[[extropy-chat] diffraction limit

[Dan Clemmensen]

Robert J. Bradbury wrote:

>On Sat, 29 May 2004, Dan Clemmensen wrote:
>
>  
>
>>I think the E-beam
>>diffraction limit is small enough for atomic precision?
>>    
>>
>
>I would suspect that to be the case but at the atomic level you
>have much greater problems such as the purity of the bulk materials
>involved.
>
>  
>
>>The industry tried to go from
>>193nm to 153nm, but this required CaF lenses, and these lenses turned
>>out to be impossible to build in production quantities.
>>    
>>
>
>Dan, Eric D. points out in Table 6.2 in Nanosystems that UV-C is from 280
>to 200 nm.  193 and 153nm are below this scale.  The energy of even UV-C
>radiation exceeds the many common bond dissociation energies (Table 3.8).
>Has anything you have read commented on the lifetime of the lenses due
>to the high flux of bond-breaking photons?
>
>Also, there was a lot of talk 3-5 years ago about the efforts to produce
>even shorter wavelength beams (effectively X-rays) and there were several
>reports if I recall correctly of capabilities of producing feature sizes
>as small as 10nm.  But I haven't heard anything about these recently.
>Do you know if these methods are still being worked on?
>
>  
>
193nm has been standard for some time now. This is about the shortest 
feasible wavelength for quartz lenses IIRC. Up to and including the 
193nm generation, Moore's law has been driven by finding a new 
technology suite for each shorter wavelength, but the lens material 
transparency was not the issue. Other lens parameters were important, as 
were resists, deposition technologies, mask creation, silicon purity, 
and many others. The hope was that 153nm (ArF lasers and CaF lenses,) 
would push the refractive optics down one more time, but it didn't work, 
so the industry is falling back on other tricks.

As far as I know, all lower-wavelength approaches will require 
reflective optics. This will require a radical change in the optical 
path in the machinery (called a "stepper"for unrelated reasons) that 
exposes the silicon through the mask. With X-rays, even standard 
reflective optics fail, and you must move to grazing-incidence lenses. 
This is even more difficult than standard reflective optics.

All of this means that the next two generations of "Moore's Law" will 
probably be achieved by using tricks that were investigated earlier but 
which were costlier than the  direct approach of using a new generation 
of reflective optics. Either the industry will use the time gained to 
implement reflective optics, or Moore's law (applied narrowly to silicon 
lithography) will come to an end. I personally feel that silicon 
lithography will be superceeded. I certainly hope so, because is it now 
very difficult to innovate at the device level, as the capital costs are 
too high. We are now innovating primarily at the system level, as a very 
small number of device-level innovations (i.e., cheaper processors and 
memory) enable a progressively wider range of cost-effective systems 
(e.g., cell phones.) I feel that global innovation will continue for 
several years at the Moore's Law rate even if there were no further 
advances in device electronics, as we continue to explore the space of 
new systems.

Ideally, we will develop a new technology that permits an engineering 
team to develop and manufacture a new device without huge capital 
investment.

Of course, I think that the existing hardware is more than capable of 
bootstrapping the singularity, given the right software,