ChipWatch: The Costly Transition To 450mm And Finding The Right "Light"
[Matt Grimshaw this week continues his look at the challenges of moving to the next major transition for the chip industry, the use of massive 450mm silicon wafers. There is a need for ever smaller wavelengths of light to make ever smaller chip features but these are costly technologies.]
By Matt Grimshaw, Editorial Director of the Semiconductor Technical Journal www.future-fab.com>Future Fab International.
I've been in the industry for a decade now and when I joined in 1999 the much touted continuation of 'optical lithography' (Extreme Ultra Violet Lithography EUV) was 'almost ready for production' (ARFP) meaning it was nearly ready to be rolled out into factories. Back then it was due for the 180nm technology cycle, which it missed.
It then went on to be 'almost ready for production' and consequently miss 130nm, 90nm, 65 nm, 45m, 32nm 22nm with the latest claim of ARFP being 16nm and even this is now looking dodgier then a $9 bill. If you have any doubts - Google "EUV ready for production" yourself and take a look.
To those not versed in what lithography is, think how photograph negatives are turned into photographs, it's pretty much the same technique but on a more extreme scale with regards to dimensions and it's far and away the most expensive single piece of semiconductor manufacturing equipment ever devised.
EUV has been described as the final extension of optical lithography, but I have contacts & friends that give me very stern looks when even describing it as such. The problem with its status as 'optical lithography' is the fact that it's so far beyond what we know as standard light it's in fact closer to x-ray then it is to light.
The major problem with the optical lithography concept itself and the reason for the push to EUV is the limitations of light itself or more specifically the frequency of light.
Currently 193nm wavelength light (called DUV - Deep Ultra Violet) is used to make cutting edge chips, at critical dimensions a small film of water is even used hence 'Immersion Lithography, but due to the scaling phenomenon commonly known as Moore's Law the features these technologies are being asked to create are in fact many times smaller then the actual wavelength of the light itself.
The industry manages to work around this issue through a whole bunch of tricks called Resolution Enhancement Techniques (RET) with equally impressive names & acronyms; Optical Proximity Correction (OPC), Phase Shift Masks (PSM) and Double Patterning/Exposure (DP/E). However these advantages of these techniques are starting to be outweighed by the disadvantages - Double Patterning for instance reduces output by as much as half therefore doubling the lithography costs. Not good in an industry where the latest lithography tools are already on the ugly side of $30 million each.
Speaking of cost, EUV tools are currently being touted at the $50-60 million mark which makes me swallow harder then a 7 year old trying to choke back a bite of liver at a dinner table. However so much money has been thrown at EUV it almost seems inevitable that it's going to happen...or does it?
When a blurring effect called Intrinsic Birefringence was discovered in the materials due to be used for the lenses for 193 Lithography's original successor; 157nm Lithography, the industry dropped the technology like a hot bottle after billions had been spent on R&D and it was extremely close to production - in fact the industry did a screeching 180 degree turn away from 157 towards the at the time unloved 193 immersion. Could the same happen to EUV?
It's difficult to make a prediction on this as there are few alternatives that those that make the decisions consider viable: E-Beam lithography, which is much cheaper but is very slow although there is some hope in a 'multi beam' approach solving this issue. Then there's Nano Imprint Lithography, which is orders of magnitude cheaper but also suffers from lack of throughput and although is seemingly finding acceptance in the hard disk drive industry.
EUV is certainly an issue that the next down-cycle will decide, if it's not decided in the short term. Can the industry afford a tool of that expense? Can it afford not to have it?
The last report I saw said that EUV tools would be ready next year (2010) but then that sounds suspiciously like the echo's I've heard for ten years...