Optical Lithography: 100 nm and Beyond
Staff -- Semiconductor International, 9/1/1998
T he debate over which non-optical technique will carry the industry beyond 100 nm continues with increased complexity. With developments in the optical arena in deep UV photoresists and high NA lithography tools, research groups are reporting similar results to non-optical methods critical dimensions (CD) of 60 to 80 nm. Therefore, in addition to X-ray, ion projection, SCALPEL and extreme UV, optical lithography can be added to the list of "alternative" lithography technologies.Despite the familiarity of optical lithography (lead photo), it too will require the infrastructure to support the resolution enhancement technologies (RET) necessary to reach production-worthy manufacturing at 100 nm design rules. For example, significant infrastructure development is required to support the lower optical wavelengths (193 nm and 157 nm) because of new lens materials and resists. Infrastructure is further required to support mask inspection and repair and the immense increase in data handling for implementation of phase shift masks (PSM) and optical proximity correction (OPC). This means greater emphasis on "image integration" leading to collaborations among stepper, mask and resist researchers and manufacturers. Resources out of International SEMATECH are currently available to virtually all resist manufacturers, such as Shipley (Marlborough, Mass.) and Clariant AZ Electronic Materials (Somerville, N.J.). In the process, all have obtained sub-100 nm lithography using optical extension technologies.
193 nm chemistries
Several 193 nm photoresist designs, all based on chemical amplification (CA), are currently being investigated. CA enhances photosensitivity and is based on a deblocking reaction catalyzed by photogenerated acid in exposed areas. The result is a solubility difference between the exposed and unexposed regions of the photoresist. Styrene polymers, typically used for DUV resists, cannot be used in 193 nm photoresists, because the aromatic groups absorb light too strongly at 193 nm. As a result, other classes of polymers must be investigated to create photoresists for 193 nm. All existing candidates contain alicyclic (aliphatic cyclic) structures to impart etch resistance to the final polymer. In general, the more rings incorporated into the polymer structure, the greater the etch resistance.1,2 The polymers being studied as potential 193 nm photoresists can be grouped into two general categories, acrylic and cyclic olefin polymers. Acrylic platforms contain alicyclic structures attached to an acryclic backbone. Cyclic olefin platforms incorporate the alicyclic structures directly into the backbone.
A research team headed by Dr. Grant Willson at the University of Texas at Austin has been working with several varieties of alicyclic polymers. By incorporating the alicyclic groups into the polymer backbone itself, these alicyclic polymers demonstrate etch resistances superior to resistances from acrylate-based polymers, Willson said. So far, the most successful of these alicyclic polymers are those referred to as alternating polymers due to the regular repeating pattern of monomers within the polymer backbone. Using conventional chrome-on-glass masks and a 0.6 NA exposure tool, these polymers demonstrated resolution capability down to 140 nm at 1:1 line:space duty cycle and down to 110 nm with a 1:1.5 duty cycle. Using an alternating PSM which alters the phase light passing through adjacent features on the mask doubles the frequency of the information allowed through the lens, thus allowing the resolution of smaller features. Observed were 80 nm features at a 1:1.5 duty cycle and features as small as 60 nm with greater isolation.
These new polymers allow light to pass through the polymer matrix and reach the photosensitive compounds. This polymer essentially combines the good transparency characteristics of acrylates with the etch resistance of styrenes, Willson said. According to the researchers, this positive-tone resist has sensitivity and speed comparable to that of 248 nm DUV resists. The 80 nm features were generated using an alternating PSM produced by DuPont Photomasks (Round Rock, Texas), on a stepper with extended half-pitch resolution of ~80 nm.
Researchers at IBM Almaden Research Center (San Jose, Calif.) are investigating both acrylics and cyclic olefin polymers, the latter being a fairly radical departure from the norm. Cyclic olefin polymers are new materials. They have the promise of significantly better etch properties than the others, said Dr. Robert Allen, 193 nm photoresist project leader. This relative newcomer is being studied in collaboration with B.F. Goodrich Chemicals. Concurrently, work underway with acrylics using alternating PSMs and a 0.6 NA exposure tool has proven successful in producing 100 nm equal lines and spaces, with a 14% exposure latitude at best focus and a 0.6 um depth of focus, Allen said. Having an even larger process latitude are 90 nm lines with a 1:1.5 pitch. With a 1:2 duty cycle, 70 nm CDs on 140 nm also have been achieved. Encouraging is a chrome-on-glass result, where no phase shifting was used: 140 nm equal lines and spaces.
