Optical Design and Engineering II, edited by Laurent Mazuray, Rolf Wartmann
`Proc. of SPIE Vol. 5962, 59620Y, (2005) · 0277-786X/05/$15 · doi: 10.1117/12.625199
`
`Proc. of SPIE 59620Y-1
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`
`
`Catadioptric Projection Lenses for Immersion Lithography
`
`Heiko Feldmann*, Aurelian Dodoc, Alexander Epple, Hans-Jürgen Rostalski, David Shafer†,
`Wilhelm Ulrich
`Carl Zeiss SMT AG, Carl-Zeiss-Strasse 22, 73447 Oberkochen, Germany
`† David Shafer Optical Design, 56 Drake Lane, Fairfield, Connecticut, 06824
`
`ABSTRACT
`
`
`Recently, the development of high NA lenses for immersion lithography turned from dioptric concepts to catadioptric
`design forms. The introduction of mirrors involves the new challenge to deal with the inevitable obscuration of either
`field or pupil. We review the strategies used in this regard for microlithography, while focussing on the two most favored
`ones, folded and inline concepts. Although the vignetting situation is more complicated for inline systems, we report
`progress in this field of optical design yielding similar system performance for inline and folded designs. Since inline
`optical systems are much easier to realize, these are the concept of choice.
`
`Keywords: Immersion lithography, High NA, Catadioptric designs
`
`
`1. INTRODUCTION
`
`
`Mirrors as a design means for lithographic lenses have a tradition of several decades. The first reason for their
`attractiveness was the simple unit magnification systems based on the concepts of Dyson and Offner [1]. Later reduction
`systems were required for lithography. Although some steps have been taken to implement reduction systems on a
`catadioptric basis [2], dioptric lenses started to dominate the development. For projection optics with i-line (365nm)
`lamps as light sources, several kinds of optical glass could be used for achromatization, and mirrors were not used for
`leading edge lithography.
`
`However, the industry demands to continuously shrink the dimensions of the electronic circuits structured by
`lithographic projection. This increasing request drives the trend to use smaller wavelengths and higher numerical
`apertures.
`
`After a long period dominated by dioptric reduction systems, mirrors were again discussed and used to enable color
`correction for 157nm lithography, where no second optical material is available and laser bandwidths can not be
`sufficiently narrowed. Also, early 193nm systems for broad-band lasers were based on catadioptric designs [3].
`
`With the advent of immersion, the circumstances changed again. The maximum NA now was extended far beyond 1.0,
`and the wavelength of 193nm can be used for the small features that previously seemed to require the wavelength of
`157nm (or even EUV). At this wavelength, dioptric systems that are made almost exclusively from fused silica have
`already demonstrated that full color correction is not imperative for the technology to work, although it may be
`economically advantageous to allow broader wavelength lasers.
`
`As a second benefit of mirrors, the textbooks on optical design mention their contribution to field curvature, which is
`inverse to the contribution of refractive lenses. This objective to control field curvature can also be achieved with a
`dioptric design using small negative and large positive lenses. However, during the first studies on extreme high NA
`immersion designs, it became clear that this dioptric concept leads to extremely large designs beyond approximately
`NA=1.1, using an extreme amount of optical material. So the most important motivation to look for catadioptric systems
`now is to find another way to satisfy the Petzval condition and in this way to obtain a compact system at reasonable
`costs.
`
`
`
`* Correspondence: email h.feldmann@smt.zeiss.com; phone +49 (0)7364 20 8388; fax +49 (0)7364 20 2334
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 1
`
`

`

`Proc. of SPIE 59620Y-2
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`In contrast to other design means like e.g. aspheres [4] or diffractive elements [5], mirrors can not be introduced
`smoothly into a design concept: they immediately pose the problem of separating the incident and the reflected beam.
`This new condition determines the basic layout of the system. It is a goal of this paper to give a coherent picture of
`catadioptric concepts, based on the principles invoked to handle the beam separation issue.
`
`
`2. SEVERAL WAYS TO REACH BEAM SEPARATION
`
`
`Probably the most frequently used strategy is to use an obscured pupil, where some small part of the optics is in front of
`a large pupil mirror. Sometimes a hole allows the image to pass through this large mirror if object and image have to be
`on opposite sides of the lens. For microlithography, such an obscured pupil would render the imaging difficult, and is
`therefore not wanted. For the rest of the paper, we discuss solutions with complete pupils.
`
`The optical designer can also try to ignore the problem of beam separation and pass it on to physics: a beamsplitter may
`be used to separate the beams. Of course, the optical design has to provide enough volume for the beamsplitter, but this
`can easily be achieved. This concept has already been studied for lithography. Two design examples are shown in Fig. 1.
`.
`One variant is to split the energy of the rays, but the energy loss of 75% is not acceptable for most applications. Another
`way is to use polarizing layers to separate the rays, using quarter wave plates to properly manipulate the polarization. In
`the latter case, the intensity loss of the system is not as severe, but only one polarization state can be used for imaging. A
`discussion of the properties of a beam-splitter coating is e.g. given by ref. [6].
`
`
`
`
`
`
`Fig. 1: Lithography lenses designed around beam splitting cubes
`Left: a large cube close to the aperture stop [7]
`Right: a design minimizing the size of the beam splitter cube [8]
`
`If the beam splitting element is close to a field conjugated plane, the system may be transformed to the second class of
`solutions: a system with off axis field, where the beam separation is facilitated by folding mirrors. The shifted image
`field now leads to a higher etendue and thus to more effort to correct the design, but the folding mirror is usually simpler
`to produce than a beamsplitter, and the system has the theoretical ability to image with less influence on polarization. In
`this sense, the folded design in Fig. 3 (Left) has been developed on the basis of Fig. 1 (Right).
`
`Working on different configurations for a folded design, a simple design strategy emerges: although it is clear that the
`used field of a folded design can not include the optical axis, the designer can try to keep the off-axis field as central as
`possible. The total etendue to be corrected is kept as small as possible. At the folding position, ideally there are two
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 2
`
`

