157nm Lithography with High Numerical Aperture Lens for the 70nm Technology Node

Transcription

1 157nm Lithography with High Numerical Aperture Lens for the 70nm Technology Node Toshifumi Suganaga*, Noriyoshi Kanda, Jae-Hwan KIM, Osamu Yamabe, Kunio Watanabe, Takamitsu Furukawa, Seiro Miyoshi and Toshiro Itani Semiconductor Leading Edge Technologies Inc., Yokohama , Japan Julian Cashmore and Malcolm Gower Exitech Ltd, Yarnton, Oxford OX5 1QU, UK Abstract157nm lithography is, being investigated for the sub-70nm technology node of semiconductor devices. Many efforts have been reported on the exposure tool, the F 2 laser, the resist materials, the resist processing and the mask materials". A critical component for the success of this 157nm lithography is the availability of high numerical aperture (NA) lenses that lead to higher resolution capability and higher process margin. In this article, we describe our recent evaluation results of a high precision 157nm Microstepper with 0.85 NA lens combined with simulation analysis of the lithographic performance. The details of the evaluation results discussed here include the resolution limit of the high NA lens and the possible effects of intrinsic birefringence upon the lithographic performance. 1 INTRODUCTION The first high numerical aperture (0.85NA) 157nm Microstepper has been operational at Semiconductor Leading Edge Technologies Inc. (Selete) since early December 2001 in the Advanced Technology Research Department at Yokohama, Japan. The Microstepper is being used by the Optical Lithography Group at Selete for 157nm resist performance testing. In addition to the primary role the performance of the Microstepper tool platform and the imaging performance of the projection optics have both been investigated and are reported here. This paper discusses first the tool platform design and the various sub-system performances and follows with a description of the imaging performance obtained which supports sub-70nm feature resolution. Finally results of early studies into the effects of intrinsic birefringence on imaging performance and a comparison between simulation and experimental results are presented. These findings may have an impact on the design and qualification of future high NA lenses as well for in-situ correction of existing systems. 2 EXPOSURE TOOL PLATFORM The exposure tool platform for the high NA lens is a 2nd generation 157nm Microstepper manufactured by Exitech Ltd, UK. This tool has been developed from their earlier MS-157 Microstepper systems. These systems that were available early 2000 operated at a maximum numerical aperture of The main specification for the tool is given in Table 1 below and this section gives a description of the Microstepper. *Correspondence: suganaga@selete.co jp;;phone:(+81) ; fax: (+81) ; http// Selete, Yoshida-cho 292, Totsuka-ku, Yokohama, , JAPAN; kanda@selete.co.jp; (+81) yamabe2@selete.co jp;phone:(+81) ; watanabe7@selete.cojpphone:(+81) kim5@selete.co.jp; phone: (+81) , furukawa2@selete.co.jp; phone: (+81) miyoshi@selete.co.jp; phone: (+81) , itani2@selete.co.jp; phone: (+81) j.cashmore@exitech.co.uk; phone: (+44) , m.gower@exitech.co.uk; phone: (+44)

2 The light source for the system is a 157nm molecular fluorine laser (Lambda Physik Novaline F1030) with dual wavelength output at and nm that emits more than 10W at 1000Hz repetition rate. The beam delivery system (BDS) that connects to the F 2 laser comprises of CaF2 beam shaping optics, high reflectivity 157nm turning mirrors and a discretely variable beam intensity attenuator. Following the BDS the radiation passes into an illuminator optic (manufactured by Corning Tropel Corp.). This unit provides a near-uniform intensity distribution at the reticle plane and is designed to illuminate the entrance pupil of the projection lens with the appropriate fill factor. Apertures can be inserted into the illuminator to provide a wide range of partial coherence factors and also allows for apodised illumination. All sections of the BDS are purged with purified dry nitrogen gas to prevent absorption of the 157nm VUV radiation by molecular oxygen and water vapour. Levels of O2 in the BDS for example are typically < 0.2ppm. An exposure dose monitor located behind the final turning mirror acts as a partial beam splitter, transmitting a small fraction of the incident energy onto the detector. The reticle is held and positioned on a stage with precise control over its small scale XYθ z position using a piezoelectric flexure stage. In addition the stage has an automated large-scale translation capability that allows a single 6" square reticle to hold 10 object fields reducing the necessity for frequent reticle exchanges. A vacuum suction ring design is used to hold the reticle in a stable position. The projection lens is a 0.85NA catadioptric objective (also manufactured by Corning Tropel Corp.), chromatically corrected over the full emission spectrum of the F 2 laser with a diffraction-limited wavefront error of < 0.05 waves rms. The wafer is held and positioned on a low contact area vacuum chuck and a set of high-resolution air-bearing stages allows very high stability positioning in the XY plane (perpendicular to the optical axis). A pair of piezoelectric flexure stages mounted on top of the main XY stage gives full 6-axis control of the wafer position (including vertical height adjustment in the Z direction) of 5nm and 2µrad resolution. The wafer height is measured in the vicinity of the exposure field with a system of 3 position sensitive detectors that comprises of a diode laser triangulation system offering 10nm focus height resolution. A 157nm sensitive photodiode detector, mounted from the wafer stages is used to track the calibration of the dose monitor detector to the exposure dose delivered to the wafer. The tool is equipped with a robotic wafer-handling interface to an in-line track and can handle both 200 and 300mm diameter wafers from either the track input or from a manual cassette loader. A position reference system at the reticle and wafer is used for field-by-field wafer alignment compatible with the requirements of overlay exposures. The reticle stage, projection lens and wafer stage are all mounted from a common granite structure that provides for an extremely stable exposure platform. The tool is enclosed in an environmental chamber with temperature and humidity control of 0.1 C and ± 1 % R.H. respectively. Proc. SPIE Vol. 4691

