High-resolution maskless lithography

Transcription

1 High-resolution maskless lithography Kin Foong Chan* Zhiqiang Feng Ren Yang Akihito Ishikawa Wenhui Mei Ball Semiconductor, Incorporated 415 Century Parkway Allen, Texas Abstract. An innovative high-resolution maskless lithography system is designed employing a combination of low- and high-numerical-aperture (NA) projection lens systems along with integrated micro-optics, and using Texas Instruments super video graphic array (SVGA) digital micromirror device (DMD) as the spatial and temporal light modulator. A mercury arc lamp filtered for the G-line ( nm) is used as the light source. Exposure experiments are performed using data extraction and transfer software, and synchronous stage control algorithms derived from a point array scrolling technique. Each exposure scan produces a field width (W) of approximately 8.47 mm with a field length (longitudinal field) limited only by onboard memory capacity. DMD frame rates of up to 5 khz (kframes/s), synchronized to the stage motion, are achievable. In this experiment, TSMR-8970XB10 photoresist (PR), diluted to 3.8 cp with PR thinner is prepared. The PR is spin-coated onto a chromecoated glass substrate to 1.0- m thickness with 0.1- m uniformity. A 0.4- m scan step is used and 27,000 DMD data frames are extracted and transferred to the DMD driver. Results indicate consistent 1.8- m line space (L/S) resolved across the entire field width of 8.47 mm. Given optimized exposure and development conditions, 1.5- m L/S is also observed at certain locations. The potential of this maskless lithography system is substantial; its performance is sufficient for applications in microelectromechanical systems (MEMS), photomasking, high-resolution LCD, high-density printed circuit boards (PCBs), etc. Higher productivity is predicted by a custom H-line ( 405 nm) lens system designed and used in conjunction with a violet diode laser systems and the development of a real-time driver Society of Photo-Optical Instrumentation Engineers. [DOI: / ] Subject terms: exposure; microlens; microlithography; micro-optics; microelectromechanical systems; liquid crystal display; optics; pattern; printed circuit board; photomask; reticle. Paper received Dec. 11, 2002; revised manuscript received Feb. 25, 2003, May 8, 2003, and May 27, 2003; accepted for publication Jun. 13, Introduction Demands in the design and manufacturing of high-density printed circuit boards PCBs, high-definition liquid crystal displays LCDs, microelectromechanical systems MEMS prototyping, biosensor applications, etc., have faltered because of the high costs associated with the making of photomasks or reticles in the lithography process. Especially for research or small-volume production requiring feature sizes in the 1.0- to m range, the costs of photomasking have become logistically unreasonable. As a result, various maskless lithography techniques have recently caught widespread attention. These maskless methods are based on different technologies, utilizing resist nanodroplets, 1 electron beams, 2,3 the atomic force microscopy AFM -based technique, 4 direct extreme ultraviolet EUV or laser writing, 5,6 or incorporating various types of spatial light modulators and optical element However, *Present address: Reliant Technologies, Inc., Palo Alto, California. kchan@reliant-tech.com. Present address: Louisiana State University, Baton Rouge, Louisiana. ryang1@lsu.edu. most of these maskless techniques are highly inefficient i.e., slow exposure time, light efficiency, or are not able to achieve sufficiently small feature size i.e., line space L/S 20 m. Currently, none of these technologies possess a combination of sufficient throughput or productivity, optical efficiency, feature size, and cost efficiency to be commercialized. Wang and Bokor 1 attempted to directly deposit photoresist droplets on wafers or substrates by microfabricating a thermal bimembrane actuator. A 6- m-scale drop was recently reported, but this technology seems far from maturity. Droplet generation was highly irregular and required more in-depth investigation. Chen et al. 6 produced array nanomirrors intended for the design of an EUV maskless lithography system. However, this modulating device is still in the early stage of characterization and refinement, and not yet available for integration with a complete optical system for lithographic experimentation. Electron-beam lithography 2,3 has recently garnered attention because of the limitations of conventional optical lithography on feature sizes less than 65 nm. Ware 3 reported an electron projection lithography EPL system comprising of a 2-D e-beam columns, covering a total exposure area of mm 2. Despite having a longer JM 3 2(4) (October 2003) /2003/$ Society of Photo-Optical Instrumentation Engineers 331

