High-Throughput Nanomanufacturing:
Assembly (with videos)

by Eric Drexler on 2009/03/01

Tin robots image, Flickr (cc) :mrMark:
Not efficient
for nanofactory automation

If you wanted to use automation to assemble an enormous number of small things, would you use robots? For throughput in the 100 ms/cycle, million-product-per-day range, a room full of robots waving their arms around might not be the best solution. A manufacturing engineer is more likely to think of using a machine like the one in the video below, shown as it assembles spray-pump parts at a rate of 500 parts per minute. Watching the first 30 seconds is enough, unless you fall in love with the machine’s rugged elegance:

In watching this, please keep in mind that the dynamics of macroscale machinery processing ~10 parts per second corresponds roughly to the dynamics of nanoscale machinery processing ~10 million parts per second. This is a consequence of elementary physical scaling laws and is the basis for the potentially high productivity of advanced nanomanufacturing.

In my previous post, I presented videos that show high-throughput systems making metal parts, and explained why manipulation of matter at the smallest scales, in the first stages of atomically precise high-throughput nanomanufacturing, should follow similar principles of operation, using sequences of stages that each perform simple, repetitive, high-frequency motions. The video above shows how the same basic principle is used to perform the basic quantum of assembly, putting two parts together.

Looking ahead to a time when we become competent at making roughly analogous nanoscale machinery, as physics promises we can, how could we use that sort of machinery in high-throughput nanomanufacturing to make products large enough to see, like a computer, car, or spacecraft? The basic idea is simple: Use simple high-throughput mechanisms to put parts together to make larger parts, then iterate, putting these together to make even larger parts, a process termed convergent assembly. A sequence of 100 pairwise assembly operations can span the gap between nanometer scale building blocks and multi-ton products.

As assemblies become larger, though, it makes sense begin using more complex machinery, as we see in ordinary factories. Simple machines pour out identical bolts, but a car is valuable enough to merit the cost and complexity of computer-controlled assembly and customization — and even the application of human hands and minds.

Here’s a video that shows more continuous-flow assembly machinery, and gives a glimpse of the process that extracts a stream of oriented parts from a jumbled heap:

You may notice similarities between these mechanisms and those shown in the nanomanufacturing video below, “Productive Nanosystems: From Molecules to Superproducts”. This is no coincidence. The animator, John Burch, is a trained mechanical engineer.

See also:

{ 8 comments… read them below or add one }

Erin March 1, 2009 at 10:42 pm UTC

[...] Molecular mills and similiar systems seem to be the more efficient of the possibilities.
So essentially once we would get past the early “primitive” nanosystems that are used to bootstrap more advanced assembler technology, then these molecular level mill and related systems would likely take over, correct?

Chris Phoenix March 2, 2009 at 5:53 pm UTC

Beware, lest people extract the wrong ideas from these videos and explanations. It might be worth emphasizing that molecular manufacturing does not involve material-forming steps like those in the bolt manufacturing process. Parts will be built in their final shape, held together in a rigid structure by inter-atomic bonds. The “assembly” steps go all the way to the lowest atomic/molecular level.

It might also be worth a discussion of the similarities and differences between the parts-orienting machine and a binding site (like those contemplated for sorting rotors). The machine has to physically move the parts, and separate the parts that fall into line from the parts that don’t.

It’s not visible in the video, but the entire drum is vibrating in such a way that the parts climb the ramp – they are shaken uphill, and shaken into line, and the ones that end up oriented backwards are dumped back into the drum.

In a binding site system, thermal noise is what shakes and moves the parts, and the binding site only attaches to parts that are in the right orientation.


Ps. typo: “a car valuable enough” -> “a car is valuable enough”

Scott Jensen March 2, 2009 at 8:17 pm UTC

How about the reverse? What about a universal disassembler? Could the above process be reversed to return products to their elements?

