Proposal for NIAC

From Wise Nano

This is the proposal we submitted to get the grant.

Large-Product General-Purpose Design and Manufacturing Using Nanoscale Modules

Chris Phoenix, Director of Research, Center for Responsible Nanotechnology

Contents 1 Abstract 2 Concept Description 2.1 Impact of the proposed manufacturing system 2.1.1 1) General Purpose Manufacturing 2.1.2 2) Exponential Manufacturing 2.1.3 3) Molecular Manufacturing 2.1.4 4) High Performance Products 2.1.5 5) Rapid Prototyping 2.2 Timeline of the proposed technology 2.3 Previous work related to the project 2.4 The Phoenix nanofactory 3 Work plan 3.1 First Month: Define the Problem 3.2 Second Month: Methods for Building Shapes 3.3 Third Month: Virtual Materials 3.4 Fourth and Fifth Months: Automated Requirement Fulfillment 3.5 Sixth Month: Wrap-up and Implications 4 Management Approach 5 Personnel 6 Special Matters

Abstract

Several major problems in space flight could be mitigated by improved manufacturing. Compact general-purpose manufacturing could allow fabrication of new equipment during long-duration missions, reducing risk and weight. Large-scale rapid prototyping could substantially aid the development of new designs. Several micro- and nanoscale general-purpose manufacturing systems have been proposed, but it is currently unclear how they would produce large products.

Molecular manufacturing, with vacuum mechanosynthesis one of several options, is a leading candidate for nanoscale manufacturing. Previous work by the PI has outlined a kg-scale manufacturing system integrating large numbers of nanoscale mechanosynthesis systems. Each system would fabricate a sub-micron cube with nanoscale features; cubes would then be combined to make larger cubes and ultimately a compressed product. That work only touched on product design issues, though it indicated the potential for extremely high performance in several areas.

A molecular manufacturing system may be expected to be compact, high-throughput, efficient, clean, cheap, and automated. If it could build flight hardware, it could serve as a substitute for spare or redundant components on long-duration missions, as well as supplying new products resulting from post-launch research. With product recycling, it could also be used to reconfigure spacecraft. It appears plausible that entire spacecraft could be built by the technology, allowing much faster, cheaper, and more aggressive design cycles and production.

I propose to investigate the design, fabrication, and performance of human-scale products constructed from aggregates of nanoscale functional units. The goal is to demonstrate that such units can be combined to form useful products, and that such products can be designed and constructed with approaches familiar to current engineering practice. Phase 1 would investigate issues such as large-scale structures, curved sliding interfaces, fractal networks, and automated construction of awkwardly-shaped products.

Concept Description

Impact of the proposed manufacturing system

Several factors combine to make the proposed manufacturing system truly revolutionary in its impact on aerospace.

1) General Purpose Manufacturing

A general purpose technology is one that is useful in many contexts, like electricity or computers. A general purpose manufacturing system will be able to make a wide range of products for many different functions. A computer-controlled fabrication system performing many simple reliable operations in a programmable sequence would satisfy this requirement, as long as the result could be aggregated into useful products. Such an automated system would minimize the labor cost of manufacturing, and if the operations were small enough (as in the systems contemplated here) the system could be quite compact.

2) Exponential Manufacturing

Several manufacturing systems have been proposed that include most or all of their structure in the set of their products. Such an “autoproductive” system could be used to produce exponential numbers of manufacturing systems, or exponentially growing integrated systems. A fully automated autoproductive system implies the easy availability of as much manufacturing capacity as desired, with the cost of the system comparable to the cost of its inputs.

3) Molecular Manufacturing

Several different families of chemistry are candidates for producing reliable, programmable, nanoscale features by selective covalent bonding. Drexler has proposed protein engineering, optionally with modified peptides[1]; a simliar proposal was recently investigated in a NIAC project by Mavroidis[2] [1]. Drexler has more recently (1992, in Nanosystems) proposed that scanning probes can be used to build solid carbon-backbone lattices (“diamondoid”) by deposition reactions in vacuum (“mechanosynthesis”)[3]. Nucleic acids fold into well-defined shapes[4]. Graphene molecules may be built in solution[5], and curved fullerene-like structures (corannulene) can also be organically synthesized[6].

Any of these chemistries, or others, may be the basis for building nanometer-scale machine components. Any of several methods, including self-assembly and direct robotic manipulation, could combine these components into machines. Because covalent chemistry is inherently digital, molecular manufacturing appears to offer easy error checking and high reliability, suitable for automated and autoproductive systems.

