Producing long-lifetime optics for cutting-edge microlithography at 157 nm has presented a major challenge to optics manufacturers. Keeping contamination and surface damage under control is key. When microlithography at 157 nm was first being investigated several years ago, optics that could survive 10 million laser pulses were the benchmark of performance. Today, pulse-count lifetimes of 40 billion pulses represent the state of the art. To achieve this, virtually every aspect of the traditional fabrication of deep-UV optics, including coating, testing and packaging processes that are used for longer-wavelength components, has had to be reviewed and refined. Progress in each of these areas has resulted in major improvements in optics lifetime and performance. The technology developed for 157-nm optics has also been used to produce better performance at other deep-UV wavelengths, such as 193, 248 and 266 nm. But producing these optics is no mean task; many practical aspects must be taken into consideration. Component cleaning The most significant issues confronted when shaping and polishing 157-nm optics are related to contamination, tolerances and materials. Contamination is of particular concern because most of the substances traditionally used for processing optics, such as polishing compounds and cleaning solvents, absorb strongly at that wavelength. Also, many common environmental contaminants (human perspiration, machine oil, airborne particulates from other polishing operations) absorb at 157 nm, even when present in trace amounts. Because avoiding contamination is imperative, surfaces must be cleaned frequently, and the cleaning process must be carefully monitored so that it does not leave its own residual contamination. The types of solvents used, their purity and handling must all be controlled. Acetone is probably the most commonly used, but combining high-purity acetone with spectroscopic-grade methanol, or using methanol alone, is usually a better solution for cleaning deep-UV optics. Other nonchemical surface-cleaning technologies originally developed for the semiconductor industry can be employed as well. Grinding and polishing Tolerances are a problem at 157 nm, partly because the wavelength is so short. A λ/10 flatness specification at 157 nm, typical for microlithography components, translates into a surface that is more than four times as flat as the same specification at 633 nm. The difficulty in achieving this flatness is compounded when working with crystalline materials, such as CaF2, which are essentially the only option for transmissive optics at 157 nm. Excellent surface quality (scratch-dig) and minimal defects are of paramount importance in achieving high-throughput, deep-UV optics. A typical scratch-dig specification for 157-nm optics is 20-10 or better. Good surface quality also delivers higher damage threshold, even at longer wavelengths. However, CaF2 is anisotropic, hygroscopic and very prone to chipping and fracturing. During polishing, it is not unusual for small particles to break loose from the part edge and be dragged over the component surface, causing scratches. As a result, achieving a given surface quality specification usually requires 50 to 100 percent longer polishing times for CaF2 than for fused silica. But a longer polishing time typically makes it more difficult to achieve a high-flatness (or surface figure for spherical or cylindrical parts) specification. Because of this trade-off, it is substantially more difficult to consistently fabricate tightly toleranced CaF2 components on a production basis as compared with harder, noncrystalline materials. In addition, the process requires a more careful balance between the various process parameters (polishing time, spindle speed, spindle pressure, etc.) to simultaneously meet both shape and surface quality specifications. These working characteristics of softer, crystalline materials necessitate modifications to the standard polishing approach. For example, the grinding process for CaF2 requires more steps and must be performed to a finer level than is typical with harder materials such as glass or fused silica. The traditional grinding and polishing tools require some modification. At Alpine Research Optics, we use a softer pitch lap for CaF2, and cut it differently to enable heavier particles to be removed faster, thus avoiding scratches. It is also important to avoid thermal shock during processing; for example, not putting cold solvents on parts that have been warmed by the polishing process. The need to scrupulously avoid any contamination also requires a change in the type of polishing compounds used. Ceria (CeO2), which is a commonly used abrasive for optics polishing, absorbs strongly at short wavelengths. It is virtually impossible to eliminate all of the residual polishing compound from an optical surface because some of it becomes embedded under the surface during processing. Therefore, alternative abrasives that have lower absorption at 157 nm must be used. We have found success with a diamond-based abrasive mixed with deionized water. The mixture is added manually during the polishing process and used only once, rather than recirculated in a slurry, as in conventional optics polishing. However, even diamond-based abrasives can contain impurities that absorb in the deep-UV. We have found it necessary to qualify our abrasives vendors because not every supplier can consistently deliver products with the required purity level. It is also important to avoid subsurface damage in the production of low-absorption, high-damage-threshold, deep-UV optics. Subsurface damage, a phenomenon first identified by researchers at Lawrence Livermore National Laboratory in Livermore, Calif., consists of fractures and scratches that occur during the grinding and polishing process and that become partially or totally covered by the polishing redeposition layer (Figure 1). This is a thin layer of material that flows while the material is being worked and that covers the surface. It is through this mechanism that residual polishing compound may become incorporated into the surface layer itself or deposited in microfractures and surface defects. Figure 1. Subsurface damage causes fractures and scratches