Eliezer S. Yudkowsky, <sentience@pobox.com>, writes:
> Mirkin and his team developed a solution to this problem that has not
> been matched by any currently available nanofabrication method. It
> required a straightforward but significant modification of the first
> experiment. Using DPN, they first drew cross-hair larger scale alignment
> marks with an MHA ink on either side of the area of gold to be patterned
> and imaged. Next, three parallel lines using the same MHA ink were set
> down at a precalculated position with respect to the alignment marks.
> The AFM tip was then replaced with a tip coated with ODT. The tip
> located the alignment marks and then, using precalculated coordinates
> based on the marks, drew three 50 nanometer ODT lines, each one exactly
> 70 nanometers to the left of an MHA line from the initial pattern. At
> that point, the entire area was imaged with an ink-free tip.
I wonder how they change out the tips? Doing that while keeping the system stable enough to find the drawing area again seems like the hard part. If you can change tips which have a different molecular coating on each tip, and find your workspace again reliably, this sounds like a good start at being able to build large molecular structures.
In his book Nanosystems, Drexler lays out a moderately detailed path to nanotech. This has not received much comment that I've seen, perhaps because it is the last chapter of a difficult and technical book. He describes a relatively simple and limited assembler as a starting point, which he calls "stage 2". Stage 1 is then the construction of that proto-assembler.
He envisions the stage 2 device as requiring approximately 400,000 atoms, built from molecules of about 20 atoms each, suggesting about 20,000 steps to build it. These would either be done by chemically synthesizing molecular building blocks with complementary binding sites which would then self-assemble into the desired positions, or else by using scanning microscopy.
With the technology above, if changing out the tips is a mechanical step that takes minutes or hours, 20,000 synthetic operations would not be practical. But if it can be automated to be done in even a few seconds, building structures of the complexity Drexler requires would become conceivable.
You'd still have to do a lot of chemistry to figure out how to actually create large devices given the limited synthetic abilities presented by the scanning microscope tip. No one seems to have made any progress on 3-dimensional assembly yet. But this does sound like a significant step in the required direction.
Hal