Fast atomically precise manufacturing is not hard sci-fi

30 posts / 0 new
Last post
NewtonPulsifer NewtonPulsifer's picture
Fast atomically precise manufacturing is not hard sci-fi

If your bonding rate is 100 nanometers per minute, and layering in 2d, a 1 cm thickness is going to take 100,000 minutes. That's 69.44 days. And assuming you can have a tip head that operates on every single possible atomic space at once. That's even not realistic - it would be more like 1000 times slower (190 years) than that (to make room for a 33x33 atom tip) if the layer is not uniform (if the layer is uniform - say 1 layer of cobalt - you can go faster).

So basically 2d layered 3d "Atomically Precise Nanomanufacturing" using an inkjet like technology would only be "fast" for very thin uniform things that could easily be assembled into larger/thicker things by chopping them up and stitching/gluing/bonding them together.

Admittedly filling a void with a liquid and letting all the individual molecules bond to each other is a different animal (and can happen concurrently) but if you need the item to be precisely laid (with a lot of variation) out down to the atom it's still going to be incredibly slow/expensive to produce.

P.S. This is a continuation of what I'll call my "singularity delayed due to faulty back of the envelope calculations" series some previous stuff at http://eclipsephase.com/fast-moving-nanites-are-not-hard-sci-fi http://eclipsephase.com/anti-nanite-weapon-flamethrower

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

Arenamontanus Arenamontanus's picture
NewtonPulsifer wrote:If your

NewtonPulsifer wrote:
If your bonding rate is 100 nanometers per minute, and layering in 2d, a 1 cm thickness is going to take 100,000 minutes. That's 69.44 days.
...
So basically 2d layered 3d "Atomically Precise Nanomanufacturing" using an inkjet like technology would only be "fast" for very thin uniform things that could easily be assembled into larger/thicker things by chopping them up and stitching/gluing/bonding them together.

Exactly. This is why you would need to do recursive assembly rather than use a glorified 3D printer.

However, for game explanation purposes calling things printers is much easier than trying to explain recursive assembly, microrobotics and the use of pre-manufactured "lego pieces" in the feedstock.

Extropian

dark.blue.nine dark.blue.nine's picture
Quote:

Quote:

This is why you would need to do recursive assembly

What does "recursive assembly" mean or imply in the physical world? I get microrobotic and object-oriented nanoassembly, but not recursive assembly. Do you have a good explanatory link or two?

I understand what "recursive" means in software programming, and find links for that when I google "recursive assembly". But I'm not seeing any links for "recursive assembly" in the context of nanoassembly.

Any help appreciated. Thanks.

The eternal silence of these infinite spaces fills me with dread. -- Blaise Pascal

NewtonPulsifer NewtonPulsifer's picture
I took it to mean having a

I took it to mean having a bunch of very small atomically precise fabricators working in 3D - say in microgravity (please someone correct me if I'm wrong).

The first group of fabricators makes cubes of say 10 atoms on a side.

The next group of fabricators takes those 10 atom cubes and bonds them together to make 100 atom on a side chunks.

The next group of fabricators takes those and bonds them to 1000 atom cubes.

If the final desktop cornucopia machine uses those 1000 atom on a side cubes, it would speed up things up perhaps around 1000 times (depending on what you're making - like I said in the OP, you can fill a void with a liquid and let it bond concurrently, so it partly depends on how many of those continuous type volumes exist in your structure). More if the cornucopia machine could also be run in microgravity (you can bond top, bottom, even sides of your structure much more easily - call it a potential 6x speedup).

There's still some "here be magic" though that creates a big problem if we're talking about possible real world atomically precise manufacturing in the future - how the heck do you take take a 1000 atom on a side cube and just bond it to another and to create a seamless thing. "Cement" joins every 1000 atoms? Made of what? Does this destroy the usefulness of what you're fabricating (so you need to build it some other way)?

There's notable exceptions to this issue, like metals in crystal form can be cold welded with pressure (most everyday metals we use are crystals but they might use alot of amorphous metal alloys "bulk metallic glass" in the future and this method then becomes quite iffy).

Another problem is with a "bottom up" atomically precise manufacturing (in terms of the EP game only), you have an economic verisimilitude problem - the final assembler only put in 1/1000th of the total nanomanufacturing tip time in. So its economic value add is tiny.