`

`IMI 1
`
`IMI 2
`
`Proc. of SPIE 59620Y-3
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`planes conjugated to the field on opposite sides of the optical axis, as shown in Fig. 2. In reality, we have to deviate from
`this ideal situation, because we do not want to have an optical surface (folding mirror) directly in a field plane. Also, the
`intermediate image is usually not well corrected in order to allow aberrations to cancel between different subsystems.
`
`
`
`Fig. 2: Ideal arrangement for folding systems: Two field planes (IMI: object, image, or some intermediate image)
`adjacent to the folding position.
`When the NA increased further for immersion lithography, the beam separation became a more and more difficult issue.
`The recipe to move the folding mirrors as close as possible to a field plane seemed to be exhausted. Now the
`development took an interesting twist: A new complete relay system was added to the designs just to follow this recipe
`even more rigorously. Surprisingly, the system gain from this step greatly outbalanced the additional effort. Today, these
`folded catadioptric lithography designs reach numerical apertures higher than NA 1.2 within the same volume that
`previous concepts used around NA 0.9.
`
` second surprise was the attempt to add again one subsystem, this time a new catadioptric system on the yet unoccupied
`side of the central folding region. With the second mirror, a better control over color, field curvature and higher
`aberrations rendered the total system approximately as advantageous in terms of dimensions and weight as the one mirror
`version. All these approaches are illustrated in Fig. 3.
`
`
` A
`
`Fig. 3:
`Left: “Traditional” folded design with one intermediate image [9]
`Center: A whole new relay system is added to ease the folding (2 intermediate images) [10]
`Right: Even the addition of a second catadioptric group can be beneficial for the total system (3 intermediate
`images) [11]
`
`
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 3
`
`

`

`Proc. of SPIE 59620Y-4
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`
`From a theoretical point of view, these designs are very attractive: an easy to understand rule “fold close to the
`intermediate images” deals with the issue of beam separation and the designer can concentrate on his main task to find an
`optimum balance of aberrations.
`
`The practical realization of the systems in this section, however, meets some obstacles. For example, additional effort is
`required to manufacture an optical system with more than one optical axis. A part of this effort is shown in [12]. In
`addition, the tilted folding mirrors have high incidence angles. These reflections lead to a degradation of the polarization
`behavior, which breaks the rotational symmetry of the system. Also, most of the folded systems have an odd number of
`mirrors and thus flip the image. Then they are not compatible with reticles for traditional dioptric lenses.
`
`For these reasons we should not be satisfied with the results so far and see if we can find configurations without folding
`mirrors. We may be encouraged by the experience that for catadioptric microlithography lenses, a seemingly higher
`effort is sometimes the better solution.
`
`
`3. THE CLASS OF INLINE SYSTEMS
`
`3.1 An early inline system
`
`
`A rotationally symmetric optical system without folding elements has one common axis of rotational symmetry for all
`optical elements. We call these “inline systems”.
`
`Catadioptric inline systems have been studied for a long time. One of the earliest designs for lithography is shown in Fig.
`4. We can use it to discuss issues and strategies to deal with clearance control for inline systems. The design consists of
`two subsystems, the first one is catadioptric and the second one is only refractive. In the catadioptric relay system, the
`first mirror reflecting the light is close to the pupil plane, and is close to a negative power refractive lens element, making
`up a Schupmann achromat. The second mirror, reflecting the light again in the direction of the image, is situated between
`pupil and intermediate image, so that the footprint of the off axis field is decentered at this mirror. This allows to cut the
`second mirror, letting the rays pass it without obscuration of the pupil. The intermediate image has approximately the
`same size as the object, but it features an overcorrected field curvature and overcorrected axial color. The main task of
`the second subsystem is then to provide the high numerical aperture on the final image side, where for lithographic
`applications the wafer is located. A large collection of extensions of this principle based on the Schupmann achromat can
`be found in [13].
`
`
`
`
`
`Fig. 4: One of the earliest catadioptric inline designs [14]
`
`While the design in Fig. 4 exhibits all the ingredients we want for an inline catadioptric projection system, there are still
`open issues: the usable section of the field height is relatively small, and we would have to scale up the design by a large
`factor to obtain an image field of sufficient size. The standard field width of a lithographic scanner field is 26 mm on the
`image side, and rectangular fields are preferred over arc shaped fields (see [15] for a discussion of ring field).
`Lithographic projection lenses already have dimensions of more than about one meter length, and they have a large
`contribution to the total cost of the microlithographic steppers. Therefore scaling up the systems is prohibited, and other
`means to allow larger image fields are required. This challenge to control the vignetting becomes even more difficult
`when the aperture is increased.
`Therefore we studied systematically the different possibilities to construct modified inline systems where better ways of
`beam separation allow a larger field size without increased system length.
`
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 4
`
`

`

`............................ :"1'
`
`........................ ,.
`
`Proc. of SPIE 59620Y-5
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`3.2 Try to separate the beams at intermediate images
`
`
`First, also here we can try to use the rule “separate the beams directly at intermediate images”.
`
`Such an attempt is shown on the left side of Fig. 5. We put one intermediate image (IMI 2) directly onto a mirror (M2),
`construct the rest of the optical system in such a way that another intermediate image (IMI 1) is nearby, on the other side
`of the optical axis. Symmetry considerations suggest that the opposite mirror (M1) then is close to a system pupil or a
`congjugated plane. Now a severe problem is encountered: how can we get the rays past this pupil mirror? Since the pupil
`mirror is necessarily centered about the optical axis, the only way is to position the whole outline of the passing by beam
`off axis, close to another intermediate image. The only free variable we have for this purpose is the curvature of the
`second mirror, the one close to the second intermediate image. But since this mirror is at a field conjugated position
`itself, it only acts on the pupil position and can not help to create another intermediate image.
`
`Optical design consists of balance and compromise. Still, we can try to optimize the position of the intermediate images
`to put them as close as possible to the mirrors as shown in Fig. 5 (Right). Now, both mirrors are in between a pupil and a
`field plane, and the usable range of field heights is already strongly increased with respect to the original inline design in
`Fig. 4.
`
`
`
`IMI 1
`
`M2
`
`IMI 2
`
`IMI 1
`
`M1
`
`?
`
`?
`
`IMI 2
`
`
`
`Fig. 5:
`Left: The method of adjacent field planes works well for one mirror, but not for the opposite mirror.
`Right: A compromise to apply the adjacent image construction for both mirrors.
`
`In doing so, we have to give up the pupil mirror, where it was easy to implement color correction as in the Schupmann
`achromat. A contribution to axial color away from the system pupil inevitably introduces lateral color. This slightly
`complicates the situation. A solution is the use of (at least) two Schupmann constructions positioned on two sides of the
`system pupil. Then their contributions to the lateral color can be balanced. The detailed attempts to introduce these two
`color correcting mirrors have brought up a new system: by adding a new catadioptric relay system, the four mirrors allow
`several combinations of color correcting means. An example of this discovery is shown in Fig. 6 (Top). Again, the
`additional effort to add a complete new relay system is outweighted by the gain in system effort, measured e.g. in mass
`or volume.
`
`The additional relay system may also be designed as a purely dioptric one, as illustrated at the bottom of Fig. 6. Here,
`however, the two mirrors have to bear the whole contribution of field curvature for the two refractive subsystems.
`
`The lesson that was surprising to us was again: the introduction of additional intermediate images, which enhances the
`system complexity at the first glance, does not increase volume or weight of the lens. In some cases, the results even
`surpass the “simpler” designs.
`
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 5
`
`

`

`I
`I
`I
`I
`I
`I
`
`Proc. of SPIE 59620Y-6
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`
`Fig. 6:
`Top: A 4 mirror concept with high symmetry allows good color correction without a dedicated pupil mirror [16]
`Bottom: Two-mirror inline system with additional relay optic. The position of the intermediate images can be
`optimized to reduce the vignetted field [17]
`3.3 Increase intermediate images
`
`
`
`
`Looking at the pupil mirror in Fig. 4, a possible second procedure to improve the vignetting comes up. The size of the
`intermediate image is a free variable in the system layout. Is it possible to make it larger, and thus shifting the inner field
`point further off axis?
`
`Indeed this approach can be implemented. A direct modification of Fig. 4, increasing the intermediate image
`magnification from approx. 1:1 to 2:1 leads to the design shown in Fig. 7. Two drawbacks are visible: Large refractive
`elements are required to capture the divergent chief rays at the intermediate image. Also, the pupil mirror is smaller, and
`a smaller contribution to axial color correction must be accepted.
`
`
`
`
`
`Fig. 7: An increase of the intermediate image size eases the vignetting control [18]
`Again, we try to add something in order to make the system simpler. Mirrors do not require transparent lens material,
`which is extremely difficult to produce in lithographic quality. Therefore it is easier to produce mirrors of larger
`diameter. An addition of two mirrors to Fig. 7 yields Fig. 8. The large intermediate image with diverging chief rays is
`kept, and two additional mirrors perform the task of bending the rays back to the axis, replacing the huge refractive
`lenses.
`
`
`
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 6
`
`

`

`Proc. of SPIE 59620Y-7
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`
`Fig. 8: Four mirror designs with large intermediate image [19] and a recent extension [18]
`
`The control of vignetting is a challenge for inline catadioptric designs, but of course the usual requirements for a good
`optical system have to be applied as well. In the end, we have to correct not only the field curvature, but at least all
`monochromatic aberrations. One advantage of designs with multiple mirrors is that in many configurations, the mirrors
`already can bear some of this correction power. A review of purely catoptric design concepts is given in [20], some
`recent advances in this field are shown in [21]. It may even be advantageous to use more than 4 mirrors for this purpose.
`Then, one can regard a high NA catadioptric system as a catoptric one with a refractive focussing group. An example is
`given in Fig. 9.
`
`
`
`
`Fig. 9: The 6-mirror catadioptric system can be imagined as a derivation from a pure mirror system, with an
`additional focussing group [22]
`
`CONCLUSIONS
`
`
`The considerations above focus on methods to deal with the challenge that is unique to catadioptric systems, the proper
`separation of ray paths incident on the mirrors and reflecting from them. For inline systems, this entanglement is stronger
`than if folding is allowed. However, a careful study of these methods has allowed us to find inline design concepts that
`can compete with folded designs both in NA and in field size, with comparable system dimensions.
`
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 7
`
`

`

`Proc. of SPIE 59620Y-8
`
`Downloaded from SPIE Digital Library on 30 Jul 2010 to 150.135.248.118. Terms of Use: http://spiedl.org/terms
`
`
`
`ACKNOWLEDGEMENTS
`
`
`The authors would like to thank Hans-Jürgen Mann and Susanne Beder for discussion, and especially Helmut Beierl for
`his support in the analysis of patent literature. We thank Marco Lukoschek for help with the preparation of the figures,
`and Winfried Kaiser and Andreas Zeiler for a critical review of the manuscript.
`
`
`REFERENCES
`
`
`1. D. M. Williamson, “Catadioptric Microlithographic Reduction Lenses”,OSA Proc. Vol. 22, 428-37 (1994)
`2. D. Shafer et. al. US patent application 4 747 678
`3. W. Ulrich et. al., “Trends in optical design of projection lenses for UV- and EUV-lithography”, Proc. SPIE Vol.
`4146, 13-24 (2000)
`4. W.Ulrich et. al., “The development of dioptric projection lenses for deep ultraviolet lithography”, Optical Review
`Vol.10, 233-240 (2003)
`5. H.-J. Rostalski et. al., “Use of diffractive optical elements in lithographic projection lenses”, this Volume
`6. R.N. Singh et. al. “High numerical aperture optical designs”, IBM J. Res. Develop. Vol. 41, 39-48 (1997)
`7. D. M. Williamson, US patent application 5 212 593
`8. W. Ulrich et. al., US patent application 2004 0 263 955
`9. D. Shafer et. al., US patent 6 665 126
`
`10. Y. Omura et. al., US patent application 2003 0 011 755
`11. D. Shafer et. al., WO patent application 2005 040 890
`12. Y. Ohmura et. al., “Catadioptric lens development for DUV and VUV projection optics”, Proc. SPIE Vol. 5040, 781-
`788 (2003)
`13. Terasawa et. al., US patent application 2002 0024741
`14. Suenaga et. al., European patent 1 069 448
`15. Y. Omura et. al. , WO patent application 2004 107011
`16. Takahashi, European patent application 1 336 887
`17. D. Shafer et. al., WO patent application 2005/069055
`18. Carl Zeiss patent pending
`19. D. Shafer et. al., US patent 6 636 350
`20. J.M. Rodgers, “Unobscured mirror designs”, Proc. of SPIE Vol. 4832, 33-60 (2002)
`21. H.-J. Mann et. al., “Reflective high-NA projection lenses”, this Volume
`22. R. Hudyma, WO patent application 2002 044 786
`
`
`
`
`
`CARL ZEISS V. NIKON
`IPR2013-00362
`Ex. 2005, p. 8
`
`