A method for evaluating RETs for advanced masks
09/01/2001
LITHOGRAPHY SPECIAL REPORT Optical lithography is being pushed an order of magnitude beyond the 1µm limit anticipated 20 years ago. Behind this is a rallying of support to shorten exposure tool wavelengths and implement resolution enhancement techniques (RETs) such as optical proximity correction and phase shift masks. A method for evaluating RETs using aerial images and an approach to incorporating them into the design flow emphasize the development of RETs as a production-worthy tool.
Mircea Dusa, Judith van Praagh, Jo Finders, ASM Lithography, Santa Clara, California, and Veldhoven, The Netherlands
Andrew Ridley, Metron International, Sunnyvale, California
 |
overview
To evaluate the efficacy of various resolution enhancement techniques (RETs), a method is proposed that gathers aerial intensity images from real masks under actual illumination conditions. Correlations to wafer-generated process latitudes are then established.
As the wavelength of exposure tools approaches the size of device features, ranking the various resolution techniques (RET) becomes intrinsically difficult. Maximizing process latitudes for half-wavelength, 130nm IC designs requires large overlapping exposure-defocus (ED) windows for critical patterns such as lines-through-pitch and end-of-lines. A multitude of RETs are currently available — from imaging optics in the exposure tool to mask-type and software to correct for critical dimension (CD) through-pitch and end-of-line optical proximity.
The minimum feature size (or CD) is a function of a process constant (k1), stepper wavelength (l), and the numerical aperture (NA):
Minimum feature size = k1 λ/NA (1)
Decreasing λ and increasing NA can reduce the minimum resolvable linewidth and improve resolution. The process constant, k1, depends on the quality of the resist, etch, optics, and mask enhancements.
Low-k1 lithography is not possible without using RETs — masks with a mix of phase-shifting structures (phase shift masks or PSMs) and optical proximity correction (OPC) treatments combined with off-axis and high-NA illumination apertures. In the low-k1 regime, the mask with its embedded RET develops into part of the illumination path. Thus, these interactions must be studied.
This article has three primary goals: 1) to evaluate a technique for fast and accurate ranking of illumination and mask enhancements in the low-k1 lithography regime; 2) to rank and compare process latitudes when RETs are applied; and 3) to evaluate the use of aerial images from real masks to predict CD results instead of exposing numerous wafers in the fab.
Traditionally, mask effects based on aerial images addressed only defect printability [1-3]. We propose extending the use of mask aerial images to predict lithography results. Aerial images were collected on a Carl Zeiss MSM100, an aerial image simulation microscope (AIMS) equipped with specially built NA and pupil shape apertures. These apertures match the shape and characteristics of the illumination path for high-NA, DUV exposure tools. In this case, an ASML PAS 5500/700 was used.
Resolution enhancements in the illumination path
Resolution enhancements in the illumination path consist of annular and multipole off-axis apertures and the numerical aperture (NA). For dense lines at half-pitch [(linewidth + spacewidth)/2] close to the resolution limit, off-axis illumination offers the unique possibility of optimizing depth-of-focus (DOF) and exposure latitude (EL) by adjusting the partial coherence settings (s, the ratio between the NA of the imaging lens and condenser lens, which describes image transfer from condenser lens into imaging lens). For semidense to isolated lines, the benefits of off-axis illumination are less straightforward and the enhancement in focus and exposure latitudes comes from enhancements of the NA and the mask. Optimum settings for DOF and EL can be different because each is determined by a different physical phenomenon. Nevertheless, both phenomena can be visualized by looking at the position of the diffracted orders in the entrance pupil of the projection lens, so-called pupil filling [4, 5].
Effects of the enhancements on the DOF and EL can be determined by looking at the position and the magnitude of the diffracted orders in the entrance pupil of the projection lens.
