🚀 go-pugleaf

Tim May has described nanolithography that uses linear arrays of 
1e4-1e5 AFM's that would scan a chip and fill in detail to 10 nm 
resolution or better.  The AFM tips could either expose a resist, directly 
etch or directly deposit material.  Nanolithography seems to me a major 
breakthrough in the road towards Drexler's vision of nanotechnology.  What 
follows includes further nanolith techniques that I have brainstormed.

A nearer-term use of AFMs or STMs that requires only writing and
reading a few atoms is to create an atomic-scale vernier that would be 
used for alignment of more traditional lithographic pattern transfer 
techniques such as ions, e-beams, x-rays, etc.  By using STM alignment 
and decreasing the distance from the radiation source to the wafer surface, 
many traditional techniques still have room for improvement to below 100 nm.

A field of great interest to nanotech is protein engineering: Brownian 
assembled proteins, structures made out of DNA, self-replicating systems 
vastly different from the natural DNA/RNA/ribosome system, etc.   Despite 
the modelling difficulties, the shear momentum in protein engineering is 
such that we may see many complex and useful protein and other organic
systems emerging during the next decade.

I suggest a class of such systems which I call _nanoresists_.
These would be proteins and other molecules that, upon specific
stimulation, assemble into geometric two-dimensional shapes
such as line segments, circles, donuts, squares, hexagons, meshes, etc.   
A primitive nanoresist would form this shape based on a simple Brownian 
assembly process upon being impinged with radiation.  Note that the 
radiation delivers only a process-start signal; the pattern emerges from 
the protein or polymer self-assembly.  Alternatively, prepatterned
nanoresist films can be placed on the wafer by STMs or molecular-tip
AFMs. More advanced nanoresists might include the following capabilities:

* Shape formed from a code based on radiation wavelength and
  start/stop pulses; it might be formed from rhodopsin or other dye 
  molecules that transmit excitons to a protein-assembling process.

* Shape formed by a short DNA code, which has been created by 
  recombinant techniques and mass-produced by the polymerase chain 
  reaction (PCR).  The DNA would form themselves into specific
  shapes, rather than going through the DNA-mRNA-ribosome-protein
  translation process.

* Specific 2D patterns produced by actual organisms distributed in
  a controlled manner on the substrate.  Some existing creatures can form 
  sophisticated hexagonal, spiral, and capillary structures; via genetic 
  engineering we might be able to make an even wider variety of patterns. 

In each case the finished resist byproducts are etched away, leaving
the skeleton which is the actual 2D resist pattern.   The resist can
be either positive or negative; that is it can serve to protect
the substrate from further deposition or from etching.  The etching or
deposition process becomes the limiting factor, since nanoresists
themselves might achieive resolutions below 1 nm.

Small-scale lithography not only improves the feature density of
existing devices but also makes possible a wide variety of new 
devices that take advantage of quantum effects: glowing nanopore 
silicon, quantum dots and wells, tunneling magnets, squeezed
lasers, etc.

Good yields are now being achieved in the deposition of diamond film, by 
seeding a silicon surface with diamond seeds and then growing the diamond
crystals epitaxially on those seeds from hot hydrogen and methane.
The diamond can also be etched by nitrogen dioxide, and does not suffer 
the crystal damage that often results from etching with fast atoms.  Thus, 
diamond etching has a better prospect of matching the resolution of 
nanoresists than traditional silicon etching with oxygen.

Diamond lithography may soon havae the following capabilities:

* Produce diamond-based ICs, vacuum tubes, and flat-screen
  display elements.
* Produce thousands of AFM cantilevers, tips, and control systems 
  on one wafer.  These AFM's can then be used to lithograph 
  further wafers, as described by May.  
* Produce many diamond mechanical parts similar to those described 
  in Drexler's _Nanosystems_, but via lithography instead of
  Drexler's molecular-machine assembly.

The latter process is the most speculative, but ultimately the
most fruitful.  Using nanoresists, and assuming that etching
or deposition can be controlled to resolutions approaching single 
atomic layers, we might create billions of mechanical diamond parts, 
of a wide variety, on a single wafer.   Many of these parts would be 
flawed on an atomic scale, but these could be eliminated by the following 
process:

First, errors produced by consistent differences between the
nanoresist shape and the desired part shape might be eliminated
by bathing the parts in a solution containing enzymes that attack
and destroy the specific atomic site errors.

Second, using advanced AFM or STM arrays, create an atomic-scale mesh.
The arrays don't have to be efficient, we only need to do it a 
few times to debug the process, like making a mask.  But they
do need to produce a series of holes perfect on the atomic scale.  

The mesh comes in three varieties.  These contain 2D holes in the
shape of the side view, top view and plan view of the 3D part.
Once we have lithographed the billions of parts, we sort them 
through the meshes on a tilted, vibrating table.  Only the perfectly
made parts get through; the others get vibrated away into
the recycling bin.  Even if we have an error rate of 99.9% on
1,000-atom parts, we can get a million useful parts per 
billion-part wafer.  Three meshes can serve to error-check
trillions of parts. 

Note that while nanolithography can greatly extend the 
capability of nanotechnology, and help bridge the gap
between current technology and Drexler's self-replicating
factories, it in no way diminishes the importance of
Drexler's mechanical-assembly approach.  Rather, the technologies
are symbiotic: lithography makes the parts, and Drexler's
machines can then assemble them together in a variety of
ways.  With nanolithography, early Drexler machines can
concentrate on assembly and leave the more complex 
diamondoid mechanosynthesis to later.

ref:
K. Eric Drexler, _Nanosystems_, John Wiley & Sons 1992
Geis & Angus, "Diamond Film Semiconductors", Sci. Am. 10/92
Tim May, personal communication