3 The design for the 157nm Microstepper tool is shown schematically in Figure 1 and a 3D CAD drawing is shown in Figure 2.

4 3 IMAGING PERFORMANCE Before reviewing the imaging performance obtained with this system it is useful to consider the interaction between the numerical aperture of the projection lens, the usable k, factor required for the 70nm technology node and beyond and the resolution enhancement techniques (RET) available. Such a comparison is shown in Figure 3. At 157nm, operating at 0.85NA, binary reticles without RET will not be suitable for the 70nm node. The addition of weak RET allows for 70nm lithography at 0.85NA but strong RET will be required for sub-70nm imaging. At the expected limit of optical systems operating at 0.9NA and with strong RET reaching the 50nm technology node is just possible. A performance comparison can be made between the imaging performance of the current 0.85NA system and an earlier 157nm Microstepper at Selete that is equipped with a 0.6NA lens. Operating with an alternated phase shifting mask (Alt- PSM) and exposing a 100nm thick layer of commercially available photoresist (Shipley XP98248) on a bare silicon substrate an ultimate resolution of 70nm US was obtained. Using Rayleigh's law the k, factor for this process is 0.27, which is at the threshold of imaging with strong RET. Applying this k, factor to scale to the ultimate resolution at 0.85NA with an Alt-PSM it is expected that 50nm dense features should be resolved. Figure 4 shows the typical imaging performance obtained with the Shipley XP98248 resist on bare silicon substrate with Alt-PSM. The expected resolution based on scaling the NA from 0.6 to 0.85 at a constant k, factor was not achieved. However, there are a number of possible factors that might limit this scaling such as limitations of the photoresist and process,

5 system vibration levels being more significant at smaller feature sizes, interaction of the high NA radiation with the photoresist layer, illumination polarisation effects and the effect of CaF 2 intrinsic birefringence. Since the numerical aperture of the 0.85NA projection lens is variable down to 0.6NA another comparison of the imaging performance of the two Microstepper systems could be made at the same 0.6NA. As an example of the effect of resist and process upon the imaging performance Figures 5 shows the typical imaging results obtained with a fluoropolymer based resist that is a 20nm thicker layer than the Shipley resist. There is a significant reduction in line edge roughness and resolution between the two resists under identical exposure conditions. The ultimate resolution of 55nm is still greater than that expected from scaling from the 0.6NA system but nevertheless these results clearly demonstrate the capability of 157nm lithography and high NA optics to image sub-70nm dense features. Figure 5. Dense line imaging performance of 0.85NA lens with Alt-PSM (σ = 0.3) in 120nm Fluoropolymer resist on silicon substrate. At 55nm LS k, = INVESTIGATION INTO THE EFFECT OF INTRINSIC BIREFRINGENCE IN CaF 2 Intrinsic birefringence at 157nm in CaF 2 was first reported by John Burnett et al. 5 of the National Institute of Standards and Technology mid Cubic crystals such as CaF 2 can to a first-order approximation be regarded as isotropic materials. However, at the short wavelength of 157nm the CaF2 crystal no longer maintains its symmetry to the electrostatic component of the radiation field and an extra non-linear contribution to the electric displacement becomes significant in magnitude. This non-linear term gives rise to a birefringence in the CaF 2 material that increases as (1/λ)². This intrinsic birefringence has a spatial distribution that is a constant function over the entire crystal and exhibits an azimuthal symmetry about the direction of propagation of the radiation. In directions along the crystal axes that possess 3-fold or 4-fold symmetry (the (111) and (001) axes) the behavior is uniaxial so there is no birefringence. In all other directions the material is biaxial and therefore a non-zero birefringence exists. The magnitude of the birefringence is determined completely by the maximum birefringence along the (110) direction. The birefringence has been measured at NIST 5 to be 11.0nm/cm along this maximum direction, which is much greater than the target requirements for high NA 157nm lenses (1nm/cm). The 0.85NA lens has its optical axis along the (111) direction but due to the high NA and the catadioptric design of the objective some of the optical rays within the lens pass close to the directions of maximum birefringence within the CaF 2 material. This results in a distribution of the birefringence about the optical axis as shown in Figure 6 giving rise to a 3-point aberration. 6

