Open Competition 3 - Electronics and Photonics

Fundamental Algorithms for Direct Tolerancing and Illumination Optimization

Sponsor: Optical Research Associates

• Project Performance Period: 5/1/2004 - 4/30/2007
• Total project (est.): $4,299,575.00
• Requested ATP funds: $1,704,957.00

The competitive pressure to reduce the size of integrated circuits expressed in "Moore's Law" ("logic density in ICs doubles every 18 months") is pressing the limits of every aspect of semiconductor manufacturing technology, but probably none more so than optics. For photolithography processes, the smallest possible feature size or "critical dimension" (CD) is determined in part by the wavelength of the light used in the imaging and in part by the numerical aperture of the optics. Current production equipment uses 193-nanometer light in the ultraviolet, and the industry is closing fast on technology for 157-nm equipment. In addition, a new path to numerical apertures in excess of 1.0 based on "immersion" technology is also rapidly developing. Past that, hopes are pinned on extreme ultraviolet (EUV) systems that operate in the X-ray range. Critical to these plans is the ability to design and build optics that can image the features of each layer of the circuit precisely, with accurate overlay, and illuminated with enough intensity that the production process can be run efficiently. Optical Research Associates (ORA®) plans a two-pronged approach to overcoming key limitations in the current generation of software used to design and build these systems. The first challenge is to improve the tolerancing for optical systems to allow as-built RMS wavefront errors of less than 10 nm RMS, particularly for 193 immersion and 157 nm systems. Tolerancing determines how well the lens must be assembled and improvements lead to opening margins for the multitude of processing parameters in semiconductor lithography that must be controlled to produce acceptable results - it is always a trade-off of time and cost against manufacturing yield. Limitations in the models now used to calculate tolerances for optical error mean that industrial designers must limit the performance of their equipment with excessively conservative estimates to ensure reliable product yields. ORA proposes to develop novel algorithms that even more accurately calculate optical performance based directly on key process parameters such as CD uniformity and overlay accuracy. This is an extremely difficult modeling challenge because the relationships involved are complex and must be calculated both accurately and rapidly enough to support interactive simulations. The company also hopes to expand on recent research in calculating optimal compensators that are used in assembly and installation of photolithography systems to adjust for manufacturing and alignment errors. The second challenge involves developing algorithms and models for illumination systems for EUV lithography. Projection optics have had computer-aided optimization software for decades, but there are no commercial optimization algorithms for illumination systems - and control of EUV sources to achieve sufficient uniform illumination at the surface of the wafer is essential to achieve commercially viable production rates. ORA estimates that meeting industry goals will require a 200 percent improvement in collection efficiency for EUV sources and that this will require optimization tools; the company plans to develop a set of second-generation algorithms called Multifunction Approximation algorithms to address the unique challenges of noisy, inaccurate merit functions in illumination optimization. Both goals involve developing efficient algorithms to solve highly nonlinear systems of equations, an extremely difficult computational problem. A small company, ORA requires ATP support for the effort because of the high technical risk of the research. The general economic downturn in the industry made it impossible to find funding elsewhere. If successful, ORA's technology will enable significantly more accurate designs for microlithography optics, increasing product yield for 193 immersion and 157 nm systems by 10 to 20 percent and extending the practical lifetime of that technology. Accurate EUV illumination modeling in turn could result in a 10 to 25 percent improvement in throughput and wafer production rate. Extending the lifecycle of 193 immersion and 157 nm fabrication lines alone could be worth $120 million to $170 million per plant in deferred capital equipment purchases, according to ORA. If that happens for at least 20 percent of U.S. plants, the savings to the industry as a whole could be at least $1.68 billion.

For project information:

Laura Mickens, (626) 795-9101
lauram@opticalres.com

ATP Project Manager

Eric Samuelson, (301) 975-6393
eric.samuelson@nist.gov