Using scanner systematic signatures to enhance OPC modeling
04/01/2008
Executive OVERVIEW
At previous technology nodes, scanner systematic signatures–including illuminator pupil-fill and polarization, lens aberration, lens apodization, and flare–could be ignored without significantly impacting optical proximity compensation (OPC) accuracy. However, these characteristics begin to play a critical role at the 45nm node and beyond. For advanced applications, it is necessary for the OPC modeling tool and the core engine of the OPC simulator to accurately model the unique scanner systematic signatures, which must be easily incorporated to be adopted in manufacturing. OPC models created with these heightened capabilities will be able to accurately predict lithographic printing effects missed with traditional “idealized” OPC models, while also reducing OPC modeling time.
Ongoing device miniaturization, coupled with decreasing k1 factors, places stringent demands on optical lithography. These added requirements are reflected in the need for increasingly accurate lithography simulations used for OPC. In order to enhance simulation accuracy beyond current levels, it is necessary to consider the realistic physical conditions of the lithography/OPC simulation tools and their associated impacts on the resulting image.
Traditionally, only scanner settings such as wavelength, projection lens numerical aperture (NA), and illuminator aperture design have been included in optical image and OPC simulation. At previous technology nodes, scanner systematic signatures–those signatures common to a scanner family such as illuminator pupil-fill, lens aberration and apodization, flare, and other such unique characteristics–were not typically incorporated due to their negligible impact on final lithography/OPC model accuracy.
Today, however, leading-edge technology is advancing to the 45nm node and beyond, and IC manufacturers are ramping =1.30 NA 193nm immersion lithography equipment into production. At these hyper-NA applications, scanner systematic signatures simply can’t be ignored due to the critical dimension (CD) error budgets for 45nm and subsequent technology generations [1].
Scanner systematic signatures
Scanner systematic signatures may vary noticeably between imaging tool generations/families (i.e. NA 0.92 vs. NA 1.3 scanner generations), and also among scanner vendors [2]. However, signatures among scanners of the same type from one manufacturer will strongly resemble each other. Therefore, averaging the measured systematic signatures across multiple tools of the same scanner type will provide that scanner type’s systematic signatures. Incorporation of the systematic signatures in the OPC simulator will enhance OPC model accuracy to meet the tightening requirements at advanced technology nodes.
Scanner illumination intensity distribution, known as illuminator pupil-fill, has typically been simplified in lithography simulation for OPC by an ideal top-hat profile with rectangular cross-sectioning where the illumination pupil is assumed to consist of either completely dark or completely bright areas. In reality, the measured illuminator pupil-fill has soft edges, with a smooth transition from bright to dark or vice versa, seen in the pupil-fill contour map. This difference between the actual illumination distribution and the ideal top-hat profile impacts the pattern transfer function from the mask to the aerial image in the lithography system [3]. This effect becomes more significant with the sophisticated layout-specific illumination optimization required for low k1 lithography. Recent work [4, 5] demonstrated that the difference between measured and idealized illuminator pupil-fill can lead to several nanometers of CD deviation depending upon the pattern layout, validating the need to incorporate the actual measured illuminator pupil-fill profile in the OPC model.
In addition to illuminator deviation from idealized characteristics, the scanner projection lens is also imperfect, contrary to typical OPC simulation assumptions. One key attribute of the scanner projection lens is aberration, which is quantified by the wave-front phase error at the lens pupil. For many years, this error has been characterized using Zernike polynomials [6, 7]. The most common wave-front phase errors are represented by 37 Zernike polynomials using the measured 37 Zernike coefficients. However, this approach represents the lens aberration only within some small portion of the projected image field, whereas full characterization of a scanner lens system requires independent measurement of the 37 Zernike coefficients at many points across the image field.
The other systematic signature of a lens system is apodization, where the lens transmission decreases near the edge of the pupil. The apodization effect is essentially due to lens imperfections that are artifacts of the manufacturing process (e.g., coating and polishing). The high diffraction orders for small pitch patterns are attenuated most severely by the apodization effect, thereby impacting the imaging and printing of patterns, especially at small pitches. Flare is another critical systematic scanner signature due to its influence on imaging performance, particularly for low k1 lithography [8]. Flare, i.e. stray light, is an added incoherent background intensity mainly caused by scattered light from lens surfaces. Most of the light scattering that occurs in the optical elements is due to lens contamination, surface roughness, inhomogeneity of the refractive index of the projection lens, and photomask blank reflectivity. Flare changes the aerial images, thereby introducing CD changes and decreasing process latitude by reducing image contrast. Flare is classified as a “global” or “DC” type when it is independent of the neighboring patterns, and as being a “local” type when it changes with the surrounding patterns.
