There's this common misconception that natural systems that use diffusion transport are fundamentally more efficient than systems that do not use thermal motion in that way. In reality though it seems machine phase system will fare better than enzymatic diffusion based systems could ever do though.
I was pointed there by this blog entry of E.Drexler:
"... (note: transporting molecules by diffusion down a concentration gradient is not cost-free — it pays the same free-energy cost as any other way of driving the motion of molecules). ...
And a lecture about "Molecular Biology of the Cell"
About the energy flows in living systems (ATP, NADH, ion gradients, ...)
What is overlooked is that while the diffusion-transport (in cell biology) is free in energy expenditure one has to expend/thermalize energy packets at critical points (waiting positions?) in the chain of reactions that are significantly bigger than kBT (kB..Bolzmann constant, T..temperature in Kelvin). This is necessary to ensure that the system is irreversibly moving in the desired forward direction (often formulated as having a defined "arrow of time"). In diffusive systems with enzymes in solution reactions are mostly unconnected* in sequence (time) and pathway (space). Because of this unconnectedness the kBT energy packets** can't be shared over a greater number of reactions.
* still more than one might think
** not a quantas!!
In non diffusive stiff machine phase system one in contrast can connect multiple mechanosynthesis mills in the background e.g. via rotating shafts (space) and shift their processing cycle slightly in phase (time) this should allow sharing of the necessary energy devaluation (as mentioned before for a defined process direction) over a much greater amount of reactions.
(By combining streams suitably maybe even many of the aforementioned waiting positions can be cheated!)
Furthermore machine phase systems can be heavily cooled making a single packet of kBT much smaller. In a diffusive system cooling would just freeze the solvent preventing the whole system from working.
This sharing of a single energy devaluation step combined with some cooling should by far outweigh the low superlubricative friction which is not present in diffusive systems.
Considering E.Drexlers comment in the aforementioned blog enty he has probably figured all this out.
It does not seem he has written it down anywhere publicly though.
I think this aspect of APM systems deserves more attention so I'm bringing it up here.
Investigation is needed: (Have you guys any ideas here?)
Needed are methods for the calculation / rough estimation for at which points in space and time there still needs to be expenditure of energy sufficiently>kBT in stiff machine phase systems?
Here are some details that may help:
First there is the parameter of allowed amount of random backward run in mechanosynthesis processes (a whole nanofactory subsystem running backward "in time").
Then there are internal flexibilities in e.g. aforementioned rotating shafts in the background. These make quantum modes of
torsion oscillations (gear interfaces will complicate matters). Those modes will likely house several kBT (as phonons) at room temperature (not too hard to check). Even with those many kBT in the interconnected-background-axle-system single units sufficiently>>kBT might be shareable due to low amplitudes of added up modes in stiff systems.
(Will spacial sharing work over the whole makro-system of a nanofactory ?! That would be way more than necessary!)
Irrelevant side-note: I doubt one will get down to the ground state at any practical temperatures. Oscillation energy quanta are usually way bigger than rotation quanta but those axle systems will contain millions of atoms.
Relevant side-note: All this only makes sense if beside reversible logic also reversible actuators are designed and applied (energy recuperation). I'am not aware on any works that target and analyze these. I think this is an interesting topic to investigate too (material for another discussion). Reversible actuators should be especially easy in the molecular mills that do not change their routine cycle. But it should also be possible in general purpose assembly since the program is known.
- Also in machine phase systems - while certainly not efficient - local spots in space or time can (if necessary) truly handle motions at the speed of sound which is rather impossible in diffusive systems.
- Despite their relative inefficiency diffusive systems seem to be incredibly useful for early steps. Efficiency is meaningless if is not buildable. E. Drexler writes in the aforementioned blog post:
"... While early systems can (and likely will) operate in fluids, considerations of efficiency will favor a move to well-ordered, fluid-free systems."
ps: This is highly uncharted territory.
There is a high probability that I'm talking complete bogus here.
If you spot anything suspicious please tell.