Two of the most important materials underpinning our industrial society are silicon and steel. Without silicon, the material from which microprocessors and memory chips are made, there would be no cheap computers, and telecommunications would be hugely less powerful and more expensive. Steel is at the heart of most building and civil engineering, making possible both cars and trucks and the roads they run on. So I was struck, while reading Vaclav Smil’s latest book, Transforming the Twentieth Century
In the year 2000, around 846 million tonnes of steel was produced in the world, dwarfing the 20,000 tonne production of pure silicon. In terms of value, the comparison is a little closer – at around $600 a tonne, the annual production of steel was worth $500 billion, compared to the $1 billion value of silicon. Smil quotes a couple of other statistical nuggets, which may have some valuable lessons for us when we’re considering the possible economic impacts of nanotechnology.
Steel, of course, has been around a long time as a material, but it’s easy to overlook how significant technological progress in steel-making has been. In 1920, it took the equivalent of 3 hours of labour to make 1 tonne of steel, but by 1999, this figure had fallen to about 11 seconds – a one thousand-fold increase in labour productivity. When people suggest that advanced nanotechnologies may cause social dislocation, by throwing workers in manufacturing and primary industries out of work, they’re fighting yesterday’s battle – this change has already happened.
As for silicon, what’s remarkable about it is how costly it is given the fact that it’s made from sand. One can trace the addition of value through the production chain. Pure quartz costs around 1.7 cents a kilogram; after reduction to metalurgical grade silicon the value has risen to $1.10 a kilo. This is transformed into trichlorosilane, at $3 a kilo, and then after many purification processes one has pure polycrystalline silicon at around $50 a kilo. Single crystal silicon is then grown from this, leading to monocrystalline silicon rod worth more than $500 a kilo, which is then cut up into wafers. One of the predictions one sometimes hears about advanced nanotechnology is that it will be particularly economically disruptive, because it will allow anything to be made from abundant and cheap elements like carbon. But this example shows the extent to which the value of products doesn’t necessarily reflect the cost of the raw ingredients at all. In fact, in cases like this, involving complicated transformations carried out with high-tech equipment, it’s the capital cost of the plant that is most important in determining the cost of the product.
Capital costs in these types of situations are mostly Manpower, Electricity and Initial Construction Costs correct? If Nanotechnology can reduce the need for the first two of these in the end, wouldn’t the cost of the end product come down in price regardless of how “hi tech” the processing is?
One thing that gets me always, is the ‘novelty’ of an items worth. Usually when you stick ‘nano’ on something, it suddenally requires respect, which raises the price. For instance, pardon if I may use a Drexler vision so lightly, but to say that a molecular assembler capable of creating steel or pure silicon molecule by molecule will in effect cost nothing more than the resources used to construct it, I must worry about the cost of the novelty and ingenuity within the design of said molecular assembler.
Will a molecularly(forgive me) assembled molecular assembler cost nothing more than the resources used to construct it, or will there be an added novelty cost to it? Or even further, the intellectual property created by humans to design the molecular assembler? I don’t mean to go off on a tangent, and I don’t mean to be stereotypical as to what the nanotechnology “dream” or vision is. 🙂
Also I just finished reading Soft Machines, I really enjoy what you had to say, and it was really easy to grasp, though I may say pictures are worth a thousand words, alot of your descriptions of how proteins assemble, or how brownian motion works would have been simpler to describe by pictures. Not to say I just read books for the pictures! 🙂
– Matt Griffith
I must say that perhaps the price of said molecular assembler would in effect act as a safety from uncontrolled building of products or perhaps worse, viruses and/or plagues. 🙂
This is the resource I use for materials production/price figures:
http://minerals.usgs.gov/minerals/pubs/mcs/2006/mcs2006.pdf
The point Drexler makes is that MNT would be able to grow the manufacturing infrastructure itself. But this assumes all manufacturing processees (slicon refining, etc) are viable MNT products. That is the speculative part.