|
|
Fig. 1. 150 nm lines with 908 sidewalls were obtained using a 193 nm resist on a Brewer ARC and exposed on a 248 nm, 0.53 NA stepper. (Source: IBM) |
According to Allen, the last generation of 248 nm single-layer resists may actually be 193 nm resist systems used with high NA, 248 nm exposure tools. This may be feasible, because 193 nm photoacid generators function at either wavelength and the polymers have a factor of five to 10 lower absorbance (increased transparency) than conventional 248 nm resists. Equally important is the anticipation of enhanced etch resistance of the 193 nm cyclic olefin polymers when used at 248 nm. The potential benefits are the ability to print thicker films while maintaining 908 wall profiles. Without the aid of phase shifting, this chemistry has been used to print 125 nm lines on a 1:2 duty cycle. Figure 1 demonstrates 150 nm lines on a 1:1.5 duty cycle and 908 wall profiles imagined in 5650 Å of resist.3 These geometries were exposed with a 248 nm, 0.53 NA stepper using a standard binary mask.
Extending 248 nm technology
The ability to image features to 80 nm with 193 nm technology is not surprising. The use of 248 nm lithography is another matter. Initially intended to be used to 130 nm CDs, results out of International SEMATECH's DELPHI project, a risk management optical resolution enhancement program, have been exceptional. In collaboration with Photronics (Milpitas, Calif.), MicroUnity (Sunnyvale, Calif.), Toppan Printing (Sai-tama, Japan) and National Semiconductor (Sunnyvale), researchers added OPC structures on alternating PSMs to achieve critical dimension control of 100 nm features using 248 nm deep UV exposures on 4X lithography systems (Fig. 2).4
|
|
Fig. 2. Using 248 nm exposures, the CD range of 100 nm printed lines improved from 55 nm to 12 nm when optical proximity correction was added to phase shift masks. (Source: International SEMATECH) |
These results were achieved with a less than state-of-the-art 0.53 NA stepper. According to John S. Petersen, DELPHI project leader and International SEMATECH fellow, if a next-generation tool with an NA of 0.68 were used, assuming linear extrapolation, this technology should be able to go to 110 nm equal lines and spaces and ~80 nm semi-dense lines. The implication of these results, noted Petersen, is the possibility of using 248 nm lithography at the 130 nm technology node. A particular technology node assumes the next three shrinks. The 130 nm technology node consists of 1:1 lines and spaces at 130 nm, 100 nm isolated lines and three subsequent shrinks.
The combined enhancement technologies address the proximity effect error seen on phase-shift reticles. As features approach one another, their diffraction patterns change, causing CD to vary significantly from the initial design. The trend toward device geometries below the wavelength of exposure light and the use of highly coherent light to strengthen the effect of phase-shifting are enhancing the need for proximity correction, which is exacerbated under these conditions. This means that the combined enhancement technologies can overcome proximity related CD errors while taking full advantage of the resolution improvement inherent in the PSM technique.
Motorola (Austin, Texas) is using alternating phase shift technology to produce 80 nm gates on ~450 nm pitch with 0.6 NA 248 nm exposure tools like MicrascanIII, according to Dr. John Sturtevant from the Advanced Products Research and Development Laboratory. While OPC has been used at multiple binary mask layers in production for the past few years, one- and two-dimensional OPCs combined with alternating phase shifting to correct for proximity and foreshortening effects are still in development. There are plans to put <120 nm gates into manufacturing within the next year and a half.
Some of these results appear to defy the limits of the Raleigh criteria. The k1 factor, a measure of the degree of difficulty in printing, is given by k1 = w x (NA/l) for a given linewidth, w. The theoretical limit for coherent light is ~0.50. Above a k1 of 0.65, printing is easy; below requires optical enhancements. At CDs of 100 nm, k1 modifiers such as OPCs and PSMs are necessary. However, the minimum k1 is tied to pitch. With image enhancements, lines and spaces can be any combination of k-line and k-space that adds to a k-pitch of 0.5 (the resolution limit of a point diffraction pattern). Off-pitch lines/spaces therefore can be used to increase resolution. Results such as those out of DELPHI, which reported a k1 of 0.21 with a k-pitch of 0.6, are therefore consistent with the Raleigh criteria.5
Beyond single layer resists
Thin layer imaging through implementation of bilayer resists, top surface imaging and the CARL process can extend single-layer resists to achieve greater resolution. Research is underway at IBM to push 248 nm lithography as far as possible by specifically investigating bilayer photoresists. The most recent resist formulations containing 10% to 11% silicon by weight are competitive with the best DUV single layer resists for imaging nested 150 nm features in thin films, said Dr. Gregory Wallraff, bilayer resist project leader. Work is in progress to optimize these formulations for potential applications with high NA KrF exposure tools for 130 nm features.