2 depth of focus than that of optical lithography, the penetration depth of an e-beam is usually shorter and hence not conducive for use on thick resists. Additionally, the exposure area is too small for large exposure areas in the hundreds of square millimeter range. A scanning probe lithography SPL technique based 4 on AFM produced a larger exposure area of 100 mm 2. This technique employs a 1-D array of 50 cantilevered tips, spaced at 200- m intervals, to obtain patterns at 26-nm line width on polymethyl methacrylate PMMA at a scan speed of 10 m/s. This technology is still in the development stage and is likely to be limited by the scan rate. However, because of the extremely small feature size it is capable of producing, it offers great potential. At the expense of small feature size, maskless lithography systems designed with higher throughput in mind have been demonstrated. Takahashi and Setoyama 8 showed 50- m feature size on a 1.2- m negative photoresist layer through direct projection of a Texas Instruments TI digital micromirror device DMD -generated pattern. Seltmann et al., 7 using a custom-designed spatial light modulator SLM similar to TI s DMD, produced 0.6- m L/S features covering an image field of m. This was done by projection through a 100:1 reduction lens having an Excimer laser ( 308 nm) as the light source. The authors employed a step and stitch method on a 4-in. wafer and estimated a throughput of 1 wafer/h. Carter et al. 9 and Gil et al. 10 demonstrated another maskless technique termed zone-plate-array lithography ZPAL. This lithographic technique used an array of Fresnel zone plates with very high numerical aperture (NA 0.95), and the researchers succeeded in producing submicrometer feature sizes reported to be 360 nm. However, it has extremely low throughput, covering only an area of 0.25 cm 2 over a period of 20 min. We have developed maskless lithography systems offering minimum feature sizes at optimized exposure conditions approaching 20, 10, and 1.5- m L/S. Productivity can be as fast as 20 mm/s for the equivalent of an area of mm over 150 s of exposure time, depending on the systems and the illumination intensity of the laser light source. Fig. 1 Illustration of the Hi-Res MLS. Four diagrams on the right (from top to bottom) describe the appearance of the DMD, the DMD inverted image coplanar with the MLA back plane, the MLA foci coplanar with the spatial filter array, and the point array projected to the substrate. These diagrams are not adjusted for image magnification or reduction. 1.1 High-Resolution Maskless Lithography System The high-resolution maskless lithography system 11,12 Hi- Res MLS consists of two major subassemblies. First, the DMD field image is projected onto an integrated optical diffractive element using a low-na 1:3 in actuality 1:2.94 magnifying projection lens (NA 0.12). The focal plane of the diffractive element becomes the second object plane. The second lens assembly consists of a high-na 5:1 reduction lens (NA 0.5), which serves to project the focal points point array from the second object plane onto the substrate. Figure 1 shows a block diagram of the lithography system. The light source is provided through an optical fiber bundle or liquid waveguide to a diffuser and a line filter if necessary, after which it is collimated before it is reflected to the DMD surface with a UV-enhanced mirror. The super video graphic array SVGA array DMD, having a pitch size of 17 m, was a product of TI. The micromirrors on the DMD can be instructed to produce a pattern by way of tilting them 10 deg off their normal position. Light reflected off the micromirrors is projected into the low-na 1:3 magnifying lens, forming a DMD image coplanar to the back plane of the integrated microlens and spatial filter array 13,14 MLSFA. This is the first image plane where the individual DMD pixel image is focused by the corresponding microlens onto a second surface coplanar with the spatial filter array. Perfectly designed microlens array foci are diffraction-limited to about 3 m. In practice, the foci or focal points are approximately 5 m. The microlens array MLA foci serve as the second object plane for the high-na 5:1 reduction lens, imaged on to the substrate as a point array to perform lithography on photosensitive materials. Background illumination due to optical scattering or noise, crosstalk, high-spatial-frequency components, as well as imperfect