Ever since 90 nm node, ArF 193 nm lithography has been in use. The previous wavelength was KrF 248 nm. The wavelength change was only about 20% but the numerical aperture has increased nearly 2X with the help of immersion, plus phase-shift masks added another factor of 2X resolution.

With 3-4X resolution enhancement available without changing wavelength. The wavelength reduction only enhanced resolution about 20% on the other hand. The consequence of reducing wavelength is a total change of photoresist chemistry, resulting in bad problems for etch pattern transfer compared to the older 248 nm wavelength.

In retrospect, we should have stayed with 248 nm instead of trying to stay with 193 nm today.

The explanation is apparently organic chemistry.

Benzene-rich polymer is more etch resistant, with its density of C=C bonds, but benzene also absorbs at 193 nm more so than at 248 nm. Generally the more tightly bound the orbital, the larger the absorption cross section, since the probability wave density is higher.

KrF resists tend to be more benzene-rich than ArF resists.

One issue I have been hearing about is resist heating. Doesn't a shorter wavelength mean more energy absorbed by the resist, which means more heating eventually?

This is most apparent for EUV, since absorption is very high, and leftover energy is significant, going into secondary electron generation. The secondary electrons start out with > 10 eV energy, then eventually slow down and thermalize, resulting in substantial heating.

It seems they must have anticipated going to immersion in order to move to 193 nm. Otherwise it would not have been worth it. Going from 248 to 193 only gives a 22% resolution boost. It would have been better to move to 193 nm from the previous wavelength (400 nm?).

It has always been the new photochemistry being the issue. The shorter wavelength is more highly absorbed. So LER is rougher at the bottom than the top. Already a problem for 248 nm, it's worse for 193 nm.

A change in wavelength must be justified by a plan to increase the numerical aperture. This helps preserve the value of the investment in the new wavelength.

At the crossover to 193 nm, the numerical aperture was around ~0.7. There was room to go up to 0.93.

Going to 157 nm would only give you 20% better resolution, but going to immersion would give 30% better resolution, and preserve the 193 nm investment better.

If immersion doesn't work, 248 nm is better to use because it gives larger depth of focus at NA ~ 1 which can be exploited with double exposure anyway.

I was reading an interesting paper "Absolute and effective cross-sections for low-energy electron-scattering processes within condensed matter" (Radiat. Environ. Biophysics vol. 37, pp.243-257 (1998)) and learned something I didn't know before. Sub-ionization-energy electrons can still apparently cause chemical damage, through an anionic dissociative attachment mechanism. These electrons can have energies of 1-2 eV for example. Such electrons can be generated by 193 nm light hitting metal or silicon (work function=4-5 eV), after going through dielectric. Since they have such low energy these electrons travel a large distance (mfp~100 nm) in dielectric too.

So I think this can be an explanation for why 193 nm resists are so soft. Some light is transmitted through the resist and underlying layer, hitting metal or silicon, and producing low-energy electrons which come up and damage the resist layer forming anions which dissociate.

Would 248 nm be safer than 193 nm light?

Probably can't be considered safer. Even less than 1 eV can still be dangerous apparently. But maybe enough material in between can block?

immersion optical lithography is just geting mature in the last two or three years. Before then, most ppl still thought the next generation lithography needed shorter wavelength. That is because the resolution is always limited by 1/NA and wavelength. A wet lithography process is very tough and still not totally satisfactory due to problems like bubbles. The advantage is all the optics remain unchanged. Optics/light source are very challenging problem for shorter wavelength lithography.
That's why most of the companies think it will surpass X-ray lithography (Intel was an exception).
Anyway, resist was not the factor pushing the optical lithography toward shorter wavelength. It's always the easiest one to find a high resolution photoresist among all the problems.

Yes I agree

Always wondered if LASIK (which uses 193 nm) had been developed in the 80's, why it hadn't been developed for lithography earlier?

Nikon pushes the envelope for 193 nm exposure technology with the NSR-S610C. The S610C begins shipping in Q4 2006.

The S610C uses an innovative multi-axial catadioptric lens design to deliver a 1.30 NA with all aspects of performance optimized for 45 nm memory requirements. The S610C lens design enables imaging below 45 nm with a superior depth of focus, and the fourth generation of POLANO polarization control provides enhanced image contrast without any loss of illumination power or throughput.

Nikon’s innovative Local Fill Technology combines the proprietary nozzle design, high water flow rate, and airless fluid handling that consistently delivers performance free of immersion-specific defects. Nikon Local Fill Technology is compatible with a wide variety of ArF resists and topcoats, enabling defect levels on par with today’s most advanced dry ArF systems at the maximum throughput.

The Nikon Tandem Stage was designed for high volume manufacturing, providing optimized performance and efficiency for immersion lithography. The exposure stage processes wafers at very high rates, while the calibration stage is used for calibrations during wafer exchange. The Tandem Stage enables throughput ≥ 130 wafers per hour, and delivers wet-dry overlay matching equivalent to dry system performance.