|
|
Fig. 3. A 193 nm bilayer resist was used to image 80 nm lines on a 1:3 pitch using alternating PSMs. (Source: Olin) |
Olin Microelectronic Materials (Norwalk, Conn.) is developing both single-layer 193 nm resists in conjunction with Lucent Technologies and has worked with International SEMATECH and the Inter-University Microelectronics Centre (IMEC, Leuven, Belgium) to develop processes utilizing bilayer 193 nm resists. Combining PSM masks with the bilayer resist has produced 80 nm features on pitches as low as 1:1.5 (Fig. 3).
Roadblocks?
Similar to non-optical lithography techniques, the potential roadblocks to optical at 100 nm and below lie in the mask technology. The attenuated materials, such as moly silicides (MoSi), currently used for the less aggressive attenuated phase shift mask blanks have many defect issues. To achieve quality and definition at these small geometries, increasingly complex optical proximity schemes will be required, translating into increased photomask write times.
While recent optical lithography performance has been impressive, to be production-worthy, final CD variation of less than 10% of mean to target across an entire field is required. Therefore, at 130 nm features, 11 to 12 nm CD variation is tolerable, and at 100 nm features, 9 nm. According to Will Conley, member of technical staff at Cypress Semiconductor (San Jose), the current difficulty in limiting CD variation to 16 or 17 nm for 180 nm design rules and the notion of improvements on the order of 2X is a concern. The loss of control as features become smaller is known as the mask error function and describes the combination of events that act to magnify mask CD errors on wafers.6 Cypress Semiconductor, National Semiconductor and others organized to address these and other production issues are part of the IMEC Industrial Affiation Program.
Phase shift mask technology becomes more critical as device geometries shrink, though it carries its own limitations, such as unwanted phase transitions and layout issues. Attenuating phase shift or half-tone masks are typically used on contact levels, while alternating PSMs provide stronger phase shifting and generally are used on more critical levels like wiring patterns and gate levels. This is not a simple technology. Phase errors can create shorts, impact image placement, center-of-focus and magnify lens aberrations. Some structures, such as T shapes, cannot be phase shifted. One approach may be to lay down the phase shift pattern in two exposure layers and integrate them into one, Petersen said. Unwanted phase transitions also can be trimmed during respective exposures. In the same vein, Numerical Technologies (Sunnyvale) markets a layout design package that removes these phase transitions.
Another challenge is contact holes. Insufficient light comes through as the size of the openings approach 100 nm. Attenuating PSM technology traditionally uses 5% to 8% transmission. Companies such as Japan's NEC Corp. have successfully demonstrated 150 nm contacts using 5% and 23% transmission PSMs with workable depth of focus.7 Studies to determine the limits of this effective, though challenging technology are ongoing.
Once thought to be a potential show-stopper, reticle inspection of PSMs will be available by year's end, said Edward Grady, vice president, measurement group at KLA-Tenor (San Jose). New algorithms have been developed to inspect alternating PSMs, which are created by etching quartz. But while particles in the etched valley can be detected, repair still remains an issue. The ion beam used for repair can add residual damage and cause river bedding and staining. These residual defects are more printable than binary defects and are often in critical locations such as in the gate region, where device performance is affected. No reticle repair technology implies the need for error-free reticles, which will mean low yields and prohibitive costs. Consortia such as the DPI Reticle Test Center (Round Rock), LLC, whose membership consists of Motorola (Austin), Advanced Micro Devices (AMD, Sunnyvale), Micron and DuPont Photomasks (Round Rock), have formed to accelerate learning for the manufacturing of next-generation photomasks.