4) High Performance Products

Diamond, graphite, and fullerenes are extremely high performance materials. A molecular manufacturing technology based on them would have strength-vs-weight perhaps two orders of magnitude better than today's metallurgy or polymers. Small cheap features should allow additional material savings through the use of e.g. fractal trusses (to convert compressive to tensile stress) and “smart” materials (to monitor and compensate loads). Nanoscale features also provide access to favorable scaling laws for power density, computation density, and speed of operation (frequency). Atomic precision should allow extremely efficient and reliable operation. Calculated performance (Nanosystems) indicates power density of 10^15 W/m^3, power efficiency >99% for many operations including some chemomechanical conversions, computation density of 10^26 gates/m^3, and computation efficiency of 10^16 instructions/second/watt. Manufacturing throughput is similarly impressive, with a self-contained macro-scale manufacturing system estimated to output its own mass of product in a matter of hours or less[7]. The orders-of-magnitude improvement in functional density would allow massive overdesign, allowing improved reliability and less exacting design restrictions.

5) Rapid Prototyping

A molecular manufacturing system as outlined above should be able to make products that are at least competitive with a wide range of today's products. The system would be programmable, able to take direction from a blueprint file without retooling or other delay. This implies on-demand manufacturing, in which there would be no difference—including cost—between a prototype and a production unit. Design iterations would be very fast and very cheap. And a single system would be able to make any product that could be designed for it; carrying general-purpose manufacturing systems into space could be quite cost effective.

Aerospace hardware construction costs could drop by many orders of magnitude. This implies that testing of concepts would be much less sensitive to failure (at least for unmanned flights), allowing more aggressive research and development.

Rapid prototyping of flight hardware would allow in-flight repair without a store of spare parts—just one manufacturing system and a small amount of feedstock. (Even the manufacturing system, being autoproductive, could be rebuilt in-flight.) This would allow less overdesign for long missions.

Timeline of the proposed technology

Informed estimates of when molecular manufacturing will be developed vary from ten to thirty years. CRN estimates that it will be developed rather quickly due to military pressure[8]. Once a basic system is developed, improvements can happen quickly due to the programmable nature of the system. Although full use of its capabilities will take significant additional time to research, even crude efforts will probably be highly competitive with other technologies.

Previous work related to the project

Josh Hall has previously described a nanoscale modular self-constructing system[9]. Macro-scale systems, capable of self-duplication from modular parts, have been built[10,11]. Chain robots have been built and tested[12]. Tihamer Toth-Fejel has recently completed an NIAC Phase 1 project describing a self-replicating system based on relatively simple modules at multiple scales[13].

Ways of producing product-like behavior from 3D aggregates of connected mobile modules have been investigated by several people (Hall, Toth-Fejel, Bishop, Michael), as discussed by Toth-Fejel[14]. These architectures have been constrained by the use of more or less homogeneous modules, requiring that large-scale motion be implemented by the coordinated motion of many small components. This requires motion-planning algorithms that are not available in today's engineering repertoire, and also imposes constraints on the performance of the system: for example, every interface must be able to exhibit both motile and rigid behavior, and modules must be able to distribute power and communication in a cooperative and piecemeal fashion.

Amorphous computing and swarm computing have been investigated, for example at MIT[15]. However, these do not attempt to implement physical properties, but merely to combine small computers into useful information networks.

The analyses by Drexler in Nanosystems constitute the best-developed nanometer-scale machinery descriptions, so are taken as a reference point for the capabilities of the nano-fabricated designs that will be taken as a starting point for the current project. As noted above, this does not imply that the vacuum mechanosynthesis described in Nanosystems is the only possible way to produce the components; the proposal does not depend on the likely-but-controversial feasibility of this specific kind of chemistry.

The Phoenix nanofactory

Drexler, Merkle, and Freitas have done major amounts of work on mechanosynthetic autoproductive systems. Phoenix has extended this work[16] and produced an architectural description of a deliberately simplistic tabletop manufacturing system, incorporating 144 quadrillion mechanochemical fabricators and producing ~1 kg/hr of product (fabricating its own mass and complexity in less than 1 day). This “nanofactory” is intended to be autoproductive, and a high-level description of fabrication of duplicates and bootstrapping was provided. The current proposal is intended to extend that work to other products.

The nanofactory is an integrated molecular manufacturing system. Its architecture incorporates very large numbers of covalent-bonding fabricators, fastened in place and working in parallel. Each fabricator creates a 200-nm cube (“nanoblock”) or partial cube (up to ~1.4 billion carbon atoms); this was calculated to require a few hours. Once the cubes are created, simple robotics push them together into larger cubes, and they are joined by simple mechanical joint mechanisms. This joining process occupies only a few seconds, resulting in a more-or-less rectilinear, more-or-less solid product which may then be expanded, rearranged, or unfolded in various ways.

