Nanotechnology Timeline

Based on our research, CRN believes that general-purpose molecular manufacturing (MM) could be (not will be, but could be) developed successfully within the next ten years, and almost certainly will be developed within twenty years at most.

Why is our timeline more aggressive than most? In part, because the incentive for development is so great.

Let's look at what's required: Maybe a hundred or so mechanochemical reactions to build the parts; some basic robotics and structural design for the fabricators and the nanofactory; a really advanced CAD program and training to design nanoscale machinery; and a nano-lithography or nano-assembly system that can build the first crude fabricator. All of this is engineering, with no need for unpredictable scientific breakthroughs.

Many of these capabilities are being developed rapidly in other nanotechnologies. Some costs are decreasing exponentially every year or two, like computers to do simulations. We don't know whether it would take a billion, ten billion, or a hundred billion dollars to do it by 2010, but almost certainly by 2020 it will be less than a billion dollars. And general-purpose molecular manufacturing, even in 2020, would be worth hundreds of billions of dollars, maybe trillions. Someone somewhere will find a way to fund it.

It appears quite possible that MM will arrive suddenly, perhaps within the next ten years, and almost certainly within the next twenty. If it takes the world by surprise, we will not have systems in place that can deal with it effectively. No single organization or mindset can create a full and appropriate policy -- and inappropriate policy will only make things worse. A combination of separate policy efforts will get in each other's way, and the risks will slip through the cracks.

By the time this technological capability arrives, we must have accomplished several things that each will take significant time. First, we must understand the risks. Second, make policy. Third, design institutions. Fourth, create the institutions -- at all levels including international levels, where things move slowly. This could easily take twenty years. If advanced nanotechnology could arrive in ten or fifteen years, then we'd better get to work.

Comments

[jim moore]

"Let's look at what's required: Maybe a hundred or so mechanochemical reactions to build the parts"

I thought that it was 6-10 tool tips was all that was needed to build the diamondiod parts. Why the ~10 fold increase in the number of reactions?

"All of this is engineering, with no need for unpredictable scientific breakthroughs"

I am not sure that I agree with that statement, Mechanochemistry may not require a fundamental theoretical breakthrough in chemistry but we are still in the situation were no one has demonstrated in the lab the atomically precise mechanochemical addition of even one carbon atom to diamond surface. Even if you just look at computer simulations we still have a lot of work to due. Fretais' dimer depostion tool only works 20% of the time. (there may be a fundamental trade off between the stability of the tool tip and its reliability in depositing carbon atoms.) I think that a good deal of science is still needed.

[Karl Gallagher]

Not that I disagree with your basic message, but the subtext of "engineering = easy" annoys me. Neither landing a man on the moon nor building a space station required a scientific breakthrough, but they weren't easy.

[SonofEris]

I was under the impression that we have nowhere near enough computer processing power to do molecular simulations of sufficient complexity. I've heard that it may require quantum computing. If so, it obviously still requires a "breakthrough".

Please correct me someone if I am wrong.

[Chris Phoenix, CRN]

Jim, better tool tip designs are being developed as we speak. There's no reason to think there's a stability/reliability tradeoff.

The increase in number of reactions is just because I wanted to be more conservative.

Karl, it's not that engineering is easy. It's that engineering can be purchased and scheduled.

SonofEris, there are some simulations we can do and others we can't. We can't come close to simulating a whole machine at quantum levels. But we can simulate a few hundred covalent-solid atoms ("tool tip" and substrate) well enough to tell us that a mechanosynthetic reaction will probably work the same in real life. At least, Freitas's work passed peer review.

Chris

[Mike Treder, CRN]

Note, once again, that we are not saying MM will be developed in the the next ten years, but that it conceivably could be. It would require a large, focused project, with plenty of money, brains, and patience behind it. This may not be likely to happen today, because at this point, only big governments can afford that kind of project, and most of them have other pressing priorities. But as the payoff for MM becomes increasingly clear, and as the costs for development continue to drop, someone somewhere will find a way to fund it. And we'd better be prepared for the consequences.