A defect sensitivity mask (DSM) design tool will allow calibration of inspection tools to detect defects on advanced masks that contain the sub-resolution features required to reach 100 nm lithography manufacturing capabilities. A current project involving MicroUnity, KLA-Tencor, AMD, Applied Materials (Santa Clara, Calif.), Photronics, DuPont Photomasks and Benchmark Technology (Lynnfield, Mass.) will result in a DSM design to produce results that give vendors confidence that high volumes of advanced reticles are feasible. According to Roger Caldwell, vice president of silicon technology at MicroUnity, this methodology, coupled with the introduction of shorter wavelength inspection tools and better inspection algorithms, should give the industry the necessary tools to bring advanced mask manufacturing into volume production.
Beyond 193 nm lithography
If these existing limitations prove to be insurmountable, the necessary approach is shorter exposure wavelengths. During the past six months, momentum has increased in 157 nm technology, a potential successor to the 193 nm ArF laser. At least one tool manufacturer, SVG Lithography (Wilton, Conn.), has indicated "serious interest" in this technology.
Lambda Physik (Goettingen, Germany and Ft. Lauderdale, Fla.) has demonstrated an industrial grade laser using a F2 excimer laser, emitting at 157.629 nm and 157.523 nm. Powers of 20 W and 500 Hz have been achieved. These numbers are comparable to current 248 nm KrF and 193 nm ArF lasers, which typically have powers in the 10 W to 20 W range and frequencies of 1 KHz and 800 Hz, respectively. Because the F 2 source has a linewidth in the picometer range, approximately 100 times smaller than that of unnarrowed KrF and ArF lasers, much of the optics and accompanying losses involved in line narrowing may be eliminated. The short wavelength and high photon energy however, are absorbed strongly by oxygen absorption bands, requiring operation in a vacuum or in an inert gas ambient. The identification of a laser source is an important beginning. Materials to correct for aberrations in the stepper optics are under investigation. These materials also must be laser damage resistant and have long lifetimes, said Dr. Heinrich Endert, national sales manager at Lambda Physik US.
With a diminishing k1 factor, shorter wavelengths such as 157 nm look more appealing, despite the short supply and expense of CaF2, the optical material needed for the reticles and lens system. The progress of 193 nm is driving the growth of the CaF2 infrastructure, which may well ease the way for 157 nm lithography, said Dr. Tim Brunner, manager of advanced lithography and metrology at IBM Semiconductor Research and Development Center (SRDC, East Fishkill, N.Y.). Resist and mask development will necessarily follow.
The bottom line
|
| Fig. 4. Amid speculation, some now believe that optical lithography could extend to the year 2015. (Source: Motorola) |
With optical lithography firmly in place, all new technology development has the potential of being implemented at some level into production. Progression on the manufacturing learning curve will see some RET technologies compatible with smaller wavelength technologies such as 193 nm, while others require significant re-engineering to make them compatible. For example, the use of sub-resolution features are receiving acceptance and are compatible with reduced wavelengths. However, attenuated PSM will require re-engineering of mask materials. As incremental RET technologies are implemented, the benefit provided by smaller wavelengths and the work required to re-engineer the established RET technologies could be the deciding factors in the cost effectiveness of the wavelength change, Caldwell said.
Optical and non-optical techniques have produced similar sub-100 nm
proof-of-concept results, and while optical lithography is mature, both
need to build and strengthen the necessary infrastructures. With four
non-optical and three optical wavelength options, a practical production
solution at the 130 nm technology node is still a long way off. For
optical, the current environment of cooperation can aid in the effort to
integrate all lithography components and extend usefulness to smaller
features. Though speculation about the end of optical lithography has
persisted for the past 25 years (Fig. 4), Sturtevant believes that with
157 nm lithography, optical methods may extend to the year 2015. 
References
- R. Kunz, et. al., Proc. SPIE, Vol. 2724, 1996, p. 365.
- T. Wallow, et. al., Proc. SPIE, Vol. 3333, 1998, p. 92.
- J. Opitz, et. al., SPIE PhotoMask, to be published, Kawasaki, Japan, 1998.
- J.S. Petersen, et. al., SPIE PhotoMask, to be published, Kawasaki, Japan, 1998.
- B.W. Smith, SPIE, Vol. 3334, 1998, p. 142.
- A.K. Wong, et. al., SPIE, Vol. 3334, 1998, p. 106.
- H. Iwasaki, et. al., SPIE PhotoMask, to be published, Kawasaki, Japan, 1998.