So your final device ends up costing 10,000 credits and your cornucopia machine added just 10 credits of final value to it - the feedstock was 9,990 credits.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

nerdnumber1 nerdnumber1's picture
I doubt most fabrication is

I doubt most fabrication is "atom by atom", rather each part is assembled with the least precision that is practical and a fabber uses a wide variety of nano and micro bots to work on as many parallel construction tasks possible. It doesn't strictly have to be a layer by layer process when multiple modules can be created at once. I'm not saying it is 100% realistic but it is good enough for my suspension of disbelief.

Nebelwerfer41 Nebelwerfer41's picture
NewtonPulsifer wrote:If your

NewtonPulsifer wrote:
If your bonding rate is 100 nanometers per minute

Where did you find this number? Is it stated in the core book somewhere?

For the record, I don't have an issue with hand-waving the fact that items can be replicated from feedstock with a nano assembler, what breaks immersion for me is that Anarchists and anti-hypercorp factions should have distributed the these "locked/verboten" blueprints the mesh long ago.

NewtonPulsifer NewtonPulsifer's picture
Nebelwerfer41 wrote

Nebelwerfer41 wrote:
NewtonPulsifer wrote:
If your bonding rate is 100 nanometers per minute

Where did you find this number? Is it stated in the core book somewhere?

No, just taking a conservative number from actual atomic bonding measurements. You can get higher than that (like 6x), but it involves highish defect rates (like 1 in 1000) or where exacting structure isn't desired (amorphous solids).

Nebelwerfer41 wrote:

For the record, I don't have an issue with hand-waving the fact that items can be replicated from feedstock with a nano assembler, what breaks immersion for me is that Anarchists and anti-hypercorp factions should have distributed the these "locked/verboten" blueprints the mesh long ago.

Re:blueprints - yeah, its actually worse than that. There's a financial interest in even *regular* hypercorps releasing blueprints as open-source if they are only using the blueprints internally (not as an end product they sell). They then benefit from the "free" updates and improvements from other users of the blueprints (which may even just be other hypercorps).

But part of the reason blueprints are a big deal in EP is because you can buy a desktop cornucopia machine or protean swarm and a blueprint and basically print unlimited money with no serious cost with rules as written. If you actually analyze it, a desktop cornucopia machine isn't feasible except to generate feedstock for fabbers (at a very slow rate at that) - and that feedstock is the vast majority of the cost of any complex item.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

NewtonPulsifer NewtonPulsifer's picture
nerdnumber1 wrote:I doubt

nerdnumber1 wrote:
I doubt most fabrication is "atom by atom", rather each part is assembled with the least precision that is practical and a fabber uses a wide variety of nano and micro bots to work on as many parallel construction tasks possible. It doesn't strictly have to be a layer by layer process when multiple modules can be created at once. I'm not saying it is 100% realistic but it is good enough for my suspension of disbelief.

But for things like:

optronics
quantum computers
smart materials
nanites
superconducting batteries
metamaterials

It really *is* atom by atom.

If I get some time in the near future I'll update this post with a calculation of roughly how much atomically precise matter you could generate in 5 hours in a desktop cornucopia machine (assuming 3d desposition in microgravity).

In the end it might simply be enough to require very high feedstock cost for complex items. Something like a desktop cornucopia machine saves you (as a percentage of final price) depeding on item cost:

Expensive - saves you 1% of the market price
High - 3.3%
Medium - 10%
Low - 33%
Trivial - 100%

Higher than expensive you could simply disallow, or allow it but it doesn't save any money and adds the "lemon" trait to the item.

EDIT1: Keep in mind the game allows for dropping 1 price tier for craptastic versions of items (no specific rules for them though), which would probably be popular for desktop cornucopia machines in this scenario due to the potentially huge cost savings. Some people don't care if the item they made is 3x the length, height, and width and takes 27x the power (and has the lemon trait) if it is only staying at home plugged in, right?

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

NewtonPulsifer NewtonPulsifer's picture
Nebelwerfer41 wrote

Nebelwerfer41 wrote:
NewtonPulsifer wrote:
If your bonding rate is 100 nanometers per minute

Where did you find this number? Is it stated in the core book somewhere?

