Simple interferometric fringe stabilization by CCD-based feedback control Preston P. Young and Purnomo S. Priambodo, Department of Electrical Engineering, University of Texas at Arlington, P.O. Box 19016, Arlington, TX 76019, young@uta.edu Theresa A. Maldonado, Electrical and Computer Engineering Department, Texas A&M University, 3126 TAMU, College Station, TX 77843-3126, maldonado@tamu.edu Robert Magnusson Department of Electrical & Computer Engineering, University of Connecticut 317 Fairfield Road, Unit 2157, Storrs, CT 06269-2157, robert.magnusson@uconn.edu A method for producing stabilized interference patterns for ultraviolet (UV) interference lithography utilizing a CCD camera as the detector element is described. Intensity data obtained from the CCD element is filtered in software to minimize speckle and detector noise effects as well as determine the relative phase of the interfering beams. A control signal is then issued to correct the fringe drift. The system allows rapid reconfiguration of the lithography setup with minimum realignment of optical components. 2005 Optical Society of America OCIS codes: 090.2880, 000.2170 1
1. Introduction Fabrication of large-area diffraction gratings with sub-wavelength periods requires recording of high-quality photoresist gratings. UV holographic interference lithography systems are routinely used for recording fringe periods ranging from ~0.2μm to ~1μm. Non-linear photoresist exposure characteristics allow only narrow latitude for exposure dosages that produce optimum photoresist grating profiles with predictable fill factors. Thermal fluctuations, air currents, and vibration are environmental sources of fringe instability during holographic exposure. Numerous passive methods are available to improve the holographic environment, but often, active fringe stabilization becomes necessary to obtain consistent photoresist exposure. [1] Single-beam lithographic interferometer setups can be quickly reconfigured to produce different grating periods, but are not easily stabilized to compensate for localized disturbances that cause fringe movement during the exposure. One additional drawback to the single-beam configuration is the inherent non-uniformity of the beam intensity over the exposure area. This effectively limits the area of photoresist exposure uniformity. The uniform exposure area for a two-beam setup is approximately double that for the single-beam, but the splitting of the available exposure energy dictates longer exposure times over that of the single-beam lithography system. Exposure times are also greatly increased when recording grating periods approaching 200nm. To reduce blurring and loss of contrast during exposure, active fringe stabilization is required to produce photoresist grating patterns having optimum fill factor and good sidewall profiles. Traditional fringe stabilization methods include locking to discrete analog photodiode signals [2-4], mixing of substrate reflection beams [5], and digital heterodyne beam modulation [6]. The simplest of these methods described in [2-4] utilizes two photodiodes placed alternately in one peak and one null of a projected interference pattern. The methods of [5, 6] require additional optical components and sophisticated control equipment. Video-based fringe stabilization systems have also been implemented as fringe detectors for use in metrology applications such as fringe shift interferometry and electronic speckle pattern interferometry (ESPI) [1, 2, 7] for measurement of surface deformations. Applications in ESPI and fringe-shift interferometry are typically performed with visible wavelength lasers allowing simplified setup of the fringe detection system. 2
A primary complexity in setup of the UV lithographic interferometer is detection of the stabilization fringe pattern. The proper fringe spacing for stabilizing a UV interferometer is difficult to detect when the beam intensities are weak and the interference pattern from nearparallel beams cannot be visually observed. Properly aligning two discrete detectors significantly increases the time necessary to reconfigure a UV lithographic interferometer. The use of a CCD camera based system allows simplified interferometer setup as well as provides the fringe detector for the active feedback system. In this work, a hardware configuration similar to that described in [2] is implemented as a fringe stabilization system (rather than as a fringe-shifting interferometer) with PC-based hardware and software that allows data acquisition and processing at the camera frame rate. Main advantages of this approach are simplicity in setup and use. Figure 1 shows the basic twobeam UV grating exposure system. In the case reported here, collimating optics are not used after the spatial filters thus the beams are highly expanded to provide uniform illumination over the substrate area. The system is easily reconfigured to produce different grating periods. The beam sampling mirror(s) near the substrate allow the CCD camera to be positioned at any convenient location in the setup. A PC-based frame grabber processes the 8-bit CCD video intensity data at a 25Hz sample rate while the piezoelectric positioner is driven by an inexpensive 14-bit D/A converter. 2. Fringe stabilization method In typical configuration, approximately 3-5 fringes are projected across the CCD detector element using a beam expander. Speckle noise from the beam expander optics and CCD shot noise have high spatial frequencies compared to the fundamental frequency of the interference fringes. Figure 2a shows a typical full frame of the CCD video data. Typical spatial noise artifacts introduced by imperfect optics are indicated at points A. These spatial noise patterns remain spatially fixed for any given setup. Point B is on an arbitrarily-selected single line of the CCD video frame data extracted to determine the relative position of the fundamental interference fringe pattern. A Discrete Fourier Transform (DFT) is performed on the 8-bit intensity data over the video frame width allowing decomposition into spatial harmonics. 3
