Hi,
For replacing the expensive custom photomask creation step of prototype wafer production, what about using a simple reuseable mask that has a single 20nm wide circular hole in it, and then have a simple pulsed laser light source send light through this hole and hit the wafer. The wafer itself can be placed on a piezoactuated XY stage with 20nm resolution to move the wafer in a XY grid to do the photoetching steps.
cheers, Jamie
What would the advantage of that be over standard e-beam lithography?
Best regards, Spehro Pefhany
Hi,
I think the lack of secondary electron generation is the main advantage of using light instead of electrons in lithography. The piezo-laser setup might be slower than an electron beam and would probably require constant XYZ feedback for XY etching grid alignment and Z beam focus distance.
I don't know if the light coming through a 20nm hole would be bright enough to trigger the photomask reaction though at non-ionizing wavelengths.
cheers, Jamie
I think a simpler idea would be to make an IC that is a grid of LED's, ideally 20nm x 20nm LEDs :) Then put that IC right onto the photoresist and turn on the LED's that are required to activate the photoresist.
cheers, Jamie
Jamie wrote in news:j1hqpk$i84$ snipped-for-privacy@speranza.aioe.org:
diffraction is the main problem,IIRC. edge effects.
Visible light has a wavelength of a few hundred nm. Diffraction will result in a 20nm hole acting as a point light source.
This is why electron microscopes exist; they can resolve features much smaller than is possible with light.
Hi,
Yes I was thinking the sheet with the 20nm hole would have to be within about 20nm of the photoresist to make it possible to work. If a XYZ piezostage with position feedback was used that could technically be possible.
cheers, Jamie
prototype
The
steps.
=20
And what wavelength would these LEDs emit at?
Ideally as high as possible below the ionization level of the photoresist, but if there was a ~20nm max thickness fluid immersion or direct contact between the LED's and the photoresist I think the wavelength wouldn't be important.
cheers, Jamie
Oops, I was thinking about that more, actually it would probably be extreme UV wavelength due to the small LED pixel size, which would be ionizing to the photoresist I guess.
cheers, Jamie
Jamie schrieb:
Hello,
there are a lot of problems mentioned already, but there is one more: try to calculate the plotting time needed for a full wafer with 20 nm resolution in both directions.
Bye
Non-performers are always full of "bright" ideas ;-) ...Jim Thompson
[On the Road, in New York]
Hi,
A typical piezoactuator adjustment frequency ie:
ranges from 500Hz to 2kHz maximum, so to plot a single layer for
1billion grid coordinates might take 138hours at a 2kHz rate. So it wouldn't be reasonable for a full modern wafer but could be useful to plot small areas of silicon at high density.cheers, Jamie
Yep, if someone told me I'd have to work on all my crazy ideas I would quickly lose my imagination :)
cheers, Jamie
One of the things i find fascinating is step and repeat mask aligners. Steps of near centimeter in some cases, post alignment resolutions of nanometers. Using 20 cm wafers that is 8 magnitudes fine resolution (about 29 bits).
?-)
Hi,
That makes sense, so you only have to make the small mask then move it, kind of a fractal mask strategy. Another potential way to produce a
20nm resolution custom photomask quickly could be to use a cheap mass produced IC with an XY grid structure of capacitors like flash memory, except instead of storing data with capacitors at each memory location, the IC would be designed so the memory elements act as a sparkgap on the nanometer scale, so that once a flash bit was charged up to a certain voltage it would trigger a release of electrons and light and cause a chemical reaction in the photomask.cheers, Jamie
What wavelength of laser do you plan to use here? 20 nm is quite a bit smaller (like 20 times) than visible light. So, you'd need a deep UV laser to get any light at all through a 20 nm hole. And, unless the wavelength was about 5 nm, the scattering would be greater than the light that passed straight through. Also, trying to get any beam to pass through a 20 nm hole would be insanely inefficient, so that only a few % would make it through. The rest would melt the aperture.
Finally, if the above could be made to work, you'd be exposing one 20 nm dot of the wafer for every laser shot. Figure out how many such dots it takes to compose even a simple chip, and you'd take hours, maybe WEEKS, to expose a SINGLE chip!
By the way, if you DO have a 5 nm wavelength laser that doesn't require either an atom bomb or a synchrotron to power it, DARPA REALLY wants to hear from you!
Jon
Appreciable amounts of light will not pass through apertures smaller than the wavelength. To pass through a 20 nm hole, the "light" needs to be X-rays, which are, by definition, ionizing. Photomasks WORK by the ionizing effects of the light.
Jon
Hi,
Thats obsolete, the new idea is to use a modified flash memory IC placed directly on the photomask, and use each memory cell as a sparkgap, or a passive matrix "mini-ion engine" (ie a microscopic grid of electron beams)
cheers, Jamie
Have something to add? Share your thoughts — no account required.
Ask the community — no account required