EUV lithography
07/01/1997
FOURTH IN A SERIES
EUV lithography
Andrew M. Hawryluk, Natale M. Ceglio, David A. Markle, Ultratech Stepper, San Jose, California
The origin of EUVL can be found in a 1988 paper [1]. It proposed that recent advances in "high" reflectivity soft x-ray mirrors could enable an all-reflective, projection lithography system operating at wavelengths around 13 nm. The authors called their proposed technology Soft X-ray Projection Lithography (SXPL) and suggested that it could be used to print features at 0.1 ?m.
In 1989, companies such as IBM, AT&T, Ultratech Stepper, and Tropel expressed enough interest in the technology to justify a three-day symposium on SXPL [2]. This conference launched a number of independent research programs funded by DARPA, the national laboratories, and industry. In 1991, SXPL became the centerpiece of the DOE`s Technology Transfer Initiative Program, through formal CRADAs (Cooperative Research and Development Agreements). When the process was completed, three national laboratories (Lawrence Livermore National Laboratory, Sandia National Laboratory, and Lawrence Berkeley Laboratory) and eight US companies (AT&T, Ultratech Stepper, Intel, Jamar Technology, AMD, Tropel, Micrion, and KLA Instruments) were engaged in a cost-sharing agreement to pursue SXPL to a scientific feasibility demonstration in 1996.
In 1993, the technology`s name was changed to Extreme Ultraviolet Lithography (EUVL). In April 1996, a two-day review of the milestone accomplishments of the EUVL scientific feasibility program, held at Lawrence Livermore National Laboratory, sparked formation of a consortium of US companies to carry EUVL forward. In October 1996, the DOE suspended funding of the EUVL program and EUVL went forward as an industry consortium-funded, advanced R&D program with a goal to produce an alpha machine in 2-3 years.
|
Figure 1. An artist`s drawing of an EUV lithography system. A laser generates EUV radiation, which is collected by a multilayer coated condenser system. The radiation is incident upon a multilayer coated reflection mask, which is imaged and demagnified onto a resist-coated wafer.
The EUVL concept
Figure 1 shows an artist`s rendering of an EUVL stepper [3]. As currently envisioned, a compact, high-average-power, high-repetition-rate laser impacts a target material, producing broadband radiation with significant EUV emission. An all-reflective condenser optical system collects the EUV radiation and projects it onto a resonant-reflective reticle. The reflected EUV radiation carries the IC pattern to the all resonant-reflective imaging system, a series of four very-high-precision mirrors. The optical system projects a demagnified image of the reticle pattern onto a resist-coated wafer. The entire reticle pattern is exposed onto the wafer by synchronously scanning the mask and the wafer, i.e., a step-and-scan exposure.
Critical challenges in EUVL
Transforming the EUVL concept into a manufacturing technology for 0.1-?m-design-rule ICs will require mastery of a number of critical technology challenges.
EUV source. A modestly priced, highly reliable laser driver must be developed with output power of a few kilowatts, repetition rate of a few kilohertz, and a beam capable of accurate pointing and focusing so that the position and pulsed intensity at the target can be accurately maintained. Since the EUV source radiation is produced by impacting the laser beam on a target, the conversion must be done within the bandwidth of the mirrors without producing debris that could coat and/or damage the collection optics.
Precision optics. Fabrication and metrology of the reduction imaging optics are the most difficult optics problems in EUV. The components of the reduction system will be aspheres and will probably be made from a low-thermal-expansion material such as Zerodur. According to the Rayleigh criteria, the aspheric figure (i.e. the optical prescription) must be accurate to a few angstroms. The surface polish should have a rms roughness of less than a few angstroms at the mid-spatial frequencies (in order to preserve system MTF) and of order 1? at the high spatial frequencies (in order to preserve mirror reflectivity). To fabricate such components, optics producers will need new metrology techniques to measure figure and surface roughness to the required accuracies.
EUV coatings. The catalyst for EUV lithography was the development of multilayer coatings, which provided normal incidence reflectivities in excess of 60%. Multilayer coating technology must be further extended, though, since normal incidence reflectivities >70% will be required for high-throughput, cost-effective lithography [4]. Multilayer deposition control must match the coating resonances on all the mirrors and achieve precision gradients of the multilayer period across at least some of the reflectors. The coatings must be stable and resistant to EUV radiation damage and thermal effects over long periods of time.