By performing an inverse DFT (IDFT) while retaining only the low spatial frequency components associated with the fundamental fringe pattern, a smooth waveform with discrete peaks and minima is obtained. Figure 2b shows the reference frame data before and after filtering. The two curves are offset for clarity. The unfiltered reference data retains 256 harmonics while the filtered fringe pattern is constructed from the lowest 35 spatial harmonics of the video frame width. On the start of phase correction during exposure, an initial pattern of filtered data is established as the zero-phase reference for the remainder of the exposure. Subsequent frames are processed in a similar manner, revealing lateral shifts in the fundamental fringe pattern irrespective of any noise superimposed on the acquired intensity waveform. The software then simply tracks the minima and maxima of the most recently acquired waveform relative to the reference, and issues an error signal through the D/A converter to correct the fringe shift. Since the IDFT places the filtered intensity data in bins (memory locations) indexed according to the original reference frame pixel data, the fringe minima and maxima are followed within an accuracy of one pixel. For five fringes projected across a 512 pixel element CCD detector, one pixel corresponds to ~1/100 fringe. For ~2.5 projected fringes, the detection and correction accuracy is ~1/200 fringe. This accuracy is a significant improvement over conventional photodiode-based fringe stabilization systems. 3. Results Figure 3(a) shows the typical uncorrected fringe shift during a 150 second exposure. The corresponding stabilized fringe shift is given in Figure 3(b). The results indicate that the stabilized fringe shift is typically less than 1 degree with maximum deviations less than two degrees over the duration of the exposure, a vast improvement over the uncorrected case. The sample rate of 25Hz is fast enough that feedback positioning of the piezoelectric element is performed by direct control rather than the proportional / integral / differential (PID) method implemented in [7]. PID control is not necessary for the current UV lithography system, but could be easily implemented in software if necessary. The excellent fringe stability provided by the CCD-based system has allowed consistent fabrication of sharp photoresist grating features with periodicity near 200nm (100nm line features) over a 20 cm 2 area. Figure 4 shows photoresist grating structures that are routinely fabricated with the UV lithography system. Figure 4 illustrates the sensitivity of the grating fill 4
factor to the photoresist exposure dosage. The exposure of Figure 4b is approximately 10% greater than that of Figure 4a, showing that the grating fill factor decreases from ~60% to less than 50%. The CCD-based fringe stabilization system allows consistent reproduction of these results. 4. Conclusions After frequency domain removal of the high spatial frequency intensity noise components, the CCD-based active fringe stabilization system provides fringe control better than +/- 1/100 fringe using inexpensive off-the-shelf components. The UV lithography system can be rapidly reconfigured to produce different photoresist grating periods with minimum realignment of the optical components. Performance is largely dominated by the CCD camera frame rate. Since only a single line of the frame data is utilized to determine the exposure fringe position, a line scan CCD element with an increased refresh rate would allow fringe correction to be performed at 500Hz rates or higher with a currently available microprocessor hardware. Acknowledgments: This work was supported, in part, by a State of Texas Advanced Research Program Grant #003656-0159-2001. The authors thank Ophir Optronics, Inc. for providing the camera and PCbased frame grabber hardware and software driver information. 6. References 1. C. Guest and T. Gaylord, Phase stabilization for holographic optical data processing, Appl. Opt. 24, 2140-2144 (1985). 2. P. Hariharan, B. Oreb, and N. Brown, Real-time holographic interferometry: a microcomputer system for the measurement of vector displacements, Appl. Opt. 22, 876-880 (1983). 3. G. Saxby, Fringe Stabilization, in The Manual of Practical Holography, (Oxford [England]; Boston: Focal Press, 1991), pp. 423-432. 4. J. Odhner, Stabilock II Active Fringe Stabilizer, (Stabilock Application Notes, 1993). 5
5. C. Lima and L. Cescato, Mixing of the reflected waves to monitor and stabilize holographic exposures, Opt. Eng. 35, pp. 2804-2809, (1996). 6. R. Heilmann, P. Konkola, C. Chen, G. Pati, and M. Schattenburg, Digital heterodyne interference fringe control system, J. Vac. Sci. Technol. 19, 2342-2346 (2001). 7. F. Hrebabetzky, A. Albertazzi Jr., Camera-based active phase stabilization for electronic holography, in Laser Metrology for Precision Measurement and Inspection in Industry 1999, Proc. SPIE 4420, 155-161 (2001). UV Argon Laser λ = 364nm or 351nm Portions of main exposure beams used for fringe stabilization Expanded fringe CCD pattern Camera Beam Expander Mirror Shutter Main exposure beams Stabilized fringe pattern recorded on substrate Attenuator Spatial Filters Mirror 60:40 Beamsplitter Piezoactuated Mirror Mirror (placed above plane of substrate shadow) Substrate holder Control PC D / A Converter Figure 1. Schematic layout of interferometer and fringe control system. 6
Figure 2. Video frame and spatial intensity data. (a) Full frame of acquired interference pattern acquired from CCD camera. Points A show typical fixed spatial intensity noise. Inset B shows the line of data extracted to obtain the current phase reference. (b) Unfiltered (top curve omitting highest-order spatial harmonic) and filtered (bottom curve) intensity data reconstructed from single line of data indicated in the full frame. 7
Figure 3. Measured uncorrected and corrected fringe shift in degrees versus time for a 150 second exposure. (a) (b) Figure 4. 250nm photoresist grating structures with variable exposure energy. E = 71mJ / cm 2 in 170 seconds (a). E = 77 mj / cm 2 in 190 seconds (b). 8