SAN FRANCISCO—Foundry company Globalfoundries Inc. will install a production extreme ultraviolet (EUV) lithography tool in the second half of 2012 and be using EUV in volume production by 2014 or 2015, a company executive said Wednesday (July 14).
Delivering a keynote address at the Semicon West tradeshow here, Gregg Bartlett senior vice president of technology and R&D at Globalfoundries, said his firm has chosen to forgo the purchase of a pre-production EUV tool. Several chip companies, including Intel Corp., Samsung Electronics Co. Ltd., Toshiba Inc. and leading foundry Taiwan Semiconductor Manufacturing Co. Ltd. have ordered pre-production EUV tools from ASML Holding NV.
Bartlett said 193-nm immersion lithography using double patterning—a technique the industry has had to use to extend 193-nm lithography to advanced production nodes—has become so expensive that, to his surprise, he is hoping that EUV lithography will reduce manufacturing costs.
"We feel very strongly that the industry does need a cost-effective EUV lithography solution for volume manufacturing going forward," Bartlett said. "Failure for the industry is not an option."
Globalfoundries (Sunnyvale, Calif.) will install the production EUV tool at its Fab 8 in Malta, N.Y., Bartlett said.
Based on a 13-nm wavelength technology, EUV is not targeted for production until the 15-nm node, though some believe it won't be in production until 2016 or later. Originally, EUV was targeted for the 65-nm node, but the technology has been pushed out due to the lack of power sources, resists, defect-free masks and other technologies.
The shot noise problem will be severe for 2x nm in mask patterning. With a 12 uC/cm2 dose, and 80 nm holes on a 4x mask, there are only 3800 electrons used per hole, which is a 10% noise value in a population of a billion contact holes.
I just read the keynote slide 21, it reads:"Immersion lithography will get us through 22/20 nm node and beyond, but with significant complexity and cost challenges." Okay, but later when describing EUV's main issues, "Cost of ownership" is clearly highlighted.
So in actuality, cost is not yet a good case for EUV.
It is kind of consistent with Intel recently changing its tack for EUV at 11 nm, to more of a complementary role. From the way the analysis goes, the dose needs to increase with each new generation. That puts lots of weight on improving the source power, and incurs additional delay.
Shot noise also has to be considered for e-beam, so it's something that needs to be addressed even for the other NGL's.
In fact Intel published recently that ASML's dose level would cause too much shot noise at 22 nm. So I guess they (ASML) would have to increase it up an order of magnitude at least, to more than 100 mJ/cm^2.
ASML recently quoted the dose to be used for EUV to be 15 mJ/cm^2. That is 10 EUV photons per square nm. Even for 30 nm minimum resist thickness, that is much less than 1 photon per nm^3. I don't think classical optical predictions can be extrapolated with such a low photon density.
Their EUV tools are now spec'd (resolution) based on assumptions of the resist performance (like dose sensitivity, diffusion length, etc.). It is no longer the usual classical dependence on wavelength and numerical aperture. This makes the value of the tool kind of questionable if it depends on something not controlled by the vendor.
One thing I heard after this article was published was that ASML only plans to make six pre-production tools, and they are all spoken for. (I haven't had a chance to confirm this yet).
Globalfoundries says "We want to get the production-level EUV tool into the fab as soon as possible so we can move development from the lab to the real production environment. This is the same thing we did with immersion, which allowed us to be an industry leader in high-volume implementation with that technology. It wouldn’t be appropriate for us to comment on the offerings of an individual tool vendor."
Join our online Radio Show on Friday 11th July starting at 2:00pm Eastern, when EETimes editor of all things fun and interesting, Max Maxfield, and embedded systems expert, Jack Ganssle, will debate as to just what is, and is not, and embedded system.