Reticles. Defect-free reticles are perhaps the most demanding challenge for EUVL. While techniques exist for repairing errors in the reticle absorber pattern, no methods are known or expected for repairing point defects in the reflective multilayer coating. Thus, EUVL reticles require defect-free deposition of multilayer coatings over large areas. Inspection techniques for unpatterned reticle blanks (i.e., multilayer coated substrates) must also be developed.
|
Figure 2. The sensitivity of normalized wafer exposure costs to several parameters is illustrated. The wafer exposure costs are most sensitive to EUV mirror reflectivity [4].
Resist. EUV wavelengths require either surface imaging resists or bilayers with a very thin top layer. For commercial viability [5], the resist material and resist process must be compatible with high-volume manufacturing.
Stepper technology. The EUVL stepper will have to operate in a multivacuum environment. A total overlay budget of order 25 nm will be required to produce devices at 0.1-?m CD design rules. Reticle to wafer alignment techniques will need to operate in the multivacuum EUV environment with membranes [6] separating the source from the reticle and the imaging optics from the wafer.
While the technological challenges to achieve a viable EUVL production capability appear overwhelming, significant progress has been made over the past few years.
EUVL cost of ownership and systems analysis
The costs associated with a modern IC manufacturing facility have been growing at an exponential rate [7]. Economic, not technical issues are likely to limit device size reductions. Any advanced lithography candidate must justify itself on the basis of cost-effectiveness. Careful attention to wafer exposure costs has been an integral part of the development of EUVL [8-10].
Our cost of ownership (COO) analysis considered an IC with a minimum feature size of 100 nm (corresponding to a 16-Gbit DRAM device). End-users, tool integrators, and original equipment manufacturers helped establish the technical specifications, performance requirements, and percent utilization for all components in the EUVL tool and estimate the capital equipment costs, maintenance costs, installation costs, and the cost of consumables [4]. Reflective reticle costs were estimated from a separate reticle factory model [11].
|
Table 1 illustrates the baseline values for some of the "EUVL-specific" technical parameters, including multilayer mirror reflectivity, laser power, conversion efficiency, etc. Table 2 summarizes the nominal performance requirements for the scanner (which are mostly technology independent).
|
The calculated wafer exposure cost depends on the estimated total capital equipment cost, its throughput, and its utilization. The capital equipment cost estimate for the EUVL scanner (about $10 million) is derived from an estimate of the component costs for the tool; the capital equipment costs for the track (about $4 million), silylation, and etch tools are estimates from equipment manufacturers and industry experts. The throughput depends on the calculated exposure time/wafer-which in turn depends on the technical performance specifications for the source, the mirrors, and the resist sensitivity-and the overhead time/wafer (load, alignment, step, leveling, etc.). Assuming the parameters listed in Tables 1 and 2 and in Ref. 4, the calculated wafer exposure cost/critical level is approximately 7.5 cents/cm2.
|
Figure 3. The ability to obtain uniform mirror reflectivity from curved optics is very important. A series of experiments produced very high reflectivity coatings with very accurately controlled multilayer coatings. The mirror in this measurement had a reflectivity and period uniformity of 0.7% and 0.4%, respectively.
A detailed COO and systems analysis helps determine the sensitivity of wafer exposure costs to systems parameters. Figure 2 [12] illustrates the normalized wafer exposure cost as a percent of nominal value for source power, mirror reflectivity, and resist sensitivity [4]. This analysis shows which parameters most influence the economic competitiveness of the candidate technology.
EUVL coating technology
No single breakthrough can claim greater credit for launching EUV lithography than the development of precision multilayered coatings with soft x-ray reflectivities in excess of 60%. These coatings are the short wavelength analog of the quarter wave stack, the resonant dielectric coatings used to produce high reflectivity at visible wavelengths. The multilayer reflection is wavelength, angle and polarization dependent, satisfying the Bragg-like equation
where l is the resonant wavelength, L is the effective multilayer period, and q is the angle of incidence (90? = normal incidence). This equation places particularly stringent demands on the design of an optical system and on the control of the multilayer period to ensure wavefront correction and illumination uniformity at the wafer plane.