focusing by the microlens array are eliminated at the foci with a corresponding array of spatial filters. This array filtering technique 12,14 increases the system contrast and produces better lithographic results. With the 5:1 reduction lens, the focal points or point array carrying the DMD pattern is projected onto the surface of the substrate. Since the DMD pixel images are focused into tiny dots point array by the MLSFA, they are disjointed during each exposure data frame. Hence, a point array technique 11,12,15 having unique formulations of overlaying array of dots is used to link lines and other patterns by flipping the DMD micromirrors at very high frame rates up to 5000 frames/s while moving the substrate on the stage. As indicated in Fig. 2, this technique requires the DMD field to be rotated at a small angle relative to the scanning 332 J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October 2003

3 the scanning direction coincides with four-point arrays, indicating a K value of 4. Given N 20 in Fig. 2, N/K 5, and hence To expose one uninterrupted nondisconnected line orthogonal to the scanning direction, the minimum horizontal separation x among the point array must be small enough given a finite point array size or diameter w Fig. 2, where x w. Thus, as long as x w, the minimum horizontal separation x is usually chosen as at least 20% of the smallest feature size a lithography system is capable of producing. For example, if the lithography system is expected to produce 2- m L/S features, then x 0.4 m. The minimum horizontal separation determines the horizontal overlaying density of the point array, shown in Fig. 2 as x d sin, 2 Fig. 2 Illustration of the point array technique. Each data frame of point array or DMD micromirrors is turned on according to the data extraction algorithm, given a set of input parameters such as the K value and the scan step, where K is the number or repetition of point array exactly overlapping one another at a coordinate position along a line drawn parallel to the scanning direction, d is the pitch size of the point array, is the discrete rotational angle, w is the point array size, x is the minimum horizontal separation, M is the number of point array pixels in the long axis, and N is the number of point array pixels in the short axis. A virtual row is added beyond the DMD field dimensions along the short axis to facilitate computation, resulting in N N 1, and thus K K 1. Note that the expression K K 1 is true only if N/K is a positive integer with an M N array larger than 1 1. In this case, K 4, highlighted with black dots to indicate four-point array repetition along the scanning direction. In this figure, for illustration purpose only, N 20 and K 4 (K 5), or N/K 5, hence, direction of the moving stage or substrate. A K value determines the discrete rotational angle. This relation can be written as K arcsin 2 2 1/2 N 2, 1 K where N is the number of point arrays along the axis having the smallest incidence angle to the stage scanning direction, as shown in Fig. 2; is termed the discrete rotational angle because for reasons of convenience in handling frame data extraction, we have required that N/K be a positive integer, hence restricting to certain discrete values; and K is the K value defined as the number or repetition of point arrays exactly overlapping one another at a coordinate position along a line drawn parallel to the scanning direction. For instance in Fig. 2, each virtual line drawn in parallel with where d is the point array pitch size, and is the discrete rotational angle. Equation 2 indicates that x influences the selection of K value through. In addition, a stage step-size or scan step during lithographic exposure along the scanning direction must also be smaller than the expected performance of the smallest feature size the lithography system is capable of. For example, if the lithography system is expected to produce 2- m L/S features, then a stage scan step of at least 20% of the feature, or 0.4 m, is recommended. Scan step determines the vertical overlaying density of the point array along the stage scanning direction, and each scan step is synchronized to its corresponding DMD frame data. The horizontal separation x and scan step of at least 20% or 20% of feature size is a criterion that was determined experimentally. Through our observation, this criterion resulted in feature sizes that are 10% of the intended pattern. For example, a 1.5- m L/S design pattern will result in a m line/1.35- m space or m line/1.65- m space on the substrate, which we deem acceptable. In our experimental setup, the Hi-Res MLS has a