193-nm litho conference challenged by insertion

12/01/1998

193-nm litho conference challenged by insertion

With likely insertion at the 130-nm technology node in 2003, details at the International Symposium on 193-nm Lithography showed that lithographers still face the challenges of simultaneously inserting resolution enhancement technologies, new resists, lasers, and optics.

ASML`s new PAS 5500/900 193-nm exposure, 0.62-NA full-field step-and-scan system, as revealed at the symposium, is capable of 11-nm critical dimension (CD) uniformity at 160 nm with single-machine overlay of 31 nm. Using off-axis illumination and a conventional mask, 1:1 line-space patterns were printed at 110 nm in Olin bilayer resist and isolated lines were printed over the full field down to 60 nm using a strong phase-shifting mask. Similarly, Nikon revealed lens performance showing line-space patterns with 150-nm CDs with a full field >300 nm depth of field.

Cost of ownership of 193 nm, however, is still 2.5-3 times that of a DUV scanner. Present ArF lasers produce ~5 W, many times the performance of a few years ago, but much less than that of KrF lasers used today. Symposium reports by laser suppliers, particularly Cymer, described systems with >5 W of sustained output and <12% pulse-to-pulse variation. Pulse rates are moving upwards beyond 1 kHz and the total cost of operation is moving below $0.10/1000 pulses.

The more complex optical systems transmit less illumination, requiring fast (10 mJ/cm2) resist to achieve 60 wafer/hr throughput. Olin`s bilayer resist requires 23 mJ/cm2 to print at 160 nm, reducing throughput to a still-substantial 46 wafers/hr.

Symposium papers seemed to indicate that the peculiar challenges of 193-nm production are straining the creativity of resist chemists. "With CDs less than a wavelength, the projected image quality and chemical nonuniformities of the thin resist films seem to cause edge roughness for the developed structures of all resist types," said Susan Palmateer of Lincoln Labs. "No one knows the exact effect of this roughness on circuits, but it is hard to believe it is entirely benign."

A thin-layer resist stack consisting of plasma-polymerized methyl silane on a baked novalac planarizing layer allows all-dry processing, but, so far, lacks sufficient photospeed. More conventional Olin bilayer resist has now been optimized for 50-nm linewidths; the lesser molecular weight of the photo-chemically active layer in this type of resist may reduce edge roughness.

Cyclic olefin polymers have proved dramatically better than acrylics for oxide etch applications, according to Bob Allen of IBM, and can be manufactured inexpensively with transition metal catalysts, as reported by Saikumar Jayaraman of B.F. Goodrich. Also, the post-development CARL silylation process may allow shifting the process window after exposure to position the isofocal exposure at the desired CD, according to Stefan Hien of Siemens, AG. Very small contact holes may be produced using CARL without edge roughness as the glass transition temperature of the silylated material is below room temperature.

The 20-million guilder projected price for the ASML system reflects the enormous cost of high-quality fused silica and calcium fluoride optical materials required in 193-nm projection lenses and illuminators; progress has been slow in obtaining the necessary optical quality and stability. One presentation revealed a new version of Suprasil 1 quartz that provides low initial light absorption and avoids sudden (high) absorption transition, but uncertainty remains about damage after 1011 laser shots. A report from Intel described a transient ~1.5% absorption appearing in 25 cm of quartz in the first few minutes of UV exposure, disappearing after the light is turned off; this new effect will likely require real-time compensation to maintain a nominal exposure dose. One supplier has calcium fluoride lens blanks up to 233 mm in diameter with homogeneity <2 ppm, roughly twice that required of top optical material, but radiation stability and economic viability are yet to be proven. Some relief may come with laser management; a new silica compaction formula takes laser pulse lengths into account, indicating that long-pulse lasers increase the lifetime of quartz lenses proportionally to the square-root of the pulse length and rep-rate increase.

John Wiesner of Nikon Precision noted a disconnect between what lithography developers at the symposium believe the new technology will cost and what ultimate users say it must cost. "For insertion, costs/wafer will have to be lower than DUV, not 3? higher. Until this technology achieves that, it will remain an R&D topic," he said. - M.D.L.