By Stephane Dana, ASML Netherlands B.V.
Significant advances have been realized in the development of 157-nm optical-lithography exposure systems as well as in the associated technology infrastructure. This time last year, the key technical issues included purging and contamination control, calcium-fluoride (CaF2) birefringence and purity issues, coating robustness, and developing a pellicle solution.1 Those issues, considered by many as roadblocks to 157-nm lithography, have now been resolved. projection optics
The projection lens of ASML's first 157-nm production tool is specified with a numerical aperture (NA) of at least 0.85 to be competitive with 193-nm systems and a field size of 26 x 33 mm to reach the imaging performance for the 65-nm node. Despite some improvements in narrowing the laser bandwidth, the lack of a glass material in addition to CaF2 to compensate for chromatic aberrations makes a catadioptric design the only possible choice for the projection optics.
Thanks to new annealing procedures, CaF2 material is now available with levels of  stress birefringence better than 1 nm/cm, and a refractive index homogeneity that reaches the performance specifications set for both  and  orientations. Nevertheless, growth yield improvement in CaF2 material and additional investments will be necessary to reduce the costs and have sufficient material quantity available for the volume production of 157-nm exposure systems.
Intrinsic birefringence, which has been confirmed at 11 nm/cm for 157 nm,2 can be efficiently compensated for in the lens design by combining and clocking lens elements of  and  crystal orientations, yielding a residual level of 0.3 nm/cm below the level of the stress birefringence. The lens design has been further optimized to minimize the total CaF2 content as well as the amount of -oriented material, due to its higher-stress birefringence and lower-yield growth, thereby improving both the high-NA imaging contrast loss performance and the cost of the lens.
Antireflection coatings have been developed by Carl Zeiss (Oberkochen, Germany) with a transmission value larger than 99% in production. Lifetime tests showed no degradation for at least 5 gigapulses of exposure at 25 mJ/cm2 energy density. The material degradation reported by the Massachusetts Institute of Technology (Cambridge, MA) was not observed, nor did coating stability require a low-water-vapor environment.3
Initial concerns of low system efficiency due to poor optical transmission are being resolved with better illumination systems, improved optical materials, and higher-power lasers. Tools for 157-nm tools should meet the throughput of previous 248-nm or 193-nm generation tools. purge and contamination control
Purging of the reticle and wafer stages has been verified on prototypes.4 The concept of local purge allows immediate access to the machine in case of service action or operator intervention. The relatively small purge compartments greatly minimize the system downtime during maintenance of the tool, with a nitrogen flow chosen to optimize both the optical performance and cost of operation of the tool.
Because optical absorption for contaminants such as oxygen, carbon dioxide, and water vapor at 157 nm is three orders of magnitude higher than at 193 nm, the maximum allowable concentration of these contaminants must be in the sub-parts-per-billion range. A tool has been developed to measure the concentration of hydrocarbon compounds in a nitrogen-purge gas flow to the level of parts-per-thousand by using thermal desorption and gas chromatography. The unit's portability and sub-30-minute response time make it ideally suited to monitor and ensure contamination- free operation in a production environment.
ASML will offer the Micrascan VII in 2003 as the industry's first 157-nm full-field scanner. Its 0.75-NA projection optics and field size of 26 mm x 34 mm will provide imaging capability for sub-100-nm structures. Early system illumination before final setup has better than 1.25% uniformity across the field, including purge.
This tool's development has been instrumental in driving the industry's CaF2 suppliers and will be used to enable resist and process development. Experience will be gained in purging and contamination monitoring, lifetime degradation of components, and optical coatings. Learning in catadioptric design, aspheric lenses manufacturing, and CaF2 bulk and surface scattering will lead to better optical performance of the production systems. pellicles and reticles
Soft pellicles remain the industry's preferred solution for a volume-manufacturing environment. Although transmission of polymer membranes has improved to better than 98%, the present lifetime only accommodates fewer than 100 wafers per reticle. This figure must be improved by more than two orders of magnitude to accommodate high-volume manufacturing; the fundamental mechanisms responsible for photochemical darkening remain to be understood.
