>In article <5k7n6d$ivj@foglet.rutgers.edu>
> soreff@VNET.IBM.COM "Jeffrey Soreff" writes:
>>Please excuse the short reply here, I don't have time to comment on each
>>point in http://www.stellar.demon.co.uk/nano2.htm individually.
>>The largest mistake in the URL is to claim that
>>"Every time the arm swings around to pick a chemical and place it at the right
>>place to synthesize an exotic chemical, it spends billions of ATP energy
>>equivalents in doing mechanical work."
>>The energy dissipation calculated in Nanosystems is 100 maJ (section 13.4.1.f).
>>I don't have the figures for ATP hydrolysis immediately at hand, but since it
>>is strongly favored thermodynamically it must be at least a few kT, perhaps
>>10 maJ. An arm motion takes the equivalent of perhaps 10 ATP equivalents,
>>*NOTHING* like billions. Mill type operations can be more efficient, with
>>dissipations from 100 maJ down to 1 maJ (section 13.3.7.b). For comparison,
>>a C-C bond energy is over 500 maJ. Joe, you are claiming that the arm
>>dissipation calculation is optimistic by around 8 orders of magnitude. Where
>>do you claim this energy is dissipated? How do you calculate the amount of
>>energy dissipated?
>There are some qualifiers inherent in the arguments which does not permit
>you to put such an argument forward. When its saying "every time the arm
>swings" it means the total energy spent in the ENTIRE process of adding one
>atom / molecular sub unit to what you have already sythesised - and it
>means that in the generic sense - not a specific system.
I agree that the important energy is the energy dissipated by the whole system per unit output product. To make this a bit more concrete, assume that each degree of freedom in one of Drexler's robot arms is driven by a mechanical D/A, which in turn is driven by probe sensors from a halogenated polyethylene tape to hold the information (as you yourself suggest in another note). The arm has less than a (100 nm)^3 work envelope, so if positioning is done to 0.1 angstrom precision, we can conservatively use log2(100/0.1) roughly 10 bits per DOF, or 60 bits total per arm movement. Reversible logic technique can be used to keep dissipation for reading these bits quite low, but even if we make the conservative assumption that each bit requires a dissipation of ln(2)kT, the total energy cost for the information is still only 166 maJ.
In addition to information, the material feedstock(s) must be brought into the system. The cost for this is calculated in 13.2.2.c: 0.5 maJ for frictional losses, 2 maJ for some mixing entropy effects, and 20 maJ for the cost of extracting a pure material from a 1% solution of feedstock. Roughly speaking, if the feedstock consists of small molecules, we'll get a few atoms of product per feedstock molecule. Assume that we only get one atom of product per feedstock molecule, and that we need roughly an arm movement per atom of product.
The grand total for the system is 288 maJ per arm movement.
>One specific system is the AFM method.
>If you took the AFM method of adding a single atom, then its not
>just the energy of moving the probe but also the computer attached(!)
>Without it, there is no information flow to the assembly point.
I don't agree that describing the energy needed for a 1997 AFM to place an atom establishes a valid criticism of MNT. Generic arguments are important when they establish bounds. If you could show that _any_ placement system must dissipate billions of ATP equivalent, this would be a fatal criticism of MNT. If you exhibit one example of a placement system, you establish an upper bound on the energy dissipated, but we don't have a _lower_ bound on the energy, and the latter is the important bound in arguing against feasibility. You need to show why the projected cost for the best MNT system is unaffordably bad.
>Even in a cell, adding 'an atom' is not just about that particular reaction,
>its the entire information delivery system to the assembler as well.
>Remember, an assembler must receive its instruction stream from somewhere,
>somehow and the goods to assemble from.
>For the assembler to be useful, it can't just be limited
>to creating specific chemicals - thats called chemistry - not nanotechnology.
Roughly speaking, I agree with you here. MNT is valuable largely because of its broad programmability. It is somewhat tricky to phrase this precisely. Only a finite (though huge) number of compounds exist with any given mass. If a system can build "only" 100^(6*10^23) specific chemicals from a gram of material, it's got them all.
>Good examples are polymerization - make any length of chain, any
>amount of cross bonding, any number of side chains molecules - but
>very limited to the chemistry of that system.