finite point array size w of approximately 1.0 m the MLSFA foci are 5.0 m, which are reduced by the 5:1 lens to point array sizes of 1.0 m, and a point array pitch size d of 10.0 m the DMD pitch size is 17 m, expanded to 50.0 m by the 1:2.94 lens, and then reduced to 10.0 m by the 5:1 lens. Assuming that the Hi-Res MLS can resolve at least a 2.0- m L/S feature, we set x to 0.4 m using the 20% rule, which also satisfies the criterion x w. From Eq. 2, we see that sin The TI SVGA DMD has micromirrors; N is 600. When N and are substituted into Eq. 1, we obtained K Since N/K must be a positive integer, we selected K 20 for convenience of data extraction, and thus N/K 30. Using Eq. 1 again, becomes deg. The final minimum horizontal separation x is now m. Because the Hi-Res MLS stage has a precision step size of 100 nm, it is prudent to set our scan step at 0.4 m while noting the slight difference in overlaying density between the horizontal/orthogonal and vertical/scanning directions. To facilitate data extraction, and synchronization between frame data and stage motion, M, N, K, d, w, scan step, and their derivatives i.e., and x are important input J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October

4 Fig. 3 Series of diagrams showing the progress of an exposure based on the point array technique. The exposure progresses from the left column (top to bottom) gradually to the right columns. The diagrams are labeled according to temporal increment from T 0, T, T 2,..., to T n, where n is an integer, and with the stage moving forward with a fixed scan step. An intended outline of the line patterns (white) is shown on the stage. As the stage scans underneath the DMD (or the Hi-Res MLS), some point array is activated as they pass within the intended pattern, exposing points that eventually connected and filled in to complete the line patterns (black) at T n. parameters in the software algorithm for data extraction and data transfer. The discussion of software manipulation is beyond the scope of this paper, but can be referred from Mei 12 and Zhou et al. 16 Figure 3 illustrates a sequence of point array data frames in the temporal domain to perform an exposure using the Hi-Res MLS. Based on the input parameters in the previous paragraph, we performed a 1-D simulation of cross-sectional overlaying intensity of 2.0-, 1.5-, and 1.0- m L/S features. Figure 4 shows the ideal input profiles and the corresponding simulation results of 334 J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October 2003

5 Fig. 4 Comparison between (a), (b), and (c) ideal input profiles and (d), (e), and (f) their corresponding simulation results of 1-D cross-sectional overlaying intensity profiles for 2.0, 1.5-, and 1.0- m L/S. The simulations were performed using (M,N) (848,600), d 10 m, w 1.0 m, K 20, and scan step 0.4 m as input parameters. The results indicate 2.0- and 1.5- m L/S features (d) and (e) are resolvable on photoresists with high-contrast characteristic curves treated with the appropriate preparation and development processes. The overlaying intensity contrast ratio for 1.0- m L/S, as seen in (f), is reduced, suggesting that 1.0- m features may not be easily produced. intensity dosage as a consequence of overlaying for the three cases of feature sizes. Results indicated that the contrast ratio of overlaying intensity in the case of a 1.0- m L/S feature is reduced compared to those of the 1.5- and 2.0- m L/S features, suggesting that the 1.0- m L/S features may not be achievable in actual experiments. The 2.0- and 1.5- m L/S features are, however, possible when using photoresists with threshold high contrast characteristic curves Methods 2.1 Exposure Experiment The performance of the Hi-Res MLS was evaluated by examining the best possible L/S pattern it was capable of resolving. The lithography system subassembly was done using propriety technologies, ensuring alignment accuracy between the DMD image and the MLSFA with tolerance less than 1/20 of a pixel for the whole field. The first portion of 1:3 projection lens and MLSFA were then aligned to the high-na 5:1 reduction lens. The assembled lithography system was then aligned to the stage scanning direction based on the desired or K value. The focal plane of the system was aligned to the substrate surface. The scanning/ scrolling and stitching not used in this experiment stages used in this experiment had a precision step size of 0.1 m on the horizontal x and y axes scrolling and stitching directions, 0.01 m onthez axis work distance adjustment, and 0.01 deg on the rotational stage. Any stitching misalignment if used can be compensated through software control since the stage absolute position is highly repeatable. Chrome-coated glass wafers were used as substrates in this experiment. An in-house spin-coater was used to uniformly spread a layer of photoresist PR on the substrates. TSMR-8970XB10 PR was prediluted to 3.8 cp with thinner. The substrates were initially spin-coated with OAP an adhesion-promoting agent between substrate and photore- J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October

