Technoscientific development is difficult to direct and nearly impossible to predict.  Because of this – not in spite of it – panel discussions like “How Do We Get There From Here?” are crucial: they allow us to speculate, imagine, and contest claims and predictions about emerging technologies.  However, planning and debating are only part of the picture.  As Dr. Hall aptly noted it’s not dispassionate calculations but “serendipity: the way science always works.”

Here’s the full report:

Getting from here to there: aiming at a moving target in a changing landscape?
Dave Conz, PhD

Note: the views and opinions expressed below are the author’s own and do not necessarily reflect those of the Center for Nanotechnology in Society at Arizona State University, Arizona State University, or the National Science Foundation.  The author would like to thank Chris Peterson and the Foresight Institute for providing him access to the Society of Manufacturing Engineers Nanomanufacturing conference at no cost.

I received an email invitation from Chris Peterson at the Foresight Institute a few weeks ago to attend the Society of Manufacturing Engineers Nanotechnology and Microtechnology conference in Mesa, Arizona. The generous offer was too sweet to pass up: attend the conference, write a blog, and your registration fees will be waived. I jumped at the chance.
Here I offer my reflections on some of the highlights of the presentation by Dr. J. Storrs Hall of the Foresight Institute, entitled “Feynman’s Pathway to Nanomanufacturing,” and the panel discussion that followed, “How Do We Get There from Here?”  Discussions such as these are crucial opportunities to reflect on – and potentially shape – emerging technologies whose destinies are often left to be determined by “market forces.”

Dr. Hall began with an intriguing argument: Feynman’s top-down approach to reaching the nano scale in manufacturing, achieved through a step-down method of replicating and miniaturizing an entire, fully-equipped machine shop in 1:4 scale over and over would yield countless benefits to science, engineering, and manufacturing at each step. These microscopic, tele-manipulated master-slave “Waldos” (named after Heinlein’s 1942 story “Waldo F. Jones”) would get nanotechnology back on track by focusing on machines and manufacturing, since most of our current emphasis is on science at the nano scale.  Feynman’s top-down approach to nanoscale manufacturing is missing from the Foresight Institute’s roadmap, according to Hall, “for political reasons.” This raises a fundamental point: science and technology cannot develop independent of the political and social spheres, which pose as many challenges as the technology.  Many would argue that social and technological processes are inseparable and treating them otherwise borders on folly.  I commend Dr. Hall for offering his argument.  It soon became clear that the panelists who joined him after his presentation disagreed.

Technoscientific development is difficult to direct and nearly impossible to predict.  Because of this – not in spite of it – panel discussions like “How Do We Get There From Here?” are crucial: they allow us to speculate, imagine, and contest claims and predictions about emerging technologies.  However, planning and debating are only part of the picture.  As Dr. Hall aptly noted it’s not dispassionate calculations but “serendipity: the way science always works.”

On the panel, Tihamer Toth-Fejel (from General Dynamics Advanced Information Systems) countered Hall by arguing that no funding exists for the incremental steps proposed by Feynman and the knowledge or potential products and processes gained from each stage would not be sufficient to finance the next step down.  He took the decidedly “engineer” approach to problem-solving as the driving factor toward the nano scale and argued that we should try to accomplish the goal of nanoscale manufacturing by traversing the shortest distance between top-down and bottom-up.  He argued that we must be able to demonstrate to investors that we can accomplish an 18-month payoff if we are to have any hope of funding this adventure.  Joining the debate, David Keenan (Small Technology Consulting, who had, earlier that morning, given an overview of recent or imminent market breakthroughs in nanotechnology in his presentation) made a cogent argument for the need to restructure our current pedagogical approaches toward educating future engineers if we are to tackle the “hurdles, bottlenecks, and next generation of cross-disciplinary challenges” around nano.  This tack is especially salient if we are to shape the way we think about approaching engineering at the nano scale.