Figure 1. Bossung plot and top-down SEM images show through focus at best energy (25 mJ/cm2) for 130-nm dense lines exposed on a PAS/750 using annular illumination (NA = 0.70, σ = 0.85/0.55; top). and through focus at best energy (32 mJ/cm2) for nominal 130-nm isolated lines overexposed to 120 nm on a PAS/750 using annular illumination NA = 0.70, σ = 0.85/0.55; bottom). All data are obtained using a reticle with an 800-µm hard pellicle.
The design of ASML's first 157-nm production system is compatible with both soft membranes (about 1-µm thick) and hard pellicles (800-µm thick). Using a 248-nm system, ASML engineers have demonstrated that the optical effects introduced by a hard pellicle are tolerable both in terms of overlay distortion and imaging, as demonstrated by the Bossung curves (see figure 1); the process windows are the same with or without hard pellicles. Errors due to the pellicle frame mounting to the reticles can be corrected within the machine.
These feasibility results are in agreement with conclusions of the thick-pellicle working group of International Sematech (Austin, TX), which defined a common set of specifications for the modified fused silica pellicle material and frame.5 Pellicle non-flatness, which, due to gravity and mounting nonuniformity causes errors in feature placement, is expected to be solved. reticles and resists
Hydrocarbon contamination affects reticle transmission and hence the critical dimension uniformity. These adsorbates can be removed using UV light in combination with oxygen in an in situ cleaning unit. Following this cleaning step, the reticles must be stored in a clean nitrogen environment. The space between reticle and pellicle, itself held with a porous pellicle frame, is purged with nitrogen in a separate module, making it possible to achieve the required oxygen and water concentration levels of less than 1 ppm without affecting the overall tool productivity.
Resist suppliers are expected to develop commercial resists based on a new high-transparency fluoropolymer with 157-nm optical densities of less than 1 µm-1 and good imaging characteristics in 150- to 250-nm-thick films on low-reflective substrates.6 Several resist developments are making significant progress in achieving a profile comparable to current production systems at a realistic film thickness and dissolution contrast.7-9 Only rudimentary process integration has been conducted so far, however, and issues with etch resistance, line-edge roughness, and excessive amine sensitivity need to be solved before the processes are ready for production.
ASML is on track to deliver its first-generation, high-NA 157-nm production tool in 2004, in line with insertion of 157-nm optical lithography technology at 65-nm technology node. This lithography allows the use of moderate k1 factors, eliminating the need for reticle enhancement techniques such as alternating phase shift masks that are required with 193-nm technology.
The author would like to thank Hans Jasper for his contribution, Martin Brunotte from Carl Zeiss SMT, Steve Hansen, Harry Sewell, Bruce Tirri, and Willem van Schaik for their input, Jan Mulkens and Paul van Attekum for their insightful comments, and Steve Hansen and Mahesh Shah for reviewing the manuscript.
1. S. Dana, oemagazine, March 2002, pp. 20-22.
2. J. Burnett, et al., Phys. Rev. B 64, 241102, 2001.
3. V. Libermann et al., 3rd International Symposium on 157 nm Lithography, September 2002.
4. S. Dana, Proceedings SEMATECH 157 nm Technical Data Review, Dallas, Texas, May 2002.
5. Final Report, International SEMATECH (ISMT), A. Grenville et al., July 2002.
6. S. Kodama et al., Proc. SPIE 4690, 76 (2002).
7. W. Conley et al., Arch Interface Conf., Sept. 2002.
8. R. Dammel et al; S.Nur et al., 3rd International Symposium on 157 nm Lithography, Antwerp, Sept. 2002.
9. K. Turnquest et al., 3rd International Symposium on 157 nm Lithography, Antwerp, Sept. 2002
By Phil Ware, Canon USA
Although initially targeted for insertion at the 100-nm node, 157-nm lithography faced many potential showstoppers. With little time to find solutions, industry researchers found themselves struggling with the most basic aspects of lithographic science, such as developing optical materials with adequate transmission, pellicles that lasted more than a few thousand pulses, and a viable photoresist process. It soon became clear that the 100-nm insertion target could not be met, and efforts were refocused to bring 157-nm tools to market in time for the 65-nm node.
Meanwhile, dramatic increases in lens NA, coupled with advancements in resolution enhancement technology (RET), promised to push traditional optical lithography below 100 nm with krypton-fluoride excimer-laser sources (KrF, 248 nm) and all the way down to 65 nm with argon-fluoride lasers (ArF, 193 nm). Although using the longer wavelengths with extreme RET will be an expensive proposition at best, at least it provides some much-needed breathing room for 157-nm development efforts. Over the past two years, 157-nm showstoppers have largely been reduced to engineering challenges, but several technology hurdles remain daunting nonetheless.