Illumination aperture, pupil filling, and DOF
The shape of the two off-axis illumination apertures — annular and multipole — for a 0.70NA, σouter = 0.85, and σinner = 0.55 settings are depicted in Fig. 1a. The position of the 1st order diffracted light was calculated for 130nm lines at 325nm pitch (1:1.5 pitch), NA = 0.70, and λ = 248nm. Close to the resolution limit, the aerial image is formed by two-beam interference of the 0th and 1st order diffracted light. This is schematically illustrated in Fig. 1c and is true irrespective of the illumination mode, annular or multipole. The left-hand part of the figure shows the pupil filling when using multipole illumination. The more symmetric the 0th and 1st order rays with respect to the main axis of the projection lens, the larger the DOF [4-6].
For maximum DOF, the illumination settings should be chosen such that the distance between the 0th order and the projection lens axis (d0) is equal to the distance between first order and projection axis (d1). This is accomplished by adjusting the radial placement of the illumination pole, that is, the sigma (s) center. In fact, using optimum illumination settings (d0 = d1), the 1st order from one pole and 0th order from another will overlap exactly (pole 0 and +1 in Fig. 1c). It should be stressed that although these poles will overlap, they will not interfere, since there is no phase correlation between the four multipole incident poles. Using the constraint d0 = d1 and simple diffraction theory, an analytic expression for the optimum pole position can be derived for multipole illumination [4-6] as follows:
 |
null
where σcenter designates the center of the illumination pole while σinner and σouter are the coherence settings that control both position and size (Fig. 1). In Eqn. 2, LW represents the linewidth to be printed (as half-pitch), NA is the numerical aperture of the projection lens, and λ is the wavelength of the exposure light. For example, for 1:1 pitch, LW = 130nm; for 1:1.5 pitch, LW = 162.5nm; for 1:2 pitch, LW = 195nm and so on. In this article, λ = 248nm and NA was 0.63 and 0.70.
Illumination aperture, pupil filling, and EL
Exposure latitude for dense lines depends on the coherence settings, but also depends significantly on NA. To optimize EL, the aerial image contrast has to be maximized [5, 6]. Aerial image contrast is a function of the amount of 1st order diffracted light that is captured by the projection lens, hence the need for high NA and high s. Decreasing the pitch will lead to a reduced amount of 1st order light captured by the lens and, thus, to a decreased EL. Again, the pupil filling can be used to depict this schematically. From Fig. 1b, we can derive the EL as follows:
 |
null
where A1 represents the amount of 1st order light captured by the lens and A0 corresponds to the total amount of 0th light. Both values can be determined by computing the respective areas in the pupil-filling diagram. The correlation factor C takes into account resist and process contrast. Therefore, this method can be used alternatively to determine the relative contrast and subsequent EL of one resist relative to another.
Instead of implementing resist parameters, simulation can be used to obtain the normalized image log-slope (NILS) to compare and optimize aerial image contrast and EL [4, 5]. In this case, lithographers calculate NILS and EL for lines-through-pitch, from fully dense to completely isolated lines using annular and multipole apertures with NA = 0.70 and the highest possible coherence settings. For feature size approaching the resolution limit, multipole illumination is the preferred setting, whereas annular illumination provides the best aerial image contrast at the resolution limit. Therefore, multipole was determined preferable when imaging dense lines with pitches in the 260-460nm range, as indicated by tests on aerial images gathered on AIMS and verified on wafers exposed on a DUV scanner for 150nm design.
Resolution enhancements on the mask
Further improvement in lithography process latitude may be achieved from reticle resolution enhancements. We used two types of enhancements, halftone (attenuated) PSMs and OPC treatments [7, 8]. Effects of these reticle enhancements on the DOF and EL can also be explained by looking at the position and the magnitude of the diffracted orders in the entrance pupil of the projection lens. The main diffraction effect of using halftone PSMs is a decrease in the 0th order light and an increase in the 1st order light. These effects amplify with increased mask transmission. There is no change in the position of the diffracted orders within the lens pupil, which are only determined by feature half-pitch.
