65-nanometer technology, reducing the cost per function by 25 percent per annum. Keep in mind that technology scaling has been paced by advances in lithography, including the steppers/scanners, the photoresist and the mask-making tools. This year, Dataquest projects that the overall lithography sector will have about $5 billion in revenue, out of about $19 billion in fab equipment revenue. It also predicts that the $5 billion spent in 2002 will more than double, exceeding $11 billion in both 2004 and 2005.

The lithography companies will need every penny of it, because that sector of the industry is being asked to do a lot.

The 157-nm lithography solution is going to be late-and expensive. The alternative, using 193-nm scanners with hard-phase-shift masks-essentially two pieces of glass-may be not quite good enough to get the industry fully into the end zone of 65-nm design rules. And hard-phase-shift masks are very expensive.

Joe Mogab, a director at Motorola's advanced-products research and development lab (Austin, Texas), said that the 193-nm lenses used for Motorola's HIP-8 (90-nm) process have a 0.75 numerical aperture (NA). "To get to HIP-9 [65 nm], we have to go to a significantly higher NA, if possible, because in my view the 157-nm tools are lagging," Mogab said. "I have a hard time believing that adequate resists can be ready for 157-nm lithography by then."

Mogab pointed to the 157-nm pellicles (which protect the mask and correct for optical aberrations) as another possible source of delay. Organic pellicles can't withstand the 157-nm radiation, and inorganic pellicles-a special form of fused silica-lose transparency unless they are made very thin. A pellicle of 800 microns (less than a millimeter) in thickness could suffer from gravitational sag, Mogab said. "The scary prospect is that the pellicles won't all be exactly alike, and I'll have to do a requalification each time I get a new pellicle," he said.

Companies would like the 19 percent shorter wavelength that 157-nm lithography would bring. The more likely prospect is that enhancements to the lenses and masks will allow the industry to limp along to quasi-65-nm design rules, or move to a half node at 75-nm design rules.

DRAM and microprocessor vendors might be able to justify the higher mask costs, while foundries may opt for a half node for their smaller-volume, fabless customers.

Send feedback to dlammers@cmp.com.

http://www.eetimes.com/