Now in its third year of development, Canon's 157-nm FPA-5800FS1 beta tool is on track for completion in the fourth quarter of 2003. A high-volume production tool, FPA-6000FS2, will follow in 2005. To cope with the aggressive development schedule, the FS1 will be based on a hybrid platform that combines high-performance scanning hardware from the current FPA-5000 series with a new purging system for the optical column, illumination system and wafer, and reticle handling systems. The FS1 beta tool features 0.8 NA projection optics with a 5X reduction ratio and a 22-mm-wide slit that is scanned over a 26-mm field. Fabrication of lens elements for the FS1's projection optics began in mid-2002, and assembly and tuning are targeted for completion in early 2003. optical challenges
Of the few materials that might be candidates for 157-nm optics, calcium fluoride (CaF2) appears to be the only one actually suitable for lens fabrication. This poses a problem for lens makers because an all-refractive 157-nm lens made from a single material would require extreme line narrowing (0.1 to 0.2 pm) of the fluorine (F2;157 nm) excimer laser, which does not appear to be economically feasible. As a result, designers have adopted a catadioptric lens approach for 157-nm scanners. Canon has developed a novel catadioptric 157-nm lens design that preserves much of the learning from previous generations of KrF and ArF refractive lenses and allows continued use of its wavefront engineering techniques. This is critical because wavefront aberrations for 157-nm lenses must be held to even lower levels than what today's best lenses offer.
After individual elements are precisely polished to the required tolerance, finished surfaces are characterized with a Canon-built surface metrology tool that characterizes imperfections in terms of the Zernike coefficients. Based on the measurement data obtained from large batches of finished lens elements, computer simulation automatically selects optimum groupings of elements that would likely yield the lowest residual errors in the final assembly and tuning process. Next, the total wavefront error of the assembled lens is measured using a phase measurement interferometer (PMI). At this point, an iterative fine-tuning procedure reduces the residual wavefront error to the lowest possible level.
Variations in CaF2 quality, such as stress birefringence and index homogeneity, have been a serious concern since the beginning, but last year researchers at the National Institute of Standards and Technology (Gaithersburg, MD) gave the industry a scare when they discovered the existence of a relatively high level (Δn = 11 to 12 nm/cm) of intrinsic birefringence (IBR) when CaF2 is illuminated with 157-nm radiation (see oemagazine, March 2002, page 23). By incorporating multiple pairs of lens elements with different crystal orientations ( and ) that are precisely clocked to minimize the IBR effect, optical designers can effectively restore the imaging performance. Canon's own evaluation confirmed that, compared to an ideal case, the contrast loss in the first FS1 lens due to IBR is less than 1%. However, the availability of CaF2 of sufficient quality and quantity still remains a critical issue for 157-nm lithography. To help mitigate this problem, Canon has invested more than $40 million to enhance CaF2 production capacity in its wholly owned Optron subsidiary.
Most materials degrade rapidly under prolonged exposure to 157-nm radiation. This was especially the case with early 157-nm optical coatings. Recently, optical-coating durability has been dramatically enhanced through deposition process improvements. High-reflection and antireflection coatings applied with Canon's new high-density deposition process exhibit virtually no degradation after 60 million 22 mJ/cm2 pulses. However, the long-term durability of optical coatings must be closely monitored and improvements made as necessary. contamination and pellicles
Figure 2. This diagram of a beta stepper shows a modular purging technique designed to minimize N2 consumption.
Since oxygen (O2) strongly absorbs 157-nm radiation, it is necessary to purge the entire beam path with nitrogen (N2) or helium. The FS1 features a N2 circulation system that controls the wafer and reticle stage environments independently. The design minimizes the consumption of N2 while keeping O2 and water concentrations below 100 ppm in the wafer and reticle loading and stage chambers. N2-purged load locks are provided for wafer and reticle transfer. The environments at the top and bottom of the projection lens near the wafer and reticle are provided with additional N2 purging to maintain the active exposure area at 10 ppm or less (see figure 2). The projection optical assembly is purged with helium to hold concentrations of O2 and water to the target level of 1 ppm inside the lens. The illumination system is maintained at the 1-ppm level with an N2 purge. Chemical filters in the N2 circulating system reduce organic and inorganic contamination. In addition, as with existing scanners, the gas flow is temperature controlled and unidirectional to maintain temperature stability throughout the critical areas of the tool.