that occur during the grinding and polishing process to become partially or totally covered by the polishing redeposition layer. Using a sequence of successively finer grinding and polishing steps minimizes subsurface scratching and ensures that each step removes sufficient material to eliminate any damage caused by the previous step. When working with soft materials, it also may be necessary to use lighter-weight grinding tools, as opposed to traditional cast-iron tools, because heavier implements have a greater tendency to create small fractures in the material. The need to avoid contamination also means that any deep-UV fabrication equipment should be devoted solely to that purpose. Furthermore, the production area should be environmentally isolated to prevent problems from airborne particulates produced by other polishing operations. Thin-film coating There are many potential contamination sources in the optical thin-film coating process. As in all traditional coating operations, parts must be cleaned first. Once the deep-UV optics are in the coating chamber, additional surface cleaning takes place. Because the chamber itself can be a source of contaminants, it must be made hydrocarbon-free by using oil-free, dry mechanical pumps and cryopumps. The coating process involves heating substrates to a high temperature, which also heats the interior of the coating chamber and can cause the residue of any materials from previous coatings to evaporate and redeposit on the optics. Obviously, regular cleaning is necessary, but it is best to avoid applying other coatings that would leave difficult-to-remove residues. It may not be practical for a manufacturer to devote a coating chamber solely to deep-UV coatings. However, using the same chamber for production of both deep-UV and infrared coatings must be avoided. Another potential source of impurities is the coating material. Rather than assaying the chemical composition of all incoming coating materials, the most practical procedure is to work with vendors to achieve a consistent supply of high-purity materials. The tiny number of coating materials available for use at 157 nm places limitations on both the design and manufacture of thin films. Only fluoride materials are transmissive at this wavelength, and these are limited to delivering coating layers over a relatively narrow refractive index range. For a typical λ/4 stack, high-reflection coating, the coating bandwidth is directly related to the index ratio between the high- and low-index layers. Thus, a smaller index split results in a narrower-bandwidth coating (Figure 2). Figure 2. The reflectivity of high-reflection coatings for 157 and 248 nm are plotted to nearly the same scale, in terms of percentage. The 248-nm coating has a bandwidth of approximately ±6 percent, while the 157 nm has a bandwidth of only about ±1 percent. This makes it more difficult to reliably produce a coating with a narrower bandwidth. This means that layer thickness must be more closely controlled during deposition, because to meet specification, a narrow-bandwidth coating must be accurately centered on its operating wavelength. However, the difficulty in achieving this accuracy is compounded by the fact that the layers themselves are very thin: A λ/4-thick layer at 633 nm is four times thicker than a λ/4 layer at 157 nm. Therefore, achieving a particular percentage accuracy in the layer thickness requires four times greater accuracy. Figure 3. Because oxygen absorbs deep-UV radiation, performance testing must be performed in a pure, dry nitrogen atmosphere. This deep-UV spectrophotometer measures the transmission characteristics of coated parts. The use of only fluoride materials also affects the mechanical characteristics of the coatings. High-reflection coatings for longer wavelengths usually alternate materials such as HfO2 and SiO2, which produce layers under tensive and compressive stress, respectively. By alternating layers with opposite stress characteristics, the net result is a thin film that is in mechanical equilibrium. In contrast, a fluoride-only material coating will be tensive and could be prone to crazing. This limits the number of layers that can be deposited, which again reduces coating bandwidth and peak reflectivity. It is possible to minimize coating stress by performing deposition at the highest-possible temperature because this produces denser, more mechanically stable layers. However, when such a coating is cooled to room temperature, any mismatch in thermal expansion coefficient between the coating layers and substrate can introduce stress, making the film mechanically unstable. This trade-off, which must be optimized through practical experience, involves combining accurate process data logging for each coating run with actual lifetime results from end users. Participation in independent testing, such as that performed in the Massachusetts Institute of Technology’s Sematech program, is an important component in supplying feedback for this process. Packaging For deep-UV optics to maintain their performance after delivery to a customer, the same care in avoiding contamination must extend to their final destination. For 157-nm optics, traditional part packaging, which utilizes lens tissue, plastic bags and foam packing material, may not be adequate. All packaging materials should be inert and not outgas. The latter is important because the product may be exposed to extremes in temperature and even pressure (e.g., in an airplane hold) during shipping. Part containers made of inert materials are available that securely hold the part by the corners or outside of the clear aperture, eliminating the need for wrapping and packing material (Figure 4). Figure 4. Modern packaging methods for deep-UV optics eliminate the use of wrapping and packing materials that may outgas. Further refinements in materials and ongoing process improvements will continue to push this performance forward. The results will benefit microlithography users as well as applications at other deep-UV wavelengths in biomedicine, electronics packaging and industrial micromachining. Meet the authors David Collier is president of Alpine Research Optics in Boulder, Colo. Wayne Pantley is the company’s sales manager.