5
\$\begingroup\$

The current lithography machines use 193 nm light, and for the past 10 years we've been waiting for extreme ultraviolet machines, that emit 13.5nm light.

It's pretty obvious that the this challange is far greater than anticipated — ASML was talking about shipping it in 2010 (!).

So why take such a huge leap? We're talking order-of-magnitude jump in wavelength. Why not go to 100nm light first? Or 50nm? Wouldn't it be easier to develop this EUV technology more iteratively and gradually improve and fix all the problems (light source, power, availability, pellicles, mirrors, etc.) one by one as they arise when stepping between wavelengths?

\$\endgroup\$
  • \$\begingroup\$ They already do what you think is easier, so why ask the question? At present, there is no great pellicle material for EUV ( detonation, acetylene etc) so down time is significant. When they can approach resolutions with DUV by upgrades it doesn't make sense to start EUV at 100 nm \$\endgroup\$ – Sunnyskyguy EE75 Mar 3 at 17:49
  • \$\begingroup\$ Maybe you want to ask a more specific question? What's the point of 50nm EUV if DUV can already achieve this as illustrated in my answer. There is an energy density charge transfer explosive (deflagration to detonation) physics issue with the dielectric boundary of the pellicle interface. \$\endgroup\$ – Sunnyskyguy EE75 Mar 3 at 18:01
  • \$\begingroup\$ They had already achieved 15nm in 2013 with 80 W UV but they say they need 115W to achieve feasable wafer rates which is the problem with energy density and pellicle detonation from any contaminant particles staticwww.asml.com/doclib/productandservices/images/…. So what you are implying is false. The light wavelength does not match the feature resolution. \$\endgroup\$ – Sunnyskyguy EE75 Mar 3 at 18:15
  • \$\begingroup\$ sciencedirect.com/topics/materials-science/… \$\endgroup\$ – Sunnyskyguy EE75 Mar 4 at 19:22
  • \$\begingroup\$ @SunnyskyguyEE75 of course wavelength doesn't determine the resolution, but with a shorter wavelength it'll be easier to get a smaller feature size probably? Like less number of patterning \$\endgroup\$ – phuclv Apr 16 at 14:06
7
+50
\$\begingroup\$

As the feature size decreases, the wavelength of the source and lithography system must also decrease. The problem is after 193nm materials and air are significant absorbers of UV which happens around 157nm. In addition, excimer lasers no longer work well as a source after 157nm.

So if a new lithography tech requires vacuum, mirrors and a new source, you might as well do it right and jump down to 13.5nm. This step has taken 15 or so years and required entire industries to be developed to achieve 13.5nm lithography, including two EUV steps would be very costly due to the costs of developing an EUV source alone.

enter image description here

Source, Slide-5 OR Extreme UV lithography - Electrical and Computer Engineering waterloo

Some other interesting notes are lenses are too absorbent at EUV wavelengths so mirrors must be used, and 13.5nm sources will need about 1MW to get 200W of EUV light.

\$\endgroup\$
2
\$\begingroup\$

Why take such a huge leap?

Laser wavelengths between this range to date have not been possible.

enter image description here

Which leads to the next question, why not?
There is no UV photon between {F2 or ArF} and EUV carbon photons.

\$\endgroup\$
  • 1
    \$\begingroup\$ Are there ANY lasers in that gap? The wiki lists does not a single one. en.wikipedia.org/wiki/List_of_laser_types \$\endgroup\$ – DKNguyen Apr 19 at 18:25
  • \$\begingroup\$ Exactly my point all along.... \$\endgroup\$ – Sunnyskyguy EE75 Apr 19 at 18:26
  • 1
    \$\begingroup\$ Well you said "suitable lasers" which left the possibility open that lasers do exist for those wavelengths, but none that are suitable for whatever reason. \$\endgroup\$ – DKNguyen Apr 19 at 18:27
  • \$\begingroup\$ Yes "nothing in this range" perhaps a review of excitation energy of electrons in different atoms required to emit this short wavelength shows why only carbon with EUV energy levels is suitable to leap over 1 decade \$\endgroup\$ – Sunnyskyguy EE75 Apr 19 at 18:33
1
\$\begingroup\$

I'm not a semiconductor production / market expert, and this is something almost certainly defined by market forces / money availability.

So, you'll get my best guess: instead of developing very expensive intermediate wavelength processes, we ride the current light source technology as long as feasible – and that technology got another large extension by going from cleanroom air between lightsource, masks and steppers to liquids with higher refractive indices than air; I think that's called immersion lithography.
Since the speed of light in e.g. water is lower than in air, so is the wavelength of any given photon energy. My guess is that manufacturers will rather push for higher index materials for a while before switching to even higher photon energy light; basically, for EUV you need to use high-vacuum chambers, and that's probably quite a bit harder than using extremely pure liquids.

\$\endgroup\$
  • 2
    \$\begingroup\$ "almost certainly defined by market forces" — I do wonder though if in an alternate history timeline things would have gone a different way. It seems that only one company (ASML) is developing EUV machines, and given that it's been "just around the corner" for 10 years, there hasn't been a point in developing intermediate tech. IF we had known it would take 10+ years though… maybe it would make sense to invest in that. (Just a guess) \$\endgroup\$ – radex Mar 3 at 17:40
  • \$\begingroup\$ The client side is based on capital return on investment. The R&D side depends on those sales so upgrades were more feasible to improve resolution before clients could afford to made quantum leaps in their budget with risks of higher maintenance costs and tradeoffs to wafer production rats with higher density litho. But ASDL's public charts indicate a growth on new systems \$\endgroup\$ – Sunnyskyguy EE75 Mar 3 at 21:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.