Pellicles are another headache for 157-nm tool vendors. To date, the search for an optically thin organic pellicle material durable enough for use at 157 nm has proven fruitless. Traditional pellicle materials quickly darken or rupture after only a few thousand pulses of 157-nm laser radiation. Operation without a pellicle is considered impractical because the cleanliness of the tool interior environment would have to be improved significantly, causing as much as a two-year delay in tool availability. Fortunately, an alternative is the so-called hard pellicle, which is an ultra-flat 800-µm-thick sheet of fused silica. At 800 µm, however, the pellicle becomes an optical element whose effects must be compensated for in the lens design. Canon designed its FS1 projection optics to be field-adjustable to accommodate either a traditional organic pellicle or a hard pellicle, just in case a durable organic pellicle material is developed some day.
Targeting high-volume production from the 65-nm to 45-nm nodes, the FS2 will feature a lens NA of about 0.9, a reduction ratio of 4X, a field size of 26 mm x 33 mm, and a synchronized wafer scan speed of 500 mm/s. Since the 6000 platform is expected to take optical lithography to at least the 45-nm node, it has been optimized for very low k1 lithography. In fact, all versions of the 6000 platform will feature low levels of lens aberration and scanning synchronization errors (MA and MSD), enhanced real-time focusing and leveling, improved alignment accuracy, a wide array of NA and illumination conditions, and a short reticle change time for double exposures. Canon will launch the 6000 platform initially with very-high-NA KrF and ArF lenses in the early part of 2003.
By Akikazu Tanimoto, Nikon Corp.
As recently as a year ago, many problems plagued 157-nm lithography efforts. Today, we can confidently say that the roadblocks to a successful, production-worthy 157-nm lithography tool with good cost-of-ownership (CoO) have been removed, and the normal engineering development activities continue now in pursuit of the end goals. Nikon's goal for its 157-nm lithography program is to intercept the International Technology Roadmap for Semiconductors (ITRS 2001) 65-nm node milestone in 2007 with early delivery of full-field tools in 2004 and production-worthy systems in 2006 or earlier.
Nikon's target specifications for the first F2 excimer-laser machine include a catadioptric projection optics design with an NA of 0.85 and a 4X reduction ratio, a 40 W source with a bandwidth of 1.0 pm (FWHM), a 26 mm x 33 mm field size, and throughput of more than 90 wafers/hour (300 mm wafers, 10 mJ/cm2 resist). The target application for this machine will be the ITRS 65-nm node in 2007. The imaging performance is expected to allow printing of 65-nm lines/spaces (1/2 pitch) with binary masks and 35-nm gate features with phase shift masks (PSMs).
The first tool is planned for delivery in 2004. Nikon also plans to perform detailed assessments of machine and imaging performance on the first several tools and to work upgraded design elements into the machines for the 2006 tools. These upgrades will address residual concerns with lens-element lifetimes, wafer and reticle handling, and general operation of ultra-high-NA operation with the new wavelength. materials and contamination
With the early 2001 identification of the so-called intrinsic birefringence effect in CaF2 throwing an unexpected complexity into the optical design, Nikon's engineers immediately grappled with correction methodologies. By the summer of 2001, designers had demonstrated that almost complete correction was feasible.1,2 By the end of 2001, Nikon had established design principles focusing on CaF2 quality and coatings.
Figure 3. Annealing can reduce the severity of stress birefringence effects in -oriented (left) and -oriented (right) material.
Besides intrinsic birefringence, birefringence arising from residual stress from the CaF2 crystal-growing process must be reduced to tolerable levels. Although the CaF2 manufacturers attempt to reduce the stress, Nikon has found ways to "super-anneal" the material to further reduce the stress (see figure 3). While the improvement in  material is not the same as for  material, Nikon expects that the current results are likely sufficient, with expectations of improvement from the ongoing progress now underway.