[Michael Vassar]

I agree that engineering can be purchased and scheduled to a point, but I think that where that point is is more or less an unknown. Look at all of the military, industrial, and aerospace engineering projects that get purchased and scheduled, and re-scheduled, and go past 3x the initial budget, and get re-scheduled, and on and on in this manner until the final produce, the space-shuttle or space-station for instance, comes in way over budget, years late, and with performance specifications so far short of the original goal that it fails to deliver essentially any of its promise. I know that Chris's standard answer to this observation is that the capabilities of diamondoid MNT are so great that we can throw away orders of magnitude of performance with respect to almost any measurable specification and the product will still be revolutionary, but it's not clear to me that this is a sufficient answer. Loss of an order of magnitude or two in speed relative to Drexler's proposals would not change the basic conclusions, but a reduction in precision, flexibility, etc, very well might.
I bet that a sufficiently well-run and sufficiently well funded engineering project could with >75% certainty produce MNT within 10 years, but currently there don't seem to be any large projects to develop MNT, the fraction of large projects of this sort that are extremely well run seems to be substantially lower than 20%, so even if several of the largest governments, the nuclear club for instance, all started MNT manhattan projects today I would have little confidence in any of them pulling it off within 10 years. On the other hand, so little effort has been put into MNT development so far that I don't think we are in a position to state with any great confidence that a team of 50 or fewer researchers couldn't pull MNT off for <$50,000,000 and in less than 5 years.

[Tom Craver]

I think it might be useful to try taking a better swag at how much MM development will cost. E.g. for a range of 10 to 100 mechanosynthesis reactions to be developed, with on average 1 primary researcher with 3 grad students on a grant of X dollars over a period of Y months. Then assume a number of parallel efforts to apply different combinations of multiple reactions in sequence, starting with just 2, then 3, then 5, etc - how long and how much investment for that?

Suppose it's 16 reactions, each one taking 2 years to develop, maybe 8 teams - call it 5 years total. And four funded efforts that each take a year to develop multi-reaction test systems to apply 2, then 3, then 5, then 9, finally 16 reactions - 5 years more. Maybe $500K/yr grants for each team, in a university setting - $30M over 10 years.

In parallel, how much effort to make a lab-production unit for automating the application of reactions and testing to see if each step "took", to build up 2D patterns of atoms. Then an effort to extend that to building up layers of atoms in patterns and dealing with problems like leaving holes and later building over them, and building loose parts (such as a rod loose to slide).

Figure this effort takes three principle researchers/engineers with 10 assistants and a lot of equipment - call it $20M up front plus $5M/yr. Figure getting the first unit starts 2 years after the above "reaction" developers, and takes 3 years to get working reasonably reliably. Figure that system is used by the "reaction" teams for their multiple reaction work, while this team goes on to develop greater reliability and 3D layering (another 3 years) and develops techniques to create useful 3D structures (another 3 years). About 9 years, or $65M to produce the first crude one-atom-at-a-time assembler.

Assume that design and construction of a nanotech assembler doesn't start until the above effort is done (due to lack of certainty as to what capabilities the lab-scale system will have), and is done in parallel with upgrading the lab-scale assembler to work with all the reactions needed. Figure this is a difficult task, but after 4 years more, produces the first nanoscale assembler capable of producing a copy of itself. Another year to refine that and expand it to arrays working in parallel with enough capability and capacity for MM to be considered "real" and ready to start emerging from the lab. Assume this effort took 20 highly specialized engineers/scientists, funded at $5M/yr - $20M.

So net cost of $115M over 16 years to the point of MM emerging "for real".

OK - I'm NOT saying "And that's how it'll be done" - I'm saying that something like the above could be done better than this quick hack at it. There could be a "crash program" scenario, a "slow but steady" scenario (a bit like the above), and maybe a "random advances until critical mass is reached, followed by multiple competitive crash programs" scenario. For each, have a range of assumptions - what if it's only 8 reactions, or if it's 100? More difficult? Less via some clever hack? What factors might make the effort cheaper if done slowly over time, more expensive if rushed?

[Mike Treder, CRN]

Good comments, Michael and Tom! I'd like to ask you to post them as appropriate over at Wise-Nano too.

[Karl Gallagher]

Karl, it's not that engineering is easy. It's that engineering can be purchased and scheduled.

BWAHAHAHAHAHAHA!

Chris, you clearly have not been paying attention.

You can make a decent cost and schedule estimate for a version 2.0 product. Want another car, another PC, another jetliner? Yeah, your budget and schedule will be reasonably accurate. And you'll get a regular state-of-art, or slightly behind the art, product.

Now if you're making something new, that's R&D. And for R&D it's target performance, deadline, budget, pick two. Apollo spent money as fast as it could and wasted a good chunk just get hardware in the air as fast as they could. Give a bunch of smart guys some time and money and they'll produce something, but there's no way to tell if it'll be what you want.

There's a very simple reason for this--when you're doing something that's never been done before, you can't tell how hard it'll be until you actually do it. And lots of people have tackled something that looks doable and found that it's not, at least with the resources they have available.

So, no, you cannot purchase and schedule the engineering to produce an MNT device. Not until you've built some. Feel free to talk about simulations, but I make my living in an industry that's been burned by lots of them. Sims won't convince me--or anyone with money.