6 Fig. 5 Exposure results of vertical lines (left column), horizontal lines (center column), and 45-deg angled lines (right column) relative to the stage scanning or point array pattern scrolling direction. Vertical lines were generated in the direction of the stage motion or point array scrolling. The top row shows scanning electron microscope (SEM) images of 1.5- m L/S, while the bottom row shows those of 1.8- m L/S. The magnified SEM on the top left corner shows the 1.5- m L/S in more detail. It is important to evaluate results of line patterns in different orientation to verify the system s alignment and stage synchronization during an exposure run. sist, by Tokyo Ohka Kogyo Co., Ltd. surface wetting at 3000 rpm for 25 s. Later, the diluted PR was spin-coated on the substrate at 3000 rpm for 30 s. On evaluation, the PR thickness was determined to be approximately m. The substrates were then prebaked at 120 C for 18 min. In this experiment, a total of 27,000 DMD data frames were generated using data extraction software algorithm, given (M,N) (848,600), d 10 m, w 1.0 m, K 20, and scan step 0.4 m as input parameters. The data were then transferred to local memories within the DMD driver. A diffuser was used to uniformly illuminate the entire field of the DMD surface, while a G-line filter was used to extract the optical power at a 436-nm wavelength from the low-power mercury arc lamp. Because of the relatively low illumination intensity, only a 0.04 mm/s scroll or scan rate was used. The Hi-Res MLS was then instructed to expose a total area of 8.47 (W) mm L. 3 Results 3.1 Maskless Lithography Exposure Results Figure 5 shows typical exposure results of pattern generated on a 1.0- m PR-coated chrome-coated glass substrate. Fig. 6 Exposure results of two star patterns on a glass substrate taken with SEM. The star pattern on the left (a) shows 1.0- m lines, whereas that on the right (b) shows 2.0- m lines. V-edges in (a) form a less perfect circle than that in (b), indicating synchronization error (this case) and/or discrete rotational misalignment limiting feature sizes of the Hi-Res MLS to 1.5 to 1.8 m. 336 J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October 2003

7 Fig. 7 Exposure results of (a) a circular-ring pattern on a glass substrate. The magnified SEM image on the right (b) shows a partial arc of 2.0- m L/S from the inner ring of the SEM image in (a). Results indicated full-field exposure of 1.8- m L/S on all line and space orientations. Patterns of 1.5- m L/S, however, were observed only in several locations where light uniformity and exposure conditions were optimized. The star pattern in Fig. 6 provides convenient evaluation of the performance of the Hi-Res MLS because of its diverging lines at many angles relative to the stage motion or point array scrolling direction. Any stage misalignment, discrete rotational misalignment, or synchronization error will result in nonuniform linewidths of the star pattern at different angles. Manifestation of the limitation of the point array technique in the Hi-Res MLS can be seen in Fig. 7. The magnified SEM image Fig. 7 b shows an arc-shaped 2.0- m L/S pattern after zooming in from the original circular-ring pattern Fig. 7 a. The arc-shaped pattern has pointed edges that were attributed to the nature of data generated and the finite size of the point array inherent in the point array technique. Such pointed edges may be minimized by reducing the synchronization scan step and/or by refining the focal size of the point array. 