by Debra Vogler, Instant Insight Inc.
Mention EUV lithography (EUVL) at any industry event, and the likely comments will be about the status of source productivity improvements. Later this year, ASML will be shipping its production version EUV scanner, the NXE:3300B, which is expected to have a throughput of 125wph at 15mJ/cm². The source power is a bit of a moving target, as the company’s roadmap shows improvements throughout 2012 from about 20W, to 50W, and ultimately, somewhere in the 105-250W range to support a throughput of 60-125wph. EUV source suppliers Cymer and Gigaphoton are both working on a number of key productivity enhancements, with new data coming into play seemingly every few months.
According to Michael Lercel, senior director of EUV product marketing at Cymer, the company has been able to demonstrate 50W of open loop average power running pre-pulse for periods of time at high duty cycle on an internal HVM phase I source (Fig. 1). Lercel noted that with pre-pulsing and at a moderate duty cycle (not the full operating duty cycle used for the 50W source), the source can get up to 90W, which is a significant increase in conversion efficiency over what the ~10W sources are running in the field today.
Figure 1. LPP with pre-pulse: capability up to ~50W average power at high duty cycle demonstrated at Cymer on an internal HVM I source. SOURCE: Cymer, 2012 SPIE Advanced Lithography Conference, paper #8322-53
Increasing the drive laser power to get to the 250W HVM source (i.e., HVM phase II) naturally involves design changes to the drive lasers, making them larger and more powerful by adding more stages of amplification, which Lercel says is a fairly well known technology. “The challenge is maintaining the optical quality as you keep increasing the power,” said Lercel. “You’re putting quite a bit of load on the optical components.” Still, he characterizes these as more engineering challenges rather than new physics challenges.
Two major changes that Cymer had to make to increase the source availability from 70% to 95+% involved the collector mirror and the droplet generator. The collector mirror had to be designed so that it can be removed and replaced much more quickly (without opening the vacuum chamber), and the collector mirror lifetime was extended by using improved coatings to enable it to run for about 8-10 weeks of operation in the field. The company cannot comment on improvements to the droplet generator except to say that the lifetime was increased by a factor of 8x.
Phil Alibrandi, director, sales and global account management at Gigaphoton USA, told SEMI that conversion efficiency is influenced by both the size of the initial tin droplet and the ability of the pre-pulse laser to properly target and expand the material into a uniform “mist.” “The drive laser’s pulse must hit the “mist,” create the plasma, and unleash 5% of the potential energy,” said Alibrandi.
At SEMICON West, Gigaphoton released news that it has been able to show a maximum of 5.17% conversion efficiency with an average of 4.7% (Fig. 2).
Figure 2. The plot shows the resulting EUV power at IF as the conversion efficiency (CE) and main laser power increase. By holding the CO2 laser power constant, the data illustrates the leverage of 5% CE vs. any lesser value. SOURCE: Gigaphoton
According to Hakaru Mizoguchi, CTO of Gigaphoton, the company’s proprietary pre-pulse laser, which uses a shorter wavelength than the CO2 main laser, is vital to achieving these results. “A shorter wavelength, pre-pulse laser creates the initial, uniform thermal expansion of the tin droplet with insignificant loss of potential energy,” said Mizoguchi. “Combined with the proper spot size of the main laser pulse, more energy is released from the plasma.” The company also said that it can reduce the size of the droplet to much less than the typical 30-35µm size. As a result, Mizoguchi explained that no tin fragments are created and superconducting magnets “sweep” away the remaining ionized material. “Significant amounts of debris never get close to the collector mirror,” he said.
Gigaphoton is also pursuing an “on-demand” tin droplet generator. This “on-demand” alternative to the current continuous stream droplet generator will use less tin material and eliminate any unused raw tin that could be propelled through the chamber.