A viable coating technology must:
produce coatings with high reflectivity,
control the multilayer period across the mirror surface,
control the period of the multilayer from mirror to mirror in order to produce "matched sets" of mirrors,
deposit high-reflectivity, defect-free coatings on flat reticle substrates, and
control the stress in multilayer coatings so that the figure of the precision optics is not compromised.
Figure 3 [13] illustrates the high reflectivity and the degree of spatial deposition control achieved with Mo-Si coatings. Figure 3a shows a narrowband reflectivity of 66% at 13.4 nm. Figure 3b shows extraordinary uniformity both in reflectivity and multilayer period across a 4-in.-dia., f/3 mirror: the measured reflectivity and period are 65 ? 0.5% and 69.5 ? 0.3?, respectively. Condenser optics (and some imaging optics) will require tailoring of the multilayer coating period across the optic. Experimental studies have matched an intended gradient profile to within 0.2 ? [14].
Mirror reflectivity with peaks as high as 70% at 11.5 nm for a Mo-Be multilayer coating have been achieved. However, Mo-Be coatings could theoretically achieve reflectivities exceeding 75% at this wavelength [15].
Optics
Optical system designs. In order to print smaller feature sizes, lithographic systems operate at shorter exposure wavelengths and use larger numerical apertures. This trend is expected to continue down to l = 193 nm exposure for the production of 180-nm structures. Unfortunately, fused silica, the only glass useful in this region, begins to absorb strongly at shorter wavelengths. Commercial lithographic tools with refractive imaging systems are unlikely to operate at wavelengths much shorter than 170 nm, making reflective optical systems necessary.
Since large pupil obscurations degrade image quality, any lithographic system must have a very low pupil obscuration. Thus, any reflective lithographic system will most likely be an off-axis system to eliminate this obscuration in the lens. Aberrations in an imaging system generally scale as the third (or higher) power of the numerical aperture; therefore systems with a small numerical aperture are desirable. Since high resolution at small numerical apertures requires short exposure wavelengths, the best choice for an all-reflective optical system is an off-axis system with a low numerical aperture, using the shortest wavelength for which efficient reflectors can be made.
The EUV designs with the best image quality are those with the largest number of mirrors. In general these designs have the largest field size, the best correction over the field, and the smallest aspheric departures on their components. However, since the best reflector we can reasonably expect in the EUV is about 70-75% efficient [16], each additional mirror (for a given source power) reduces the energy reaching the wafer and lowers the throughput. EUV lithography requires a trade-off between the imaging performance over large fields and the ease of fabrication of optics, vs. system throughput.
A 0.1 numerical aperture, two-mirror system is possible, but the very narrow ring field width of this system constrained its usefulness [17, 18]. The width of the low distortion ring was less than 1 mm. Various two-mirror, off-axis Schwarzschild systems [19, 20] have demonstrated EUV imagery, but these typically have corrected fields too small to be useful for production.
Various three-mirror EUV imaging systems have been proposed. A 1X Offner ring field system with a 0.085 numerical aperture delivered diffraction limited imagery over a 0.1-mm-wide slit [21]. Three mirror reduction designs come close to meeting the requirements for production lithography [22, 23]. Unfortunately, all three mirror systems, and all systems with an odd number of mirrors, have the mask and wafer in roughly the same place, where they can interfere with one another.
|
Figure 4. An example of a four-mirror, ring field imaging system that may be suitable for EUV lithography [25]. The system has a numerical aperture of 0.085 and has a 2-mm-wide slit at the wafer.
Four-mirror reduction systems have been designed [24]. Generally speaking these ring field systems have slit widths of at least 1 mm at the wafer and numerical apertures of at least 0.08, which is sufficient to resolve 0.10-?m features. The choice among them will depend primarily on the variation in incidence angles on the mirror elements. Since EUV mirrors contain a large number of layers, they operate efficiently only over a narrow range of wavelengths and incident angles. With tapered coating thicknesses, different parts of a mirror can efficiently reflect different incidence angles; but large variations in incident angle at the same point in a mirror inevitably degrade reflection efficiency. Some four-mirror designs that otherwise appear to be promising have one or more optical elements with an unacceptably large variation in ray angles at the same point on the mirror.
Other requirements for an EUV imaging system include an accessible pupil and mirror sizes small enough to fit inside an EUV coating tool. Once these and other practical requirements have been met, ease of fabrication will determine the optical system. An example of a four-mirror EUV projection system is shown in Fig. 4 [25].
