MICROELECTRONICS PROCESSING - Lithography at 157 nm gains momentum
Much has been written over the years about the imminent demise of optical lithography. At one time, the technology was thought to be limited to the forming of integrated-circuit (IC) features of 1 µm and larger. Now, however, semiconductor-fabrication facilities are making IC chips with features of 0.18 µm in production quantities using optical lithography. This steady progress toward smaller linewidths has been made possible through the continuous improvement of all portions of the
John Wallace, Senior Editor
Much has been written over the years about the imminent demise of optical lithography. At one time, the technology was thought to be limited to the forming of integrated-circuit (IC) features of 1 µm and larger. Now, however, semiconductor-fabrication facilities are making IC chips with features of 0.18 µm in production quantities using optical lithography. This steady progress toward smaller linewidths has been made possible through the continuous improvement of all portions of the IC manufacturing process-from photoresists to wafer-alignment techniques to metrology equipment. But advances in optics and light sources have been key to maintaining the IC industry's leading edge.
It is generally agreed that current optical-lithography techniques will reach a limit. For this reason, other forms of lithography are being developed, including those based on extreme-ultraviolet (EUV) radiation, SCALPEL (scattering with angular limitation projection electron-beam lithography), x-rays, or ion beams. However, even as the IC manufacturers consortium Sematech (Austin, TX) chose EUV and SCALPEL as the candidate technologies most likely to carry IC chip makers past the 100-nm hurdle in feature size, optical lithography has continued to strain against its limits. According to Sematech's Rich Harbison, it now appears that optical lithography based on the 157-nm wavelength emitted by the fluorine (F2) excimer laser will allow the production of feature sizes down to 70 nm. To make this possible, scientists and engineers are tackling the considerable problems related to 157-nm technology.
Excimer defines wavelength
The design wavelength of a lithographic projection lens is defined by its light source. A decade ago, this source was the mercury arc lamp, with emission lines at 365, 405, and 436 nm. As IC feature sizes pushed down below 0.5 µm and shorter wavelengths became necessary to create them, the excimer laser became the definitive source for leading-edge lithography-first the krypton fluoride laser at 248 nm and then the argon fluoride laser at 193 nm. This relationship is so strong that nonexcimer lasers intended for lithography must be designed to emit at excimer wavelengths (see Laser Focus World, Feb. 1998, p. 135). The 157-nm-emitting F2 laser is simply the latest in this series of excimers.
Such a laser is commercially available from Lambda Physik Inc. (Ft. Lauderdale, FL). Heinrich Endert, international marketing manager, explains that a lithographic laser is not simply a scientific laser with another name. "It must meet a coordinated set of standards developed by Sematech," he says. The commercial laser emits 10-mJ pulses at a 600-Hz repetition rate, with a 1-kHz rate available soon. A similar 1-kHz experimental laser produces an average power of 20 W. As for difficulties related to the 157-nm wavelength, Endert says that "coming up with energy monitors is a challenge."
Researchers at Cymer Inc. (San Diego, CA), a company whose sole product is lithographic excimer lasers, have developed a 20-W F2 laser that operates at a 2-kHz repetition rate, with an estimated gas lifetime of 25 million pulses using periodic F2 injections. The laser operates either in a broadband mode, emitting at two transition lines, or in a lower-power single-line mode with a full-width-half-maximum line width of 1.14 pm. Although this linewidth is narrow enough for catadioptric projection lenses, a fully refractive lens would require a laser with line narrowing to a width of 0.2 pm, due to the chromatic dispersion of 157-nm optics. "There is a lot of interest in the potential for line narrowing," says Gerry Blumenstock, 157-nm program manager at Cymer. "For the [wafer-stepper and scanner] manufacturers already using refractive optics, the infrastructure is in place to build all-refractive systems." A commercial version of the laser is in development, says Blumenstock.
Calcium fluoride predominates
Just as optics intended for 193-nm lithog raphy are made primarily from fused silica, optics for 157-nm lithography are based almost entirely on calcium fluoride (CaF2). Excimer lasers, with their high-energy pulses, are hard on optical materials and coatings, with shorter wavelengths tending to cause more damage. With funding from Sematech, a group at Massachusetts Institute of Technology Lincoln Laboratory (Lexington, MA) is subjecting ultraviolet (UV) optics and coatings to excimer-laser radiation over long periods of time (see Fig. 1). Headed by researchers Mordechai
Rothschild and Vladimir Liberman, the group tests optics submitted by wafer-stepper and scanner manufacturers as well as by optical-element and coating producers.
"We obtain samples from suppliers, test lifetime, then share the data with the supplier," says Rothschild. The bulk materials tested by the group for 157-nm use fall into two areas: CaF2 intended for lenses and windows and material intended for photomask substrates. Selected grades of CaF2 have survived 300 million pulses, Rothschild notes, although he cautions that lithographic optics will see 10 billion pulses a year in operation.
A recent development appears to make the construction of 157-nm photo masks much more practical. At 193 nm, fused silica-hard, stain-resistant, and with a low thermal-expansion coefficient-is the standard photomask substrate. The initial thinking was that CaF2 was the only available material for a 157-nm photomask. But a photomask must remain extremely dimensionally stable while a wafer is being lithographically exposed, or else the overlay of each layer of features patterned on the wafer relative to the previous layer is lost. Absorptive heating of CaF2 during exposure, combined with the material's high thermal-expansion coefficient, is likely to cause unacceptably large image-placement errors. Now, re searchers at Corning Inc. (Corning, NY), Nikon Corp. (Tokyo, Japan), and elsewhere have modified the standard ultraviolet-grade fused silica to be transparent at 157 nm by lowering the OH- content and doping with fluorine.