Lithography simulations produced by Synopsys’ SolidE and using a 0.92 lens NA determined that each of the scanner systematic signatures (illuminator pupil-fill, lens aberration, apodization, and flare) has a non-negligible impact on optical proximity effect. CD errors in imaging predictions exceeding 1nm (and potentially up to several nanometers) will result when the corresponding scanner signature is excluded from the lithography modeling. However, pure lithography simulation tools such as SolidE cannot generate silicon data-calibrated, full-chip OPC deployable process models. This limitation necessitated development of a new methodology to account for these systematic signatures within the OPC modeling tool.
Optical accuracy and scanner systematic signatures
Adopting the scanner systematic signatures in an OPC modeling tool first required verification that the OPC simulator Progen could accurately model the scanner signatures. This work was performed for a 193nm Nikon NSR-S308F lithography scanner, with a 0.92 lens NA, annular illumination (σin = 0.57, σout = 0.85), 6% att PSM, and full film stack (resist/barc/substrate). Six categories of test patterns (Fig. 1), including 1D and 2D features, were selected to calculate the aerial image CD using a constant threshold modeled under the nominal optical condition (without scanner systematic signatures), and then modeled under the realistic optical condition (with these signatures). Results obtained using the Progen OPC simulator were then compared against those calculated by the more rigorous and computationally-intensive SolidE lithography simulator.
 Figure 1. The six test patterns used for aerial image CD calculation in Progen and SolidE. Both simulators were anchored to the same point of CD vs. pitch OPC. |
The CDs calculated by Progen and SolidE for the six test patterns under the nominal optical condition agreed very well (Fig. 2a shows L/S results). Only slight CD differences between the simulators were observed, with the largest CD error seen for the isolated line end feature without hammerheads or serifs, showing ~1nm difference between SolidE and Progen. These small isolated line end features without hammerheads or serifs are not used in real mask layouts and are known to have very low aerial image slopes. Under such conditions, similar minor CD differences can be observed even between different traditional lithography simulators simply due to subtle differences in numerical implementation. SolidE and Progen results agreed remarkably well, given that their numerical implementations are substantially different, and that only a limited number of kernels were retained within Progen after transmission cross-coefficient (TCC) decomposition in order to enable high speed, full-chip calculations.
After verifying that Progen and SolidE matched well under nominal conditions, a measured pupil-fill map was added into Progen using the kernel import facility (KIF); the illuminator pupil-fill was imported into SolidE, and the respective aerial image CDs were calculated. Progen and SolidE results again converged well using the measured illuminator pupil-fill. In Fig. 2a, the primary y axis shows the L/S pattern aerial image CD (nm) calculated through pitch in Progen (illum_P) and in SolidE (illum_S) using the measured illuminator pupil-fill. The secondary y axis shows the CD difference as calculated by the two simulators (illu_diff), also in nm. Next, using the Progen and SolidE CD data obtained at nominal condition along with the measured illuminator pupil-fill condition, the CD impact of measured illuminator pupil-fill, relative to the nominal condition, was calculated and visualized for the six different test patterns. The CD impact observed by Progen and SolidE matched very well, confirming the accuracy of the Progen OPC tool when modeling measured illuminator pupil-fill. Figure 2b demonstrates the L/S pattern aerial image CD impact (nm) of using the measured illuminator (CDscn) relative to the nominal condition (CDnom) as observed by Progen (illum_P) and SolidE (illum_S).
The first 37 Zernike coefficients were then experimentally measured on the NSR-S308F scanner at three locations in the projection image field: 2S1, 2S2, and 2S3. These Zernike coefficient sets describing projection lens aberrations were included in CD calculations with Progen and SolidE, and the CD impact was calculated for the 2S1, 2S2, and 2S3 field locations. Progen (labeled as 2s1_P, 2s2_P and 2s3_P) and SolidE (labeled as 2s1_S, 2s2_S and 2s3_S) results were also in excellent agreement in this aberration modeling. Fig. 3a shows L/S pattern aerial image results through pitch (nm), comparing the measured CDscn and nominal CDnom condition. Similar CD and CD impact calculations were performed using the measured lens apodization data, with strong convergence between Progen (apod_P) and SolidE (apod_S) modeling. Fig. 3b shows L/S pattern aerial image results through pitch (nm) comparing the measured CDscn and nominal CDnom condition. Global flare simulations also confirmed agreement between the two image simulators, once again validating the robustness of the Progen OPC simulator engine.
