Water, water everywhere—well, maybe not

193nm-i Adoption Trends Not a Good Sign for EUV 

The industry is split over when—or whether—to change to immersion lithography. And while that may seem rather obtuse for much of the semiconductor industry, the implications are potentially huge.

At the IEEE International Electron Devices Meeting in December, IBM and AMD announced plans to utilize immersion lithography to produce 45nm microprocessors. The companies said they expect products to be available in mid-2008. 

That should be no surprise. IBM has been very public about its R&D investments and successes in utilizing immersion beta tools. In fact, it looked like IBM would be an early adopter of immersion tools in its 300mm facility.

Nor was IBM alone. Months earlier, Texas Instruments CTO Hans Stork said it was very likely TI would be using immersion at 45nm. Other companies that will most likely convert to 193i at 45nm include the usual suspects, i.e. the Common Platform participants such as Chartered, Infineon, and Samsung. Even Flash manufacturers might be a candidate for 193i at 45nm if IMEC is successful with double patterning techniques at 32nm for both flash and logic.

But the decision to turn on the spigot is far from unanimous. TSMC was just as public about touting its successes with immersion, but the company now says it will continue with 193nm dry scanners at 45nm. And Intel, which generally is known for being one of the most efficient, high-volume, bleeding-edge manufacturers, also has opted to remain dry.

Intel and TSMC have their own mask shops, put out very high volumes and claim to have superior DFM tools. And they’re both squeezing out another generation of products on their existing scanners, postponing the investment in new equipment. This isn’t just any equipment, though. It’s the most expensive equipment in the fab.

So here’s the tradeoff, and what’s really going on behind this decision process:  Immersion lithography at 45nm does provide better resolution than conventional dry lithography and reduces the amount of trick adjustments with OPC and RET. But there’s a downside, too. It requires companies to buy new, very expensive equipment, and with all new equipment you have to work out the kinks, which most likely means a rise in defects. 

Regardless of which choice companies make, something very fundamental has changed in the semiconductor manufacturing world. Timing of equipment adoptions is critical to the success of equipment companies with new technologies. We saw what the delay of 300mm wafer adoption did to the industry. The same will be true for EUV. It’s generally agreed that EUV will not be ready for 32nm. There are some skeptics who feel the 193i liquid will be a problem at 32nm. We believe a high-index immersion fluid will be found and 193i will be utilized at 32nm, and that the industry will benefit from the experience of the early adopters of immersion technology.

But the future is far from clear. If Intel adopts 193i for 32nm, for example, would it only use it for one generation? It certainly doesn’t make sense to adopt a new scanner technology for only one generation. Can Intel extend the life of 193i to 22nm? The answer is probably yes. Will EUV be ready for 22nm? If it is, the timing of introduction will be unfeasible in terms of economics. There might be a couple of companies interested in switching to EUV at 22nm but EUV vendors beware. Don’t expect any substantial EUV sales activity until 2013 at the earliest.

–Joann Itow, Managing Director, Manufacturing, Semico Research Corp.