The Lincoln Laboratory group is testing these materials. "Results look quite encouraging," says Rothschild. "Their transmission appears to be high enough for photomasks, although not for lenses."
In use, lithographic photo masks are usually protected from contamination by a pellicle permanently mounted 6-10 mm away from the patterned side. But the thin fluorocarbon pellicles developed for 193-nm use are not transparent enough at 157 nm, according to Rothschild. "However, there are recent reports that newer materials have 157-nm transmission in the 90% range," he says. "This still needs to be confirmed, and it still may be too low, but the direction is definitely positive." Plans at Lincoln Laboratory include the testing of pellicles.
Alpine Research Optics (ARO; Boulder, CO), a producer of UV optical coatings, will be working with the Lincoln Laboratory researchers in the next round of tests on coatings. David Collier, president of ARO, describes some of the difficulties faced by the coating industry when working at 157 nm versus 193 nm. "There is a very limited pool of available coating materials that are not too absorbing. Their spread in refractive index is small, so you need more layers. And the layers are very thin and can crack."
Acton Research Corp. (ARC; Acton, MA) will also be participating. In 157-nm tests at Lincoln Laboratory last year, an ARC antireflection coating withstood 50 million pulses at 3 mJ/cm2 without damage, while the company's high-reflection coating survived 200 million pulses at 5 mJ/cm2 (see Fig. 2). Ultimately, such coatings must endure billions of pulses, says Michael Case, vice president of research and de velopment. He notes that one priority of ARC is to independently life-test each material that goes into a multilayer coating.
Because 157-nm light is absorbed by oxygen, carbon dioxide, and water vapor, the entire optical path of a 157-nm projection system must be purged with nitrogen (at 193 nm, gas purging-though often desirable-is not always necessary). Another purpose for a gas purge, according to Rothschild, is the elimination of hydrocarbons that can contaminate optical surfaces. In fact, a new optical element that is adequately purged will show a rapid initial rise in transmission due to laser cleaning of pre-existing surface contaminants, he explains. Purging of the optical path becomes especially tricky at the photomask and wafer, both of which must be removable and therefore cannot be sealed into the system.
The Lincoln Laboratory group has built a 157-nm "microstepper" to pursue development of photoresists. Containing a commercially available Schwarzschild reflecting microscope objective that serves as a projection lens, the wafer stepper has been imaging 80-nm features in photoresist for two years. The 0.5 numerical aperture of the lens prevents the stepper from achieving smaller linewidths, says Rothschild. The researchers want to use the stepper to determine which photoresist polymers are most effective; criteria include sensitivity and image contrast.
IC industry interested
Wafer-stepper and scanner manufacturers, though all interested in 157-nm lithog raphy, have publicly indicated varying degrees of commitment. Perhaps the most aggressive pursuer of the technology is Silicon Valley Group (SVG) Lithography Inc. (Wilton, CT). The company currently produces a wafer step-and-scan exposure system that includes a catadioptric lens built around a large beamsplitting cube, a design that shifts much of the optical power to a reflective surface to allow a relatively large bandwidth. The lens is already incorporated into 248- and 193-nm versions of the scanner.
FIGURE 3. Polished CaF2 surface has an irregularity of 30 Å root-mean-square (rms) and a roughness of <1 Å rms over a 40-mm clear aperture, as seen in this Twyman-Green surface interferogram.
In describing SVG's commitment to 157-nm lithography, Jim McClay, vice president of the company's 157-nm program, says that two questions had to be answered before engineering work could begin in earnest. "Could we get a qualified light source, and could we get CaF2 of high-enough quality for our beamsplitting cube-these were what we had to find out," he says. With a laser from Lambda Physik fulfilling the first requirement and improvements in CaF2 manufacture satisfying the second, SVG is now planning to introduce a full-field scanner by the end of 2001. Its 5 x 26-mm lens field will be scanned to produce an exposure field of 26 x 34 mm.
To enable development of 157-nm photoresists and wafer-process techniques, SVG plans to introduce what it calls a "mini-scanner" by the fourth quarter of 2000. The scanner will incorporate a CaF2 lens with an on-axis broadband catadioptric design that is being designed and built by Tropel Corp. (Fairport, NY).
Tropel has produced a similar lens based on fused silica that is now used in several locations for 193-nm photoresist development work (see Laser Focus World, Feb. 1997, p. 75). Tropel's 157-nm lens will have a numerical aperture of 0.75 and a 6-mm field diameter, with a 4-mm-square portion to be used in the SVG machine. Scanning of the lens field will create a 4 x 22-mm exposure field.
Jim Webb, lens-design manager at Tropel, describes CaF2 as "not at all easy to work with." As well as being much softer than fused silica, CaF2 is a crystal and thus has anisotropic polishing properties, he explains. Proper orientation of the crystal during lens polishing is important, with best results obtained when the optical axis of the lens element coincides with one of the crystal's axes of symmetry. In addition, the very large thermal-expansion coefficient of CaF2 makes tight thermal control necessary during polishing. With care, though, high surface quality can be achieved (see Fig. 3).
Along with the 157-nm lens for SVG, Tropel will supply an illuminator containing a light integrator. Consisting of thin mirrored strips combined to form a long, hollow rod and intended to create uniform illumination across the lens field, the integrator presents another challenge, according to Webb. "The gas purge is much more difficult at 157 nm than at 193 nm," he says. "The integrator must be purged for much longer with a high flow of clean gas."
The major feature of 157-nm lithography that makes it so attractive to IC equipment manufacturers, says McClay, is that its development is one of "evolution, not revolution." The light sources, the overall optical and mechanical form of the system, and even the lens-design details are similar to existing excimer-laser-based steppers and scanners. But it is a fast-moving technology. "New developments today are old news tomorrow," says McClay.