For the record, I don't have an issue with hand-waving the fact that items can be replicated from feedstock with a nano assembler, what breaks immersion for me is that Anarchists and anti-hypercorp factions should have distributed the these "locked/verboten" blueprints the mesh long ago.

From my perspective hand-waving the arbitrary bonding of feedstock cubes/chunks/polyhedrons together "somehow" puts nano-fabrication solidly into the realm of "soft" sci-fi not "hard" sci-fi" for rapid versions of it (hence the original title of the thread).

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

Madwand Madwand's picture
I'm sure there will be

I'm sure there will be significant engineering challenges that Eclipse Phase unrealistically glosses over for the sake of exploring the consequences of a transhuman setting more fully (and the fact that no one really knows). However, reasonable people can disagree on the real difficulty of atomic manufacturing. Take a look at this essay with much more optimistic projections:

http://e-drexler.com/p/04/04/0507molManConvergent.html

Eclipse Phase (reasonably, perhaps?) stands between the two extremes of time required given in the above essay and the OP.

nizkateth nizkateth's picture
Huh

I didn't know nano-fabbing in EP was supposed to be an atomic-level process. I thought it was more like restructuring larger bits of matter at closer to a cell-like scale or larger. So a cloud of nanites would work from every angle to assemble an object from tiny parts instead of printing it layer-by-layer.

Reapers: Do Not Taunt Happy Fun Ball.
My watch also has a minute hand, millenium hand, and an eon hand.

NewtonPulsifer NewtonPulsifer's picture
nizkateth wrote:I didn't know

nizkateth wrote:
I didn't know nano-fabbing in EP was supposed to be an atomic-level process. I thought it was more like restructuring larger bits of matter at closer to a cell-like scale or larger. So a cloud of nanites would work from every angle to assemble an object from tiny parts instead of printing it layer-by-layer.

It really depends on what you're building. If you're wanting to exploit the "cool" technologies like room temperature superconductor based batteries that require diamond strength structures to hold them together, metamaterials, carbon nanotube structures etc. you're going to be using things like carbon, which even the most optimistic proponents of Drexlerian mechanosynthesis don't believe can be done better than a couple of atoms at a time.

Otherwise you're stuck with bolting, screwing, riveting, or epoxying blocks together which radically reduces their strength (and thus usefulness in the first place) or structure (metamaterials).

And keep in mind any 3D manufacturing can be simplified into a 2D problem (and also that gravity complicates 3D manufacturing).

We also havent gotten a laundry list of other problems - things like the problems of positional uncertainty. The slightest jolt of a fabrictor could ruin whole batches of particles, and actually make a "desktop" version totally infeasible without some amazing (and heretofore unimagined) vibrational dampening.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

NewtonPulsifer NewtonPulsifer's picture
Madwand wrote:I'm sure there

Madwand wrote:
I'm sure there will be significant engineering challenges that Eclipse Phase unrealistically glosses over for the sake of exploring the consequences of a transhuman setting more fully (and the fact that no one really knows). However, reasonable people can disagree on the real difficulty of atomic manufacturing. Take a look at this essay with much more optimistic projections:

http://e-drexler.com/p/04/04/0507molManConvergent.html

Eclipse Phase (reasonably, perhaps?) stands between the two extremes of time required given in the above essay and the OP.

That's a design for building architectural things (all of which would be slotted/bolted/screwed together).

Besides which using a nanofabricator to build macrostructures being totally unecessary, inferior, and more expensive etc. EDIT: macrostructures meaning things like bridges, walls, doors etc.

Say I want something very basic like a 1mm thick bundle of carbon nanotube electrical wiring. How is this nanofabricator going to create that?

Here you can take a look at the input of actual experimental researchers who are more skeptical of easily overcoming big challenges to make things fast debating a nanofabrication apologist:

http://www.softmachines.org/PDFs/PhoenixMoriartyI.pdf

Here's an interview with Professor Moriarty on his progress in mechanosynthesis:

http://nextbigfuture.com/2011/03/philip-moriarty-discusses.html

EDIT 2:To further elucidate the progress, mechanosynthesis has not been successful yet with diamond, and very limited success has been had with silicon (liquid nitrogen temperatures for the silicon, hot tip). The diamond facing they are using has a lattice distance that is very small (0.154nm -as opposed to silicon - which is slightly more than 50%longer). It's not surprising silicon has been much much easier.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

nizkateth nizkateth's picture
Well

I guess I just assumed that since they'd figured out how to seamlessly transfer consciousness from one body to another, that they had overcome little challenges in nanofabrication. It all seemed to fit the overall level of advancement.

