Lasers Increasingly, if Indirectly, Used as Semiconductor Lithography Light Sources

Hank Hogan

Laser makers are being confronted with novel semiconductor lithography technology. Dubbed extreme-UV lithography for its photon wavelength of 13.5 nm, the technique is being pursued because it will allow semiconductor feature sizes to continue to shrink.

The most advanced lithography tools available use excimer lasers operating at 193 nm to pattern features that are 65 nm and below. However, extreme-UV tools will generate photons from plasmas. The leading contenders have been discharge-produced plasmas, with laser-produced plasmas almost out of the running because of inadequate laser performance.

In the laser-produced plasma approach to generating extreme-UV photons, light strikes a tin droplet and creates the plasma. One of the advantages of the laser method is that several beams can be combined through either space- or time-multiplexing to achieve the needed power. Courtesy of Powerlase Ltd.

Vivek Bakshi, a senior member of the technical staff at research consortium Sematech in Austin, Texas, noted that the situation has changed. “We have seen really good progress this year from laser suppliers.”

According to Bakshi, that fact was highlighted at a recent workshop sponsored by the consortium, where both CO₂ and Nd:YAG laser makers demonstrated that they are getting closer to delivering the 10 kW of average power, 10- to 20-ns pulse widths and the 7- to 10-kHz frequency believed to be necessary for extreme-UV operation. Multiple lasers, for example, can be used in conjunction to achieve the power requirement. Bakshi noted that these lasers operate at a 10 percent duty cycle, although suppliers expect to ramp up that ratio.

Improvements on the laser-produced plasma front will be monitored at twice-yearly Sematech extreme-UV gatherings. The next is slated for November in Sap-poro, Japan.

If lasers do become a principal extreme-UV light source component, the prospect could be significant in more ways than one. The Extreme Ultraviolet Lithography System Development Association in Japan estimates that the introduction of the new technology into high-volume manufacturing after 2011 will begin to cannibalize sales of the existing laser lithography market.

In addition, Bakshi noted, new light sources will be needed in more devices than simply lithography scanners. “Besides scanners, you will be using low-power extreme-UV sources in a lot of metrology and resist development related to extreme-UV lithography. There is a huge market there.”

He added that 13.5 nm was selected because mirrors that reflect these photons are more readily made. The industry wants shorter-wavelength sources because performing finer lithography with current sources may require high-refractive-index fluids, high-numerical-aperture optics, and double-patterning or other expensive optical tricks. A more than tenfold decrease in wavelength will make lithography of smaller features simpler and more forgiving.

The cost of this simplification is the development of an entirely new lithography infrastructure, a quest the industry has been on over the past decade. For economic reasons, the new scanners must achieve throughput of 100 wafers an hour. That requirement, in turn, places certain restrictions on the source and on the lasers behind it.

Discharge-produced plasma starts cold and is magnetically compressed to make it hot enough to generate extreme-UV photons. Currently, such sources are more powerful, more energy-efficient and less expensive than the laser-powered alternative. It is little wonder, then, that they are being incorporated into the first scanners being rolled out.

However, plasma generated by firing a laser at a sliver of tin offers potentially significant advantages. Plasmas churning out extreme-UV photons also produce a lot of heat and debris, both of which can seriously damage expensive scanner components. With laser-produced plasma, collector mirrors can stand farther off, and, therefore, the systems can mitigate these unwanted by-products more effectively.

Of the lasers being investigated, the pulsed CO₂ type has generated the most power. One company involved in this work is Gigaphoton Inc. of Oyama, Japan. At the SPIE Advanced Lithography 2007 conference held in February in San Jose, Calif., researchers from Gigaphoton — who are working as part of an Extreme Ultraviolet Lithography System Development Association program — reported on a system with 6 kW of power, a 22-ns pulse width and a repetition rate of 100 kHz. Those figures equate to 40 W of extreme-UV power, and that translates into throughput of 40 wafers per hour.

Powerlase Ltd. of Crawley, UK, is betting on Nd:YAG lasers, having recently introduced a 2-kW version based on a product used in flat panel display manufacturing. According to Samir Ellwi, vice president of strategic innovations, the company has a patented approach that combines five or six of these lasers to achieve the required 10 kW. “We are capable of doing temporal or spatial multiplexing.”

Ellwi pointed to work done in collaboration with the University of Central Florida in Orlando, where the addition of a second beam more than doubled output from 10 to 23 W. Such a ganged approach, he added, has its advantages. When a beam goes down, the entire scanner is less productive but keeps going. Under the same scenario, a single-beam system would cease operation until repairs were completed.

Lucas van Grinsven, a spokesman for scanner maker ASML Netherlands BV in Veldhoven, said that the company uses a discharge-produced plasma extreme-UV source in its current alpha-stage tools. “But it is likely we will also have to consider laser-produced plasma as a light source.”

In describing what type of source ASML is looking for, van Grinsven noted that intensity is key, with cost also important. Laser-produced plasma sources, he added, promise higher intensity, but discharge-produced plasmas are more cost-effective.

Which, if any, laser will power extreme-UV sources has yet to be determined. It could be, as Sematech’s Bakshi pointed out, that one source will be used for alpha-level scanners, another for beta-level systems and a third when high-volume production units are deployed. In addition, various scanner manufacturers could decide to use different sources.

He noted, however, that the next decision point for the industry is rapidly approaching. “By next year, we have to have a source technology decision made for the beta level.”