EUV sources shape up for competition

03 Oct 2003

Extreme-ultraviolet sources look to have secured the key role in next-generation lithography, and are scheduled for use in full-scale microchip production by the end of the decade. Michael Hatcher reports on recent progress.

From Opto & Laser Europe October 2003

If Moore's Law is maintained, lithography tools will need to produce 32 Gbit memory chips and 20 GHz processors with feature sizes of 20-50 nm by 2009.

With a question mark hanging over the viability of 157 nm lithography, extreme ultraviolet (EUV) lithography has been thrust into the spotlight recently (see Opto & Laser Europe September p11). Intel, the biggest customer for lithography equipment manufacturers, is backing EUV technology and wants to use the method to make its microchips at the 32 nm node in 2009.

So the pressure is now on for EUV lithography component developers to resolve the key issues that might delay the introduction of the technology. Chief among these is the EUV source: finding one that can generate the power levels needed for a production-quality tool operating at 13.5 nm, as well as offering a high pulse repetition-rate and good stability, is a challenge.

The contenders EUV sources come in two distinct forms: discharge-produced plasmas (DPPs) and laser-produced plasmas (LPPs). Both methods are based on the principle of delivering energy into a target material, typically xenon gas, to create an EUV-emitting plasma. This emission is collected and directed using silicon-molybdenum mirrors rather than conventional optics, which - along with air - absorb EUV light.

DPPs are similar in principle to excimer lasers, using a pair of electrodes to excite EUV emission, whereas LPPs rely on high-power solid-state lasers to bombard the target with optical energy. Following strong development over the past couple of years, DPP-based EUV sources are now being integrated with other key components into prototype microstepper systems.

UK company Exitech builds prototype microlithography steppers and is due to deliver the first of two EUV steppers to International SEMATECH, the global consortium of semiconductor manufacturers, by the end of this year. The second stepper is destined for a leading chip manufacturer.

The steppers incorporate a DPP EUV source built by Xtreme Technologies, a joint venture between German firms Lambda-Physik and Jenoptik, that is developing both types of EUV source. The source used by Exitech emits 30 W of 13.5 nm radiation and according to Malcolm Gower, Exitech's chairman, it meets all of the stepper requirements. "The scheduled delivery to Sematech is proceeding as planned," he said.

But while this power level is acceptable for such prototype tools (Sematech will use the microstepper to develop EUV photoresists), it is well short of the output needed for high-throughput microchip production.

Production power EUV output can be judged by the in-band power at intermediate focus - that is, the power emitted at 13.5 nm that can be delivered from the tool to the wafer. For a production scanner, around 100 W is needed at intermediate focus, equating to a total output of at least 300 W.

According to Gower, none of the EUV sources currently being developed will easily scale to this power level. He told Opto & Laser Europe: "This is currently one of the biggest 'show-stoppers' that could prevent EUV lithography becoming a reality for volume manufacturing of integrated circuits."

The issue looks set to be a strong theme at the 2nd International Extreme Ultraviolet Lithography Symposium, to be held in Antwerp in early October. Xtreme's Uwe Stamm - one of the key Antwerp speakers - told Opto & Laser Europe that since the company shipped its prototype to Exitech in February, it has doubled the output power of its DPPs. Xtreme's latest source emits 10 W at intermediate focus, says Stamm, and he believes that this can be increased to 20 W or possibly even 30 W with today's technology.

Other companies are also making solid progress with DPP: the US excimer laser specialist Cymer is reported to have developed a discharge source producing a 60 W total output at a 4 kHz repetition rate. But despite this achievement, reliably producing 100 W at intermediate focus appears to be impossible with the current technology.

The problem is caused by the proximity of the DPP to the wall of the source. This makes the plasma hard to cool and limits the power that can be extracted. Pumping in more energy simply overheats the plasma, and the source soon malfunctions. "In principle, we could achieve the necessary power, but in doing so the lifetime of the source would become so short as to be impractical," said Stamm.

LPP playing catch-up Progress in LPP-based EUV sources has lagged behind DPP, and for the moment Gower says that Exitech has no plans to use LPP sources in its stepper tools. "They do not produce enough power and are not sufficiently developed for our purposes," he said. "As well as an expected much higher cost of ownership, [LPP EUV sources] currently have fewer commercial champions developing the technology compared to the discharge approach."

