Cymer Advances EUV Source to ASML HVM Acceptance

Along with a flurry of other announcements Cymer Inc . (San Diego) released around the SEMICON West 2007 timeframe was one claiming a victory for its extreme ultraviolet (EUV) lithography source. Toolmaker ASML (Veldhoven, Netherlands) has opted for Cymer's EUV source in high-volume manufacturing tools, signing a multi-year, multi-unit agreement, with the first shipment scheduled for late 2008.

Hand in hand with this ASML agreement, Cymer detailed its latest advances in its laser-produced plasma (LPP) source technology, including reaching a source power of 50 W — double the power output the company had previously announced. Cymer intends to double that output again by the end of the year, ultimately aiming for at least 200 W for volume manufacturing. The 100 W end-of-year goal was previously considered on track for a high-volume EUV scanner, but that has been a moving target. Original estimates depended on ideal resist sensitivities and mirror systems. Lacking those, power requirements could go up by a factor of 2 or 3, according to Nigel Farrar, Cymer's vice president, lithography applications marketing.

Cymer made the switch from a discharge-produced plasma (DPP) source to an LPP source about three years ago, initially using excimer as the drive laser technology and lithium as the target material.¹ The latest incarnation of the EUV source — in operation since June 2006 — uses a multi-staged CO₂ laser and a tin droplet target. Several developers have switched focus from DPP to LPP sources largely because LPP sources are considered to be more scalable and, therefore, more able to create the power needed for high scanner throughputs. They have also switched from primarily xenon targets to tin for its higher conversion efficiency.

EUV source is a laser-produced plasma (LPP) generated by a CO2 laser on a tin droplet target. (Source: Cymer)

However, tin causes increased concern with debris mitigation and, thus, the lifetime of the expensive multilayer-mirror collector. But Cymer began talking about its improved debris mitigation system earlier this year, and Farrar contends that it is a key part of making EUV technology affordable. Enabling a collector lifetime of about a year, Cymer is "pretty far ahead of the others," he said. Cymer has also demonstrated its source on actual tin droplets, rather than the larger tin surfaces that are typically being used.

In August of last year, ASML delivered two alpha demo tools (ADTs) — one to IMEC (Leuven, Belgium) and the other to the College of Nanoscale Science and Engineering (CNSE) at the University at Albany (N.Y.). Those tools are currently equipped with DPP sources from Philips Xtreme UV GmbH (Aachen, Germany). But word is the tools were unable to perform at the levels required even for research projects. In a conversation with Semiconductor International at SEMICON West 2007, IMEC's Ludo Deferm said that the DPP source resulted in a scanner output of only 0.2 wph. He said that program partners were asking IMEC to speed things up, but they were limited by the source.

Cymer set its EUV source roadmap at the beginning of last year, Farrar said, and has hit each mark since. He speaks with confidence, therefore, about his company's ability to get to 200–225 W in the not-so-distant future. The latest advances show where the improvements need to be made to get there, he said.

But now the work has begun to integrate the 10–15 different technologies that need to come together to make the EUV source ready for ASML's production scanner. One tricky aspect of the integration scheme will be to minimize the number of bounces that the EUV beam has to make off of various optical elements, meaning that the source will no longer be allowed to wind its way to the scanner from the subfab, Farrar said. Engineers must come up with an optimal illumination scheme for the source characteristics. "You certainly need the plasma source very close to the exposure tool," he noted.

With the multilayer mirrors, the best reflectivity that's been achieved is ~69%. Normal and grazing incidences are best, Farrar said, and if the beam angle is changed, that reflectivity goes down, meaning there's not much freedom to redirect the beam.

Cymer engineers will have their work cut out for them over the next 16 months or so, Farrar noted, adding, however, that they are well on track to meet ASML's schedule.