineering involves no new scientific principles.
In this connection, it is important to note that — despite great differences at the level of unit operations and system architectures — molecular machines in biological cells demonstrate the fundamental physical principles and operations that enable APM.
Here’s an outline:
| The physical basis of APM-based fabrication |
| Nanoscale devices can bind both chemically reactive molecules and larger structures. |
| Nanoscale devices can position molecules with atomic precision relative to larger structures. |
| Positioning reactive molecules with atomic precision can cause structure-building chemical reactions at specific sites. |
| Programmable devices can direct sequences of site-specific, structure-building chemical reactions. |
| Programmed sequences of site-specific, structure-building chemical reactions can build complex, atomically precise nanostructures. |
Each of these operations — though in a non-APM-like context — is demonstrated by ribosomes, which build peptide nanostructures by positioning a series of activated molecular building blocks (aminoacyl tRNA molecules) in sequences programmed by mRNA.
Predictive molecular modeling outside the biological context
For suitably selected chemical and mechanical systems, each of these fundamental operations can be analyzed quantitatively in a non-biological context.
It is important to note that systems that are compositions of components that consist of strong, stiff covalent solids, are by design easier to model than biological systems for two reasons: First, stiffer, more robust structures are insensitive to small errors in the Hamiltonians that drive simulations of molecular dynamics; second, severe conformational restrictions (e.g., as a consequence of ubiquitous polycyclic structures) avoid the familiar challenge of evaluating the free energies of protein folding and ligand interactions in water. Computational chemists will immediately see that these characteristics can greatly increase the predictive power and reduce the computational cost of simulations.
Similar remarks apply to the molecular physics of chemical transformations, where restricting translational and rotational degrees of freedom (and choosing favorable, well-understood reactions) can greatly increase the predictive power of density functional theory, as well as the utility of applying fully ab initio methods to model compounds and transition states.
These features follow from design choices made in order to avoid poorly understood phenomena. Choices made to facilitate implementation of analogous systems by means of near-term technologies will be very different.
High-throughput APM
As a basis for examining the performance of a conservatively designed class of advanced APM systems, exploratory engineering methods have been applied to characterize devices with a range of functions sufficient to implement an extensive class of mechanical systems.
As I mentioned above, the devices chosen for design and analysis are members of a class of non-biological structures that has been selected, not for optimality of some sort, or for near-term implementation, but to facilitate high-confidence modeling and analysis by means of standard physics-based methods. This objective favors studies of devices based on covalent solids, many of which can be regarded as large, highly polycyclic organic molecules; fortunately, these structures can also have favorable mechanical properties — high strength and stiffness, low sliding friction.
(Regarding a range of more accessible materials, including oxides, see this post.)
A note on methodology
The state of physical and computational technology presently forces a choice between studying devices that are suitable targets for laboratory development and studying devices that are designed for high stability and for low sensitivity to the limitations and inaccuracies of current computational models. (This is a somewhat subtle point, and a failure to understand the nature of the problem being addressed can lead scientists to mistake designs that are conservative — in a modeling-confidence context — for the opposite.)
Given this choice of materials, the mechanisms analyzed and the treatment of uncertainty were chosen with the aim of establishing conservative lower bounds on the performance parameters of devices. This is standard engineering practice, and is necessary to support confidence that a collection of components will perform adequately when composed to form a system.
System-level analysis
At a molecular level, the fabrication of small building blocks is governed by the molecular physics of chemical reactions, and must be analyzed as such, and not in terms of conventional mechanical operations. Organic synthesis and surface science provide a range of model reactions; quantum chemistry has been used to study others.
Studies of carefully selected structures at a larger scale (several nanometers) show that components and systems can be engineered to closely mimic a wide range of familiar mechanical devices. In this, several considerations are important. At interfaces between components, it is necessary to choose structures that are chemically stable and take advantage of short-range surface forces. For reliable mechanical operation, thermal fluctuations are a key concern. These impose requirements for adequate energy barriers blocking transitions to unwanted states (typically ≥ 50 kBT), and a suitable combination of mechanical stiffness and tolerance for positional displacements. Typical values for stiffness, ks, can be designed to be ≥ 20 N/m, giving r.m.s. displacements, (kBT/ks)–½, ≤ 1.4 × 10–11 m at 300 K. A typical tolerance for displacement is ~1.5 × 10–10 m.
A conservative engineering analysis shows that high-throughput APM can be performed by non-biological, factory-style systems that operate on a wide range of molecular building blocks to produce a wide range of non-biological materials, atomically precise nanostructures, and larger assemblies. These products include components and assemblies of the kinds necessary to implement these factory-style systems.
