SANTA CLARA, Calif.—Cymer Inc. Thursday (Feb. 16) revealed it has made significant progress in the development of a power source for extreme ultraviolet (EUV) lithography over the past few months.
In a paper delivered Thursday morning at the SPIE Advanced Lithography Conference here, David Brandt, senior director of EUV marketing at Cymer, detailed significant improvements in source power, availability and dose stability for the company's laser-produced plasma EUV source.
Cymer revealed it has demonstrated the capability for average power of roughly 50 watts at high duty cycle (80 percent) using a newer exposure technique that makes use of a pre-pulse on the company's HVM I source. Less than three months ago, the company was unable to demonstrate anything above 10 watts using the same equipment and technique. The 50W achievement is also a 15X improvement from Cymer's results at the time of last year's SPIE, Brandt said.
Brandt cautioned that the result was achieved in open-loop testing, minus the controls placed on systems in the field to improve stability, which typically reduce power by about 25 percent (though the company is working to implement the controls with less power loss). Cymer has not yet implemented these controls in the system using the pre-pulse technique. The pre-pulse conditions the target--causing the liquid tin target to vaporize--so that it more readily absorbs the dose.
"It's a good step forward," Brandt said.
encouraging to see that the average source power is increasing very
significantly," said Hans Meiling, director of product management for
EUV at lithography tool vendor ASML Holiding NV, during a presentation
earlier in the morning Thursday.
While the improvement to 50W average power is significant, Cymer acknowledges that it still has a ways to go. The goal for putting EUV systems in the field with adequate throughput of 60 wafers per hour is 100W average power. The ultimate goal is to achieve 250W power for throughput of 125 wafers per hour.
"Long term, we need to go to higher power levels to sustain 125 wafers per hour, maybe more," Meiling said.
The Cymer paper also reported that the availability of the five HVM I sources Cymer currently has in the field—the time that the sources were available for use as opposed to being serviced—improved to 70 percent in the fourth quarter of 2011, up from about 50 percent in the third quarter and the less than 40 percent reported at last year's SPIE conference.
Development of an adequate power source has proven the most cumbersome stumbling block to putting EUV lithography into volume semiconductor production. ASML NXE:3100 pre-production tools in the field have a throughput of less than 10 wafers per hour. Both ASML and Cymer have acknowledged that development of an adequate power source has been slower than expected, but ASML has pledged to ship production tools in the second half of this year with throughput of 60 wafers per hour by the end of this year.
Given the anxiety over the development of EUV power sources, Cymer representatives acknowledged that they weren't sure how warmly their progress would be received at SPIE. Brandt said he expected the progress on the availability to be applauded and that the progress on the average power would be welcomed by firms developing EUV processes, but that it's not yet adequate for commercial production.
Also Thursday, Marc Corthout, president of Xtreme Technologies GmbH, said the company has achieved 37W average power at 100 percent duty cycle with its laser-assisted discharge produced plasma source. Corthout said module-level testing of the various components of the source have proven operation at 50W. By the third quarter, the company should be able to deliver a source with 70-90W average power, he said. "By summertime we should be at the 100W region [at 100 percent duty cycle]," Corhout said.
Cymer is also working on HVM II, the successor to HVM I, on which the company has demonstrated 160W average power in open-loop testing at low duty cycle, using the pre-pulse technique. The company plans to begin integrating the first HVM II sources into ASML NXE:3300B scanners beginning in the second quarter.
Brandt acknowledged that the schedule for EUV source delivery has slipped a bit, in part because the company is now working on both modifications to HVM I and preparing to integrate HVM II. But he reiterated that the company is on track to deliver a source that will support production throughputs by the end of this year.
"A lot of what we did in 2011 was getting those sources out there, getting those sources operational," B randt said. "Now we are trying to catch up by building our own tools so that we can improve the performance."
Cymer has two HVM I sources undergoing testing and development at its headquarters in San Diego and is working on a third, Brandt said. "We think its fundamental to have the tools in your own factory, run them the way the chip makers run them, and learn, hopefully, at a faster rate than they do," Brandt said.
Brandt also said Cymer is well on track to exceeding the final EUV specification for dose stability of less than 0.2 percent. "The dose stability is amazingly stable," Brandt said. "When it runs well, it's incredibilty stable."
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.