That could be about to change, however. The key problem with LPP so far has been how to build lasers that can deliver sufficient energy into the plasma at a reasonable cost. With an optical-EUV conversion efficiency of 1% at best, a huge amount of laser power is needed to generate even a modest EUV output, and using expensive diode-pumped lasers further increases the price tag. But according to one developer, LPP has a number of advantages that suggest it could be scaled to produce the high-power EUV output needed for a production tool, and at a cost to rival DPP sources.

Terry Nowell is chief executive of Powerlase, a UK-based firm that is developing an LPP EUV source. In Antwerp, Nowell's team will set out to change the negative perception of LPP-based EUV sources.

Powerlase's system uses three diode-pumped Nd:YAG lasers that each fire 1 kW at a xenon target to generate the plasma. Nowell says that the Powerlase LPP source will be able to produce 24 W EUV output by the end of this year, which translates to 12-14 W at intermediate focus. According to the company's technology roadmap the total output will scale to 100 W by 2005 and 300 W for a production tool in 2008. Nowell says that with industry collaboration, the source mirrors can be optimized for more efficient light collection, further improving the intermediate focus power.

Crucially, Nowell's team believes that an LPP-based EUV source could be operated at a cost-of-ownership comparable to that of a DPP source. The current cost is estimated at more than $100 000 per watt of in-band power, but with more lasers multiplexed and a projected fall in the cost of high-power laser diodes, Nowell reckons that this figure could be cut by 90% by 2007.

Thanks to an electro-optic switching mechanism that Powerlase has developed with crystal specialists, its laser modules now deliver high-energy pulses of 7-10 ns. The duration of the exciting pulse is critical, as the resonance time of the plasma is only 4-5 ns. "Longer pulses are simply a waste of energy," said Nowell. The short pulses improve the conversion efficiency of wall-plug power into EUV emission.

The Powerlase scheme also uses temporal multiplexing, in which laser pulses from separate modules are interlaced to produce an overall laser repetition rate of 20 kHz. This in turn improves the rate of EUV output. "Ideally, we need an output repetition rate of 4 kHz, as this is the lithography benchmark nowadays," said Nowell. "Discharges would struggle to achieve such a high figure when scaling up to higher powers."

Other advantages to the laser scheme include the angle over which EUV emission can be collected, which is restricted by the electrode geometry in a DPP source. Nowell adds that LPPs are more compatible with solid (e.g. tin) and liquid (e.g. xenon jet) targets, which are expected to produce higher EUV output owing to superior conversion efficiency. If Nowell's calculations are right, it could give LPP EUV sources a serious competitive edge over their DPP rivals.

Xtreme has also been making progress in LPP source development. Using a xenon target, Stamm's team has produced around 1 W at intermediate focus using a 500 W pump laser, as well as using a few-nanosecond diode-pumped Nd:YAG laser pulsing at up to 10 kHz.

Japanese investment Despite being home to two key lithography stepper manufacturers - Canon and Nikon - Japan has been relatively slow to pick up the EUV baton. However, the EUV System Development Association (EUVA), a consortium of nine Japanese firms, is now seeking to redress the balance and is investing heavily in LPP source development.

The EUVA's aim is to produce a 10 W alpha tool by spring 2006. So far, it has built an LPP source with a 10 kHz repetition rate (the highest yet reported), although the in-band emission was just 0.6 W. It is also developing a discharge source that has produced 6.3 W EUV output at 2 kHz.

Ultimately, it will be chip-manufacturing big guns such as Intel and Texas Instruments that call the shots on the lithography production tool developers - ASML, Canon and Nikon. According to industry sources, Intel's lithography chief Peter Silverman has told stepper manufacturers that the company is prepared to pay $30 m (€26.8 m) for an EUV prototype and $20 m for subsequent production tools. ASML reckons that to achieve the necessary production throughput, a 120 W EUV source is required. Laser-generated plasma may turn out to be the only way to produce such powers.

In the near term, DPP remains the more mature technology and looks set to dominate the prototype stage. Now begins the challenging task of producing a source suitable for early production tools. For the winning technology the reward will be a contract to supply the biggest customers in the semiconductor industry.