The capital cost of high tech facilities I’m sure has many factors, but a big one must be the amortization of the R&D costs (which itself has a large manpower component). Assuming for the moment that MNT will work, it’s going to need a big R&D effort whose cost will be reflected in the price of the manufacturing infrastructure it produces. It’s difficult to imagine that the first product would be so perfect that further R&D wouldn’t be required, meaning that these costs will be ongoing. One could argue that nanotechnology would itself reduce the manpower cost of doing research. This argument has been made many times before, in connection with developments like computer simulation and rapid-throughput experimentation, but the expected savings never seem to materialise.
Ok, but lets say we all get a black box that can make just about anything we desire from a basic feedstock that is universally accepted in the industry.
What’s going to stop us from just pirating the “software”? Once something gets released, it will become essentially worthless. How would they cover costs then? Will they start to offer pricey services for complicated jobs while giving their basic code away for free, while keeping all the specialized stuff under lock and key?
I’m getting a headache trying to reconcile Market Economics with MNT. I’m starting to think that the two are incompatible in this day and age.
I wish to bring to everyone’s attentions this website:
http://www.math.temple.edu/~wds/homepage/
In particular:
http://www.math.temple.edu/~wds/homepage/chemcomp.ps
In this paper, the computational complexity of synthetic chemistry is #NP!
This does not mean that everything is hopeless, but it means that the higher limit given todays algorithms and computers is of the order of tens of thousands. The internet does not really change anything due to exponential explosion.
Note, we can SIMULATE molecules with millions of atoms, but a simulation is not a SYNTHESIS pathway!
This is the FUNDEMENTAL reason why REAL scientists, when they want to make large nonlinear molecules utilise biology!
signing out Zelah
Yes Richard, I didn’t mean to imply product development gets cheap until MNT is in a mature form. Assuming a minimal library of deposition reactions have already been modelled and then (somehow) experimentally verified you might need anywhere from a billion to a trillion dollars to get one protoassembler.
Globalization combined with the diffusion of correspondence courses might bring down some manpower costs a bit. But yeah, it doesn’t look like capital equipment costs are trending downwards.
Hi,
I tried leaving a message before under my old alias, but
I wish to bring to everyones attention this file:
http://math.temple.edu/~wds/homepage/chemcomp.ps
I believe that this makes UNIVERSAL molecular manufactoring highly unlikely due to complexity issues!
Zelah
Zelahx: why unlikely? Hypothetic “universal” molecular manufactorer should be able to do just enough basic synthetic reactions. “Complete” one should do all (that are to be included in the search space). Useful one can live with handful reactions.
Phillip, Zelah, once again my apologies that your comments got caught in my spam filter (Phillip in particular, its very puzzling what in your comment it took exception to; Zelah’s fell foul of a rule about two or more links). If this happens again please just submit a very short comment with no links saying what’s happened and I’ll fish it out of the moderation queue. The spamming rate has gone up again, so these two comments were a bit swamped amidst 187 spam comments advertising unsavoury things.
Hi Kharvryuchenko,
Okay, I will interpretate your question as saying that one should be able to replicate conventional chemistry on a chip!
I have no problem with this. Maybe a little.
However, nobody would get all that excited about Nanotech if it was just conventional chemistry scaled onto a chip.
The real point I believe is that novel chemistry will occur with millions of atoms being precisely moved about! Even for Monochemistry (i.e Carbon Nanotubes) this is highly complex. We can only do chemistry at this level stochastically.
Thanks Richard for the explaination regarding Spam!
Signing out Zelah
Zelah,
Yes, my question may be interpreted like this.
You say:
“However, nobody would get all that excited about Nanotech if it was just conventional chemistry scaled onto a chip.
The real point I believe is that novel chemistry will occur with millions of atoms being precisely moved about!”
Sorry, but I don’t see a major (comparable to getting down to atomic precise chemical reactions control) issue here. Staying within the viewpoint of the article, you’ve mentioned, to achieve a million operations computation (millions atom reaction) we have to be able to control reliably just each of the computation necessary (each of the necessary reaction).
NP-complete? Hey, that’s *easy* — predicting the performance or properties of arbitrary computer programs is provably *impossible*.
Yet we manage, because we don’t generate random computer programs and try to analyze them, but write programs out of analyzable building blocks. I think Drexler talked about this back in Engines of Creation: instead of predicting how arbitrary proteins fold, you’d try to find protein subunits whose folding you can predict and build with those.