Although Nikon's high-NA (0.85) optical designs achieve very low aberration values, actually building, testing, adjusting, and mounting these lenses represent engineering tasks that also must be accomplished. Finally, lens-manufacturing tasks such as the polishing and coating of crystal surfaces of several different crystal orientations have been undertaken, with results adequate to support the overall lens designs and machine-performance targets.
Lens designs and manufacturability issues are not the whole story, of course. Contamination is a serious concern. Even relatively low levels of water vapor, oxygen, and organics can seriously reduce the transmission of 157-nm radiation through optical systems. Furthermore, unwanted contaminant gases excited by 157-nm photons can attack and damage optical coatings, severely reducing optical-system lifetimes. Nikon engineers have been addressing these issues by system purging with high-purity, dry nitrogen gas. Computer and experimental simulations have led to designs with acceptable startup times from the start of purging to turning on the illumination. ancillary systems and requirements
Achieving good critical-dimension control and providing feature placement with low-distortion, low-aberration optics is, of course, not the whole story for a new lithography system. Use of an F2 system introduced at the 65-nm node will require a whole new kind of cooperation between the equipment supplier and the chip maker. This wavelength will be the first debuting with such high-NA lenses, small depth of focus, and immediate use of PSMs. An overlay of 23 nm (mean +3 σ) on product wafers must be achieved and bettered.
To perform lithography with such small process margins, equipment suppliers must provide exceedingly stable machines, and chip makers must develop processes and methodologies to cope with these difficulties. reticles and resists
Although Nikon does not control the reticle and resist developments, a few comments are in order. First, the issue of pellicles for reticles has not been completely put to rest. While Nikon would prefer soft pellicles, materials with suitable properties and lifetimes have not yet been identified. Although feasible, hard pellicles introduce other problems. For example, if the pellicle material is not adequately flat, with sufficiently low wedge, and mounted sufficiently parallel with the reticle pattern, various kinds of imaging difficulties arise, including pattern offsets and aberrations. It is probably not cost-effective to adjust the projection lens to compensate for optical errors introduced by each pellicle-ized reticle. Further progress with pellicles is necessary; otherwise, the industry must decide on the use of reticles without pellicles, as will be the case for extreme ultraviolet (EUV) lithography.
Of course, a mask infrastructure supporting the 65-nm node is not yet in place. Further work on mask writers and on inspection and repair tools is necessary. This is true not only to support F2 technology but also for masks for other technologies, such as advanced ArF.
Resists for 157 nm are in their infancy. Suitable, practical, high-quality production resists with a transparency permitting their use at a thickness of 200 to 250 nm and providing good resolution (vertical sidewalls, no toes, no T-topping) do not exist yet. Substantial progress is still needed here. the roadmap
There are clearly questions as to the business challenges, both for the supplier and for the chip maker, of introducing F2 technology with an apparently short lifetime. However, Nikon believes that this question should be rephrased as follows: First, can F2 technology be introduced at a point in time at which its CoO can be expected to be sufficiently low to displace existing (probably ArF) technology for critical layers? If so, then the F2 introduction will make good business sense. Even if F2 is expected only to be used for one generation of critical layers, history teaches that this outlook is almost certainly too pessimistic. Compare the expectations and history with KrF and ArF, for example. Furthermore, F2 machine CoO will almost certainly be reduced as experience with the technology grows. Hence, F2 tools will continue to be purchased and used for semi-critical wafer levels for many years after a successful introduction.
On the other hand, if F2 cannot be introduced with a CoO adequate to displace ArF and at a time prior to the introduction of an affordable EUV lithography tool, the F2 story ends here. However, this discussion is properly conducted within an EUV lithography context, which is a subject for another article. oe
The author thanks John Wiesner for preparing the English manuscript.
1. N. Shiraishi, S. Owa, et al., "Understanding and Simulation of CaF2 Intrinsic Birefringence," International Sematech Calcium Fluoride Birefringence Workshop (2001).
2. S. Owa, N. Shiraishi, et al., "Nikon F2 Exposure Tool," 3rd 157 nm Symposium (2002).
Stephane Dana is product manager with ASML, Veldhoven, the Netherlands.
Phil Ware is senior fellow, lithography strategy at Canon USA Inc., Irving, TX.
Akikazu Tanimoto is the general manager of the development department, IC Equipment Division, Precision Equipment Company, Nikon Corp., Kumagaya, Japan.