Lithography: Birefringence in fused silica and CaF2 for lithography
02/01/2000
Part One in a Series
Baoliang "Bob" Wang, Hinds Instruments Inc., Hillsboro, Oregon
Studies of residual linear birefringence in fused-silica and calcium fluoride samples indicate that optical component suppliers and microlithography tool manufacturers and users need to be more aware of this property and its effect on optical lithography progresses.
Fused silica and calcium fluoride (CaF2) are used in advanced optical lithography systems. While fused silica is the standard lens and mask material for 248nm (KrF excimer laser) generation step-and-scan systems, CaF2 is becoming important in 193nm (ArF) generation step-and-scan systems and is seemingly the only practical choice of lens material for emerging 157nm (F2) lithography [1]. As optical lithography advances, the quality requirements for these two materials become more stringent, including minimizing residual linear birefringence in optical components. (Hereafter, "birefringence" implies "linear birefringence.")
Residual birefringence occurs from mechanical stress or strain formed during the production of an optical element. It is characterized by magnitude (retardation or retardance) and orientation (angle of fast axis), as for a quarter wave plate.
Residual birefringence in an optical element is an indication of the inhomogeneity of refractive indices. A high level of residual birefringence in an optical component can lead to the aberration of light rays in an imaging system and thus affect the imaging quality of a step-scan instrument. When linearly polarized light is used, the orientation of birefringence becomes a key factor for altering light polarization.
Residual birefringence is usually measured as retardation, a relative phase shift between the two orthogonal linear polarization components of the passing light. Retardation represents an integrated effect along the path of a light beam traversing the sample.
 |
Figure 1. a) A surface plot for birefringent magnitude, and b) birefringent angle distribution from a 6-in. fused-silica photomask blank. The retardance unit in a) is nm/cm. The actual retardance of the blank, with its 0.635cm thickness, is 64% of the values in a).
Traditional methods for measuring birefringence use crossed polarizers and wave plates [2]. These methods have relatively low sensitivity and speed for measuring birefringent magnitude and do not provide sufficient information for determining birefringent angle. On the other hand, photoelastic modulator (PEM) technology [3], which is used in the Exicor birefringence measuring instrument [4], provides both high accuracy and fast speed to researchers in the semiconductor industry who are concerned with residual birefringence in optical components. PEM in the Exicor system provides simultaneous determination of both magnitude and angle of residual birefringence at high sensitivity (i.e., 0.005nm) and reliability [4, 5].
 |
The unit for birefringent magnitude used here is nm/cm (i.e., retardance normalized to the path-length of a sample), but can be converted into radians, degrees, or number of waves if desired.
Figure 2. Birefringent image of a 6-in. CaF2 photomask blank.
We have used the Exicor system to study a variety of fused-silica and CaF2 samples in our laboratory that exhibit different levels of magnitude retardance [4-6], including birefringent studies of photomask blanks, lens materials, and optical windows.
Fused-silica photomask blanks
Drawing from our detailed presentation on birefringence in photomask blanks at a recent SPIE conference [6], Fig. 1a shows the magnitude of retardance color-coded and displayed in 3-D. Figure 1b is a 2-D plot of the angle of the birefringent fast axis, where the direction of a short bar indicates the angle of the fast axis.
The data depicted in Fig. 1a show a birefringent pattern typically found in fused-silica photomask blanks that the author has measured. Note that four areas along the extreme edges of the blank have the highest birefringent magnitudes. Fortunately, these areas are at the periphery, away from the center of the photomask blank (labeled O) where a reticle pattern is typically located. Residual birefringence in these extreme edge areas may not affect the reticle image projected onto the wafer. However, there are four other regions of the blank between the center and edges (labeled A, A', B, and B') with high levels of residual birefringence; these will likely coincide with the fabricated reticle patterns, especially if the reticle contains more than one die. In this case, residual birefringence for different areas in the same die will be significantly different.