Reapers: Do Not Taunt Happy Fun Ball.
My watch also has a minute hand, millenium hand, and an eon hand.

nick012000 nick012000's picture
NewtonPulsifer wrote

NewtonPulsifer wrote:
Nebelwerfer41 wrote:
NewtonPulsifer wrote:
If your bonding rate is 100 nanometers per minute

Where did you find this number? Is it stated in the core book somewhere?

No, just taking a conservative number from actual atomic bonding measurements. You can get higher than that (like 6x), but it involves highish defect rates (like 1 in 1000) or where exacting structure isn't desired (amorphous solids).


That's for modern technology, with a single probe that rearranges the atoms, right? Eclipse Phase nanoassemblers don't work like that. There are no probes; there's a stupid number (thousands? millions?) of nanobots, each of which works simultaneously. You don't print line by line, you build the entire surface at the same time.

+1 r-Rep , +1 @-rep

Madwand Madwand's picture
NewtonPulsifer wrote:

NewtonPulsifer wrote:

That's a design for building architectural things (all of which would be slotted/bolted/screwed together).

Besides which using a nanofabricator to build macrostructures being totally unecessary, inferior, and more expensive etc. EDIT: macrostructures meaning things like bridges, walls, doors etc.

Say I want something very basic like a 1mm thick bundle of carbon nanotube electrical wiring. How is this nanofabricator going to create that?

A billion molecules at a time, in parallel. Read the page -- and the article it references -- again. It explicitly describes starting at the molecular level. The authors might be entirely wrong about the possibilities of the design, but we really can't know at this point, and it offers some hope.

NewtonPulsifer NewtonPulsifer's picture
nick012000 wrote

nick012000 wrote:
NewtonPulsifer wrote:
Nebelwerfer41 wrote:
NewtonPulsifer wrote:
If your bonding rate is 100 nanometers per minute

Where did you find this number? Is it stated in the core book somewhere?

No, just taking a conservative number from actual atomic bonding measurements. You can get higher than that (like 6x), but it involves highish defect rates (like 1 in 1000) or where exacting structure isn't desired (amorphous solids).


That's for modern technology, with a single probe that rearranges the atoms, right?

Actually that's from a modern understanding of physics, so its a more fundamental limit.

nick012000 wrote:

Eclipse Phase nanoassemblers don't work like that. There are no probes; there's a stupid number (thousands? millions?) of nanobots, each of which works simultaneously. You don't print line by line, you build the entire surface at the same time.
My calculation is assuming a bunch of 33x33 atom tools all packed side by side. It doesn't really matter if the tool is attached to a nanobot or not.

Please imagine what you're describing - a bunch of nanoscale tools operating on the whole atomic surface at once? Wouldn't that imply that each tool+worker is 1 atom thick? My picking of 33 atoms wide is actually smaller than any optimistic proponent thinks each tool can get, and me using the maximum bonding rate is also much faster than even they think is possible.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

NewtonPulsifer NewtonPulsifer's picture
Madwand wrote:NewtonPulsifer

Madwand wrote:
NewtonPulsifer wrote:

That's a design for building architectural things (all of which would be slotted/bolted/screwed together).

Besides which using a nanofabricator to build macrostructures being totally unecessary, inferior, and more expensive etc. EDIT: macrostructures meaning things like bridges, walls, doors etc.

Say I want something very basic like a 1mm thick bundle of carbon nanotube electrical wiring. How is this nanofabricator going to create that?

A billion molecules at a time, in parallel. Read the page -- and the article it references -- again. It explicitly describes starting at the molecular level. The authors might be entirely wrong about the possibilities of the design, but we really can't know at this point, and it offers some hope.

There's a lot of problems in the assumptions here. First, if you start with a single molecule (or even a larger clump) it won't be in solid form, complicating handling at room temperature (or even liquid nitrogen temperatures).

Second the convergent assembly webpage your first link references has things like:

"hile a wide range of possibile surfaces exist, one surface that might be worth further investigation is the diamond (110) surface. This surface does not reconstruct, and so bringing together two diamond (110) surfaces in vacuum might be an effective method of directly bonding two diamond blocks to each other."