In sci.nanotech newsgroup, I’ve started a thread about such a thing and described what I think might be a way of doing it. Thread title is “Nanotech garbage disposer?”.

Guy March 3, 2009 at 6:27 am UTC

When I was investigating molecular imprint lithography I was thinking about what it would be like to have a roll-to-roll cylinder that had it’s surface imprinted in that fashion. This way a finite surface could supply an infinite work area. Each cylinder could stage a single step in a process. Successive cylinders would roll against each other, picking up and compositing increasingly more complex parts. Feedstock molecules for each stage would be supplied from a solution basin or porous “wick” source. It’d still be self-assembly and probabilistic but the probabilities at each stage would be fairly tightly constrained.

Eric Drexler March 3, 2009 at 8:59 pm UTC

@ Scott Jensen — Molecular synthesis steps usually aren’t directly reversible. We can take a lesson from biology here: Making complex structures like proteins is done by intricate machines; taking them apart is done by simple digestive enzymes. I used the term “disassembler” in my 1986 book to refer to a class of instruments for analyzing a structure in molecular detail, a process which is very different from digestion or recycling. Unfortunately, the term suggests a false analogy, and has helped fertilize the ground for science fiction written with sloppy science. (By the way, and for the record, I also explained that the idea of a genuinely “universal assembler” is mistaken.)

@ Guy — Yes, nanoimprint lithography and micro-contact printing are interesting processes, and provide a nice demonstration that direct interactions between physical structures can produce effects with higher spatial resolution that can be achieved using large, expensive, state-of-the-art machines that aim beams of photons or charged particles at a surface. (This is not to say that the contact techniques can replace them, of course.)

Will Ware March 9, 2009 at 7:54 pm UTC

Assembly lines offer a tradeoff between speed and flexibility. Many of today’s assembly lines are inflexible in the sense that retooling to produce a different product is a costly and time-consuming affair. The bolt-producing machine in the video is an example of an assembly line that is optimized for speed, but it can make only one specific product.

Just as I want my microwave to heat many different kinds of food, and my home computer to run lots of different programs, I’m hoping my desktop nanofactory will produce a wide variety of products. One thing in the nanofactory video that’s rarely mentioned but I think fairly prescient is that there is a nice balance between fast inflexible assembly lines in the early assembly stages and robot arms later on, where it makes sense, having built little blocks with different little functions, to connect the blocks together in different configurations. (Chris Phoenix’s paper from a few years ago, discussing nanoblocks, is also prescient in this regard.)

Potentially useful mental models here are programmable logic chips such as FPGAs, which offer a huge amount of flexibility in a standardized package, and the software engineering practice of profiling, which is the measurement of code execution time on a function-by-function basis, and which is the only reliable way to decide which code is worth the trouble to hand-optimize. There is probably an assembly line equivalent of software profiling, just as Chris’s nanoblocks can be thought of as nanomechanical equivalents of FPGAs. (At least that’s how I think of them.)

Eric Drexler March 10, 2009 at 8:13 pm UTC

@ Will — Yes, that’s a good way to think of the tradeoffs and how to balance them.

Jonathan Banks December 28, 2010 at 12:10 am UTC

Hello Eric I read Engines back in 96 when none I spoke to knew anything about nanotechnology (right up to around 2004) but it certainly fired me up.

So are you saying now that the original idea of having the nano equivalent of a seed or something larger than a desktop nanofactory to create large scale artifacts like cars, jets, spacecraft etc isnt feasible?

Will things that big have to be assembled in normal style factories from nanofactured building blocks?

What about buildings?

If nature can build a giant tree from a seed why couldnt we eventually do something similar? Just in terms of using nanofacturing to take the artifact all the way from beginning to completion no matter how large it is.

Also is the concept still looking feasible that nanofactured artifacts can be artificially intelligent and alive in the sense of models developed in the field of artificial life?

{ 7 trackbacks }

Leave a Comment

Previous post:

Next post: