Projection. optical lithography

Transcription

1 Projection optical lithography by Mordechai Rothschild Projection optical lithography has had a remarkable history and, most probably, it will have an equally successful future for at least another decade. To date, it has met all the major challenges posed by the semiconductor industry roadmap. In order to do so, it has undergone important transformations, and has greatly expanded the frontiers of the science and engineering of optics. This paper will review the most recent developments, including transitioning to the short wavelengths of 193 nm and 157 nm, moving toward ultrahigh numerical apertures facilitated by liquid immersion, and incorporation of a range of resolution-enhancing techniques such as optical proximity correction and phase-shifting masks. Lincoln Laboratory, Massachusetts Institute of Technology, Lexington, MA 02420, USA rothschild@ll.mit.edu Projection optical lithography has been the mainstream technology in the semiconductor industry for more than two decades. Despite prognostications to the contrary, it has been able to keep up with the expectations of Moore s Law and the semiconductor industry s roadmap 1. In fact, it has enabled the shrinking of critical dimensions to the sub-100 nm region, and current trends will keep it on course to at least 45 nm, i.e. to at least the year It is plausible to expect that new inventions will keep optical lithography as the dominant process well into the next decade. While other lithographies may face challenges in throughput, yield, or cost, optical lithography has faced its toughest hurdles in enabling high resolution. By its very definition, optical lithography employs photons in the optical regime, and the wave nature of light dictates that diffraction limits the patterning resolution to a certain fraction of the wavelength. This appears to be a fundamental limit that cannot be overcome. In light of this argument, the continued success of optical lithography is indeed remarkable 2. This paper will focus on the means that optical lithography has employed to enable ever-increasing resolution and on the prospects to continue doing so in the future. As will be noted throughout, materials issues have played a key role in the evolution of optical lithography. The essential components of projection optical lithography are represented in Fig. 1. As in any lithography, the goal is to generate a desired pattern in a thin layer of resist on the Si wafer. In order to do so, a related pattern is 18 February 2005 ISSN: Elsevier Ltd 2005

2 illumination, and patterns on the photomask. Diffraction theory places a lower limit of 0.25 on k 1. It is important to note that in eq 1, LW is half of a dense pair of lines and spaces. This pattern geometry is the more challenging from the optics standpoint. For semi-isolated lines, i.e. when the pitch is large, LW can be made significantly smaller by incorporating special process steps with equivalent k 1 of 0.05 and less 3,4. With these considerations in mind, eq 1 identifies the directions in which optical lithography must evolve in order to keep up with the demand of reducing LW: decreasing the lithographic wavelength λ, increasing the numerical aperture NA, and decreasing the parameter k 1. All three approaches have been implemented to various degrees, and are discussed below. Fig. 1 Schematic of the optical projection system. The main components are the laser, illuminator, photomask, projection lens, and photoresist-coated wafer. formed on a photomask, and a sophisticated imaging system projects the photomask pattern onto the photoresist. Typically, the pattern on the photomask is larger than the final pattern in the photoresist, and the projection lens system provides a demagnification ratio; 4x in most systems. In addition to the photomask, the projection lens, and the photoresist, at least two other components must be included in any analysis of an optical lithographic system: the radiation source and the illumination geometry. A convenient if not completely accurate expression that describes the ultimate resolution in optical lithography is: LW = k 1 λ / NA (1) In eq 1, LW is the printed linewidth, λis the exposure wavelength, and NA is the numerical aperture of the projection lens, defined as nsinθ (n is the refractive index of the medium above the photoresist, and θ is the largest angle of converging rays subtended at the photoresist). Since the optics are usually in air, n is 1, and therefore the upper limit on NA is also 1. The parameter k 1 represents the composite effect of such disparate subsystems as photoresist response, Wavelength The choice of lithographic wavelength is primarily determined by the availability of powerful radiation sources. Until the late 1980s, optical lithography was performed with highpressure Hg discharge lamps operating at 436 nm (G line) and 365 nm (I line). Transitioning to shorter wavelengths (eq 1) was first performed with Hg discharge lamps operating at ~250 nm, but was soon replaced with more powerful and monochromatic excimer lasers. The first wave of excimerbased lithographic systems operated with krypton fluoride lasers at 248 nm. These so-called deep-ultraviolet (DUV) systems have been the backbone of the semiconductor industry for over a decade until very recently. In the beginning, they were used to print 350 nm lines and spaces, and have been employed in patterning down to 130 nm lines and spaces. However, excimer lasers encompass a whole class of discharge gases and corresponding emission wavelengths. Therefore, it is natural to explore the feasibility of lithography at 193 nm, the wavelength of argon fluoride excimers 5. The last two years have indeed witnessed the introduction into manufacturing of the first-generation 193 nm systems. These are used in patterning 90 nm lines and spaces, and are expected to provide the platform for 65 nm and even 45 nm lithography 6-9. The transition from 248 nm to 193 nm lithography has not been straightforward, however. A number of challenges had to be resolved, mostly related to materials issues. The first challenge involved the choice of material for the lens elements. At the short wavelength of 193 nm, the February

3 available transparent materials are limited to high-purity synthetic fused silica (SiO 2 ) and several crystalline fluorides, the leading candidate being CaF 2. While fused silica is preferred for practical reasons 6, it has been observed that prolonged 193 nm irradiation causes two unwelcome changes in the material 10. One is reduced transmission because of the formation of E color centers at ~ 213 nm. While this effect can be reduced through careful purification and preparation of the material 11,12, the second phenomenon has proven more troublesome. Complex photoinduced changes in material structure appear to cause birefringence, leading to quasirandom time- and space-varying changes in the state of polarization of the transmitted wave. Since the projection lens must be near-perfect and aberration-free, such laserinduced birefringence is unacceptable. The underlying mechanisms for the formation of birefringence are still not fully understood, but they involve at least two stressinducing processes: laser-induced compaction and laserinduced rarefaction 13,14. At present, it is still uncertain to what degree these processes can be reduced or eliminated. Consequently, most 193 nm lithographic systems include a number of elements made of CaF 2, especially in places where the laser intensity is highest. The short wavelength of 193 nm can induce undesired photochemistry in other materials as well. An important element in projection optical lithography is the use of thin transparent membranes or pellicles, which protect the photomask from particulate contamination. Pellicles are critical to maintaining high manufacturing yield in the semiconductor industry. They must satisfy a number of technical requirements, including photochemical stability under extended irradiation. At 193 nm, they are made of fluoropolymers, and certain variants have shown unacceptable rates of photochemical darkening. Still, several recent varieties have met most requirements 15,16. A third wavelength-related challenge has been the design of new classes of photoresists. The chemical mechanism underlying the operation of 193 nm and 248 nm photoresists is similar. It involves the photochemical generation of acids, which then catalytically deprotect the polymer host to make it soluble in aqueous base developers. The difficulty at 193 nm lies in the fact that the photoresist cannot be too thin if it is to be used in pattern transfer into underlying layers. On the other hand, the traditional 248 nm photoresists, which are based on aromatic polymers, are much too opaque at 193 nm. If they are to be used, their thickness is limited by their low transparency to impractically small values. Therefore, large efforts have been applied to the development of new, more transparent photoresist platforms at 193 nm 17. The most common resists today use acrylatebased polymers. These are transparent enough, but their plasma etch resistance is not as high as that of 248 nm photoresists. As mentioned above, the difficulties with 193 nm lithography have been largely resolved, or at least circumvented, in the last few years. It is therefore logical to investigate the next shortest excimer laser wavelength of 157 nm, that of the molecular fluorine laser 18. At this wavelength, even atmospheric oxygen is too absorptive and the optical path must be purged with an inert gas such as nitrogen or argon (the term vacuum ultraviolet (UV) is a misnomer, as there is no need for vacuum). The materials issues encountered at 157 nm are the familiar ones dealt with at 193 nm, except that they are now greatly amplified. Fused silica is eliminated from the class of transparent lens materials, and CaF 2 is the only practical option 18. CaF 2 does not undergo photoinduced compaction or rarefaction. However, since it is crystalline and not amorphous, it has an intrinsic birefringence 19. Fortunately, the intrinsic birefringence is deterministic, and can be compensated for by preparing lens elements with different crystalline orientations 20. An important practical difficulty is the relatively low yield of the growth process for high-quality, large-diameter CaF 2 crystals, such as those required in 157 nm lithographic lenses. Since fused silica is opaque at 157 nm, the traditional 6.3 mm thick transparent substrate of photomasks must also be replaced. Indeed, it has been shown that if UV-grade fused silica, which contains ~1000 ppm OH groups, is prepared instead with a certain amount of F groups, it is transparent enough to serve as a photomask substrate 21. The development of fluorine-doped fused silica has been a major achievement. It is limited, however, to photomask substrates. Wider use as a lens material is probably impractical because of the large rates of laser-induced compaction (Fig. 2) 22. The 157 nm photochemical darkening of fluoropolymers for pellicle applications has proven to be a major obstacle in developing lithography at this wavelength. The transmission of 0.8 µm thick pellicles must remain constant for hundreds 20 February 2005

4 developed where the polymer matrix is based exclusively on fluoropolymers 26. The photoacid-induced chemically amplified deprotection is still operative, but the polymer matrix has to be re-engineered 27,28. Patterning results have shown good progress with such photoresists 28,29, but more development will be required to reduce the observed line edge roughness 30,31 and to increase the plasma etch resistance The cumulative difficulties encountered by 157 nm lithography availability and cost of lens material, lack of organic pellicles, and slow progress in photoresist development have conspired to dampen the enthusiasm for this technology. The years 2003 and 2004 have witnessed a decrease in resources allocated to solving the materials challenges of 157 nm lithography, while an alternative approach, that of increasing NA, has gained momentum. Fig. 2 Laser-induced compaction in two grades of fluorine-doped fused silica, as evidenced by 633 nm stress birefringence. (a) Irradiation at 157 nm causes birefringence much more rapidly than at 193 nm. (b) Spatial mapping of 157 nm induced birefringence, showing that the highest stress occurs in a ring-shaped region surrounding the irradiated area. Numerical aperture All optical lithographic systems operate in a gaseous ambient, either air or nitrogen, whose refractive index n is very close to 1. Therefore, increasing NA has meant increasing the acceptance angle of the lens, θ. This process has had to of Joules per square centimeter. Despite all efforts to date, all candidate materials have displayed degraded transmission within <10 J/cm 2 (Fig. 3) 23. The photochemistry causing these changes is understood, in general terms, as being a result of radical formation followed by recombination 24. Still, no practical way has been found to reduce the magnitude of this effect by the required one to two orders of magnitude. Alternative approaches have been explored, including the use of inorganic pellicles. These would be relatively thick, ~0.8 mm, sheets of fluorine-doped fused silica, the same material used in photomask substrates. While the transparency requirement is satisfied with the inorganic pellicles, there are significant engineering challenges that must be overcome, such as mounting and gravitational sag. Although these issues can be resolved 25, they involve considerable effort and resources. Not surprisingly, the photoresists used at 248 nm and 193 nm are too opaque for 157 nm use 18. Indeed, any hydrocarbon-based polymer is too absorptive at this wavelength. Therefore, new photoresists have been Fig. 3 Durability test results of candidate organic pellicle materials at 157 nm. (a) A typical in situ transmission plot of a thin film (left) is converted into thicknessnormalized absorbance units and the dose sufficient to induce a 1% (or 10%) change in the transmission of a 0.8 µm thick membrane is defined as the 1% (or 10%) lifetime (right); (b) Compilation of the 10% lifetime obtained in this fashion for nearly 50 candidate materials; none reached even 10 J/cm 2, while the minimum required value is at least ten times higher. February

5 overcome significant challenges in optical design and fabrication because the lens must be near aberration-free and the image size must be kept large, ~4 mm x 26 mm. Despite these difficulties, the NA of projection systems has grown steadily, from 0.5 in around 1990 to over 0.8 in 2004, with plans to exceed 0.9 in the future 6,7. However, increasing NA in this manner has become more and more difficult with diminishing returns in resolution. A more radical approach has been gaining acceptance in the last few years, namely changing the refractive index n by 40-50% in one step. This can be achieved by replacing the gas ambient with a condensed medium, such as a transparent liquid. The concept of liquid immersion to enhance resolution is more than a century old, but until now it has been applied only to the field of microscopy and only at longer wavelengths. Work toward its implementation in lithography at 193 nm (or 157 nm, for that matter) started in earnest in Liquid-immersion lithography has not been introduced into manufacturing yet, but it is widely accepted as the next major milestone in the evolution of optical lithography. Immersion at 193 nm can be accomplished with high-purity water, whose absorption coefficient is small enough for this purpose 35. Its refractive index at nm is (measured at 21.5 C) 36, so that for a given sinθ the NA is increased by 44%, with a parallel increase in resolution (eq 1). Although the potential benefit of liquid immersion is significant 37, the concept of filling the space between the photoresist and the lens is radical enough that many engineering issues must be resolved first. Of course, new optical designs must be developed. Then, the whole subsystem of liquid dispensing and retrieval must be engineered so as to avoid the formation of bubbles or temperature and pressure gradients Furthermore, the liquid cannot impact on the performance of the photoresist, and therefore leaching of photoresist components, dissolution, swelling, and other phenomena must be monitored carefully 41,42. The liquid cannot interact with the lens either. Most optical designs use CaF 2 as the last optical element because of the relatively high fluence there (see previous section), and this material is readily attacked by water. New protective coatings must be implemented to prevent such reactions from taking place. Fig. 4indicates that, at present, such coatings are susceptible to laserinduced damage, as observed with ellipsometry 43, and therefore further improvements will be required. Finally, the yield of the lithographic process cannot be reduced by the introduction of new types, or larger numbers, of defects. It should be noted that the density of liquids is ~1000 times higher than that of atmospheric gas. Purity control of the liquid must therefore be 1000 times more stringent to prevent contamination of, or particle formation on the wafer. Despite the concerns listed above, and many more not mentioned here, all the initial experimental and simulation results indicate that 193 nm, water-immersion lithography will be relatively easy to implement. By the end of 2005, several full-field prototype systems around the world will be generating large amounts of yield data. Unless major effects are discovered by then, the first manufacturing systems may be in production by The news to date has been encouraging enough that efforts are being made to extend immersion lithography beyond water at 193 nm. New liquids are being investigated, both for higher-index 193 nm use and for 157 nm immersion 44. These activities are preliminary but, if ultimately successful, they may enable further extensions of optical lithography. Resolution-enhancing techniques Eq. 1 indicates that, in addition to changes in λ and NA, resolution can also be enhanced by reducing the k 1 factor. Indeed, k 1 has been decreasing steadily, from ~0.8 in the 1980s to less than 0.4 in the more advanced processes today. Values of 0.3 can be expected in a few years. This trend can be attributed to the introduction of aggressive forms of resolution-enhancing techniques (RETs), including off-axis Fig. 4 Spatial mapping of the ellipsometric angle of a protective coating irradiated at 193 nm in a water ambient for a total dose of 4.5 MJ/cm 2. At this stage of irradiation, no other effects on the coating are observed. This is not surprising, since the laser-induced effect is relatively small, and ellipsometry is a very sensitive technique. Nevertheless, the unmistakable change in may represent the incipient stage of more pronounced damage that may occur after longer irradiation. 22 February 2005

6 illumination, phase-shifting masks (PSMs), and optical proximity correction (OPC) 45. To obtain the highest resolution, illumination of the photomask is not performed by a disc-shaped source. Rather, the angular distribution of the illumination beam may have a complex structure, such as an annulus, a set of off-axis circles, or even a continuously varying profile 46,47. The specific illumination geometry is designed to enhance the contrast in the wafer plane of the photomask features whose dimensions are most critical 48. The introduction of off-axis illumination is coupled with complex structures on the photomask. The pattern on the photomask is no longer just an enlarged version of the desired pattern on the wafer. That approach was satisfactory when the lithographic wavelength was much smaller than the ultimate pattern. However, in the 1990s optical lithography crossed the threshold of subwavelength lithography, and the more advanced systems today print at half-wavelength dimensions 2. The wave nature of light must be included in planning a lithographic process. Complex computational models have been developed that calculate the inverse problem of diffraction. Given the desired aerial image in the image plane (the photoresist), they calculate the pattern that must exist in the object plane (the photomask). Often, the resulting pattern on the photomask offers only scant resemblance to the final image. Various serifs, elbows, and subwavelength-assist features are fabricated on the photomask, whose purpose is to counter the loss of contrast that diffraction imposes on the image (Fig. 5) 2,46. This OPC process has proven to be a powerful device in enabling Fig. 5 Schematic of OPC. The spatial-frequency effect distorts the image patterned on the photoresist by rounding sharp corner features and shortening narrow line ends. The addition of photomask correction features such as mousebites, hammerheads, serifs, and scattering bars produces an image closer to the desired result. subwavelength lithography, in effect reducing the value of the k 1 factor. The challenges facing OPC are largely in the realm of developing fast and effective computational algorithms for arbitrary geometries PSMs provide an additional lever to enhance the resolution of optical lithography. These are photomasks with structures that manipulate not only the amplitude of the transmitted waves but also their phase. Controlling the phase enables constructive or destructive interference at desired locations in the image plane, thus sharpening or dulling the contrast as desired. There are numerous variants of PSMs, each designed to provide more process latitude in printing specific features 52. Some involve the fabrication on the photomask of three-dimensional structures in transparent layers, in addition to or without absorptive patches ; while others employ semitransparent thin films which attenuate the amplitude while retarding the phase in specified ways 56,57. The challenges encountered here are numerous. They include new fabrication processes for the photomasks, new materials as absorbers or phase retarders, and the development of robust mask inspection and repair methods The cumulative effect of off-axis illumination, OPC, and PSMs has been dramatic. The k 1 factor has been halved and photoresists are routinely patterned at half-wavelength dimensions, while the lithography industry maintains its high manufacturing yield, reliability, and throughput. There has been a price to pay, however, in the dramatic increase in cost of advanced photomasks, which is directly traceable to the complexity of preparing them and the relatively low yield of the mask-making process. Many applications require a set of photomasks for a small number of wafers. For these, state-ofthe-art photomasks may become prohibitively expensive. In the future, their practitioners may explore new variants of optical lithography, including the use of templated masks 61 and even maskless lithography 62. Summary and future outlook Projection optical lithography has shown unsurpassed vitality for several decades. It has retained its leading role in the semiconductor industry, and it probably will continue doing so for at least several more years. It has adapted to new and difficult requirements, and in order to meet them it has adopted new strategies: the wavelength has been reduced into the deep, deep UV, and in the future it may extend into the vacuum UV; it is on the verge of introducing liquid February