As far as I know that just will not work, period. Graphitic bonds are the preferred bonds thermodynamically, so if you take two diamond (110) surfaces, strip off the passivating hydrogen bonds somehow, and just stick them together, you get mostly graphite bonds between the two diamonds *not* diamond bonds. The issues with the massive energy released from that bonding scenario is also glossed over (a huge deal for tiny diamonds, not a big deal for bigger scale ones as the surface area to volume ratio is vastly less). Diamond starts to graphitize at like 800 degress Celsius, so it is a big issue.

I'm not saying relatively rapid (compared to the multi-year example in the original post) bottom up nanofabrication is impossible. Its just looking like it will be:

1. Not atomically precise in an absolute sense of the entire structure (only in a relative sense between certain internal structures).
2. Using ionic and hydrogen bonds for the vast majority of it.

What you have above is basically super advanced "wet" and "bio" nanomanufacturing. The "dry" and "robot-like" is still going to require billion credit factories to have specific toolings to get around engineering issues to rapidly/profitably create things.

If you find me an example of two micron sized diamond (110) surfaces that bonded cleanly then I'm wrong.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

nerdnumber1 nerdnumber1's picture
It is definitely a non

It is definitely a non-trivial engineering problem to create fast nano-fabrication, but I'm not entirely convinced that atom-by-atom, manual construction is necessarily the optimal way to construct these special materials. You can make molecule-thick graphene with a pencil and tape or electroplate one metal with another with salt water and a battery. It is possible that nano-bots can use "short-cuts" that we don't foresee to speed up material creation (heck, there might be fabber bricks with pre-made, or partially pre-made advanced materials). I believe that making a process a few orders of magnitude faster is not enough to disqualify a setting from "hard" sci-fi when we believe the slow version is not too far off. You obviously can't have all the answers for how a science fiction setting's tech works or you'd just invent/patent the tech and it wouldn't be sci-fi.

NewtonPulsifer NewtonPulsifer's picture
nerdnumber1 wrote:It is

nerdnumber1 wrote:
It is definitely a non-trivial engineering problem to create fast nano-fabrication, but I'm not entirely convinced that atom-by-atom, manual construction is necessarily the optimal way to construct these special materials. You can make molecule-thick graphene with a pencil and tape or electroplate one metal with another with salt water and a battery. It is possible that nano-bots can use "short-cuts" that we don't foresee to speed up material creation (heck, there might be fabber bricks with pre-made, or partially pre-made advanced materials). I believe that making a process a few orders of magnitude faster is not enough to disqualify a setting from "hard" sci-fi when we believe the slow version is not too far off. You obviously can't have all the answers for how a science fiction setting's tech works or you'd just invent/patent the tech and it wouldn't be sci-fi.

This is not just a matter of an unsolved engineering problem,though. Its a physics problem.

Let me give you an example - somebody posts a detail about their new hard sci-fi game. "In my hard sci-fi game, the internal combustion engines are so much more advanced that you could take a 2013 car, replace the old engine with the super advanced future internal combustion one (same approximate size and weight) and get 10 times the miles per gallon."

If the engine you replaced runs at 25% thermal efficiency that just isn't possible to get it to 250% - that's over unity.....and it's not hard sci-fi if you just hand wave it.

Trusting that if you bond two reactive surfaces (e.g. diamond) that on average the bonds won't take the most thermodynamically favorable bond (e.g. graphitic bonds) is analogous example.

If you're not careful throwing in "a few orders of magnitude" results in over-unity problems.

Also pushing the complexity problem to "fabber bricks" made "somewhere else" pretty much relegates desktop cornucopia machines from your EP setting (which I'm totally fine with - they never made economic sense) to adding very little "value add" to any fabrication process. That is, the price of the feedstock to create an item will be very very close to the final value. That at least solves the cornucopia machine infinitely re-building itself to infinite wealth over a period of two weeks problem.

Obviously I'm not sold on the arbitrary bonding of strong covalent bonds of two "fabber bricks" - not because of current engineering limitations, but because of current physics ones.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

Wyvernjack Wyvernjack's picture
I'm glad I'm not smart enough

I'm glad I'm not smart enough to make this an issue for me.

