1998 SPIE Lithography Symposium
05/01/1998
1998 SPIE Lithography Symposium
M. David Levenson,
Contributing Editor
The recent SPIE International Symposium on Microlithography showed the maturation of the lithography field. The presentations had a more practical view than in previous years, with even the "Emerging Lithographic Technologies" conference focusing on progress toward implementing actual systems. Even though the focus of the conference was nominally on the future, it proved that, in the semiconductor industry, the future is now.
Fulfilling the 1997 Roadmap. Much of the symposium, which met in Santa Clara, CA, concerned the SIA Roadmaps for semiconductor manufacturing: the effects of the 1994 Roadmap and the steps needed to fulfill the 1997 version.
William Siegel, group VP and chief scientist of Advanced Micro Devices (AMD), perhaps captured the tone best in his plenary address on the "Intended and Unintended Consequences" of the 1994 Roadmap. While the committees compiling the Roadmap meant it to be a needs statement for their suppliers, the semiconductor companies each tried to "beat the roadmap" and gain proprietary advantage. This had the unintended consequence of accelerating Moore`s Law so that the production the 250-nm generation began one year early. The 1997 Roadmap took this acceleration into account, and predicted a two-year interval between nodes, not the traditional three years. This unintended acceleration stresses the entire current infrastructure and may not be sustainable.
Siegel pointed out that lithographic manufacturing is a systems problem in which a semiconductor company must coordinate 10-12 suppliers, each with its own agenda. While 200-nm gates are in production, and optical methods appear adequate down to 100 nm, the dizzying array of materials innovations needed for sub-250-nm circuits and the traditional long latencies for new technologies (DUV took 15 years) complicates the future. As difficult as the technology may become, it might prove easier than the "social engineering" needed to implement such complex inter-related innovations, according to Siegel.
Potential for 157-nm lithography. Mordechai Rothschild of Lincoln Laboratories (MIT) reviewed the progress of 193-nm lithography and the potential for 157-nm "optical" exposure. The accelerated Moore`s law implies that new systems will be introduced with large (>0.60) numerical apertures, leaving little room for future refinements. While fused silica is useable as an optical material at 193 nm, most "good" material available today cannot assure a 10-year lifetime against compaction. That may not be a problem if other methods supersede 193 nm as rapidly as required by the Roadmap. However, Rothschild noted that the recent success at the University of Texas in printing 80-nm line-space patterns using 193-nm exposure, and a phase-shifting mask, implied a longer technology lifetime. Plausible candidates exist for pellicle and resist materials at 193 nm, and most stepper manufacturers plan to have prototype large-field tools available in late 1998.
A follow-on optical lithography based on the 157-nm F2 laser and fluoride optics seems "a reasonable option" according to Rothschild. The only nonbirefringent optical material with decent physical properties is CaF2, although MgF2 or sapphire might be useable as mask substrates. A microstepper has been built and has shown 80-nm resolution using a PSM.
Don Eigler of the IBM Almaden Research Center finished the plenary session with a discussion of what might be achieved if atoms could be moved individually, as they are in his laboratory. However, he disowned most of popularized "nanotechnology" as fantasy, unlikely to be implemented within the next 20 years. In the mean time, he recommended that scanning-tunneling microscope technology be used to understand the world, not pattern it.
Emerging Lithographies Symposium. This symposium highlighted the progress made in EUV technology with 14 papers, most by members of the Virtual National Laboratory. The design of the Engineering Test Stand (ETS), capable of 100-nm half-pitch exposure, is well along, and a sensitivity analysis for predictable imperfections shows promising results. The 13.5-nm light source will be based on a recycled high-pressure jet of xenon gas, excited by a 1.7 kW laser. This source suppresses most of the debris that might contaminate condenser optics, but experiments by Glenn D. Kubiak, et al., indicate that another order of magnitude improvement is needed. A collaboration of Japanese and American researchers presented an EUV system based on synchrotron radiation and a 3-mirror aspherical projection-lens design that avoids the debris problem and may prove more suitable for industrial applications. Other papers described developments in x-ray proximity printing, e-beam lithography, interferometric lithography, and other traditionally emerging areas.
Optical lithography. The Optical Microlithography symposium began with a series of five papers describing the advantages and difficulties of patterning chips using phase-shifting masks (PSMs). Hua-Yu Liu of Hewlett Packard ULSI Research Laboratories expanded upon her report of sub-15-nm full-wafer CD control at 150-nm linewidth (k1 = 0.237) using PSMs designed by co-authors from Numerical Technologies Inc. Line-space patterns with 450-nm pitch also showed similar linewidth control. Regina Schmidt and her co-workers from AMD reported a detailed analysis of the impact of aberrations on CD control with multiphase masks, which concluded that "symmetric" mask structures reduce the impacts of coma and other antisymmetric aberrations. Takashi Nakabayashi and his co-authors from Matsushita`s Kyoto ULSI Development Center described a new way to print contact hole arrays using specially-tuned illumination and a three-phase PSM.