One provocative statement by Dr. Hall during his earlier talk was, “If we would have taken Feynman’s advice back then, we’d have nanobots today.”  This is akin to stating that, in my humble view, had we followed Feynman’s advice we’d have the contents of the Library of Congress written on the back of a postage stamp.  Both might be possible but statements like this perplex me: On the one hand they are moot and and on the other, meaningless without a line of additional follow-up questions.  Yes, we are capable of nanowriting but in 2010 with networked digital computing and e-readers who among us would opt for an electron microscope and Feynman’s postage stamp library?  The problem is we are aiming at a moving target in a changing landscape.  That said, it can be valuable (if not necessary) to imagine alternatives and create space for multiple possibilities as we create new devices, processes, and knowledge.  In this light, we could ask “If we had followed Feynman’s advice and accomplished his ‘step-down’ scaling, how might we be better and/or worse off than we are today?”

Let’s discuss.

IBM Research in Zurich has demonstrated a new nanoscale patterning technique that could replace electron beam lithography (EBL). The demonstration carved a 1:5 billion scale three-dimensional model of the Matterhorn, a 4,478 meter high mountain lying on the border between Italy and Switzerland, to show how their technique could be used for a number of applications, such as creating nanoscale lenses on silicon chips for carrying optical circuits at a scale so small that electronic circuits are inefficient…

The demonstration also sculpted a relief map of the world that measured 22 by 11 micrometers. According to the IBM press release the scale of the map is so small 1,000 of them could be drawn on a single grain of salt. IBM says the current technology can go as small as 15 nanometers, but in the future could go even smaller.

See also the abstract from Science.  Impressive top-down nanotech: who knows how small they will go?  —Chris Peterson

Nicw round-up with videos of the latest in the Rep-Rap world.

Vibratory Parts Feeder

One way of reducing friction at the macroscale is vibration, as in this vibratory parts feeder. Now nanotechnologists at IBM have harnessed vibration to reduce wear at the nanoscale:

In their paper, published in the September issue of Nature Nanotechnology, IBM scientists solve this challenge by “demonstrating the effective elimination of wear on a tip sliding on a polymer surface over a distance of 750 meters by modulating the force acting on the tip-sample contact.” By applying an AC voltage between the cantilever—the mechanical arms on which the tips are attached and over which they are controlled—and the sample surface, the cantilever can be excited at high frequencies of one Megahertz. The cantilever bends and the tip vibrates with an almost imperceptible estimated amplitude of one nanometer.

This is over a thousand times faster vibrations than in macroscale machines, but then that’s what one would expect…

[“Dynamic Superlubricity and the Elimination of Wear on the Nanoscale” by M.A. Lantz, D. Wiesmann, and B. Gotsmann,  Nature Nanotechnology 4:9 (September 2009)]

First is simply telerobotics, as in Feynman Path manipulation.  We want the feedback to help develop an intuitive feel for mechanism at all the scales from here to molecular.

The second is for robotics/AI. A haptic telerobot control means that all the signals, motor and sensory alike, have to be handled and transmitted.  It’s all too common in robotics to try to get away with too little feedback, and it’s much worse in the upper levels of AI. Any trend in full-duplex control software is likely to contribute to the development of robust AI, IMHO.

Purely coincidentally, we at Foresight have been discussing self-replication in the context of the Feynman Path and I came up with an example that shows just how counter-intuitive self-replication, if you try to view it as a capability, can be.

Self-replication is a poor criterion to use to judge risk, either of autonomous runaway or hijackability.  Consider, for example, two versions of the Drexler/Burch nanofactory:

1) as shown: the input is pressurized cannisters of fairly pure acetylene and possibly other refined chemical feedstocks.

2) the input is cassettes of nanoblocks, as output by the next-to-last stage in (1).

Now I claim that isn’t too hard to make (2) self-replicating.  All it does is slap nanoblocks together in the right patterns; maybe 10% of the total functionality of (1).  And it’s a lot more likely you can design a machine that does that entirely out of nanoblocks.  Bingo, a self-replicator.

On the other hand, it’s quite difficult to make a self-replicating version of (1).  From the lowest, mechanosynthetic levels of (1), it’s a hardwired, cast-in concrete gadget that builds nanoblocks.  To build all the gadgetry in (1) as well, it’d take probably 100 times as much mechanism.

Now to us, (1) is the much more capable machine.  After all, look at all it’s doing.  But to the user, (2) is much more capable.  Both machines require the user to go out and buy feedstock containers — pressurized acetylene pods don’t grow on trees.  Cost difference between pressurized cylinders and cassettes would be minimal: given the technology, it would be about as cheap to run the feedstock through a nanoblock maker and packer as to pump it into cylinders.

But machine 2 could make copies of itself and machine 1 could not.

And yet we know that not only does machine 1 do more stuff, but the range of outputs for the two machines is exactly the same! (Note that machine 1 can make a machine 2. Neither can make a machine 1.)

And yet which is more dangerous?  Consider which one would do the government of, say, Iran, today, the most good in terms of bootstrapping itself to full nanotech capability if one of each fell into its hands?  Obviously (1).

So I claim that self-replication is essentially worthless as a criterion by which to judge risk of accidents or abuse.

Some highlights (from the chapter on manufacturing):

Vignette 2: One-of-a-kind, discrete-part manufacture and assembly

A small job shop with 5 employees primarily catering to orders from medical devices companies is
approached by an occupational therapist one morning to create a customized head-controlled input
device for a quadriplegic wheelchair user. Today the production of such one-of-a-kind devices
would be prohibitively expensive because of the time and labor required for setting up machines
and for assembly. The job shop owner reprograms a robot using voice commands and gestures,
teaching the robot when it gets stuck. The robot is able to get the stock to mills and lathes, and runs
the machines. While the machines are running, the robot sets up the necessary mechanical and
electronic components asking for assistance when there is ambiguity in the instruction set. While
moving from station to station, the robot is able to clean up a coolant spill and alert a human to
safety concerns with a work cell. The robot responds to a request for a quick errand for the shop
foreman in between jobs, but is able to say no to another request that would have resulted in a
delay in its primary job. The robot assembles the components and the joystick is ready for pick-up
by early afternoon. This happens with minimal interruption to the job shop’s schedule.

For a manufacturing timeline:

5 years: Achieve ability to set up, configure and program basic assembly line operations for new products
with a specified industrial robot arm, tooling and auxiliary material handling devices in under 24 hours.
10 years: Achieve ability to set up, configure and program basic assembly line operations for new
products with a specified industrial robot arm, tooling and auxiliary material handling devices in one 8
hour shift.
15 years: Achieve ability to set up, configure and program basic assembly line operations for new products
with a specified industrial robot arm, tooling and auxiliary material handling devices in one hour.

Autonomous navigation:

5 year: Autonomous vehicles will be capable of driving in any modern town or city with clearly lit and
marked roads and demonstrate safe driving comparable to a human driver. Performance of autonomous
vehicles will be superior to that exhibited by human drivers in such tasks as navigating through
an industrial mining area or construction zone, backing into a loading dock, parallel parking, and
emergency braking and stopping.
10 years: Autonomous vehicles will be capable of driving in any city and on unpaved roads, and exhibit
limited capability for off-road environment that humans can drive in, and will be as safe as the average
human driven car.
15 years: Autonomous vehicles will be capable of driving in any environment in which humans can
drive. Their driving skill will be indistinguishable from humans except that robot drivers will be safer
and more predictable than a human driver with less than one year’s driving experience.

Dextrous Manipulation:

5 years: Low-complexity hands with small numbers of independent joints will be capable of robust
whole-hand grasp acquisition.
10 years: Medium-complexity hands with tens of independent joints and novel mechanisms and
actuators will be capable of whole-hand grasp acquisition and limited dexterous manipulation.
15 years: High-complexity hands with tactile array densities approaching that of humans and with
superior dynamic performance will be capable of robust whole-hand grasp acquisition and dexterous
manipulation of objects found in manufacturing environments used by human workers.

Nano-Manufacturing:

5 years: Technologies for massively parallel assembly via self-assembly and harnessing biology to
develop novel approaches for manufacturing with organic materials.
10 years: Manufacturing for the post-CMOS revolution enabling the next generation of molecular
electronics and organic computers
15 years: Nano-manufacturing for nano-robots for drug delivery, therapeutics and diagnostics.

So it is mainstream to be calling for nano-robots in the mid-20s.  My guess is that the kind of nano-robots they have in mind are not the fully-capable kind implied by the Feynman Path, however, but something more like present-day MEMS.  In that case, generally-capable manufacturing nanobots by 2030 may still be on the optimistic side; and there is plenty of headroom for the Feynman Path to accelerate the process.

It’s the 20th anniversary of the first Foresight Conference this year. Over the intervening two decades, one of the most common questions of Foresight members and supporters has been, “What can I do to help with the development of nanotech?”  Foresight has had many useful programs, and encouraged development in many ways (notably with the Feynman Prizes, in the spirit of the prizes Feynman himself offered for developments leading along his pathway).  But we have never taken a hand in the direct development of nanotech per se.

I feel to some extent that this may have contributed to the lack of focus the field of nanotech has had in its course of development. But that can change. The Feynman Path initiative is a specific, concrete proposal — but more, it’s one that can be done in an open-source way, for at least the first, roadmap, phase. Anyone can contribute design ideas.

Moreover, anyone can begin to experiment with a macro KSRM model. Getting past that giggle factor and having a real, physical machine that people can watch as it copies itself could cause a sea change in attitudes and the orientation of research.

There’s absolutely no need to have just one model of a KSRM. I’ll be trying to build one myself, and blogging about the details, but this should be a community of free and open ideas. Many actual machines, variants of the original design, can be built, as the RepRap community shows.

There’s a fundamental similarity between a Feynman Path machine (FPm) and a RepRap, obviously, in their orientation to self-replication.  This includes the fact that both schemes require a human to be actively involved in the replication process, in the FPm by teleoperation.  But there are some fundamental differences:

• Attitude to cost: a RepRap is intended to be a means to cheap manufacturing, so it’s oriented to using the least expensive materials available.  An FPm has much less concern about that: each successive machine in the series uses less than 2% the material of the previous one. It would be perfectly reasonable to design an FPm that had to carve all its parts out of solid diamond, once past the millimeter scale, for example. The goal is to understand principles, not supplant the economy (at least until the nanoscale is reached).
• Attitude to closure:  RepRap assumes human assembly labor, but an FPm has to provide its own manipulating capabilities. RepRap allows exogenous parts that are widely available and inexpensive; an FPm allows parts that are available at all scales.
• Assembly time vs accuracy: As a consumer-goods production machine, RepRap has at least some concern for how long it takes to do its job.  An FPm has much less concern about time, but much more about accuracy, since it has to improve its product’s tolerance over its own by a substantial factor.

Given that, however, there’s no reason that there shouldn’t be a free flow of ideas between the projects.  RepRap took a long time getting off the ground, and so may the Feynman Path project — but if you want to help, in any capacity or form whatsoever, let me know!

There hasn’t been a lot of work on self-replicating workcells. There’s been plenty on robotic workcells that don’t replicate, but almost all of this falls into the “more complex than what it makes” category. The basic idea goes back to Waldo: imitate a machine shop and the person servicing the machines / assembling the parts.

Back in part 3 I wrote:

It seems clear that a major step toward the Feynman Path would be to work out a scalable architecture for a workable KSRM that actually closed the circle all the way. A reasonable start would be a deposition-based fab machine, a multi-axis mill for surface tolerance inprovement, and a pair of waldoes. See how close you could get to replication with that, and iterate.

I used the idea of a general-purpose manipulation robot in my Architectural Considerations for Self-replicating Manufacturing Systems paper:

but this probably more complex than would be needed for the Feynman Path. The difference is that the system in the paper was geared for producing significant amounts of product as well as reproducing/extending itself. We could probably get away with just a pair of arms on a rotating base with access to a few machines and a workbench (for assembling more machines).

A Feynman Path workcell actually avoids the problem that a standard solid-freeform-fab (SFF) design has with building something its own size, because it’s building a copy that’s smaller than itself! Even so, it will have to have general manipulation capability — and the target system, at the nanoscale, will have to build more copies at its own size, so we can’t go with just a simple SFF design.

Even so, SFF is the key to giving the replicating system a manageably small size. It’s the main technology that wasn’t here in Heinlein’s and Feynman’s day.

The key to using SFF in a scaling sequence is to understand what kinds of depositions could be done at different stages. One of the most straightforward at the macroscale is melting the substance of interest and allowing it to cool as you deposit essentially drop by drop on the workpiece. This works for materials ranging from wax to titanium. Scaling works both for and against us here — the super-fast dissipation of heat at smaller scales means you have greater control in time, but less in space, of what melts.

At smaller scales, electrodeposition (and electro-removal, as in EDM) will likely have to be used. At the smallest scales, the processes used in electroplating, but controlled at the near-atomic scale, are good candidates.