The Nanofactory paper includes (Sec. 3.2.1) a description of a mechanical joint system that could be used to fasten nanoblocks together. It contains only two moving parts and requires no force to assemble or power to activate; just push the two faces together (a one degree-of-freedom operation), and they become strongly fastened as Van der Waals force pulls interlocking pieces irreversibly into the locked configuration. (The joint, once formed, is much stronger than Van der Waals force; it preserves much of the strength of the solid diamond material.) Placement of the joint components can be used to prevent misaligned joints. Adding one actuator, the joint can be reversible.

In contrast with the work on products composed of sliding modules which have many aggregate degrees of freedom (lattice or KCA), the nanofactory's “nanoblock” product modules are intended to be permanently fastened together into relatively few kinematic components, producing more or less rigid structures (which may be articulated with large-scale bearings) of complex virtual or “smart” materials. These relatively simple structures should be amenable to current engineering practice and volume-filling specification. In this way, nanofactory products will be similar to the output of current rapid-prototyping systems. But current systems produce very coarse (pixellated) and simple structures. The nanoblocks, 200 nm cubes in the reference design, would have nanometer-scale features, allowing a single nanoblock to contain rather complicated machinery with up to millions of features.

In addition, each nanoblock is built individually under programmed control, and its position in the final product is known deterministically. This allows nanoblocks to be heterogeneous. Where a swarm or fog system might have trouble creating strong structural members in one place and high-bandwidth communication pathways in a different place, the programming of a nanofactory can be designed to place different kinds of nanoblocks in the appropriate places. Instead of requiring emulation of all functions with a single module type, the nanofactory/nanoblock construction method allows some functions to be physically built into the product at the time of construction.

Work plan

Overview. The goal of the project is to investigate the feasibility of designing and manufacturing a wide range of products, given a basic nanofactory. The strategy is to create a level of abstraction that is suitable for current engineering methods. This will be approached from two directions. First, useful features and functions will be listed, including structure, actuation, sliding (bearing) interfaces, and various distribution networks. Second, means of constructing or emulating these functions based on nanoblocks will be invented.

The project will start with the assumption that a system comparable to the Phoenix “nanofactory” (described above) will be available. Although there are many open questions about such a system, such as a set of mechanically driven covalent bonding operations that can build rigid carbon-lattice parts, most of these questions are more in the realm of engineering than architectural exploration or system creation. Earlier work[17] has specified architectures for the major components of a nanofactory in a fair amount of detail. What remains largely unexplored is the means by which useful, easy-to-design products can be derived from aggregates of tiny modular cubes.

Previous work has addressed this problem in various ways. As noted above, the heterogeneous output of the nanofactory allows a component-based rather than a swarm-based approach at all scales. Thus the product should be designable with traditional engineering methods. The purpose of the current proposal is to create a level of product design that hides the nano-built nanoblock aspect, allowing product designers to use familiar methods on familiar-seeming structures and functions. This will require 1) identifying these familiar methods, structures, and functions and 2) Designing a bridge, by a combination of construction and emulation, from nano-built components to familiar design territory.

First Month: Define the Problem

The first month will be spent on defining the desirable engineering components to simplify as much as possible the design of products. A range of representative products will be selected, with emphasis on aerospace hardware including life support. Design methods for those products will be studied, with emphasis on converting requirements into blueprints (as opposed to developing the requirements in the first place). Thus the process of specifying requirements and high-level functional design can continue unchanged, minimizing the need to reinvent bureaucracy.

Once the products are selected and the current mapping between function and blueprint is listed, the process of designing new mappings for each function can begin. For example, a function might be to hold two pieces a certain distance apart. The blueprint might call for a bar of a certain alloy. The current mapping, then, is from a structural requirement to a shape and material specification. The new mapping may result in a volume to be filled with a certain density of diamond foam—this will be determined later in the project. But knowing that an important part of product design is converting structural requirements to structure will tell us what questions to ask. Thus, the first month's work lays the foundation for the rest of the project.

The classes of requirement to be studied, in addition to structural requirements, will include actuation; energy storage and transformation; distribution of materials, energy, and information; optical and textural surfaces/displays; various sensing modalities; useful material properties such as transparency and massiveness; and computation. Other classes may be identified.