Optics fabrication requirements. Using the Rayleigh criteria, an acceptable peak-to-peak wavefront error of 0.2 l at 13 nm corresponds to a peak-to-peak mirror surface error of ~ 0.6 nm*. The corresponding rms figure, about 0.15 nm, is about a factor of 30 beyond present commercial practice. This requirement applies to low-frequency fabrication errors in the optical elements and typically corresponds to approximately 0.1 of the diameter of the optic.
Optical metrology. Extending the accuracy of figure measurement by a factor of 30 over the current state-of-the-art seemed to be a hopeless task until the invention of the Sommargren phase-shifting diffraction interferometer [26]. Traditional interferometers measure the quality of an optic by illuminating it with coherent radiation and comparing the reflected (or transmitted) beam with a reference beam. The measurement is limited by the quality of the reference beam, which often limits the tool to a measurement accuracy of l/20-l/50 (l ~ 633 nm) and a precision of approximately l/200. EUV lithography requires optics with an accuracy+ of ~ l/1000. Sommergren created measurement and reference wavefronts by a natural process, namely diffraction, producing wavefronts with excellent fidelity (Fig. 5).
|
Figure 5. The residual map error in the clear aperture of an EUVL optic as measured with the Sommargren phase-shifting diffraction interferometer. This optic has a measured rms deviation of 0.95 nm.
Other issues
Part 2 of this article will appear in a future issue of Solid State Technology. The authors will continue their review of EUV lithography, discussing reticle, resist, and source technologies, as well as their conclusions and predictions.
References
1. A.M. Hawryluk, L.G. Seppala, J. Vac. Sci. Tech., B6, 2162, 1988.
2. Proceedings of the First Technical Symposium on Soft X-ray Projection Lithography, ed., N. M. Ceglio (Lawrence Livermore National Laboratory, Conf-9001104), 1990.
3. N.M. Ceglio, A.M. Hawryluk, G.E. Sommargren, Front-End Design Issues in Soft X-ray Projection Lithography-Applied Optics, Vol. 32, p. 7051, 1993.
4. A.M. Hawryluk, N.M. Ceglio, "EUV Lithography Cost of Ownership Analysis," OSA Proceedings on Extreme Ultraviolet Lithography, 1994, Vol. 23, Optical Society of America, F. Zernike and D. Attwood, eds.
5. A.M. Hawryluk, N.M. Ceglio, "EUV Lithography Cost of Ownership Analysis," OSA Proceedings on Extreme Ultraviolet Lithography, 1994, Vol. 23, Optical Society of America, F. Zernike and D. Attwood, eds.
6. N.M. Ceglio, A.M. Hawryluk, "Soft X-ray Projection Lithography System Design and Cost Analysis," SPIE Proceedings, Vol. 1547, 1991.
7. C.R. Barrett, "Silicon Valley, What Next?" MRS Bulletin, July 1993, (Materials Research Society, Pittsburgh, PA).
8. N.M. Ceglio, A.M. Hawryluk, "Wafer Cost Analysis for a Soft X-ray Projection Lithography System," Journ. Vac. Science and Technology, Vol. 3, No. 3, June 1992.
9. K. Early, W. Arnold, "Cost of Ownership for Soft X-ray Projection Lithography," OSA Topical Meeting on EUV Lithography, Monterey, CA, A.M. Hawryluk, R. Stulen, eds., 1993.
10. A.M. Hawryluk, N.M. Ceglio, "EUV Lithography Cost of Ownership Analysis," OSA Proceedings on Extreme Ultraviolet Lithography, Vol. 23, Optical Society of America, F. Zernike, D. Attwood, eds., 1994.
11. A.M. Hawryluk, G. Shelden, P. Troccolo, "EUVL Reticle Factory Model and Reticle Cost Analysis," Trends in Optics and Photronics, OSA TOPS on Extreme Ultraviolet Lithography, Vol. 4, G. Kubiak, D. Kania, eds., 1996.
12. A. M. Hawryluk and N. M. Ceglio, "EUV Lithography Cost of Ownership Analysis", OSA Proceedings on Extreme Ultraviolet Lithography, Vol. 23, Optical Society of America, F. Zernike, D. Attwood, eds., 1994.
13. D.G. Stearns, R. Risen, S.Vernon, "Multilayer Mirror Technology for Soft X-ray Projection Lithography, Applied Optics, Vol. 32, p. 6956, 1993.
