News and New Products

Immersion-lithography road map hits dead end

By Ron Wilson, Executive Editor -- EDN, 7/20/2006

As the promise of EUV (extreme-ultraviolet) lithography fades, the old war-horse, 193 nm, is again rising up to fill the gap. By immersing the wafer and the front element of the lens in water, stepper vendors have been able to edge up the numerical aperture and the depth of field of their optical columns just enough to make 65-nm fabrication ready for production and probably enough to make the 45-nm node feasible. However, the next step, to 32 nm, could be costly and may be a step too far.

The lithography heavyweights, ASML and Nikon, were at Semicon West in force this month with their advanced 193-nm immersion systems. Canon, which appears to have lost ground to the leaders in the race toward 45 nm, was discussing a development platform rather than a production stepper. Everyone voiced a certain amount of confidence that the systems, which pump highly purified water through a puddle that contacts both the objective lens and the wafer, will also form the basis of 45-nm production.

Immersion lithography helps because the light leaves the objective lens into a fluid—water—with a higher refractive index than air. To increase the depth of field, numerical aperture, or both for 32-nm work, developers must find a fluid with an even higher refractive index. The refractive index of pure water is simply too low for 32-nm work. According to Ludo Deferm, vice president of business development an European research consortium IMEC, all of the fluids that vendors have so far tried either are too viscous to remain uniform as the wafer passes under the objective at high speed or are chemically reactive with the objective lens coatings, the photoresist, or both. So far, no one has been able to solve the materials problem, leaving some unpleasant alternatives. Critical layers will depend on double-exposure techniques using two masks, according to Deferm. You expose each mask using an off-axis lighting pattern to produce the desired image on the resist. This technique appears to be able to produce the feature size that the 32-nm requires, but it demands additional masks and additional stepper time, both of which are hits to cost.