Our instrument uses a He-Ne laser as the light source, measuring birefringence of a sample at 633nm. To obtain the birefringence at 193nm, we use a conversion factor of 1.5, which was obtained from previous studies on stress-optic coefficients of quartz and silica [7, 8].
 |
Figure 3a. Birefringent images of CaF2 "windows": a 40mm dia., 6.35mm thick sample.
Figure 1a then leads to a calculation of about 10nm (17°) at 193nm for the residual birefringence in the areas of A, A', B, and B'. If the illuminating light is linearly polarized in the horizontal direction (bisecting OA and OB), the residual birefringence at areas around A, A', B, and B' will act almost like a collection of 1/20 wave plates oriented at different angles. Passing through these four areas, the incident light will become elliptically polarized and can be decomposed into a linear polarization component (bisecting OA and OB) and a circular polarization component. A 10nm retardance oriented 45° from the incident polarization will result in a circular polarization component of ~10% at 193nm.
Figure 1b indicates that regions A, A' and B, B' have birefringent angles approximately aligned at 45° and -45°, respectively. Therefore, the two sets of diagonal quadrants, AA' and BB', will lead to opposite parities of circular polarization (i.e., "right and left handedness"). If certain components in the optical train of a lithography tool that follow the mask are sensitive to either linear or circular polarization, a specific light-intensity pattern would be projected onto the wafer. This intensity pattern introduced by the mask blank would affect the quality of the image projected onto the wafer. However, if the residual birefringence in a photomask is controlled below 2nm, the circular polarization component will be <0.5% and will lead to an even smaller light-intensity inhomogeneity on the wafer.
A given level of residual birefringence in a photomask will affect all-refractive and catadioptric step-scan systems [9, 10] in different ways. It is beyond the scope of this article to discuss in detail the relationship between polarization inhomogeneity in the light beam and image quality projected onto the wafer for different step-scan systems.
 |
Figure 3b. Birefringent images of CaF2 "windows": a 90mm dia., 35mm thick sample.
We were surprised to find such a high level of residual birefringence in some of the fused-silica photomask blanks that we measured. Our caveat here is that residual birefringence in fused silica used for photomask blanks deserves more attention. Our experience in using fused silica for building PEM instruments is that the level of residual birefringence in our optical elements is <<0.5nm/cm [3, 4].
CaF2 optical components *** Measurement of the magnitude and angle of residual birefringence depicted together in Fig. 2 in a 6-in. CaF2 mask blank revealed a level comparable to that present in most fused-silica blanks. However, to date, we have only measured one CaF2 blank because they are not readily available and are not likely to be used as a photomask substrate due to this material's much higher thermal expansion coefficient. Further studies on more samples will be required to reach a more meaningful conclusion.
CaF2 "windows," samples from 40-205mm in diameter and from ~6-55mm thick (used for manufacturing optical components), are more readily available:
- A 40mm dia., 6mm thick CaF2 window manufactured under optimal conditions showed randomly oriented angles of birefringence and a few data points with a magnitude above 1.3nm/cm (Fig. 3a). This CaF2 sample perhaps represents the highest quality level in terms of residual birefringence currently available in the industry.
- A 90mm dia., 35mm thick CaF2 window showed birefringence values below 4nm/cm, except in some peripheral regions (Fig. 3b). This sample also showed a regular distribution in the angle of birefringence, indicating a possible annealing problem during its manufacture. Our evaluation was that the quality of this CaF2 window was less than that in Figure 3a.
- A ~205mm dia., ~55mm thick CaF2 sample (Fig. 3c) was measured in quadrants. (The area exceeded the scanning limit of our current instrument.) These data show strong local regularities in the angular distribution of residual birefringence, and the overall level is also high. This sample had the lowest quality of those we studied.
