Fulfilling the manufacturing potential of immersion lithography requires tool designs that deal successfully with flare, lens heating, overlay, throughput and defectivity issues.

S. Owa, A. Hazelton, Nikon Corp., Tokyo, Japan, H. Magoon, Nikon Precision Inc., Essex Junction, Vermont

Immersion technology has been used for many years to improve resolution in microscopy. Applying it in mainstream semiconductor patterning, however, requires more in-depth characterization and development. Although immersion lithography provides enhanced resolution and expanded depth of focus (DOF), the technology is not justifiable if the overlay accuracy, productivity, or defectivity level are sacrificed. For immersion lithography to be successful in high volume manufacturing, it must deliver superior imaging performance without negatively impacting yield.

Imaging

The main motivation for the transition to immersion lithography is to enhance image resolution by increasing lens numerical aperture (NA). Simulations have shown [1] that a numerical aperture of 1.30 is required to achieve sufficient DOF for 45nm line/space patterns in high volume manufacturing, with vertical and horizontal features imaged simultaneously. Litho scanner suppliers have been moving quickly in this area, and designs for such lenses are already completed.

While all-refractive optics have been employed in the past due to their simple structure and minimal flare, their practical NA is limited by the lens diameter that can be manufactured. To further increase NA to meet 45nm requirements, catadioptric lens designs, employing both refractive lenses and mirrors, are required.

Catadioptric lenses are far more challenging to design than fully refractive systems because manufacturing imperfections, such as the surface roughness of the mirrors, inherently contributes to local flare, that is, unwanted background resist exposure. Generally speaking, the reflection at an imperfect mirror surface causes about 13 times as much light scattering as a similarly imperfect lens surface [2]. One strategy for limiting the amount of local flare is to limit the number of mirror surfaces in the optical system.

In addition, the surface roughness and mid-spatial frequency roughness of the lens elements, particularly the mirror surfaces, must be carefully controlled. Techniques for polishing spherical and flat surfaces have matured over the past several hundred years. Because these techniques use large area pads, they have an averaging effect and tend to result in very good surface finishes. Polishing techniques for aspherical surfaces, on the other hand, use small pads and result in local distortions of the surface. Immersion scanner manufacturers are able to minimize flare and deliver low aberration levels (see table) by reducing the use of aspheric mirror elements. In addition, advanced and proprietary polishing techniques, such as those learned through EUVL development, greatly reduce the surface roughness that leads to local flare.

Several design options were considered for the catadioptric lens. These options can be separated into two main groups: uni-axial [2, 3], where the optical elements are in a single barrel, generally aligned with the lens optical axis, and multi-axial [2, 3], where some optical elements are in a separate barrel, with an axis not aligned to the lens optical axis. The main disadvantage of a uni-axial design is the optical elements need to be carefully arranged to prevent obscuring the optical path. Because larger NAs require wider collection angles of the lens elements, obscuration tends to be a significant issue for high-NA lenses. To compensate, uni-axial designs usually require fairly compact light bundles within the lens, leading to concentrated heating of lens elements and resulting in thermal aberrations. Uni-axial lenses also require aspherical elements to reach the 1.30 NA required for 45nm production, leading to excess lens flare as discussed previously.

For these reasons, Nikon selected a multi-axial catadioptric lens, as in Fig. 1a. This catadioptric lens design is able to deliver the required 1.30 NA without the use of aspherical mirrors, enabling minimized aberrations and reduced local flare. The multi-axis configuration is also preferable due to better chromatic aberration control which, in turn, can reduce the effects of finite laser bandwidth. NSR-S610C data has demonstrated chromatic error ~50% less than dry ArF systems, with laser bandwidth requirements relaxed by 25% [4]. The NSR-S610C design incorporates sophisticated infrared aberration control (IAC) to reduce thermal aberrations. In addition, a three-mirror multi-axial design can also deliver transmission efficiency >77%, while requiring less laser power (60W), and minimizing the image height from the lens axis to control thermal and field aberrations.

Polarization

Polarized illumination is required for hyper-NA (NA >1.0) lithography [5], due to its significant influence on image contrast, which affects line-edge roughness (LER) and the mask error factor (MEF), as well as process latitude. In the case of catadioptric lens designs, there is the added concern that polarization error may be induced by the folding mirror as a result of its wide and very high ray reflection angle [6]. Such polarization issues have been avoided through development of a new coating technology which enables more precise control of the coating’s reflection characteristics for polarized light over a wide range of angles. Figure 2 shows how the NSR-S610C folding mirror system self-compensates for variation in the angle of incidence, eliminating any concerns about polarization errors at the mirror surfaces.

In fact, the ratio of specific polarization (RSP) degradation where RSPH = IH/(IH + IV) and RSPV = IV/(IH + IV) [1, 7] caused by the folding mirror is even smaller than that due to birefringence of dioptric lenses [6]: >0.995 for the folding mirror vs. 0.985 for the conventional lens used in the NSR-S609B. The 1× reflective image relay, such as in the NSR-S610C, enables a self compensating setup to minimize aberrations, apodization, and diattenuation. Such a carefully balanced polarization design delivers RSPs greater than 0.99-for both the pupil center and edge. Further, the symmetry of a multi-axial catadioptric design preserves the illumination polarization through the scanner lens, enabling all of the advantages of such mature polarization technologies as Nikon POLANO which maintains the selected polarization throughout the illumination system.