4 Discussion 4.1 Discussions on Exposure Results It was difficult to detect discrete rotational misalignment and synchronization error based on the 1.5- and 1.8- m L/S results in Fig. 5. That is partly because 1.5 and 1.8 m were well within the performance capability of the Hi-Res MLS. In Fig. 6 a, however, the discrepancies between discrete rotational angle of the system relative to the stage motion and synchronization became more apparent on a 1.0- m star pattern. The horizontal lines in Fig. 6 a seemed thinner and weaker than the vertical lines, and the tips of V-edges toward which the lines converged did not form a near circle as in the case of Fig. 6 b where the lines were 2.0 m wide. Thus, Fig. 6 a shows a higher degree of error in synchronization. Because of synchronization error between the stage scan step and DMD frame data, the activated point array spreads over a wider crosssectional area along the scanning direction, resulting in a weaker dose of average intensity, and thus a thinner/weaker line pattern with a shallower profile. This is in contrast to the vertical line, which has a strong and prominent line profile. If discrete rotational misalignment were more serious than synchronization error, the reversed scenario would occur; the vertical line would be weaker and less prominent than the horizontal line. Analyses of discrete rotational misalignment and synchronization error prior to exposure test on a substrate can be performed. An imaging device is placed on the scanning stage and the point array images during an exposure trial images formed by changing DMD frame data as the stage is being moved are observed on the monitor. We assume, for instance, that the vertical line is parallel to and the horizontal line is perpendicular to the stage scanning direction. If there were significant synchronization errors, the horizontal line would gradually drift upward or downward along the scanning direction as the frame data changes. Synchronization errors may be limited by tolerance in lens magnification that resulted in slight error in the desired point array pitch size d or the scan step accuracy of the stage. The errors can be compensated by refining the input parameter d and reextracting the frame data using the software algorithm. However, if the vertical line gradually drifts left or right, there is discrete rotational misalignment of which is a result of imperfect adjustment of the discrete rotational angle for the corresponding preset K value. Discrete rotational misalignment and synchronization errors are deemed acceptable only if the drift is within 10% of the smallest intended linewidth of the lithography system. For example, if the maskless lithography system is designed for exposure feature sizes of 2.0- m L/S, the error or drift must be within 0.2 m when analyzing the frame images during a trial exposure. The limitation of the point array technique can be seen in Fig. 7 b, where pointed edges appeared on 2.0- m L/S over an arc when magnified from a circular-ring-shaped pattern. These edges were a result of the finite focal size of the point array limited by Rayleigh s criterion, as well as the software algorithm used for data extraction inherent in J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October

8 Fig. 8 Portion of a photomask fabricated using the Hi-Res MLS, shown here after chrome etching. The smallest resolution on this photomask sample is 8- m L/S, easily achieved by this lithography system. The photomask is being used in testing some of our process technologies. the technique. To minimize this effect in the future, smaller point array sizes, reduced scan step, as well as modifications in the process of data extraction may be necessary. Our results indicated consistent exposure of 1.8- m L/S, with marginally acceptable results at 1.5- m L/S and yet unacceptable results at 1.0- m L/S. Actions are being taken to enhance the Hi-Res MLS performance to at least 1.0- m L/S as our target specification under the best conditions for the appropriate substrate handling and processing pre- and postexposure. These future improvements are discussed in the next section. Fig. 9 Small portion of a large 2-D array of concave microlens taken with a SEM. As shown, the pitch size of this microlens array is 17 m. The role of the Hi-Res MLS was critical in making this unique microlens array fabrication process possible. 4.2 Examples of Various Applications The Hi-Res MLS has been a workhorse within our research laboratory in the development of various technologies. We expect to further enhance the capability of the Hi-Res MLS at several fronts, which include faster data extraction and transfer rate, higher optical power and more efficient wavelength for illumination, and maximization of the DMD s capability. We currently manufacture standard 300-mW and 1-W violet diode ( 405 nm) laser systems for use in the design of the next generation 18 Hi-Res MLS. These laser systems are already being used for some of the company s products. Under development is also a high-speed driver for real-time data extraction and data transfer subsystem, which will tremendously improve the Hi-Res MLS s performance. In addition, future maskless lithography system design may employ DMD structures requiring much less data management 19 or even eliminating the DMD entirely. 