This overall conclusion follows from design and analysis at both the component and system levels. The key results reflect the relevant chemistry and physics, and can be summarized as follows:
| The physical basis of high-throughput APM-based fabrication of large products: |
| Atomically precise nanostructures can implement components that provide a full range of ordinary mechanical functions: These include motors, bearings, gears, conveyors, and so forth. |
| This range of component functions is sufficient to implement machinery like that found in high-throughput assembly systems. |
| At and above the nanometer-scale building-block level, chemical and intermolecular interactions can play roles like those of fasteners, and can bind and align surfaces to form atomically precise interfaces. |
| Assembly of small atomically precise components can therefore yield larger atomically precise components, and this can continue through micro- and macroscopic scales. |
| All operations above the molecular level can be assembly operations (hence no molding, machining, casting, coating, etching, heat treatment, polymer curing, etc.). |
| The mechanical stiffness of suitably designed nanoscale machinery can limit the r.m.s. amplitude of thermal fluctuations to ~10–11 m at 300 K. |
| Tolerances for positional offsets during assembly can be large enough to limit error rates induced by thermal fluctuations to (for example) ~10–15 per operation. |
| Energy dissipation per molecule processed and incorporated into a product structure can be limited to a small multiple of the energy dissipated in conventional chemical processing of materials. |
| The natural motion speeds of mechanical devices are independent of scale, hence the natural frequencies of mechanical operations are inversely proportional to scale. |
| On a throughput per unit mass basis, the natural productivity of assembly machinery is proportional to the frequency of operations, hence inversely proportional to scale. |
| The size and frequency ratios of conventional manufacturing machinery in comparison to machinery based on atomically precise nanoscale components are, respectively, approximately 1 : 10–6 and 1 : 106. |
| The natural productivity of nanoscale manufacturing machinery on a per unit mass basis can therefore exceed that of conventional machinery by a factor of ~106. |
| In a full production system, throughput is limited by the speed of assembly of the largest (end-stage) components that result from convergent assembly of smaller components. |
| The natural productivity of a full APM system, from raw materials to products, is therefore similar to the productivity (mass-based throughput) of final assembly operations in conventional manufacturing. |
| The broader physical and engineering analysis supporting the above indicates that necessary system-level constraints not mentioned above can be satisfied. These include providing means for power supply, cooling, feedstock flow, feedstock purification, binding of input molecules to mechanical components, component and molecular transport, matching of mass throughput at successive scales of assembly, provision of control signals, and fault-tolerance adequate to provide reliable system operation in the presence of manufacturing defects, background radiation, and failures of nanoscale components. |
There’s a long way yet to go to develop the tools needed to build tools of this complexity and performance, but we’re already many steps along the path. Every extension of the scope and applications of atomically precise fabrication helps to build the next-level technology platform. Contributing research can be found in almost every area of nanotechnology.
Revised and expanded, 23 August 2010
See also:
- Nanotechnology Roadmap
for Atomically Precise Nanofabrication and Productive Nanosystems
-
-
-
-
-
- Subscribe
Tagged as: fabrication, molecular manufacturing, nanomachines
{ 5 comments… read them below or add one }
Mr Drexler
I´d like to know
How much does this nanofactory´s cost to be done???
Is it possible to accelerating it in next five year with massive investiments ?
Please answer to me
Imagine that the year is 1970 (when the 4-bit microprocessor was still in the future), and the question is how much it would cost to develop something like a Pentium. If it had been clear that Moore’s law progress would go far enough, this question would have made some sense, but what would have been the right answer? Not the cost of developing a Pentium from the immediately preceding generation of chips, nor the total investment in semiconductor technology starting from the Intel 4004. The development process paid for itself along the way, as with most technologies. That’s a natural scenario here.
It isn’t the only scenario, of course; there have been large, focused programs that weren’t driven by incremental commercial rewards — the Apollo Moon-landing program, for example. The development of increasingly advanced atomically precise fabracation technologies will likely be an intermediate case, as it has been so far.
The necessary technology base has been developing with both commercial and governmental support. It could move faster with greater support, but at the moment, the main opportunity is to coordinate research objectives to develop a stronger technology base for implementing a wide range of complex nanosystems. Framework-directed self-assembly, in particular, is at the threshold of becoming a very powerful technology. My recent post on 3D-SDN points to a recent milestone.
Dr. Drexler:
I joined the Foresight Institute as a freshman in college, in 1991, and have followed your work with interest for the past 18 years. Most criticisms of MNT have been trivial and failed to understand or address the proposals. Richard Jones has recently made more pointed critiques: namely that stiff diamandoid structures will spontaneously reconfigure and also cannot catalyze reactions the way flexible proteins can. Therefore nano will not progress beyond “protein machines” or “soft machines”. It seems to me that an advanced nanomachine can have a combination of soft & hard parts, that fullerenes are very stable, and that “soft” protein-like systems can form multiple peptide- like bonds to build multiple covalant bond “hard” products. Has anyone made a point by point counter-argument to Jones? Thank you again for your work.
Gus K.
I know Dr. Drexler
This questions important to me, because here in Brazil, the government will spend 208 billion reais or 104 billion of dolars to extract pretoleum
It´s so ridiculous, there are air cars, hydrogen cars, eletric cars, water cars
With 10 % of this, i think you , perphaps, can advance in molecular manufacturing in 20 years
There is no way you can amass that kind of investment unless a big entity, like a government, is very interested in your project. Governments are more interested in getting turnaround on things in the short term than long term, that is why they pump so much into extracting petroleum.
{ 21 trackbacks }