Optical Lithography Lives On

The long-anticipated demise of optical lithography was postponed again this week. A group of researchers from the University of Texas, Austin, in work presented at the SPIE Microlithography conference, printed 80 nm (0.08 micron) lines and spaces using 193 nm light. The announcement challenges the conventional view, that non-optical exposure methods such as x-rays or electron beams will be needed for features below 100 nm.

Leading edge production currently prints 180 - 220 nm features using 248 nm light, the wavelength of a KrF excimer laser. Since the features are smaller than the wavelength, diffraction effects distort the image. Large lens apertures (NA) help, but reduce the depth of focus. Lithographers currently use a combination of off-axis illumination, optical proximity correction, and phase shift masks to overcome diffraction effects. With off-axis illumination, the incident light strikes the wafer plane at an angle.

Resolution improves, but less light reaches the wafer. Optical proximity correction (OPC) adds sub-resolution features to the mask to correct for printing and etching errors. Phase-shift masks (PSM) use destructive interference between adjacent features to cancel diffraction effects. Advanced mask designs require more complex fabrication, and closer cooperation between designers and manufacturing. For instance, conventional inspection tools will mark OPC features as defects.

Resolution enhancement techniques can't repeal the laws of physics, either. At some point, probably around 150 nm or 130 nm, a new exposure wavelength will be required. The 1997 Semiconductor Industry Association (SIA) National Technology Roadmap for Semiconductors predicts that 150 nm features will be in production in 2001.

Most observers view 193 nm as the leading candidate to succeed 248 nm. Unfortunately, conventional lens and resist materials absorb strongly at this wavelength. Equipment makers will need to consider reflective and catadioptic (both reflective and refractive elements) designs instead of purely refractive lenses. Resist chemists will have to find an alternative to the DNQ/novolak formulations in use since 1972. New materials and new designs will require huge research and development expenditures.

Worse, 150 or 130 nm features will still be smaller than a 193 nm exposure wavelength. Resolution enhancement techniques will still be necessary. According to the conventional wisdom, these techniques will be insufficient beyond 100 nm, which the SIA Roadmap predicts will reach production in 2006. Thus, the industry would have to recoup its investment in 193 nm in a single device generation, and would then have to make another enormous investment in some other technology.

This week's announcement casts a ray of hope on this grim scenario. Grant Willson, Kurt Patterson, and coworkers appear to have identified a resist formulation that will print features down to 80 nm. In fact, Willson said, "it appears that the process latitude is there to generate smaller features yet."

The resist, based on an alternating cyclo-olefin polymer (Fig. 1), uses a triphenylsulfonium nonaflate photoacid generator, with added base inhibitors and dissolution promoters. According to Larry Thompson, chief technical officer at Integrated Solutions Inc.(ISI), the material has good etch resistance and is relatively easy to synthesize. Thompson told Semiconductor Online that this development could make 193 nm lithography last more than one generation, and could avoid the need for a 0.7 NA lens.

Figure 1

While the resist is an important breakthrough, substantial work remains before 193 nm lithography can be used in production at all, much less for such small features. The researchers used Sematech's ISI ArF MicroStep (Fig. 2), a small-field, high numerical aperture system designed for resist development. The optical design is not scalable to a full field production stepper, Thompson told Semiconductor Online.

Figure 2

Optics are the principal difficulty for 193 nm exposure tools. The ArF laser has inherently broadband emission, and relatively low power (10 -12 watts), Thompson said. Line narrowing the beam to eliminate chromatic aberration, a normal procedure in 248 nm steppers, would throw away too much of the already limited power. Instead, the lens design must correct for chromatic aberration. Only two known materials, fused silica and CaF, can be used for 193 nm lenses at all, since other materials absorb strongly at that wavelength. According to Thompson CaF elements will be needed to correct chromatic aberration, and CaF simply is not available with high enough purity in large enough pieces for stepper lenses.

Second, because ArF lasers have such low power, throughput is not likely to exceed 30-40 wafers per hour (WPH). For comparison, ASML recently announced an i-line (365 nm) step and scan tool with 96 WPH throughput (See "I-Line Step & Scan" in Semiconductor Online's Product Showcase), and 248 nm tools delivermore than 80 WPH. Low throughput would dramatically increase the cost of ownership for 193 nm lithography.

Exposure masks are also critical to any technology that requires resolution enhancement. The University of Texas group used an alternating aperture PSM (See fig 3.), supplied by Dupont Photomasks. While this design, also known as a Levenson PSM, provides strong phase correction, it is best suited to periodic features. Unwanted extra phase-shift lines in non-periodic features print as dark image areas.

Figure 3

These masks are also difficult to fabricate. Gil Sheldon, Dupont Photomasks' director of engineering, explained that PSMs are the first masks to require second level alignment. Conventional chrome masks only have one layer, and only need to align that layer to the edge of the mask. In a PSM mask, the phase features are etched into the quartz blank, and must be placed to within 100 nm. Moreover, the quartz etch must be controlled to within
< 5° of phase, or ± 3%. Even though the MicroStep uses a 10x reticle, rather than the conventional 4x, Sheldon told Semiconductor Online the required feature sizes were comparable to the most advanced masks Dupont is making.

According to Sheldon, defects are the most serious problem with phase shift masks. No tools are available to inspect masks at the exposure wavelength, so it's very difficult to identify phase defects. The etch process prevents 180° phase errors, but smaller errors, while less serious, are more difficult to detect.

In summary, rumors of the death of optical lithography have been greatly exaggerated. Nonetheless, the patient remains on life support. As Thompson pointed out, non-optical technologies such as SCALPEL become more viable as the cost of ownership of optical lithography increases.

By Katherine Derbyshire

For more information:

University of Texas, Austin, www.utexas.edu/coe/
SPIE, www.spie.org
Semiconductor Industry Association, www.sia.org