A particularly important aspect of the Feynman Path is that not much more than halfway down to molecular scale in part size, we already hit atomic scale in tolerance. That’s within a generation or two from our likely starting point at 1/1000 scale. A micron-sized part really needs atomic-scale tolerance to be considered high-precision. Thus much of the work in that size range will be aimed at surface forming or re-forming. Even so, there will be a pressure to design machine elements where bearing surfaces are flat (as in a thrust beating or slider) so they can follow crystal planes, until such time as it becomes possible to construct strained-shell circular bearings (simple example: MWCNTs aka nested buckytubes).

Motors

It seems very likely that the motors we use will be electrostatic steppers. Virtually every micro- and nano-scale motor built so far has been an electrostatic stepper (or at least what we might call “capacitive-synchronous”).

Both the motors themselves, and the distribution of power and control to them, will require the SFF to be able to lay conductive paths in non-conductive structure. Given that the motors will be being fabbed in place, it will be easy to integrate high-ratio reducers into them for extremely fine angular control:

and build them directly into the manipulator arms and the leadscrews of the SFF and finishing machine(s).

In the last post we suggested that finding the appropriate starting point was one of the critical items to address in forming a Feynman Path roadmap, and that is true. A thorough survey of available techniques should be made, and recent advances in machining, nanomanipulation, and so forth taken advantage of.

However, as a point of reference, at least one experiment has been made, in a sense, which suggests that a 1/1000-scale system might be achievable (as compared to the desktop-scale prototype with finger-size parts). To quote from Freitas and Merkle’s encyclopedic Kinematic Self-Replicating Machines (full text available online):

[I]n 1994 Japanese researchers at Nippondenso Co. Ltd. fabricated a 1/1000th-scale working electric car. As small as a grain of rice, the micro-car was a 1/1000-scale replica of the Toyota Motor Corp’s first automobile, the 1936 Model AA sedan. The tiny vehicle incorporated 24 assembled parts, including tires, wheels, axles, headlights and taillights, bumpers, a spare tire, and hubcaps carrying the company name inscribed in microscopic letters, all manually assembled using a mechanical micromanipulator of the type generally used for cell handling in biological research. In part because of this handcrafting, each microcar cost more to build than a full-size modern luxury automobile. The Nippondenso microcar was 4.8 mm long, 1.8 mm wide, and 1.8 mm high, consisting of a chassis, a shell body, and a 5-part electromagnetic step motor measuring 0.7 mm in diameter with a ~0.07-tesla magnet penetrated by an axle 0.15 mm thick and 1.9 mm long. Power was supplied through thin (18 micron) copper wires, carrying 20 mA at 3 volts. The motor developed a mean torque of 7 x 10^-7 N-m (peak 13 x 10^-7 N-m) at a mean frequency of ~100 Hz (peak ~700 Hz), propelling the car forward across a level surface at a top speed of 10 cm/sec. Some internal wear of the rotating parts was visible after ~2000 sec of continuous operation; the addition of ~0.1 microgram of lubricant to the wheel microbearings caused the mechanism to seize due to lubricant viscosity. The microcar body was a 30-micron thick 20-milligram shell, fabricated with features as small as ~2 microns using modeling and casting, N/C machine cutting, mold etching, submicron diamond-powder polishing, and nickel and gold plating processes. Measured average roughness of machined and final polished surfaces was 130 nm and 26 nm, respectively. The shell captured all features as small as 2 mm on the original full-size automobile body. Each tire was 0.69 mm in diameter and 0.17 mm wide. The license plate was 10 microns thick, 0.38 mm wide and 0.19 mm high.

Nippondenso Micro-Car

In other words, it’s pretty clear that the technology exists today to manipulate micron-scale parts, and to make parts with a few tens of nanometer roughnesses (not the same as tolerance, but just as important in many cases).  It’s important to note that the Nippondenso Microcar had relative tolerances more like MEMS than high-precision machining. However, given the techniques developed since 1994 in aid of mainstream nanotechnology, it’s very likely that considerably finer tolerances (and roughnesses) are possible today.  If so, we could start the Feynman Path halfway down — at 1/1000 scale.  Another factor of 1000 and we have flat-out nanotech.