6 Figure 6. Schematic showing the orientation of the crystallographic axes in a CaF 2 lens element and the distribution of birefringence about the optical axis. An investigation was carried out to attempt to quantify the magnitude of the 3-point aberrations in the lens and therefore to gain an understanding into whether intrinsic birefringence might have a serious impact on the imaging performance of the system. This is particularly relevant in the case of this high NA system since the lens was qualified before delivery at Coming Tropel using phase-measuring interferometry (PMI) only at a wavelength of 244nm. Therefore the PMI data collected is far less sensitive to the CaF 2 intrinsic birefringence that at the 157nm wavelength of operation. A sensitive metric to 3-point aberrations that can be derived from resist-image measurements is the line width difference (LWD) between the outermost features of a small collection of dense line structures. A schematic of the type of pattern used on a binary reticle is shown in Figure 7. The LWD is measured from the CD difference between line 1 and line 5. Figure 7. Schematic of the binary mask pattern used to measure the coma and 3-point aberration in the 0.85NA lens. The orientation of the patterns was 0, 45, 90 and 135. The LWD parameter is also sensitive to coma aberrations but the influence of coma and 3-point can be distinguished from each other by measuring the LWD as a function of the angle of rotation of the pattern about the azimuthal direction. The feature size chosen for these patterns was 180nm US using an illumination partial coherence value of 0.3. The LWD data was collected through a range of focus at constant exposure dose and is displayed in Figure 8. The data has been normalised so the best focus position is centered at 0µm.

7 Figure 8. Evaluation results of LWD measurements (CD difference) made through focus and as a function of the angle of rotation of the pattern. From an examination of the variation of the LWD with the angle of orientation of the pattern it is possible to make a conclusion regarding the existence of a significant 3-point aberration contribution. If the variation in LWD were solely attributable to coma aberrations then the quadrature sum of the LWD at the 0 and 90 orientations would be approximately equal to the LWD at the 45 orientation as follows: Applying this condition to the evaluation data in Figure 9 reveals that there is a residual LWD of approximately -40nm that confirms that there is a significant 3-point aberration contribution to the LWD. It is now possible to model the variation in LWD with orientation of the pattern as an expansion series of the orientation (0) containing terms that describe coma(x) and coma(y) (functions of θ) and 3-point(X) and 3-point(Y) aberrations (functions of 3θ) as follows: A plot of the measured LWD at the best focus position as a function of the pattern orientation is shown in Figure 9 together with a simulated best fit curve that describes the LWD variation. The magnitudes of the expansion coefficients A to D (in nm) are related to the magnitude of the particular aberration.

8 Figure 9. Graph of the measured LWD variation together with the best fit modelled variation as a function of the angle of rotation of the pattern. Having established the magnitudes of the effects of the coma and 3-point aberrations it is possible to derive approximate values for the actual aberrations themselves in terms of the Zernike coefficients in the expansion of the transmitted wavefront. The Zernike coefficients that describe low-order coma are Z7 (X) and Z8 (Y) and low-order 3point are Z10 (X) and Z11 (Y). A simulation was carried out using PROLITH/3D on exposures of the same 5-bar pattern used for resist exposures. The simulation conditions are detailed in Table 2. At each orientation of the 5-line structure the CD difference between the first to the fifth line was calculated (LWD) for varying amounts of each of the Zernike coefficients under investigation. The results of the simulation over a range of Zernike coefficients Z8 from 0.1λ to 0.3λ and Z11 from 0.2λ to 0.8λ are shown in Figure 10.

