Harder Uv Source For Lithography

[Enthalpy]

Hello semiconductor technologists!

I'm thinking at a source of extreme UV for semiconductor lithography, so what would be your wishes, as compared to the first figures that emerge from my rantings?

I take 30nm wavelength. Would less be better? I fear fluorescence hampers lithography.

Right now I estimate half a watt of light is produced, continuous and monochromatic and coherent, initially in a narrow beam of small divergence. Could that be enough?

This absolute silver bullet would cost several millions. Is that any worry?

Thanks!
Marc Schaefer, aka Enthalpy

 

[berkeman]

Hello semiconductor technologists!

I'm thinking at a source of extreme UV for semiconductor lithography, so what would be your wishes, as compared to the first figures that emerge from my rantings?

I take 30nm wavelength. Would less be better? I fear fluorescence hampers lithography.

Right now I estimate half a watt of light is produced, continuous and monochromatic and coherent, initially in a narrow beam of small divergence. Could that be enough?

This absolute silver bullet would cost several millions. Is that any worry?

Thanks!
Marc Schaefer, aka Enthalpy

I don't think your post translated very well. Is this question for your work, where you are putting together an IC fabrication line? Or are you thinking of putting together a small IC fabrication setup as a hobby (and have the several million dollars to spend on it)?

 

[carlgrace]

Hello semiconductor technologists!

I'm thinking at a source of extreme UV for semiconductor lithography, so what would be your wishes, as compared to the first figures that emerge from my rantings?

I take 30nm wavelength. Would less be better? I fear fluorescence hampers lithography.

Right now I estimate half a watt of light is produced, continuous and monochromatic and coherent, initially in a narrow beam of small divergence. Could that be enough?

This absolute silver bullet would cost several millions. Is that any worry?

Thanks!
Marc Schaefer, aka Enthalpy

The technical bottleneck in practice isn't the light source. The bottlenecks are the lens and the OPC patterning of the masks. Besides, the trend is increased reticle size so a narrow beam isn't necessarily a good thing.

If you really want a silver bullet (and we're talking pie-in-the-sky anyway) you should invent a much faster e-beam lithography technique.

Or, figure out an OPC algorithm that doesn't require a supercomputer to calculate in a reasonable time.

 

[Enthalpy]

Thank you!

The light source isn't the bottleneck? I thought channels 22nm long were patterned with excimer lasers at 193nm wavelength or a bit less, and this fundamentally unhealthy situation puts stress on the lenses. Shorter waves shall ease this.

http://en.wikipedia.org/wiki/Excimer_laser#Major_applications
corresponds to what I read here and there.

A narrow beam is easily broadened.

 

[carlgrace]

Thank you!

The light source isn't the bottleneck? I thought channels 22nm long were patterned with excimer lasers at 193nm wavelength or a bit less, and this fundamentally unhealthy situation puts stress on the lenses. Shorter waves shall ease this.

http://en.wikipedia.org/wiki/Excimer_laser#Major_applications
corresponds to what I read here and there.

A narrow beam is easily broadened.

Nope. Not the bottleneck. Of course all aspects of the system need to be improved but the killers are the lens and the OPC. Right now it takes something like a CPU-century to calculate OPC for a typical mask.

 

[Enthalpy]

Ask someone else for the computer. It boils down to assembling many Cpu or Gpu, I'm not that interested.

The lens is difficult because the half-wave is too big for the small pattern features. This also leads to shorter and shorter waves, and fiddlings like double exposure, but no progress has been made since the excimer lasers, and indeed IC shrink ever slower.

Such things tell a lot about the present cost and power of light sources:
http://www.cymer.com/XLR600ix/
it doesn't look cheap neither, but it produces 90W at 193nm.

 

[carlgrace]

Ask someone else for the computer. It boils down to assembling many Cpu or Gpu, I'm not that interested.

The lens is difficult because the half-wave is too big for the small pattern features. This also leads to shorter and shorter waves, and fiddlings like double exposure, but no progress has been made since the excimer lasers, and indeed IC shrink ever slower.

Such things tell a lot about the present cost and power of light sources:
http://www.cymer.com/XLR600ix/
it doesn't look cheap neither, but it produces 90W at 193nm.

Well, I would say that it boils down to improving the algorithms, since if you shrink down the process the compute problems grows faster than our ability to assemble gpu clusters.

True, a shorter wavelength UV source would relax requirements on the lens, but would also require retooling the whole system. Could work, though!

 

[Enthalpy]

Found some answers.

-----

All actors consider wavelengths like 13.5nm or 11nm for the next sources.

The synchrotron and undulator I considered is already an old idea, the current advances being fluorescence in various forms, which provides more power than the half-watt typical of synchrotrons and undulators.

Lenses are a difficulty at 13.5nm (100eV!) because materials are opaque. Mirrors aren't much better, achieving 50% reflectivity at normal incidence.

-----

I wanted to give the accelerating cavities the same wavelength as the undulator, and now I'm confident this already exists.

I wanted to add a Perot-Fabry cavity around the undulator to increase the output power. This has been considered but is difficult at such a wevelength because neither mirrors nor light guides are efficient.

Apparently I won't bring anything new nor useful in this topic and give it up.