 Figure 2. Aerial intensity images gathered from halftone phase-shifting mask: a) -400nm defocus, b) best focus, and c) +400nm defocus. |
According to Eqn. 3, the expected lithographic effect when using halftone PSMs is an increase in EL without significant gain in focus latitude. Increased EL also means an increase in aerial image contrast, NILS, which translates to lower mask error factor (MEF) and, thus, improved CD control. MEF is the parameter that describes how much the CD variation of the mask is transferred into that of the wafer. Mathematically, this can be written as: wafer CD range/mask CD range/4, where four is the projection lens magnification.
Different changes in diffraction pattern occur with OPC treatment of a mask feature. Depending on the OPC mode, the position of 1st order light may change if assist features are applied to the main feature, making an isolated line look denser. In our study, we used two halftone PSMs with 6% and 18% transmission. Both masks include line and end-of-line patterns with groups of line sizes ranging from 110-150nm. For each line size, the pattern density (pitch) varies from 1:1 to fully isolated. Each feature size and each pitch has several levels of OPC treatments, such as positive bias (increased linewidth) or negative bias (reduced linewidth), single or double assisted features. For end-of-line patterns, OPC treatments consist of serifs and hammerhead structures with several levels of overlap and size.
 Figure 3. Total window area (TWA), the experimental evaluation metric to rank the efficiency of the resolution enhancement techniques, is the area under the EL-DOF curve. |
Experiments
Our experimental work started with gathering aerial images through focus using AIMS equipped with specially built pupil illumination apertures that emulate the ASML 248nm scanners: NA apertures of 0.70 and 0.63; and pupil shape off-axis apertures, annular and multipole, with inner/outer maximum s of 0.55/0.85. First, aerial images were collected though focus from 130nm lines at 1:1.5 pitch and the threshold value needed to image 130nm linewidths at the isofocal position determined. Aerial images were also collected through focus from 130nm line patterns at three other pitches, 1:1, 1:2, 1:3, and for fully isolated lines. Figure 2 shows AIMS-generated through-focus mask aerial images, indicating measurement locations for both line and end-of-line patterns. The aerial images clearly show defocus-generated differences in feature shape, size, and image contrast. The same procedure was used to gather aerial images through focus from end-of-line (space) patterns with four different levels of OPC correction: no OPC, two types of serifs, and one hammerhead.
Finally, exposure defocus curves (ED, a plot of exposure energy vs. defocus values for ±10% CD variation from the process target) were generated to calculate DOF and EL. From the ED curves, the area under the EL-DOF curve was calculated in order to generate the total window area (TWA) [9]. Figure 3 illustrates the concept, where the TWA metric uses arbitrary units of %-µm. Our results show an almost linear correlation wherein 10 units of TWA equal a 10% exposure latitude at a DOF of 0.6µm. Thus, the TWA units can be correlated to the more conventional lithography metric, EL at DOF, as follows:
10 TWA units (% - µm) = 10% EL @ 0.6µm DOF
After gathering the aerial images, both reticles were exposed on the high-NA 248nm scanner under the same illumination conditions used to capture aerial images: 0.63 and 0.70NA, annular and multipole, with 0.55-0.85 inner/outer s settings.
Prediction of ED latitudes based on aerial images
Beginning with line patterns through-pitch, aerial images were analyzed and combined with NA and pupil shape apertures emulating the real scanner illumination path in the form of TWA as an indication of total process latitude. This, therefore, provides a metric to rank the efficacy of RETs. Figure 4a shows the ranking of RETs in the illumination path, while Fig. 4b shows the ranking of mask RETs.
 |
Figure 4. Comparative ranking of illumination enhancements mode based on maximum overlapping process latitudes of 130nm line patterns through-pitch. a) Compares illumination apertures in shape, annular to Quasar, and size, 0.63-0.70NA; and b) compares mask types — 6-18% transmission of halftone phase-shifting mask.
Figure 4a illustrates a factor of 2x increase in process latitudes (TWA) for 130nm lines at >1:1.5 pitch using multipole illumination — a conclusion that confirms theoretical predictions from the pupil filling discussed earlier. Further analysis follows diffraction theory predictions, where increased process latitudes are expected at high NA (0.70) for 1:1 fully dense line patterns and annular illumination. Altogether, multipole illumination provides the largest process latitudes (TWA) for semidense to isolated pitches and equals annular illumination for 1:1 pitch.