20 Figure 8 shows a portion of a photomask made using the Hi-Res MLS on a chrome-coated glass substrate. We succesfully applied this system to make photomasks to further enhance the capability of our technologies and improve the understanding of our fabrication processes. Such goals can be achieved swiftly because of the short turn-around time facilitated by the Hi-Res MLS to produce new or modified patterns. In general, the exposure field length in the scanning direction is limited only by the amount of memory allocated in our electronics hardware. However, the horizontal field dimension perpendicular to the scanning direction is limited by the DMD cross-sectional field length. To expand the exposure field in the horizontal direction, multiple scans are performed adjacent to or overlapping one another, depending on the software algorithm used; thus termed a stitching function. Figure 9 illustrates a small section of a large 2-D array of a microlens measuring mm. In this case, the stitching function was applied. The Hi-Res MLS was used to generate patterns on fused silica or other optical glasses that have undergone preexposure preparation. After lithographic exposure with the Hi-Res MLS, the substrate was further treated through a series of semiconductor processes before the concave microlens array was produced as shown. 13,14 5 Conclusion The Hi-Res MLS has potential for numerous applications. These include but are not limited to high-definition LCD screen manufacturing, MEMS/microopticalelectromechanical systems MOEMS prototyping, photomask fabrication, high-resolution PCB production, biomedical device design, optical component fabrication, academic or institutional research, etc. The Hi-Res MLS is particularly suited for fast turn-around designs, device testing, modifications, and product improvement. We have successfully applied this technology to fabricate various inhouse components to enhance our company s competitive edge in other areas, be it semiconductor electronics or optical device fabrications. The market opportunities for the Hi-Res MLS are enormous. We are currently promoting the use of the Hi-Res MLS for academic and institutional research to expand the bases of applications in different science and engineering disciplines. Acknowledgments Funding for this research was provided by Ball Semiconductor Inc. and its shareholders. The authors would like to thank Mr. Akira Ishikawa for his dedication and support. The DMD used in this paper was a product of Texas Instruments, Inc. Also, many thanks to the time and efforts put in by the Exposure Systems Division team members to make this a successful project. References 1. Y. Wang and J. Bokor, Maskless lithography with nanodroplets, 2. C. S. Whelan, D. M. Tanenbaum, D. C. La Tulipe, M. Isaacson, and H. G. Craighead, Low energy electron beam top surface image processing using chemically amplified AXT resist, J. Vac. Sci. Technol. B 15 6, P. Ware, Removing the mask, OE Mag. 2 3, Mar C. Quate, Sub-micron lithography with the atomic force 338 J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October 2003

9 microscope, group/home/ HomePages/Presentations/U98part1a.PDF. 5. B. Bae, O. Park, R. Charters, B. Luther-Davies, and G. R. Atkins, Direct laser writing of self-developed waveguides in benzyldimethylketal-doped sol-gel hybrid glass, J. Mater. Res , Y. Chen, Y. Shroff, and W. G. Oldham, Modeling and control of nanomirrors for EUV maskless lithography, in Technical Proc. Int. Conf. Modeling and Simulation of Microsystems, pp , San Diego, CA Mar R. Seltmann, W. Doleschal, A. Gehner, H. Kuck, R. Melcher, J. Paufler, and G. Zimmer, New system for fast submicron optical direct writing, Microelectron. Eng. 30, K. Takahashi and J. Setoyama, An UV-exposure system using DMD, J. Inst. Electron., Inf. Commun. Eng. J82-C-II 3, D. J. D. Carter, D. Gil, R. Menon, M. K. Mondol, and H. I. Smith, Maskless, parallel patterning with zone-plate array lithography, J. Vac. Sci. Technol. B 17 6, D. Gil, R. Menon, X. Tang, H. I. Smith, and D. J. D. Carter, Parallel maskless optical lithography for prototyping, low-volume, production, and research, J. Vac. Sci. Technol. B 20 6, W. Mei, T. Kanatake, and K. Powell, Maskless exposure system, U.S. Patent No. 6,425,669 B W. Mei, Point array maskless Lithography, U.S. Patent No. 6,473,237 B R. Yang, K. F. Chan, and W. Mei, Method for fabrication of shadow mask on micro lens array by self-alignment, U.S. Patent pending. 