SVG builds 157-nm bridge to EUV

Semiconductor manufacturing roadmap get reality check as 193-nm steppers go online, 157-nm systems are unveiled, and 13-nm systems continue development.

By: Katherine Derbyshire, Semiconductor Online

Dismissed as unworkable a little more than a year ago, 157-nm lithography has now emerged as the leading successor to 193 nm, according to Jim McClay, VP of Silicon Valley Group's 157-nm program. The company's roadmap, unveiled Wednesday, extends 193 nm exposure from 150-nm (0.15 µm) to 120-nm feature sizes, with 157-nm exposure entering production for 90-nm features, and extreme ultraviolet (EUV; 13 nm) required for feature sizes below 65 nm.

Both 193 nm and 157 nm tools will build on SVG's deep ultraviolet (DUV; 248 nm) Micrascan III (MS III) design. The company dominates the step-and-scan market at this wavelength, with 150 out of 204 installed units worldwide, according to company president Bill Hightower. A recent $200 million expansion program has increased manufacturing capacity at SVG's Wilton, CT facility to 16 DUV tools per month.

Design improvements
SVG's catadioptric lens design is significantly smaller than comparable refractive designs. As a result, Lithography Division VP of engineering Daniel Cote said it is less sensitive to vibrations from the scanning stage. The stage itself is magnetically levitated and rides on air bearings, transmitting less vibration to the optical path than other designs.

Both of these features appear in the company's recently-introduced Micrascan 193 (MS 193) product. The MS III and MS 193 lens designs are almost identical, relying on a beam-splitting cube to deliver polarized light to the wafer (see Figure 1). The MS 193 uses primarily fused silica optics. The elements nearest the wafer, which are exposed to the most intense radiation and therefore are most prone to compaction damage, are made from CaF₂.

High-numerical-aperture refractive designs require extremely large lens blanks, raising concerns about the availability of optical grade CaF₂ (see Calcium Fluoride Optics Rise to the Deep UV Challenge, Dateline: SPIE San Diego). The MS 193's CaF₂ elements are much smaller, McClay said. Besides consuming less material, smaller components are easier to handle and polish.

Two MS193 units are being assembled now for delivery in May and June 1999, Hightower said. The company expects to sell four units in 1999, out of a worldwide total of 16. Lambda Physik's ArF laser has been qualified for the MS193; qualification of Cymer's equivalent laser is still underway.

Optics for 157-nm systems
The next step in the planned evolution of the Micrascan platform is more controversial. At 157 nm, only CaF₂ is transmissive enough for optical elements. Though catadioptric lens designs will still be more compact than refractive designs, McClay said, CaF₂ availability and manufacturing are worrisome in any all-CaF₂ system. SVG is working with several sources to guarantee a reliable supply of CaF₂.

The beam-splitting cube at the heart of the design requires extremely precise manufacturing. The optical path lengths within the cube must be accurate to within a few nanometers. The beam-splitting layer itself is created by cutting a cube in half diagonally and polishing the exposed faces to atom-scale flatness. After deposition of a reflective coating on one face, van der Waals forces are used to "glue" the two halves back together. While SVG has not yet manufactured a splitter large enough for a full-field, 0.7 NA, 157 nm step-and-scan system, the company has made beam-splitters of this type from CaF₂ cubes several inches across.

Lambda Physik's 157-nm fluoride excimer laser is the planned light source. SVG has partnered with other supplier and customer companies for development of resists, masks, and other infrastructure. A new material for mask blanks—probably CaF₂ or F-doped SiO₂—will be needed.

On to the EUV regime
SVG's MS 193 and the forthcoming 157 nm system are evolutionary descendants of the Micrascan III. In contrast, Noreen Harned, who leads the extreme ultraviolet (EUV) effort, emphasized that EUV represents a revolutionary change. No transmissive materials are available at the 13-nm wavelength, so the optical designs must be all-reflective. The SVG optical design uses six mirrors, compared with four in the Engineering Test Stand at Sandia National Laboratory (see Deciphering Extreme Ultraviolet Lithography).

SVG has been developing EUV technology since early 1997, and signed a development agreement with the EUV LLC in January, 1998. The company is working with TRW to develop a light source with sufficient output power to meet the 40 300-mm wafers/hour throughput target. SVG expects to begin taking orders for first generation systems by the end of 1999, and to have production tools available well before the 2007 introduction date given in the SIA roadmap.