Defectivity

Aside from the design and manufacturing of hyper-NA lenses, defectivity was considered the greatest potential challenge with immersion lithography. Introducing a layer of water between bottom lens element and wafer has the potential to wreak havoc with defectivity levels. Initial defectivity concerns were focused on the possible presence or formation of microbubbles during the scanning and exposure process. Initial results from some early immersion scanners showed thousands of bubble-related defects. Watermarks, caused by the drying of droplets leaked from the immersion nozzle, and particles drawn onto the wafer surface by the water film were also significant sources of defects.

Understanding that bubbles and leaked water droplets need to be avoided at all costs, Nikon developed the Local Fill Technology nozzle. Its design (see Fig. 3) relies on surface tension to contain the immersion water. The combination of this nozzle design, high water flow rate (~300 ml/min), and airless fluid handling enable immersion scanners to deliver defect levels comparable to dry manufacturing, without impacting throughput. The systems are also fully compatible with a wide range of ArF resists and topcoats (developer soluble and developer nonsoluble), allowing wafers to be processed at maximum throughput with no water spots or backside wafer contamination.

Nikon Local Fill technology has performed free of immersion-specific defects since its introduction [9] on the Nikon Engineering Evaluation Tool (EET) in 2004. Resist topcoats, however, are needed to protect current ArF photoresists from the immersion fluid. With solvent-soluble topcoats, Local Fill produces as few as six defects per 300mm wafer at 400mm/sec scan speed (Fig. 4a), with none of them characteristic of immersion. Wafers with developer-soluble topcoats typically show nine or so defects, again of types similar to those seen in dry lithography (Fig. 4b).

Overlay accuracy

Overlay accuracy also emerged as a potential showstopper for immersion lithography. Although there are many sources of overlay errors in the immersion process, evaporative cooling of the immersion liquid and the resulting cooling of the wafer itself emerged as the key immersion-related overlay detractor. When evaporative cooling occurs, the resulting temperature change causes the wafer to shrink. Since dry exposures are not subjected to this temperature variation, the result is a significant overlay mismatch between layers exposed with immersion and those exposed on dry tools [2].

Although evaporative cooling naturally occurs on the free surface of any volume of water, the presence of flowing air near the surface greatly increases the evaporation. The presence of air bearings in the nozzle or an air curtain for containment of the immersion water is particularly problematic and was successfully avoided in the Nikon Local Fill nozzle.

Another potential source of evaporative cooling is the draining and refilling of the immersion liquid itself. Draining and refilling also affects the temperature stability of the fluid and nozzle, as well as reducing the tool throughput. To prevent these effects, Nikon introduced the Tandem wafer stage. The Tandem Stage system uses two stages with separate functions: an exposure stage to process wafers, and a calibration stage to calibrate the system during wafer exchange. By keeping one of the stages under the projection lens and nozzle at all times [1], continuous water flow can be maintained, preventing evaporative cooling and providing temperature stability.

These innovations have been successful in preserving the overlay accuracy of immersion tools. From the first mix-and-match overlay measurements using the EET in 2004 to the present, the wet-to-dry overlay has not shown any issues due to the immersion process [2]. Currently, the overlay matching between NSR-S308F dry exposure tools and NSR-S609B immersion scanners is 10nm for Y and 11nm for X. This outperforms the standard 15nm specifications for dry S308F overlay matching.

Immersion productivity

The remaining potential show-stopper for high volume immersion lithography is productivity. Although resolution, defectivity, and overlay are critical for immersion technology, in the end, economics and lithography affordability drive the semiconductor industry.

Two key points to maintaining immersion productivity are not draining and refilling the immersion water, and being able to scan at speeds comparable to dry processing. The Tandem Stage system with continuous water flow solves the first issue, and the novel Local Fill Technology of the nozzle solves the second. Immersion scanners such as those using the Tandem Stage design are more than capable of delivering throughput ≥130 wafers/hour, on track with today’s leading edge dry exposure systems.

Summary

Intensive immersion scanner development, along with collaboration between IC manufacturers and resist, track, and litho scanner suppliers has enabled resolution of the main technical challenges that had been considered potential show-stoppers for immersion lithography. Leading-edge immersion scanners will deliver not only the 1.30 NA required for 45nm half-pitch production, but also will provide overlay accuracy in single digits, defectivity comparable to dry exposure tools, and maximized productivity of more than 130 wafers/hour. Although growing pains associated with this new technology are to be expected, we think immersion lithography has justified its selection as the technology of choice for 45nm half-pitch manufacturing and beyond.

References

1. M. Okumura et al., “Mass Production Level ArF Immersion Exposure Tool,” Optical Microlithography, SPIE, Vol. 6154, pp. 61541U-1 to -8, 2006.
2. S. Owa et al., “Current Status and Future Prospect of Immersion Lithography,” Optical Microlithography, SPIE, Vol. 6154, pp. 615408-1 to -12, 2006.