Figure 4b points to the strong interactions between mask enhancement, specifically halftone PSM, and lithographic results. Higher transmission (18%) of the mask substrate produces such a large increase in EL that the total process latitude is a factor of 1.25-2.5x larger than that of the 6% halftone mask. As the pitch relaxes, multipole illumination captures more light from the 1st diffracted order light and the TWA increases significantly for the 18% halftone mask. This effect is also visible for isolated lines. It is clear that printing 130nm lines-through-pitch requires all possible RETs, from illumination path and from the masks.
Further predictions
End-of-line patterns and space-type patterns are always difficult to measure on CD SEMs. Thus, it is difficult to accurately calculate their process latitudes. In contrast, the aerial image approach appears to be accurate and much easier to use. Figure 5 examines the TWA for end-of-lines in the presence of various OPC-type RETs in comparison with no OPC treatment. We analyzed pitches of 1:1.5 and 1:2 with three levels of OPC in terms of process latitudes, two serif cases with sizes of 60 and 80nm, and a 60nm hammerhead. The TWA was calculated for 0.63NA with annular apertures and 0.70NA with multipole apertures. Shown in Fig. 5, ranking in terms of effective NA, illumination aperture shape, and optimization of OPC for end-of-line separation is summarized as follows:
- Illumination enhancement rankings point to the combination high NA (0.70NA) with multipole illumination. This combination produced a 1.5-2x larger TWA than any other NA-illumination combination.
- The best enhancement strategy appears to be multipole illumination with 0.70NA, combined with hammerhead as OPC, and 260nm (1:2 pitch) end-of-line for a design rule.
- Annular illumination marginally discriminates different EOL corrections, though this and 0.63NA work better for 260nm end-of-line separation (1:2 pitch) than for 195nm (1:1.5 pitch).
- Multipole illumination shows better discrimination between different end-of-line corrections and can identify smaller differences between 260nm and 195nm end-of-line separation. This indicates that a multipole aperture allows more freedom in the selection of a particular type of OPC correction.
- A combination of 0.70NA with multipole illumination can illustrate robust process latitudes for the smallest end-of-line separation (195nm), a conclusion very useful for IC designers when they look into shrinking cell designs.
Calibration of predictions with resist results
The final step in ranking RETs is the calibration of the predicted aerial image process latitudes, with resist-measured process latitudes. The calibration between aerial image prediction and real resist patterns has always been a concern and is becoming even more critical in the lithography regime where device geometries approach half the wavelength of the exposure tool. We suggest two ways to calibrate aerial image predictions.
First, calibration can be done by correlating two EL vs. DOF curves — one curve is generated from aerial image analysis, the other from CD measurements on actual wafers exposed on a 248nm scanner. In our example, both curves were generated from the same mask (18% halftone) and using similar illumination conditions, 0.70NA, and multipole illumination. Figure 6a shows the same shape for both EL-DOF curves. However, the predicted aerial image ED latitudes are smaller than the actual measured latitudes. Though the reason for this underprediction is not clear, stray light present when aerial images were gathered, as well as the nonlinear defocus response of the AIMS Z-stage, may be the causes.
The second method of calibration correlates CD through-pitch of aerial image predictions to wafer measurements. Such calibration is not easy, because the CD through-pitch proximity is a nonlinear function of all resolution enhancements.
 |
Figure 6. Correlation between lithographic results predicted from aerial intensity images and wafer measurements of resist CDs. Wafers were exposed on ASML 5500/700 scanner, and the same mask and similar illumination settings were used. a) Comparison of aerial image predicted process latitudes and wafer-generated process latitudes; and b) correlation between aerial image predicted CD proximity and wafer-measured CD proximity through-pitch. Aerial image CDs were calculated based on threshold to size at each particular pitch.