The reader will notice that the round images, especially in Fig. 3b, show a few gray data cells without angle indicators on the edge of the windows. With our instrument, when <2% of the light intensity reaches the detector because it is reflected, blocked, or diffused by the edge of the sample, the software labels the corresponding data point invalid and displays it in gray. Practically all data points around the edge of a sample must be interpreted with caution, since several cases may lead to false conclusions about the birefringence magnitude [4]; a good practice would exclude data points within 0.5mm of the outer edge of the sample.
 |
Figure 3c. Birefringent images of CaF2 "windows": one quadrant from a 205mm dia., 55mm thick sample.
The data that we gathered from these CaF2 samples seems to indicate that a larger CaF2 optical element is likely to have a higher level of residual birefringence. Accordingly, this drives the need to carefully monitor optical quality during the manufacture of optical elements from such material.
Irradiated CaF2 optical element
Fused silica compacts when it is irradiated with 193nm excimer-laser energy [11], but CaF2 seemingly does not because it is a stable, single crystal [12]. With our ability to detect small variations in residual birefringence, we looked for 193nm irradiation effects on CaF2. Our sample was a rectangular prism of CaF2 (20mm x 40mm x 80mm) with all six surfaces polished. This prism had been irradiated through the central ~5mm dia. of its 20mm x 40mm surface with a fairly large dose of exposure (>2 billion pulses at ~1mJ/pulse). We imaged birefringent magnitude and angle using the Exicor instrument over the central 18mm x 18mm area on the prism's 20mm x 40mm surface (Fig. 4); birefringent results in Fig. 4 are normalized over the light-path length of 80mm.
 |
Figure 4. Birefringent image of a 193nm excimer-laser-irradiated CaF2 prism.
The data in Fig. 4 show variations of both birefringent magnitude and angle across the entire region scanned for this CaF2 element. However, such variations are perhaps due to residual strain formed during production. If the 193nm excimer-laser irradiation had any appreciable effect on residual birefringence, there would be some regular change in birefringence around the center of the image shown in Fig. 4, which we did not observe.
Conclusion
As optical microlithography moves to finer resolution, the requirements for high-quality optical components used in microlithography systems become more stringent. Residual linear birefringence in fused silica and calcium fluoride is an important, but much neglected property that affects the quality of optical components in leading-edge applications. The semiconductor industry's attention to this property should lead to products with improved performance.
Acknowledgments
The author thanks many colleagues and friends for generously providing samples for birefringence measurements. Exicor is a trademark of Hinds Instruments Inc.
References
- D. Collier, W. Pantley, Semiconductor Online, May 5, 1998.
- H. Aben, C. Guillemet, Photoelasticity of Glass, Springer-Verlag, Berlin, pp. 114-115, 1993.
- J.C. Kemp, J. Opt. Sci. Amer., 59, 950, 1969.
- B.L. Wang, T.C. Oakberg, Rev. Sci. Instrum., 70, 3847, 1999.
- B.L. Wang, T.C. Oakberg, P. Kadlec, SPIE Proceedings, 3754, 197, 1999.
- B.L. Wang, P.M. Troccolo, "Photomask Technology and Management," SPIE Proceedings, 3873, September 15, 1999.
- N.K. Sinha, Phys. Chem. Glasses, 19, 69, 1978.
- A.J. Barlow, D.N. Payne, IEEE J. Quantum Electronics, QE-19, 5, 1983.
- H.J. Levinson, W.H. Arnold, Chapter 1 in Handbook of Microlithography, Micromachining, & Microfabrication, Vol. 1, ed. P. Rai-Choudhury, pp. 11-138, 1997.
- J.A. McClay, A.S.L. McIntyre, Solid State Tech., pp. 57-68, June 1999.
- W.G. Oldham, R.E. Schenker, Solid State Tech., p. 95-102, April 1997.
- V. Liberman et al, Opt. Lett., 24, 58, 1999.
Baoliang Wang is an applications scientist at Hinds Instruments Inc., 3175 NW Aloclek Dr., Hillsboro, OR 97124; ph 503/690-2000, fax 503/690-3000, e-mail [email protected].