9 Figure 10. Simulation results from PROLITH analysis of the effect of coma(y) and 3-point(Y) aberration on LWD variation as a function of the angle of rotation of the pattern. Comparing the simulation results with the best-fit curve for the LWD variation as a function of the pattern direction that was derived from the experimental results allows the Zernike coefficients for coma and 3-point aberrations to be determined. These results are summarized in Table 3. The specification for the wavefront error over all 37 Zernike terms for this objective is less than 0.05 waves rms. With a measured rms wavefront of waves the PMI data collected at 244nm and scaled to 157nm supports this specification. One possible explanation for the discrepancy between the 244nm PMI derived Zernike data and the results in Table 3 may be due to the intrinsic birefringence of the CaF 2 lens material. Although aberrations such as coma and 3-point were measured and corrected for using 244nm PMI measurements those components of these aberrations that are specifically dependent on the intrinsic birefringence could not be fully corrected using only 244nm wavelength interferometry. However, there are other possible reasons for the presence of coma and 3-point in the lens which include induced mounting stresses in the lens elements or non-uniformity in the internal refractive index within the lens.

10 5 CONCLUSIONS We have presented details and imaging performance results of a high NA 157nm Microstepper system at Selete that is capable of supporting the needs of photoresist research and development at the 70nm technology node and beyond. Resist image-based measurements have also indicated that there is a possible influence on the imaging performance from the intrinsic birefringence of the CaF 2 lens material. The Exitech MS-157 Microstepper, which is designed for operation with both 200 and 300mm wafers possesses ultrahigh stability wafers stages and 10nm autofocus system resolution that are important requirements to achieving the ultimate imaging performance of the system. It is also the first exposure tool at this wavelength to be equipped with a 0.85NA projection lens that is capable of sub-70nm resolution. The feasibility of 157nm high NA lithography in combination with strong resolution enhancement techniques to address the 70nm technology node has been demonstrated. An ultimate resolution of 55nm US was obtained with an alternated phase shifting mask using a 120nm thick fluoropolymer resist. It is now well known that the intrinsic birefringence of CaF 2 is a significant effect at 157nm with measurements indicating over an order of magnitude larger value than the 1nm/cm target requirement for high NA 157nm lithography lenses. This lens was designed and fabrication was mostly completed before this knowledge became available. Intrinsic birefringence was not apparent from the 244nm interferometry data used to qualify the lens. Imaging in resistbased measurements and simulations were carried out to measure the effects of the intrinsic birefringence. Measurements were made of the line width difference between the outermost features of a 5-line pattern as function of the orientation of the pattern. Matching this behavior to a simulation of the effect of coma and 3-point aberrations derives values for the respective Zemike terms that are considerably larger than those measured during the 244nm interferometry lens qualification e.g. we measured Z10 which represents 3-point (X) aberration as waves rms compared with the 244nm PMI measurement of the entire wavefront error of waves rms. One possible conclusion is that such differences are attributable to the effect of intrinsic birefringence although there are other possible influences such as mounting induced stresses and non-uniform birefringence. Future resist based experiments are planned to measure and confirm the effect of birefringence in this lens at 157nm. This work has highlighted the desirability of high NA lens qualification by actinic wavelength interferometry whereby the effects of intrinsic birefringence can be measured and compensated for during the lens design, build and testing phases. 6 ACKNOWLEDGMENTS The authors would like to thank the SELETE member companies for their cooperation as well as and Corning Tropel Corp. for their expert work in designing and fabricating the 0.85NA lens used in this work. 7 REFERENCES 1. M. Toriumi, S. Ishikawa, S. Miyoshi, T. Naito, T. Yamazald, M. Watanabe and T. Itani, Proc. SPIE 4345, 371 (2001) Satou, M. Watanabe, H. Watanabe and T. Itani, Proc. SPIE 4345, 361 (2001). 3. T. Matsuo, T. Onodera, T. Itani and H. Morimoto, Proc. SPIE 4186, 268 (2000). 4. T. Onodera, T. Matsuo, T. Itani and H. Morimoto, Proc. SPIE 4346, 61 (2001). 5. J.H. Burnett, Z.H. Levine and E.L. Shirley, International SEMATECH CaF 2 Birefringence Workshop, San Francisco, July J. Webb, International SEMATECH CaF 2 Birefringence Workshop, San Francisco, July 2001.