Madwand Madwand's picture
Ultimately, I don't think

Ultimately, I don't think this forum is the right place to be trying to convince people of the arguments against fast molecular manufacturing. You might be right, and it's impractical. You might be wrong. Very few people here have the expertise to argue one way or another, and most just want to enjoy playing a game.

The right way of arguing these topics is a scholarly paper. Gather a few knowledgeable co-authors and explain your arguments. Publish them in a reputable, peer-reviewed conference or journal paper. Give experts (not us!) a chance to rebut your arguments. I would personally very much like to read such a paper, particularly a good rebuttal of the paper I cited. This is all fascinating stuff.

NewtonPulsifer NewtonPulsifer's picture
Madwand wrote:Ultimately, I

Madwand wrote:
Ultimately, I don't think this forum is the right place to be trying to convince people of the arguments against fast molecular manufacturing. You might be right, and it's impractical. You might be wrong. Very few people here have the expertise to argue one way or another, and most just want to enjoy playing a game.

The right way of arguing these topics is a scholarly paper. Gather a few knowledgeable co-authors and explain your arguments. Publish them in a reputable, peer-reviewed conference or journal paper. Give experts (not us!) a chance to rebut your arguments. I would personally very much like to read such a paper, particularly a good rebuttal of the paper I cited. This is all fascinating stuff.

That's kind of the reverse of the normal process, though, right? Normally someone will publish a paper in a reputable peer reviewed journal, then people will take shots at it.

I'd be analyzing fringe work that has never been published in any such journal. The onus is really on them to get published first by passing basic peer review, right?

If you simply want to appeal to the authority of already existing experts, the consensus so far is "yeah....not gonna work" (The Royal Society, Gordon E. Moore of "Moore's Law" fame etc.)

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

Madwand Madwand's picture
You wouldn't make an analysis

You wouldn't make an analysis of the paper I linked the primary focus of such a paper, only part of it. The main thesis of your paper could be something like "What are the fundamental limits of molecular manufacturing?" Rather than a rebuttal paper, you are asking a more fundamental question about the limits of physics when it comes to how fast it is possible to manufacture items in a "Cornucopia Machine" or future "3D printer".

I've seen papers on the fundamental limits of computing power. Maybe you can do something similar for molecular manufacturing:
http://arxiv.org/abs/quant-ph/9908043
https://www.cise.ufl.edu/research/revcomp/physlim/PhysLim-CiSE/PhysLim-CiSE-5.pdf

Find a materials science professor, or physicist somewhere with an interest in these topics and strike up a conversation. See if they want to publish a paper with you. Maybe contact the authors of the above papers to see if they can offer any insight as to how you might conduct such an analysis. Hopefully you already have some experience with publishing papers yourself, or this will be a long road. You might have to pursue (at least) a Masters degree in an appropriate field in order to get people to take you seriously. I encourage you to do so if you haven't already.

Yaginor Yaginor's picture
Madwand wrote:Ultimately, I

Madwand wrote:
Ultimately, I don't think this forum is the right place to be trying to convince people of the arguments against fast molecular manufacturing.

Just like discussions on bicycles in microgravity, I think we should cherish threads like this one, instead of trying to shoo them away. It's informative about the real physics and engineering problems behind nano-fabrication, and can be interesting to know when designing house-rules or setting hacks.
Madwand Madwand's picture
Yaginor wrote:Just like

Yaginor wrote:
Just like discussions on bicycles in microgravity, I think we should cherish threads like this one, instead of trying to shoo them away. It's informative about the real physics and engineering problems behind nano-fabrication, and can be interesting to know when designing house-rules or setting hacks.

Oh, I definitely don't want to shoo anyone away. I hope he does the research and keeps us updated with the results here. I'd really like to see them. But, I also don't think there are going to be many people here who are qualified to argue with him. If he really seeks an intellectual discussion on the physical limitations of molecular manufacturing (and it certainly seemed like it to me), he would probably gain more from a discussion with actual experts. The world in general would also gain from published results that prove something about those limits. He's already part way to a paper, why not go the rest of the way?
NewtonPulsifer NewtonPulsifer's picture
Madwand wrote:I'm sure there

Madwand wrote:
I'm sure there will be significant engineering challenges that Eclipse Phase unrealistically glosses over for the sake of exploring the consequences of a transhuman setting more fully (and the fact that no one really knows). However, reasonable people can disagree on the real difficulty of atomic manufacturing. Take a look at this essay with much more optimistic projections:

http://e-drexler.com/p/04/04/0507molManConvergent.html

Eclipse Phase (reasonably, perhaps?) stands between the two extremes of time required given in the above essay and the OP.