Roughly speaking, agreed. Again, there are some quibbles around the edges. The ability to make polypropylene with selected steriochemistry has only limited use, but the ability to make _any_ arbitrary isomer of Cn in pure form would give you any diamond machine part, any buckytube wires, and any graphite bearing surfaces that you wanted.
[DNA snipped]
>There is a huge amount of purpose behind this scheme.
>The equivalent of the robot arm swinging around getting its information
>and attaching a sub unit is the equivalent of moving three base pairs
>extending the RNA by one unit. So information delivery to the assembler
>does not involve a huge computer and all the energy spent in that.
>Also, the total energy required is the integration of ALL the systems
>that are in place to ensure reliable synthesis.
See system comments above.
>That is where it turns from 10 ATP equivalents to perhaps millions
>of ATP equivalents. However the cell and perhaps half of ALL its
<mild irritation>
1) No, the energy cost doesn't go up that high. See calculation above.
2) Why did you switch from billions to millions? Three orders of magnitude
matter.
3) Will you _please_ justify either figure with some sort of numerical
argument?
</mild irritation>
>functionality is designed to address this weakness. Wherever
>energy is spent, it is reclaimed by some other process. Not just any
>old process - a deliberate process that is critical to ensuring that
>that the energy drain is not one way.
>>In the absence of this claim, the rest of the argument falls apart. For
>>example, the claim that "The result is an energy requirement to synthesise
>>concrete that is way beyond the energy required to make concrete from existing
>>ingredients. For this reason, bulk materials will never be synthesised using
>>nano technology methods." depends crucially on the energy estimate.
>I afraid the argument does not fall apart.
>You read the book by Stryer on the cell, and look at the
>glorious technicolour details of DNA synthesis, and jot down what
>you need to do to realise even half the story. Its just immense!
I'm not following something here. How does the complexity of DNA synthesis support the independence of the claim that MNT will never synthesize bulk materials from the claim that MNT requires billions of ATP equivalents per arm movement?
>> "However, enzymes have to be developed that co-exist with other
>>enzymes and other chemicals. In nature, this is achieved through
>>millions of years of evolution where the right chemicals have been
>>found to do the right job through natural selection pressures. Beyond
>>that, compartmentalisation is used where chemicals cannot co-exist
>>through their design. The compartmentalisation also requires various
>>molecules to transport materials through membranes separating the
>>compartments. All these operations require a huge diversity of
>>chemicals that have to be researched and perfected so that they can
>>co-exist with the previous set of chemicals.
>>Time Restrictions
>>To perfect such systems require an unreasonable amount of effort on
>>behalf of a nano technologist to search out all combinations. It
>>requires considerable effort even now to research just one chemical in
>>all its glorious working detail let alone combinations of chemicals in
>>a system."
>>You seem to be saying that all of the chemical species in a system must
>>be designed so that they can tolerate potential reactions with each other,
>>"co-exist", (except when separated into a few separate compartments).
>>The designs in Nanosystems don't permit diffusion of reactant moieties
>>(except at the system input and output ports). They are moved on controlled
>>trajectories, not diffused through liquids. They don't come within reactive
>>distances except by design. If two groups are 5 nanometers apart, then no
>>one need consider their combinations, because they can't react.
>>If you can show that mechanical transport is energetically unaffordable,
>>or otherwise unworkable, and that the only viable transport mechanism is
>>diffusion, then you would be correct to claim that all combinations must
>>be examined. Can you show any problems with the belt and roller
>>mechanical transport system described in section 10.7.6 of Nanosystems?
>We are now in different territory - and once again, there are a lot
>of things that do not permit you to put such arguments forward.
>The purpose of making materials is either they are the end products,
>or they are precursors, or they are part of a structure that has an end use.
Basically agreed.
>If you are just making end products or precursors, then on the balance of
>probabilities I call all that chemistry. There is probably more than
>a hundred ways of making something.
Huh????? This is a very different division of the subject than I have seen people use. You seem to be saying that, for example, if the end product was hemoglobin, and it was built by positionally controlled synthesis by a robot arm, this would count as chemistry rather than MNT.