CD impact of scanner systematic signatures
Progen and SolidE results confirmed that the CD impact from scanner systematic signatures is indeed significant, relative to the very tight CD error budget at the 45nm node and beyond, with CD impact depending upon the particular layouts being patterned. Figure 4 shows the imaging CD impacts (calculated in SolidE (Fig. 4a) and Progen (Fig. 4b) on line/space features through pitch due to the measured systematic scanner signatures (illuminator, aberration, apodization, and flare). Both Progen and SolidE indicated that illuminator pupil-fill and lens aberration have the largest CD impact: up to 4nm for pupil-fill impact through pitch, and up to 3nm for the lens aberration impact through pitch. This was followed by lesser, although still significant, effects due to flare and lens apodization.
 Figure 4. Delta CD impact of L/S feature through pitch in nm from scanner systematic signatures as calculated by a) SolidE and b) Progen. |
null
Conclusion
The effects of scanner systematic signatures were simulated using Progen and SolidE. The simulations agreed well between the two imaging simulators, confirming the accuracy of the Progen simulator engine. This work also confirmed the need to include scanner systematic signatures within the OPC models, demonstrating CD impacts on the order of several nm depending upon the systematic scanner signature and the pattern being imaged.
Collaboration between Nikon and Synopsys Inc., creators of Synopsys Proteus OPC software, has resulted in a newly developed “manufacturing-aware” OPC solution that allows Proteus users to automatically access Nikon proprietary scanner information contained in the Nikon scanner signature file (NSSF). It includes such higher-order lithographic effects as polarization, flare, synchronization, and various aberration data for leading-edge dry and immersion Nikon scanners. OPC models created with this enhanced methodology, incorporating the scanner systematic signatures, can accurately predict lithographic printing effects previously missed with traditional “idealized” OPC models. Therefore, this new integration will significantly increase OPC modeling accuracy and minimize crucial modeling time.
Acknowledgments
Progen, Proteus, and SolidE are trademarks of Synopsys Inc.
References
- Q. Zhang, J. Tyminski, K. Lucas, “Modeling Scanner Signatures in the Context of OPC,” Proc. SPIE, Photomask Technology, 2007.
- S.P. Renwick, H. Nishinaga, N. Kita, “Characterizing a Scanner Illuminator for Prediction of OPE Effects,” Proc. SPIE, 6154, 2006.
- T.C. Barrett, “Impact of Illumination Pupil-fill Spatial Variation on Simulated Imaging Performance,” Proc. SPIE, 4000, 2000.
- J.K. Tyminski, Q. Zhang, K. Lucas, L. Depre, P. VanAdrichem, “Scanner-characteristics-aware OPC Modeling and Correction,” Proc. SPIE, 6521, 2007.
- J.K. Tyminski, T. Nakashima, Q. Zhang, T. Matsuyama, K. Lucas, “The Impact of Scanner Model Vectorization on Optical Proximity Correction,” Proc. SPIE, 6607, 2007.
- T.A. Brunner, “Impact of Lens Aberrations on Optical Lithography,” IBM Jour. of Research and Development 41, p. 57-67, Jan/Mar 1997.
- D.G. Flagello, B. Geh, “The Influence of Lens Aberrations in Lithography,” Microlithography World, Winter 1998, p. 11-16.
- C.A. Mack, “Measuring and Modeling Flare in Optical Lithography,” Proc. SPIE, 5040, 151, 2003.
- C.T. Bodendorf, R.E. Schlief, R. Ziebold, “Impact of Measured Pupil Illumination Fill Distribution on Lithography Simulation and OPC Models,” Proc. SPIE, 5377, 1130, 2004.
- Q. Zhang, H. Song, K. Lucas, “Polarization Aberration Modeling via Jones Matrix in the Context of OPC,” BACUS 2007.
Qiaolin (Charlie) Zhang received his MS and PhD in electrical engineering and mechanical engineering from the U. of California at Berkeley and is a staff R&D engineer at Synopsys Inc., A1.233N, 700 E. Middlefield Rd., Mountain View, CA 94043 USA; ph 650/215-8480, e-mail [email protected].
Jacek K. Tyminski received his MS in technical physics from the Technical U. of Gdansk, Poland, and his PhD in physics from Oklahoma State U. He is a principal engineer at Nikon Precision Inc., 1399 Shoreway Rd., Belmont, CA 94002 USA; ph 650/413-8252; e-mail [email protected].
Holly H. Magoon received her BS in chemistry from St. Michael’s College and is a marketing manager at Nikon Precision Inc.
Kevin Lucas received his PhD in electrical and computer engineering from Carnegie Mellon U. and currently manages the Application Deployment Team in the Lithography Technology Group at Synopsys Inc.
Hua Song received his PhD in computational and applied mathematics from Rice U. and is senior staff R&D engineer at Synopsys Inc.