[Tom Craver]

Karl -

So it would be worth asking which components of the development effort are to be truly novel, and which are "2.0" as you put it, or "incremental" as I'll call it (and I mean incremental given likely precursor technologies, not necessarily incremental by today's standards).

My guesses as to some possible examples of each:

*Figuring out a range of reactions, some subset of which might be sufficient to bootstrap MM using a *modified AFM - novel.
*Developing the ability to do those reactions reliably, repeatably, precisely, at an AFM tip - incremental.
*Developing an AFM based lab assembler to do precisely positioned reactions - incremental.
*Developing the ability to do patterned reactions to create 2D patterns of atoms - incremental?
*Going from 2D to 3D patterning of atoms - novel?
*Devising techniques to create a variety of 3D structures - incremental.
*Designing a nanoscale assembler based on capabilities of what 3D structures you can build up an atom at a time with a lab-scale assembler - novel.
*Building the first nanoscale assembler - incremental.
*Improving nanoscale assemblers to the point of practicality - incremental.

Such a list would be useful in identifying the hard tasks that need the best minds applied to them, and which might go over-budget as unexpected difficulties arise.

It also gives an idea of the stages that development might go through, with novel technologies that might be profitable to develop even if they weren't on the path to MM.

[Tom Craver]

Chris - I'll try to get something on cost/time estimation up on WiseNano - suggestions regarding where to put it?

I'm thinking to organize it with a general header page, and scenario pages linked from brief descriptions of the scenario.

[Karl Gallagher]

Taking one example:

*Designing a nanoscale assembler based on capabilities of what 3D structures you can build up an atom at a time with a lab-scale assembler - novel.
*Building the first nanoscale assembler - incremental.

Those aren't separate steps. You'll come up with a design, build it, discover something didn't match the theory, analyze the implications, do a new design, repeat until there are no more discoveries. Lots of R&D produces a long series of prototypes before getting to a usable design. Going from a design to a hardware implemenation of it can be incremental, but since you'll do that step an unknown number of times it's still pretty hard to budget.

[Tom Craver]

Kurt -

I don't think the fact that there might be iterative steps in the novel process of designing and testing a nano-sized assembler, including building prototypes, changes my meaning.

As an analogy: if you want to write innovative software, you need a compiler that you can trust to simply work most of the time, or your project is doomed. Writing your new code - with all the iterations you do - novel. Compiling a particular instance of your code - incremental.

In the case of making a nano-scale assembler, you'd better have worked out some design tools and tested them on other nano-scale devices before you even attempt to build your first nano-scale assembler design.

[Chris Phoenix, CRN]

Karl-
"And for R&D it's target performance, deadline, budget, pick two."

I think we agree. The crash-project scenario assumes budget goes out the window. If I get to pick target performance and deadline, that's enough.

If performance and budget are picked, I really like Tom's analysis approach. I've tried to guesstimate the numbers for a laid-back development program, and came up with something on the order of $100 million. That's based on a few data points, people in labs working on projects that seemed to be measureable fractions of the total effort required.

Tom, I just looked through the Wise-Nano site map, and couldn't see a good place to put your cost estimate work. Oops! I've been regretting that site map since about a week after I wrote it. Too large and too inflexible. Might be worth starting a new top-level category called "development." Or something like that. It's your work; do what you think best. If you want more suggestions than that, email me; I don't always have time to read this blog anymore.

Chris

[jim moore]

Chris and Mike,

( this comment is in the spirit of constructive criticism, it is not an attack)

I think you guys have misnamed your organization. The focus you have is not on the Responsible Development and Usage of Nanotechnology ( with nanotechnology seen in its broadest terms) but rather on the Responsible Development and Usage of General Purpose Exponential Manufacturing.
If you change from CRN to CREM the added precision may avoid confusion about the risks and benefits of nanotechnology in general versus the risks and benefits of exponential manufacturing.

Again let me thank you two for creativity and hard work you put in on this topic.

jim

[jim moore]

Oh I just thought of how you could keep the CRN call letters, just change nanotechnology to nanofactory. The Center for Responsible Nanofactories.

[Chris Phoenix, CRN]

Jim, to a lot of people, nanotechnology = nanobots. To those people, our name is accurate, and any more technical name would be confusing. And those people are our intended audience.

We have been talking to nanoscale technologists; but if not for them first grabbing the word and concept of "nanotechnology" and then badmouthing MNT, we wouldn't be talking to them at all because what they're doing has very little relation to MNT.

We did discuss lots of names like "Center for Responsible Molecular Nanotechnology" but they were just too unwieldy. Finally I said, "Why not just Center for Responsible Nanotechnology?" and we knew immediately that that was the right name.

Chris