|
No litho for old men: Random EUVL notes from Day 1 at SPIE |
|
|
|
Feb 26, 2008 at 08:15 AM |
Among the faithful congregating at the annual SPIE Advanced Lithography event in San Jose this week, one gentleman's out-of-control throwback hair style is momentarily en vogue, an eerie follicular facsimile of the big bad 'do sported by Javier Bardem in his Oscar-winning supporting actor role as stone-cold killer Anton Chigurh in "No Country for Old Men." But the Coen brothers would be hard pressed to script anything as dramatically and technically challenging as making EUV lithography a production-worthy technology in the next few years.
Yet plenary presentations by two EUVL proponents, Micron president/COO Mark Durcan and ASML's exec VP Martin van den Brink, both provided reasons for a few photons of optimism for supporters of the next-gen litho tech currently championed by many in the industry.
Durcan said that despite the hard challenges of EUVL, he is "a big believer" in the technology, although "we're not gonna hit 2009, I don't think.
I have confidence that EUV is coming at the end of the roadmap."
He showed that despite the historical trend of relentlessly increasing patterning costs (measured in the total patterning costs per layer), that the extreme litho solution might reverse the trend, achieving better projected cost at the 22-nm half-pitch node than any of the previous several generations, including 32-nm EUV and high-NA double-patterning immersion. That is, of course, if the various and not insubstantial EUVL issues involved with throughput, source quality, defectivity and the like are remedied in a timely manner.
Van den Brink (who was also presented with the Frits Zernike Award for Microlithography this year) showed a slightly different version of the trend chart, which nonetheless found the total litho costs per layer significantly less for 32- and 22-nm EUVL single-exposure approaches than any of the 32- and 22-nm double-patterning immersion schemes under consideration.
He also revealed an EUV roadmap, touting the tech's extendability down to at least 11 nm, along with mutiple mirror designs currently being researched that will help the push down the technology path. (Glad to see that "doing it with mirrors" has its place in EUV development!)
He also shared brand-new images depicting the first full-field test devices made with the alpha tool installed as part of the EUV program at the IBM/AMD/et al. Albany Alliance, a 32 x 22-mm piece of silicon that includes a SRAM cell and working transistor. From the IMEC EUV efforts, crisp SEM micrographs of 55-nm dense and isolated contacts made without OPC reflected significant progress on the patterning front. Noting that "the cost of EUVL is one-to-one driven by source requirements," he also pointed to substantial progress on the source roadmap critical path.
Van den Brink announced that ASML plans to ship five EUV production systems to both memory and "multiple" (as in more than one?) nonmemory customers starting in late 2009. (In a postplenary conversation with Micron's Durcan, he confirmed that his company will NOT be one of those initial customers. More of my interview with Durcan will appear next week in Chip Shots.)
Whether EUV does indeed become a viable litho solution in 2010, 2011, or 2012, the drama---and the development costs---will continue to escalate until the technology can prove it has the star (and source) power to make it on the big screen, and not end up on the cutting-room floor of failed innovations.
|
Comment by GUEST on 2008-03-03 14:28:03
EUV is going the way of 157nm, X-ray and E-beam. In each of these earlier cases, just one issue was enough to kill it. 157nm had resist issues, X-ray had mask issues, E-beam had throughput issues. EUV, being so continually thoroughly investigated, has probably accumulated enough issues (at least the sum of the above) to kill it 3 or 4 times over. | Comment by GUEST on 2008-02-27 15:18:29
A 100 nm EUV mask defect will print as 25 nm defect on wafer with 4X reduction. A 100 nm ArF immersion mask defect will not print as 25 nm under 4x reduction, because it cannot be resolved even with 1.35 NA. | Comment by GUEST on 2008-02-27 10:53:53
EUV flare is known to aggravate LWR (linewidth roughness). Flare is worse for EUV due to its shorter wavelength and higher energy. |
|
Blogs 