>Its only when you are making molecules to order that are part
>of a structure that has an end use, then I call it nanotechnology
>and its also the place where you have to worry about mixtures of chemicals.
Certainly if the final output chemicals are part of a fluid phase where they can diffuse around, then they must not react in undesired ways. For many products this isn't a difficult criterion to meet. A lot of products are solids (eg computer chips with silicon, dopants, aluminum, and SiO2). A lot of products are unreactive or only slowly reactive (eg foods, drugs in capsules or tablets...).
>Of course, then you get into a tremendous scrap. How do you manage your
>mixtures? A conveyor belt is useless for assembling anything because its
>not directed. How do you make this conveyor belt? Thats a structure.
>How do you direct it? (i.e. move it around, line it up, set the angles?)
>How do you make the assembler, how do you make all the machinery that
>puts it together and place it in position to make a simple 1 micron
>sized gear wheel?
Conceptually, the simplest way to proceed is to build up the whole structure a layer at a time by positionally controlled reactions at the surface. This is overkill, of course (like building everything with stereolithography), but it answers all of the positioning and manipulation questions. More efficient techniques include convergent assembly.
>A Horizon program I saw which was all driven by Drexler,
>had pretty graphics and wiggly worms rotating and god forbid
>reproducing as a result of this wiggling and rotating.
>The mechanics of generic as opposed to specific implementation is where
>it all begins to go wrong.
>Its very rich to talk generics and not discuss specifics or discuss
>specifics and ignore the consquences to the global picture.
>What I want to do is discuss the global picture and sort out the
>lowest levels of implementation without skipping details or trying to
>go forward with a disconnection between the two levels.
I agree that a design where the global picture relies on something which is not successfully implemented on the local level is broken. I haven't seen you show that there is such a break in MNT.
>The stuff at my web site is all about that - though its robotics.
>What I've done is taken the uncertainties
>of atomic behaviour and created a digital atom with simplest behaviour.
>This is the robotic cube. It has properties of linking (i.e. chemistry) and
>movement on command (which is indirect in atoms - i.e. the conformation
>changes). I also made it fractal to ensure that it works at different
>scales instead of just one scale, the atomic scale because that means
>we can cheat by exploiting resolution limits.
>You have to see my original description of the whole concept which is
>a chemists wish list (I had to change it unfortunately because
>at the time, nobody had a clue what I talking about! :-( )
>The robotic cubes were called monomers, they linked to make polymers on
>command. The whole material was called programmable materials because you
>can program its shape and with sensor fitted you can also program
>it properties. For example, you could make this material react against
>a pushing force at twice that force. These properties
>are difficult to implement with ordinary chemicals but not for
>a programmable material.
>I don't doubt it can all be done eventually - but that word eventually
>is what I hate most. Programmable materials was meant to be a stop gap
>before all that happened - but unexpectedly a lot more came from it
>than originally planned including the mechatronics assembler
>http://www.stellar.demon.co.uk/universe.htm which I now envision as the
>only viable commercially relevant assembler concept.
>I guess time and the market place will tell who is going to be in charge :-)
You have an interesting web page. It looks like you have a combinational chemistry system, with your small test tubes playing the roles somewhat between the resin beads and wells in well-plates in conventional combinational chemistry. You seem to be quite concerned with avoiding mixing micron sized quantities of reagents. In conventional combinational chemistry, resin bead techniques have no problems with partially synthesized peptides diffusing from one few-micron-sized bead to the next. Similarly, diffusion from one well to the next in a well plate isn't normally a problem. You suggest that "Since there is no mixing, we can use highly reactive direct chemical assembly paths." Actually, the problem with highly reactive reagents is typically the side reactions that one gets because the orientation of the molecules is random on the nanometer scale rather than the micron scale. For example, if you react nitric and sulfuric acids (fairly reactive reagents) with chlorobenzene, the products include about a 30:70 mix of the ortho and para isomers of nitrochlorobenzene. If you run a few more reactions with poor selectivity on this mixture, you can get a mixture where no single compound is present in large amounts. These mixtures do not give clear signatures to many assays, so they are not desirable in optimization techniques.
-Jeffrey Soreff
standard disclaimer: I do not speak for my employer.