Modeling and optical proximity correction (OPC) were the topics of two sessions. It is clear that OPC is becoming mainstream, even though semiempirical methods still give superior results. Four papers discussed different aspects of antireflection layers (ARLs), required by DUV processes because of the limited process window and high reflectivity of substrates. Bottom ARLs can introduce etch biases and other anomalies, especially when coated on dielectric layers, according to Ramkumar Subramanian and his team from AMD.
Shigeru Hirukawa of Nikon described that company`s plans for advanced DUV and ArF projection lenses. For sub-250-nm DUV lithography, Nikon plans a 0.68-NA lens that will operate with off-axis illumination. However, the minimum dose at a narrow dark line was found to be fully 60% of the dose-to-clear, requiring major improvements in resist technology and process control for practical production. The proposed ArF lens will have NA >0.56 and give similar performance at 150 nm with conventional illumination. Both lenses will operate as scanners with a 25-mm wide field.
Syed Rizvi of Photronics Inc. presented a moving analysis of the effect of the uncertainty as to post-optical lithography options on the mask industry. The 1997 SIA Roadmap foresees five different possibilities, each of them requiring a different mask technology. Since the mask industry lacks an R&D infrastructure, it cannot pursue even one of these in depth. According to Rizvi, the Roadmap lacks detail in the mask lithography area, leading to confusion, paralysis, and potential misunderstandings between maskmakers and users.
BACUS panel. Three papers and the after-dinner BACUS panel discussed the "Mask Error Factor" (MEF), which increases the impact of small CD variations on a mask in the low-k1 patterning regime required for future lithography by the SIA Roadmap. Alfred Wong and his team from IBM reported the results of massive simulations on the MEF, along with comparisons to experiment. Improved resist contrast and decreased diffusion tend to reduce the effect, according to Wong. Kurt Ronse of IMEC described how these seemingly small mask errors dominated over differences between state-of-the-art steppers and scanners. Nevertheless, the scanning system gave better CD performance at the edges of the focus window.
At the BACUS panel, Karen Brown, SEMATECH lithography director, defended the 1997 Roadmap saying, "It is about the industry`s needs in order to continue to observe Moore`s Law, not about solutions for meeting those needs." However, she conceded that the mask subcommittee could not agree even to the answers of such basic questions as "what is a defect." Joe Langston of Intel reported that the MEF is a big problem that had been ignored in the Roadmap. For example, he said that butting errors in the mask-writer were contributing 6-10 nm of error to the wafer CDs. Gil Sheldon of Dupont Photomasks observed that the mask-industry roadmap had not accelerated along with the semiconductor industry`s, leading to a technology shortfall. One panel member observed that much future grief could be avoided if the industry committed to PSM technology. Panel coordinator Neil Berglund summarized the panel`s conclusions: There is a technology shortfall, but decisions on addressing it would be made based on economics; also the mask industry needs an internal technology roadmap and more university research.
Other show highlights. The Metrology, Inspection, and Process Control meeting focused on ways to measure circuit structures smaller than a wavelength of light. Low voltage SEMs and atomic force microscopes seem the most promising new technologies, but C.P. Ausschnitt and Mark Lagus of IBM described an optical means to infer CD from line-end retrenchment. Their method employs an array of fine lines (called a "schnitzl") whose width can be measured accurately with overlay metrology tools.
The Resist Technology and Processing conference covered the laboriously won progress in optimizing DUV and ArF resists for production. The emphasis is on single-layer resists because of their seeming simplicity in processing and the reduced edge-roughness shown by these systems. Late-breaking developments included two reports of plasma-polymerized methyl-silane as an all-dry-processable material, and the 80-nm breakthrough in 193-nm lithography by Kyle Patterson at the University of Texas.
The equipment show seemed larger than usual and could have benefited from a third half-day. FSI introduced its Polaris 2500 wafer processing system, which is especially compact because of the use of dedicated transfer arms between hot and cold plates in each stacked bake module. ASM Lithography introduced its i-line step-and-scan system for mix-and-match applications, and claimed to have overtaken Canon for the #2 position. Canon disputed that claim. One wonders what Ultratech (who had an agreement with ASML) thinks of ASML`s entry into mix-and-match with a $4 million machine. Several OPC companies participated for the first time. For some, it may be the last time.