Okay I'll stick with high school level stuff for now. This is simply a critique of the convergent assembly thrown out there - this doesn't address at all the difficulty of arbitrarily bonding diamond cubes together.

That link you referenced has a conclusion of "Concerns that molecular manufacturing will be impractically slow seem misplaced." It references this link to support its conclusions:

http://e-drexler.com/p/04/04/0505prodScaling.html

...but overlooks power requirement scaling.

force=mass*acceleration, and power needs scale linearly with force

So lets calculate the force needed for an arm to sweep a 1 meter distance twice in one second. Well, we need to sweep a total of 2 meters in 1 second - so our average velocity needs to be 2m/sec. Assuming steady acceleration and same time to accelerate and decelerate, we need to reach a peak velocity of 4m/sec twice (4m/sec final, minus 0m/sec starting, divided by two to get average velocity of 2m/sec).

Here's a slightly harder question - what's the acceleration? Well, we need to reach peak velocity twice (two sweeps) and if you assume you slow down as fast as you accelerate, then we need to reach peak velocity (4 m/sec)in period of 0.25 seconds (twice at 0.25 and 0.75 seconds in) in the middle of one of our two sweeps. So acceleration is 16m/sec (4m/sec divided by 0.25 seconds). 16m/sec times mass of 100kg gives us 1600 newtons of force.

What if we shrink the arm down by 0.5 scale and then double the frequency (and thus average velocity). Well, we sweep a total of 1 meter in 0.5 seconds. Average velocity is 2m/sec. Peak velocity? 4m/sec. Acceleration - we need to reach peak velocity (4m/sec) in a period of 0.125 seconds. So acceleration is 32m/sec. Mass is constant at 100kg so we have 3200 newtons of force.

Uh oh. We doubled our force - that means we doubled our power needs. We need to keep it constant.

Okay, so what does the frequency need to be to keep force constant for the 0.5 scale arm? Well, it turns out you just keep the frequency at 1 no matter what the reduction in scale and force/power stays constant.

So what does this mean for throughput? Well, it also stays constant.

Well, you say, 1kg/sec is pretty awesome for a less than 2 cubic meter setup!

There's still a big problem. The 1 cubic meter long section is being fed by four 0.5 cubic meter sections (not eight). That's half the volume and mass, thus half the throughput. This problem continues down stage by stage until after 20 stages you're at approximately 1 millionth of the throughput - the smallest stage is the bottleneck.

So with that convergent assembly design in 20 stages a kilogram item will take a 1,048,576 seconds (12 days).

EDIT: 20 steps gets us starting with around 63nm cubes from a 66mm final cube. If you wish to take the steps all the way out to say a 1nm molecule that's 6 more steps (and 64 times the time - up to 768 days).

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto

nizkateth nizkateth's picture
Huh

With the psychics and cthulhu-oid cyber-monsters I didn't realize EP was trying to be really hard sci-fi. It struck me as more... relatively firm sci-fi. I've definitely seen softer.
But I was recommended EP by someone because I wanted a system/setting reminiscent of Metroid (they mentioned the gate-crashing, power armor, and superhuman augmentations).
^_^

Reapers: Do Not Taunt Happy Fun Ball.
My watch also has a minute hand, millenium hand, and an eon hand.

thezombiekat thezombiekat's picture
NewtonPulsifer wrote:

NewtonPulsifer wrote:

So lets calculate the force needed for an arm to sweep a 1 meter distance twice in one second.

Why do you believe the cornucopia machine will have such an arm at all? I suspect it comes from the fact that this is the basis upon which bubble jet printers and modern 3d printers work, that is to say every printer I and likely you have ever seen in person or on YouTube.

Cornucopia machines must work differently. I got the impression that a cornucopia machine was a box that contained a viscous semi liquid matrix comprised of several types of nanobots and standard feed stock. The constructor nanobots start building simultaneously at multiple seed points in a 3 dimensional grid, the initially separate parts are connected adding more of the substance they are comprised of into the gap (more like a perfect weld than an adhesive) or the joins are made with such atomic precision that they bond on contact (like cold pressure welding of soft mettles but better quality).