Second Month: Methods for Building Shapes

A major open question is how to make non-rectilinear structures given the rectilinear nature of nanoblocks and the simplicity of the proposed convergent assembly operation. Thus, a question that must be answered early is how to unfold, inflate, or otherwise reconfigure compact blocky products into heterogeneous volumes (with empty space in complicated patterns). An alternative way to solve this is to construct the product with empty volumes in the first place, but this may require arbitrarily complex or even impossible operations to build partially supported structures. The second month will be used to invent and develop as many means as possible of creating heterogeneous volumes. (Recall that nanoblocks can be individually and deterministically specified, so that the question of how to create mixed matter/space structures is separate from the question of what blocks to place in the matter-filled volumes.)

The goal is to allow the simplest possible engineering, hiding construction-specific details from the design process. Therefore, each unfolding method will be designed with an eye to automated derivation. Ideally, a means will be found for producing any desired configuration from a cubical mass of modules. However, we do not expect to solve such a general problem. Instead, we will find ways to create useful structures (informed by the results of the first month's investigation). This way, engineers will be able to specify standard shapes, and the means of unpacking and unfolding these shapes will be computed automatically. (The convergent assembly process used in the nanofactory does permit limited inclusion of empty volumes, and this is expected to make the packing problem tractable. Also, collapsible volumes may be quite useful, roughly analogous to sacrificial materials in semiconductor and MEMS fabrication.)

Third Month: Virtual Materials

In today's products, volumes and features are usually specified with micron or coarser precision. Products built by mechanosynthesis are expected to have nanometer features. This indicates that each monolithic particle of today's products can be made to include a billion features. (Features may be much larger and simpler if this degree of precision is not needed.) The ability to build so much machinery into such small volumes implies a substantial emulation and miniaturization ability, compensating to some extent for the limited materials assumed in the nanofactory architecture (basically, carbon-based lattices). For example: bulk properties of elasticity and energy absorption (plasticity) can be reproduced by nanoscale trusses with various hardware in the joints.

Some components in use today, such as computers and electromagnetic motors and transformers, are grossly oversized compared with the volume of nanostructured diamondoid that would provide similar functionality. It may be possible to design standard components that combine structural and functional properties. For example, a motor outputting power on a shaft might be replaced with merely a shaft of comparable dimensions with the power transformation hidden inside, freeing substantial mass and volume. (The greater efficiency implied by scaling laws indicates that cooling may not be a problem even for substantial sizes of equipment. This will certainly be taken into account.) Materials incorporating sensing, computation, and display might also be useful: instead of specifying that a display (or even paint) be attached to a surface, simply specifying the surface property directly may be sufficient if the surface structural material can be programmatically designed to exhibit the desired property.

Structural components today are monolithic and bulky, often to resist buckling from compressive force. Nanostructured components may provide elegant ways to convert compressive into tensile stress, including fractal trusses and pressurized volumes. Obviously, it is preferable to specify the structural characteristics without having to define each element of a fractal truss.

A moderate improvement in today's design could be achieved with “smart materials” that can sense impending failure. Likewise, materials with distributed built-in actuators and sensors could change shape to some extent, perhaps allowing new aerodynamic designs. Various other uses for smart materials could be easily imagined.

All these virtual materials will require interfacing between components. Some attention will be given to the bookkeeping required of a CAD program to ensure that adjacent components make proper functional connections.

Virtual materials will of course be built out of nanoblocks. The Nanofactory paper specifies some useful categories of nanoblocks. It may be necessary to specify nanoblocks in more detail. However, we will not address detailed mechanical design of nanoblock internals, but remain at the level of achievable function. With the performance implied by diamondoid materials and with over ten million potential features per nanoblock, we do not expect that there will be any question as to whether the desired nanoblock functionality will be achievable.

With nanoscale components, radiation damage is a significant engineering issue. Some redundancy will be required for any product larger than a few nanoblocks. In general, this redundancy can be implemented either within nanoblocks or at the level of a few nanoblocks (as part of the virtual material). By implementing redundancy at low levels, any desired degree of reliability can be achieved at the design level, frequently at negligible mass/volume cost. Like heat dissipation, this is a factor we will keep in mind, but is not expected to be a major constraint in most cases.

During the third month, we will be in a good position to give a status report. We will have been thinking all along about how our designs will correspond to the product features identified in the first month. Thus, we will already have some idea of how successful the remainder of the project will be, in which we will study the use of automated design to translate requirements into construction.

Fourth and Fifth Months: Automated Requirement Fulfillment

The fourth and fifth months will be used to investigate how to produce the desired features and functions given the available construction operations. Although we do not anticipate having time to write a CAD program, our intention is to outline the algorithms that would be useful for such a project. The goal by the end of these two months is to demonstrate that an engineer working to duplicate a typical product (a laptop computer, an air filter, etc) would not have to worry about the fact that his design would be built with a nanofactory. The functional/structural specification should be algorithmically translatable into nanofactory control software, so that the product can be built as soon as it is designed.