14. D.P. Gaines, S.Vernon, G. Sommargren, D. Kanie, "Coating Strategy for Enhancing Illumination Uniformity in a Lithography Condenser," OSA Proceedings on Extreme Ultraviolet Lithography, Vol. 23, 1994.
15. K. Skulina, et al., "Beryllium Based Multilayers for Normal Incidence EUV Reflectance," OSA Proceedings on Extreme Ultraviolet Lithography, Vol. 23, Optical Society of America, F. Zernike, D. Attwood, eds., 1994.
16. A.M. Hawryluk, N.M. Ceglio, "Wavelength Considerations in Soft X-ray Projection Lithography," Applied Optics, Vol. 32, No. 34, 1993.
17. H. Kinoshita, et al., "Large Area, High Resolution Pattern Replication Using a Two-Aspherical Mirror System," Applied Optics, Vol. 32, No. 34, p. 7079, Dec. 1993.
18. M. Ito, et al., "Optical Technology for EUV Lithography," OSA TOPS on Extreme Ultraviolet Lithography, Vol. 4, 1996.
19. K. Murakami, et al., "Soft X-ray Projection Imaging at 4.5 nm using Schwarzchild Optics," OSA Proceedings on Extreme Ultraviolet Lithography, Vol. 23, Optical Society of America, F. Zernike, D. Attwood, eds., 1994.
20. D. Tichenor, et al., "10x Reduction Imaging at 13.4 nm," OSA Proceedings on Extreme Ultraviolet Lithography, Vol. 23, Optical Society of America, F. Zernike, D. Attwood, eds., 1994.
21. A. A. MacDowell, et al., "Extreme Ultraviolet 1:1 Ring-Field Lithography Machine," OSA TOPS on Extreme Ultraviolet Lithography, Vol. 4, 1996.
22. Bruning, et al., US Patent 5,353,322.
23. T. E. Jewell, K. Thompson, US Patent 5,315,629.
24. W. Sweatt, "Ring-field EUVL Camera with Large Etendue," OSA TOPS on Extreme Ultraviolet Lithography, Vol. 4., 1996.
25. W. Sweatt, "Ring-field EUVL Camera with Large Etendue" OSA TOPS on Extreme Ultraviolet Lithography, Vol. 4., 1996.
26. G.E. Sommargren, "Phase Shifting Diffraction Interferometry for Measuring Extreme Ultraviolet Optics," OSA TOPS on Extreme Ultraviolet Lithography, Vol. 4., 1996.
ANDY HAWRYLUK received his BS, MS, and PhD degrees in electrical engineering from the Massachusetts Institute of Technology. He is Director of Critical Projects at Ultratech Stepper, where his responsibilities include project planning and technical analysis of strategic programs, including optics, stage technologies, and new product development. Hawryluk holds 11 patents, primarily in applications of lithography, and has received numerous awards, including the DOE Award of Excellence in 1985. He has published numerous papers on lithography, reticles, and laser technology. Prior to his experience at Ultratech, Hawryluk held a number of positions at the Lawrence Livermore National Laboratory. Ultratech Stepper, 3050 Zanker Rd., San Jose, CA 95134, ph 408/321-8835, fax 408/325-6444.
NAT CEGLIO received his BA and BS degrees in physics from Columbia University in 1966 and 1967, respectively, and his MS and PhD degrees from MIT in 1969 and 1976, respectively. He is vice president of Engineering and Product Development at Ultratech Stepper. Prior to joining Ultratech, Ceglio worked at Lawrence Livermore National Laboratory from 1976 through 1996, leading the Advanced Microtechnology Program. Ceglio directed the Laboratory`s program in Giant Magnetoresistive Head development; he also directed the Laboratory`s program in extreme ultraviolet lithography.
DAVID A. MARKLE is vice president of Advanced Technology at Ultratech Stepper. A recognized scientific innovator within the semiconductor industry, he brings more than 25 years of experience to his position. During his 19-year tenure with Perkin-Elmer Corp., he led the invention and development of the Micralign, a scanning projection aligner, and the Micrascan step-and-scan, DUV lithography system. In addition to holding over a dozen patents, Markle developed technology used in research projects such as Skylab. In 1994, he was honored by the Optical Society of America with the David Richardson Medal for his contributions to the field of applied optics.