One or more types of motile nanobot would be responsible for maintaining the stability of the suspension matrix, holding partially constructed components in place moving components to their final position (a journey measured in nanometres) and delivering additional or special feed stock to locations it is needed in, as well as removing unused components of the standard suspension feed stock.

The only time something analogous to a print arm would be advantageous would be for delivering significant quantities of specialised feed stock (or feed stock used in much greater quantity than available in the suspension matrix) to the vicinity of where it is needed this head will not be doing any atomic bonding, just dumping a wide trail of raw material for the nanobots to work with. It would probably rough out the structure of the finished product with precision comparable to a modern 3d printer in the first few minutes of the production poses.

Naturally coordinating the prose I describe would require a significant amount of computer processing power but processing power is cheap in EP.

NewtonPulsifer NewtonPulsifer's picture
thezombiekat wrote

thezombiekat wrote:
NewtonPulsifer wrote:

So lets calculate the force needed for an arm to sweep a 1 meter distance twice in one second.

Why do you believe the cornucopia machine will have such an arm at all? I suspect it comes from the fact that this is the basis upon which bubble jet printers and modern 3d printers work, that is to say every printer I and likely you have ever seen in person or on YouTube.

Check out http://www.crnano.org/Merkle.pps

That powerpoint describes the smallest tool "arm" (more of a finger) yet blueprinted. Its still like 40 nanometers wide (that'd be like 250 atoms wide if using the admittedly tight diamond spacing) at its thickest point.

thezombiekat wrote:
Cornucopia machines must work differently. I got the impression that a cornucopia machine was a box that contained a viscous semi liquid matrix comprised of several types of nanobots and standard feed stock. The constructor nanobots start building simultaneously at multiple seed points in a 3 dimensional grid, the initially separate parts are connected adding more of the substance they are comprised of into the gap (more like a perfect weld than an adhesive) or the joins are made with such atomic precision that they bond on contact (like cold pressure welding of soft mettles but better quality).
If I was going to make a hollywood sci-fi movie with nanomanufacturing, and was going to "zoom in" CSI or Fast and Furious style I'd definitely hire you before the guy that came up with the cheezy graphic on the convergent assembly essay.

However what you're describing still mathematically follows the same scaling laws as the cheezy one and thus has the same problems (it requires tiny builders to hustle at very high rates of acceleration, requiring unrealistic power densities).

As an aside, cold welding metals at the nanoscale takes something like 4 seconds for the bond to form - so that would be a big limiter on designs that rely on cold welding.

Using a viscous semi-liquid isn't feasible, as atomic scale manufacturing requires *very* exact tolerances. Nanoscale devices are profoundly blind, so they do all the building by dead reckoning. You need a near vaccuum or something like a breeze or wave clear across the building cavity moving will throw off your attempted dead reckoning operation. Also, at the atomic scale brownian motion in gasses and liquids makes things inherently "windy and wavy". You also need to keep the temperature (of the manipulator tool and the manipulated matter) cool and even also for that same reason (thermal expansion) .

thezombiekat wrote:

One or more types of motile nanobot would be responsible for maintaining the stability of the suspension matrix, holding partially constructed components in place moving components to their final position (a journey measured in nanometres) and delivering additional or special feed stock to locations it is needed in, as well as removing unused components of the standard suspension feed stock.

The only time something analogous to a print arm would be advantageous would be for delivering significant quantities of specialised feed stock (or feed stock used in much greater quantity than available in the suspension matrix) to the vicinity of where it is needed this head will not be doing any atomic bonding, just dumping a wide trail of raw material for the nanobots to work with. It would probably rough out the structure of the finished product with precision comparable to a modern 3d printer in the first few minutes of the production poses.

Naturally coordinating the prose I describe would require a significant amount of computer processing power but processing power is cheap in EP.

Whether you describe it in terms of separate "boxes" of work areas or manipulator "arms" (once again more like fingers) with a variety of detachable tips or rather have them attached to mobile tiny robots I don't see how it changes the math.

"I fear all we have done is to awaken a sleeping giant and fill him with a terrible resolve."- Isoroku Yamamoto