By Nigel R. Farrar and David C. Brandt, Cymer Inc.

Laser Produced Plasma (LPP) EUV source technology has advanced towards the specifications needed for actual chip production.

The clear leader for high-power production EUV sources is laser-produced plasma (LPP) technology. After investigating alternatives, LPP has been Cymer’s chosen technology path for the last four years. This source architecture provides key advantages of both high conversion and collection efficiency with intrinsic scalability of output power.

Source power and lifetime have been top issues for EUVL development in lithography industry surveys for the past three years. This is because the high throughput needed to provide competitive cost of ownership (COO) for EUV exposure tools is critically dependent on source power and resist sensitivity due to the low system transmission, which results from using reflective optics. Since it appears difficult to improve resist sensitivity while simultaneously maintaining resolution and line-edge roughness (LER) performance, high source power is essential for high productivity production EUVL.

LPP sources use a drive laser to heat a target material to a high-temperature forming plasma which emits at the EUV wavelength-of-interest, 13.5nm. Research into different combinations of drive laser wavelength and target material over the last few years has shown that the optimum combination is a CO₂ drive laser running at 10.6µm wavelength, and tin target material, which produces the highest conversion efficiency from laser input energy to in-band EUV output energy. This energy is emitted isotropically, but must be collected and transmitted to the intermediate focus (IF), which is the interface between the source and exposure system. It has been shown that LPP architecture allows much higher collection efficiency of output energy than alternative approaches due to the small source size and largest geometric collection angle possible.

A key challenge in LPP source design is to maintain high collection efficiency and consistent power output over long periods of time to meet COO targets. The proximity of the collector optics to the high temperature plasma exposes it to high energy ions and other debris which can damage the reflective coating and reduce collection efficiency. Minimization of damage from debris requires the use of small droplets of target material. Development of a source system that can maintain high-conversion efficiency using droplet targets implies maximizing the laser input energy that is coupled into the small droplets, which requires complex optimization and control of the laser beam delivery and focusing optics.

Cymer’s plasma chamber architecture is shown in Fig. 1. The tin droplets are emitted from a droplet generator in the wall of the vessel. Light from the CO₂ laser is introduced into the chamber through a central hole in the collector mirror. The laser beam is focused and steered to impinge on the droplets using closed-loop feedback from the droplet, targeting cameras that monitor droplet position. A plasma is formed at one focus of the elliptical mirror and light is collected and refocused to the second focus of the mirror, which is the IF position. Turbo pumps near the IF, and other containment components, prevent molecular transport from the source chamber to the scanner vacuum chamber. A beam-stop prevents laser light from entering the scanner. Any non-targeted droplets are collected and may be recycled, although if the droplet size goal is achieved, only a very small amount of tin will be required during a year of operation.

The plasma chamber will be integrated directly with the scanner body. This close coupling between the source and scanner is unlike current laser light sources and has required very close design interaction between Cymer and its direct customers. In the fab environment, the laser and support electronics will be installed in the sub-fab and the output beam will directed through the fab floor to the plasma chamber.

The initial power requirement for EUV sources is about 100W (in-band at IF). The source design parameters needed to meet this requirement are shown in the Table. Cymer has demonstrated progress towards achieving these target performance parameters over the last several years, with a goal of delivering sources for use in pre-production scanners starting at the end of 2008. The development program is on track to meeting this goal.

The first two sources for delivery have been assembled and are operational. One of these is shown in Fig. 2. The high-power pulsed CO₂ laser is based on production-proven commercial technology, capable of delivering 12kW of output power at 50kHz, well in excess of the initial target shown in the Table.

In-band conversion efficiency at the 3% target level has also been demonstrated on these systems. Figure 3 shows a demonstration of 100W burst output power over short periods of time. This has been achieved with extensive development of droplet generator technology, which produces highly repeatable droplet size and spacing, and a closed-loop laser targeting control system that ensures each laser pulse is optimally focused and accurately targeted on the corresponding droplet. Currently, development of enhanced thermal control of beam delivery and focusing optics is ongoing to extend continuous operation performance. To date, 25W average power output over 90 minutes operation has been achieved (Fig. 3). Currently, power is calculated IF value based on measurements at the plasma and using the collector characteristics shown in the Table.

The power output has ramped rapidly over the last two years, as shown in Figure 4. Initially, power scaling was focused on burst power, which was required to demonstrate that high conversion efficiency could be achieved using an integrated system of droplet targets. Burst power at low-duty cycle was scaled by two orders of magnitude over 18 months. Since mid-2007, when it was clear that burst power was on a trajectory to meet the 100W target, focus was shifted to scaling average power with a goal of reaching 100W by the end of 2008. This work has required the identification of various issues limiting high duty cycle operation, and developing engineering solutions, primarily in the area of thermal control. To date, progress is on plan, with another factor of 3 improvement required to meet the 100W goal.

Another integral part of the strategy for high-power output, shown in the Table, is the high-reflectivity, high-collection-angle mirror. This mirror will have a 5-steradian collection angle and be greater than 600mm in diameter. It will be coated with many silicon-molybdenum multilayers, similar to the scanner optics mirrors, to reflect a small band around the target wavelength of 13.5nm. Two key differences in the mirror are the need for a graded multilayer spacing from center to edge, to compensate for changing incidence angles, and a mutilayer design that will resist interdiffusion (and loss of reflectivity) at the high temperatures likely to be experienced in the source chamber. Small diameter (320mm) mirrors with such coatings have been produced and show excellent center-to-edge reflectivity uniformity, which meets the performance target shown in the Table. Larger mirrors are currently under fabrication for insertion into the prototype systems and have recently completed the polishing step, as shown in Fig. 5.

To deliver acceptable COO for the source, it is essential to maintain high power, high collection efficiency, and clean transmission over long periods of time. This requires technology within the plasma chamber that prevents degradation of mirror reflectivity from three main sources:

• Deposition of tin particle debris from the droplets
• Erosion of the mirror by high-energy ions and neutral atoms
• Deposition of tin from vapor

The first is mitigated primarily by the geometric design of the source vessel. Buffer gas significantly reduces erosion rates by mitigating the ion flux incident at the mirror surface by up to 4 orders of magnitude, and the ion energy by about one order of magnitude. Buffer gas also offers the advantage over electric or magnetic mitigation schemes in that it will also mitigate neutral atom sputtering of the mirror. Although the erosion rate is significantly reduced, the targeted collector lifetimes will also require the addition of hundreds of sacrificial multilayers during the coating step of collector fabrication. Deposition must be almost entirely eliminated because only about 1nm of deposited tin results in unacceptable reflectivity loss. Cymer has shown that debris mitigation techniques in the chamber can effectively protect the collector from reflectivity degradation and provide a reliable and robust method of preserving collector lifetime.

Cymer’s source development and product roadmap is closely aligned with the scanner manufacturers’ roadmap. First-generation prototype EUV sources are planned to be available in late 2008 for shipment to scanner suppliers. The technology will be improved during 2009, before exposure systems are delivered to chipmakers in 2010 for process development of next-generation devices. By the time EUVL is introduced into production, it is expected that the technology will meet all of the requirements of the commercial semiconductor capital equipment industry. Additionally, later systems for high-volume production are expected to operate at higher power than these initial prototypes. Further development and engineering of EUV sources is planned for many years to come, with progressively higher power and lower COO systems being delivered to support the roadmaps of both the scanner manufacturers and chipmakers through the next several device nodes.

Contact: Nigel Farrar, Cymer Inc., 17075 Thornmint Ct., San Diego, CA 92127, Tel: 858-384-5527; e-mail: nfarrar@cymer.com