March 20, 2007

Exponential production and nanofactories

Exponential manufacturing is one of the key features that magnifies the impact of molecular nanotechnology Exponential manufacturing is a nanofactory being able to produce another nanofactory. Nanofactories with this capability can double their number in the time it takes to make another copy.

When would exponential manufacturing dominate over projected conventional production ?

A nanofactory could produce the equivalent of a wafer full of chips which is equivalent to the output of a multi-billion semiconductor fabrication factory. How many nanofactories would be needed to equal the roughly 50,000 wafers produced each month by a modern fab?
About 100 wafers per hour from the fab. So 100 personal nanofactories based on some rough estimates of projected performance. If it took 1 day to make another nanofactory then it would take 8 days to have nanofactories that equalled the production of the fab. If the production performance of the nanofactories were 10 times less, then it would take 1000 PN to equal the production rate and it would take 110 days to make the 1000 nanofactories at the 10 day per doubling rate.
If doubling performance of the nanofactory stayed higher than 100 days per doubling then the impact would not be as shocking and their would be time to adapt.

The molecularly precise nanofactory would produce higher quality and performance products.


al fin said...

Can you tell me if you have worked out any approximate timeline for development of molecular fabs? First would come special purpose molecular fabs which would make only one product by molecular assembly. Over time, the fabs would become more and more general purpose--able to make more types and variations of devices with appropriate programming and raw material.

A truly generic molecular assembler which could make copies of itself, as well as many other products, would seem intuitively to be much farther away, in time.

Do you agree? To me the difference in difficulty in making special purpose vs. general purpose molecular assemblers is profound.

bw said...

I think that there will be molecular fab variants and some cost differentials. The cost differentials could persist for static molecular products versus ones with actuators. Just as the black and white versus color printing cost difference persists.

DNA nanotech/synthetic biology is here first now. DNA synthesis is still relatively low volume.

A graphene fabber (1 atom thick) could be a special purpose fab that makes powerful 2D electronics and computing devices. That could arrive before the more classically envisioned diamondoid nanofactory.

Most diamondoid materials used for nanomachinery would be constructed from the atoms of 12 elements in the Periodic Table: carbon (C), silicon (Si) or germanium (Ge) in Group IV, nitrogen (N) or phosphorus (P) in Group V, oxygen (O) or sulfur (S) in Group VI, fluorine (F) or chlorine (Cl) in Group VII, boron (B) or aluminum (Al) in Group III, and, of course, hydrogen (H).

With the 1 nanometer resolution optical microscope project (target 5 years), possible significant scale adiabatic quantum computers next year (1024 qubits Dwave systems) and maybe millions of qubits within 10 years, million artificial neuron systems now and billion neurons within 10 years, petaflops now and exaflops within 10 years, all of the control and tools we are getting for manipulating 1-30 nanomaters and possible breakthroughs with diamond mechanosynthesis (Freitas, Merkle et al), DNA nanotech now, UK ideas factory projects, directed self assembly, rotaxanes etc...

I do not see how it would take longer than 2020 for molecular fabs .

In 2020, you would have
billions of artificial neuron AI and automation systems.
Millions if not billions of qubit quantum simulators and computers.
exaflop machines, the first experimental diamond dimer placements made over 10 years before, synthesizing gram or kilogram quantities of DNA and polymers... molecular fabs should be done as well.

I am leaning towards thinking cruder but very useful systems by 2016 and then moving up the refinement cycle at a quickening pace. Versions 2 and 3 in 2017. Version 4-8 in 2018. Version 9-20 in 2019.

extended DNA nano, with bubble labs on a chip could get pretty interesting and powerful in the 2008-2015 timeframe

bw said...

Improvements from basic to advanced nanofactories could be quite short. Cost of raw materials could stay differentiated for a while.

Duplication estimates for assemblers and nanofactories (primitive to advanced)

Say you have a 10 kg nanofactory invented in an arbitrary country on January 1st, 2020. Let's say that the design is similar to the Phoenix nanofactory, in which case we'll work with the following assumptions:
The size, mass, energy requirement, and duplication time of this nanofactory design depend heavily on the properties of the fabricator. ... a tabletop nanofactory (1x1x1/2 meters) might weigh 10 kg or less, produce 4 kg of diamondoid (~10.5 cm cube) in 3 hours, and require as little as fifteen hours to produce a duplicate nanofactory.
Say that this first nanofactory is used to make a duplicate nanofactory, then both nanofactories are used to make duplicates, and so on, until you have 200 million units, ready for distribution to the majority of households in the nation. How long would this take? Under 28 duplication cycles, or approximately 18 days.
In his 1992 technical analysis of this factory approach to molecular manufacturing systems [208], Drexler outlines an architecture for a system capable of manufacturing macroscopic product objects of mass ~1 kg and ~20 cm dimensions in a cycle time of ~1 hour, starting from a feedstock solution consisting of small organic molecules. The feedstock molecules enter the system through a molecular sorting and orientation mechanism (Figure 4.34), pass through several stages of convergent assembly using mill-style mechanisms (Figure 4.35), and then pass through several more stages of convergent assembly using manipulator-style mechanisms (Figure 4.29). The full system has 10 stages with progressively larger machines assembling progressively larger components at progressively lower frequencies (Table 4.1). If the manufacturing system can manufacture all of the components of which it is itself composed, Drexler’s proposed desktop manufacturing system (system mass ~1 kg) would also be capable of self-replication in about 1 hour.

Finally, the entire factory is enclosed in a suitable casing with a mechanism to output final product without contaminating the workspace. In the highest level nanofactory layout, the overall nanofactory shape is a rectangular volume (Figure 4.59). The exterior shell consists of six flat panels, with each panel: (1) providing support to anchor the interior and prevent the working volume from collapsing under atmospheric pressure, and (2) supporting each other. Panels are hollow and pressurized, held rigid and flat using internal tension members set at a slight angle. The design is easily scalable to tabletop size, with a ~1 meter factory producing eight ~5 cm blocks per product cycle. A tabletop nanofactory measuring 1 meter x 1 meter x 0.5 meters might weigh 10 kg or less (without coolant), produce 4 kg of diamondoid (~10 cm cube) in 3 hours, and could require as little as twelve hours to produce a duplicate nanofactory