7 immersion into the lithographic process; and it has implemented a range of illumination and imaging techniques to enhance the resolution beyond that enabled by classical optics. During the 1990s, optical lithography started patterning at subwavelength dimensions. In the near future, it will pattern at the recently unthinkable one-third wavelength, and possibly even one-fifth wavelength dimensions. This means that a combination of short wavelength, high-index immersion fluids, and aggressive RETs can satisfy the expected industry needs to at least the year 2013 (the 32 nm node). New lithographic strategies, such as multiple exposures and lithography-on-a-grid, may extend its utility even further. Optical lithography has demonstrated a unique combination of attributes that have kept its leading position in the semiconductor industry. Besides high resolution, it also provides high throughput and high yield in a costeffective manner. Alternative technologies will have to demonstrate a similar combination if they are to supplant it in the future. MT Acknowledgments I wish to thank my colleagues at MIT Lincoln Laboratory for their results and concepts presented in this paper. This work was sponsored by the Advanced Lithography Program of the Defense Advanced Research Projects Agency under Air Force Contract F C and by a Cooperative Research and Development Agreement between MIT Lincoln Laboratory and SEMATECH. Opinions, interpretations, conclusions, and recommendations are those of the author, and do not necessarily represent the view of the US Government. REFERENCES See also: Rothschild, M., et al., Lincoln Lab. J. (2003) 14, Yost, D., et al., J. Vac. Sci. Technol. B (2002) 20 (1), Fritze, M., et al., IEEE Circuits Devices Magazine (2003) 19 (1), Rothschild, M., and Ehrlich, D. J., Proc. SPIE (1988) 922, Matsuyama, T., et al., Proc. SPIE (2004) 5377, Namba, A., et al., Proc. SPIE (2004) 5377, Hsu, S.-H., et al., Proc. SPIE (2004) 5377, Fuard, D., and Schiavone, P., Proc. SPIE (2004) 5377, Rothschild, M., et al., Appl. Phys. Lett. (1989) 55 (13), Levy, D. H., et al., J. Appl. Phys. (1993) 73 (16), Smith, C. M., et al., Appl. Opt. (2000) 39 (31), Allan, D. C., et al., Proc. SPIE (2004) 5377, Algots, J. M., et al., Proc. SPIE (2004) 5377, Rothschild, M., and Sedlacek, J. H. C., Proc. SPIE (1992) 1674, Grenville, A., et al., Proc. SPIE (1998) 3236, Allen, R. D., et al., IBM J. Res. Dev. (1997) 41 (1-2), Bloomstein, T. M., et al., J. Vac. Sci. Technol. B (1998) 16 (6), Burnett, J. H., et al., Phys. Rev. B (2001) 64, (R) 20. Sewell, H., et al., J. Photopolym. Sci. Technol. (2002) 15 (4), Liberman, V., et al., J. Vac. Sci. Technol. B (1999) 17 (6), Liberman, V., et al., Laser-induced birefringence in modified fused silica: 157 nm and 193 nm irradiation, Presented at 1 st International Symposium on Immersion and 157 nm Lithography, Vancouver, BC, Grenville, A., et al., Proc. SPIE (2002) 4691, Lee, K., et al., Proc. SPIE (2004) 5377, De Bisschop, P., et al., Proc. SPIE (2004) 5377, Kunz, R. R., et al., J. Photopolym. Sci. Technol. (1999) 12 (4), Crawford, M. K., et al., Proc. SPIE (2003) 5039, Houlihan, F. M., et al., Proc. SPIE (2004) 5376, Li, W., et al., Proc. SPIE (2003) 5039, Yamazaki, T., et al., Proc. SPIE (2003) 5039, Otoguro, A., et al., Proc. SPIE (2004) 5376, Kawaguchi, Y., et al., Proc. SPIE (2003) 5039, Yamada, S., et al., Proc. SPIE (2003) 5039, Switkes, M., and Rothschild, M., J. Vac. Sci. Technol. B (2001) 19, 2353; Switkes, M., and Rothschild, M., Proc. SPIE (2002) 4691, Switkes, M., et al., Proc. SPIE (2003) 5040, Burnett, J. H. and Kaplan, S. G., J. Microlith. Microfab. Microsyst. (2004) 3 (1), Lin, B.-J., Proc. SPIE (2004) 5377, Owa, S., et al., Proc. SPIE (2004) 5377, Streefkerk, B., et al., Proc. SPIE (2004) 5377, Honda, T., et al., Proc. SPIE (2004) 5377, Hinsberg, W., et al., Proc. SPIE (2004) 5376, Kishimura, S., et al., Proc. SPIE (2004) 5376, Liberman, V., et al., Long-term 193 nm laser irradiation of thin-film-coated CaF 2 in the presence of H 2 O, Presented at 1 st International Symposium on Immersion and 157 nm Lithography, Vancouver, BC, French, R. H., et al., Structure-composition-property relations for 157 nm immersion lithography fluids, Presented at 1 st International Symposium on Immersion and 157 nm Lithography, Vancouver, BC, Wong, A. K., Resolution Enhancement Techniques in Optical Lithography, SPIE Press, Bellingham, WA (2001) 46. Fritze, M., et al., Lincoln Lab. J. (2003) 14, Heil, T., et al., Proc. SPIE (2004) 5377, Zhang, G. and Hansen, S., Proc. SPIE (2004) 5377, Cobb, N., and Granik, Y., Proc. SPIE (2004) 5377, Schacht, J., et al., Proc. SPIE (2004) 5377, Choi, S.-H., et al., Proc. SPIE (2004) 5377, Levenson, M. D., Jpn. J. Appl. Phys. (1994) 33, Conley, W., et al., Proc. SPIE (2004) 5377, Hotta, S., et al., Proc. SPIE (2004) 5377, Levenson, M. D., et al., Proc. SPIE (2004) 5377, Chang, C.-H., et al., Proc. SPIE (2004) 5377, Liberman, V., et al., Proc. SPIE (2002) 4691, Bingham, P. R., et al., Proc. SPIE (2004) 5375, Heumann, J. P., et al., Proc. SPIE (2004) 5375, Amano, T., et al., Proc. SPIE (2003) 5256, Fritze, M., et al., J. Vac. Sci. Technol. B (2001) 19 (6), Watson, G. P., et al., J. Vac. Technol. B (2004) 22 (6), February 2005