In the middle of the fifth month, our status report will list our success for each product requirement. Success will be measured in three increments: 1) Algorithms for implementing the requirement are known. 2) Algorithms are unknown, but are expected to be findable with further effort. 3) There is reason to think that the requirement cannot be met with the planned system architecture of nanofactory, nanoblock, and limited post-manufacture reconfiguration. We do not actually expect that any requirements will end up in the third category.

Sixth Month: Wrap-up and Implications

The sixth month will be used for review, improvement of the designs, and final writeup. Individual pieces of the work will have been reviewed by others during the project, but review of the integrated vision will probably show some open questions. We do not anticipate the need for major revisions, but additional questions may need to be handled explicitly.

Success of the project will be measured by how complex a product can be designed, with all properties decomposable into volumes of specified virtual materials that can be constructed and assembled without human intervention. A truss? A computer? An attitude thruster? An entire spacecraft?

The final deliverable will include five sections: 1) The progress report from the fifth month, updated as appropriate, showing which product features can be designed and built. 2) Analysis of which classes of products can be designed within these limitations. 3) Analysis of the performance that can be expected from these products, with emphasis on mass, size, cost, and time advantages vs. mainstream fabrication methods. 4) Analysis of the expected utility of a nanofactory-like fabrication system for ground-based and in-flight manufacturing of research and production hardware. 5) Recommendations for future work.

Management Approach

The small size of the project and the demonstrated self-sufficiency of the investigators indicate that an project management will be done on an informal basis.

Communication and coordination between Phoenix and Toth-Fejel will consist of phone and email. Problems will be assigned as they occur, with substantial brainstorming anticipated.

Informal collaboration with a number of other researchers in the molecular manufacturing community is also anticipated, though not formalized or committed at this time. This is expected to be as-needed and unfunded. Past work experience, personal acquaintance, and informal conversations about this proposal indicate that significant help will be available on request, although the project is planned to be completely accessible to the two named researchers.

Personnel

Chris Phoenix, the PI, has studied molecular manufacturing for fifteen years since taking a class from Eric Drexler at Stanford University. He obtained a BS in Symbolic Systems and an MS in Computer Science there, then worked for six years as an embedded software engineer in a groundbreaking company (Electronics for Imaging) that created the color desktop publishing industry. He has spent the last three years focused full-time on nanotechnology, co-founding the Center for Responsible Nanotechnology in 2002 to promote discussion of the consequences of advanced nanotechnology and molecular manufacturing. His articles have been published in the Federation of American Scientists' Public Interest Report and on several prominent technology websites. His peer-reviewed technical publications include “Vasculoid: A Personal Nanomedical Appliance to Replace Human Blood” (primary author: Robert Freitas) and “Design of a Primitive Nanofactory.”

Tihamer Toth-Fejel has been interested in self-replication since finding out about it while working on it in his EE Master’s Thesis, “Self-Test: From Simple Circuits to Self-Replicating Automata”. While working on this thesis, he obtained a pre-publication of the self-replication section of the 1980 NASA summer study from Robert Freitas, and immediately saw the importance of self-replication. For his second employer after graduate school, he worked on hypertext networks based on the Internet, so when the HTML standard came out, he was quick to initiate and complete a project to transfer the NASA summer study onto the web. Toth-Fejel’s 1996 article “LEGOs to the Stars” anticipated much of the work that appeared after 2000 (especially that of Chirikjian and Suthakorn), and appropriately was referenced in the NASA Ames’ 1998 summary on aerospace applications of nanotechnology. Working as a research engineer on a wide variety of projects, many on which he has published in technical publications. Toth-Fejel has recently completed a Phase 1 NIAC study, “Modeling Kinematic Cellular Automata: An Approach to Self –replication.”

Special Matters

This project does not require any environmental statement to be filed; no human subject or animal care provision statements need to be made, etc.

World Care is a 501(c)3 non-profit corporation founded in 1996; its mission includes education. Center for Responsible Nanotechnology (CRN) is a project / affiliate of World Care. CRN's bookkeeping is done by World Care staff; its operations and strategy are planned in consultation with World Care's president; its technical research function is essentially independent.

Chris Phoenix, the PI for this proposal and Director of Research for CRN, has produced over 100 pages of peer-reviewed technical analysis and architecture of modular (Vasculoid) and monolithic (Nanofactory) systems based on molecular manufacturing. Tihamer Toth-Fejel has recently completed a NIAC Phase 1 project on modular self-replicating systems.