Figure 6b shows the linear regression correlating the aerial image threshold to the printed CD through-pitch at their nominal size with actual measured CDs. This plot covers 130nm lines from 1:1 to fully isolated. A 0.93 regression coefficient illustrates excellent correlation.
Conclusion
This study extends the use of mask aerial images to predict critical issues in low-k1 lithography: CD through-pitch proximity, end-of-line, and overlapping exposure defocus latitudes. The proposed method allows the lithographer to decide when and how to use RETs to minimize the level of OPC aggressiveness. Addressing these issues closes the loop between IC design and optical lithography decisions in selecting an OPC strategy for a specific design and to decide the RET combination generating the largest process latitude. Efficiency and accuracy of the proposed procedure were demonstrated by calibrating the predicted CD performance with measured wafer CDs.
Acknowledgments
The authors thank Mark Manslow from KLA-Tencor/Finle Division for developing the total window area software module; David Witko, Ted Paxton, Murali Guntu, and Anita Pici for wafer exposures and CD metrology; and in particular, Stephan Sinkwitz, ASML-Veldhoven, for his valuable comments, suggestions, and continuous support.
AIMS is a registered trademark of Carl Zeiss Jena GmbH. Quasar is a registered trademark of ASML.
References
- R. Budd, J. Staples, D. Dove, "New Tool for Phase Shift Mask Evaluation: The Stepper Equivalent Aerial Image Measurement System," SPIE, Vol. 2087, pp. 162-171, 1993.
- R. Budd et al., "A New Mask Evaluation Tool: The Microlithography Simulation Microscope," SPIE, Vol. 2197, pp. 573-583, 1994.
- A. Martino et al., "Application of the Aerial Image Measurement System, AIMS, to the Analysis of Binary Masks and Resolution Enhancement Techniques," Optical/Laser Microlithography, SPIE, Vol. 2197, pp.130-139, 1994.
- F. Finders et al., "DUV Lithography for 130-nm Using Off-Axis Illumination and Assisting Features, Semi Technology Symposium, Semicon Japan, December 1999.
- J. Finders et al., "KrF Lithography for 130nm, Optical Lithography XIII," Proceedings of SPIE, Vol. 4000, March 2000.
- D. Nam et al., "Lithography of 180nm Designs for 1 Gbit DRAM, SPIE, Vol. 3334, p. 117, 1998.
- J. Fung Chen et al., "Practical Method for Full-Chip Optical Proximity Corrections," SPIE, Vol. 3051, pp. 780-803, 1997.
- J. Fung Chen et al., "A Practical Technology Path for Sub-100µm Process Generation Via Enhanced Optical Lithography," MaskTools, internal report, 1999.
- A. Wong et al., "Level-Specific Lithography Optimization for 1-Gb DRAM," IEEE Transaction on Semiconductor Manufacturing, Vol. 13, No.1, pp. 76-87, Feb. 2000.
Mircea Dusa received his masters degree in electrical engineering and his PhD in applied optics from the Technical University of Bucharest, Romania, and worked in National Semiconductor's advanced R&D group. He is a senior imaging scientist at ASML's Technology Center, where he integrates the imaging components for sub-half wavelength lithography into a complete low-k1 imaging system. ASM Lithography, Parkway Towers, 4800 Great America Parkway, Ste. 400, Santa Clara, CA 95054; ph 408/855-0500, fax 408-855-0549.
Judith Van Praagh received her bachelors degree in chemistry from the University of West Brabant in The Netherlands in 1995, and served as an engineer in the laser group at the University of Eindhoven. In 1997, Van Praagh joined ASML, where she is an application engineer.
Jo Finders received his MS and PhD in physics from Aachen University of Technology, Germany. He worked as a member of the Micro-Patterning Group at IMEC, where he was active in resolution enhancement techniques, CD control, and CD metrology. Finders joined ASML in 1997, and serves as the project manager for advanced resolution enhancement techniques.
Andrew Ridley received his bachelors degree in physics from the University of California at San Diego, and served as a product specialist for the Micro Electronics Group at Carl Zeiss. He is a senior applications engineer for Metron Technology.