14. R. Yang, K. F. Chan, and W. Mei, Design and fabrication of microlens and spatial filter array by self-alignment, Proc. SPIE 4985, W. Mei, T. Kanatake, and A. Ishikawa, Moving exposure system and method for maskless lithography system, U.S. Patent No. 6,379,867 B X. Zhou, T. Kanatake, W. Mei, and K. F. Chan, System and method for lossless data transmission, U.S. Patent pending; Application Pub No. US 2002/ H. J. Levinson, Principles of Lithography, pp , SPIE Press, Bellingham, WA J. Zhai, W. Mei, and K. F. Chan, High power incoherent light source with laser array, U.S. Patent pending; Application Pub No. US 2002/ W. Mei and K. F. Chan, Light Modulation Device and System, U.S. Patent No. 6,433,917 B K. F. Chan, W. Mei, J. Zhai, and A. Ishikawa, Integrated laser diode array and applications, U.S. Patent pending; Application Pub no. US 2002/ Kin Foong Chan received his BS, MS, and PhD degrees, all in electrical engineering, from the University of Texas (UT) at Austin in 1996, 1997, and 2000, respectively. Between 1994 and 1998, he worked on electronics design for data acquisition and instrumentation control with National Instruments Corporation in Austin, Texas. He has been an optical research engineer with Ball Semiconductor, Inc., since 2000, and has been an appointed adjunct professor in biomedical engineering at UT Southwestern Medical Center since 2002, both in Dallas, Texas. He joined Reliant Technologies, Inc., Palo Alto, California, in August 2003, to develop advanced laser and optical technologies for clinical dermatology. His research interests include biomedical optics and instrumentation, electro-optical devices, lithography, and optical nanotechnology. He has published numerous journal and proceedings papers related to laser-tissue interaction and microlithography, and has several issued or pending U.S. and foreign patents. Dr. Chan is a member of Tau Beta Pi, Eta Kappa Nu, SPIE, and the IEEE. working under Dr. Therese M. Cotton on the electron transfer property of redox reactive proteins. Since 1997 Dr. Feng has been with Ball Semiconductor, Inc., in the Ball Lithography Group as senior R&D and process engineer. From 1997 to 2000 he was a corporate researcher with Tohoku University of Japan, working in Dr. Masayoshi Esashi s Lab on the application of microelectromechanical systems (MEMS) technology for Ball new devices. He is now with Ball s Exposure System Division working on the development of a highresolution exposure system and microlens array fabrication. Dr. Feng was a member of the Electrochemistry Society of Japan from 1994 to 1996 and is a member of the American Physical Society and the IEEE. Ren Yang attended Tsinghua University, majoring in optical precision instruments, and received his BS engineering degree in optical precision instruments in May 1996 and his MS degree in optical engineering in March Beginning June 1999, he furthered his graduate studies in microelectromechanical systems (MEMS) with Louisiana State University, where he received MS degree in engineering science in May Mr. Yang became an optical engineer with Ball Semiconductor, Inc., in September Akihito Ishikawa joined Ball Semiconductor, Inc. (Ball), when the company was founded in October 1996, as a photolithography process engineer, and led much of the development activities for various resist-coating techniques and equipment for the spherical devices. Prior to joining Ball, he founded AtoZ Systems Performance Motoring (AZSys) in 1992, while pursuing his BS degree in mechanical engineering at University of Texas, Arlington. AZSys served the auto racing industry, supporting international race teams in vehicle testing, vehicle preparation, and logistical services. Since his appointment as the Marketing Director in January 2001, he has been promoting the maskless exposure and spherical device technology products at Ball Semiconductor, Inc. Wenhui Mei holds a PhD degree in opticoelectronic engineering from the Beijing Institute of Technology (BIT). He is the president of the Exposure Systems Division and was elected a vice president of Ball Semiconductor, Inc., in March He worked in design engineering for GL Automation Inc. and USA Display L.L.C. before coming to Ball Semiconductor, Inc., in May 1998 as an optics and mechatronics engineer. Dr. Mei is a member of the Optical Society of America and the SPIE and was formerly a professor at BIT. Zhiqiang Feng received his BS degree in physics from Fudan University in During the following 6 years he was a research associate in physics with the Shanghai Institute of Mechanical Engineering. He entered the Yokohama National University of Japan as a research student in 1991, where he received his MS degree in 1994 and his PhD degree in physical chemistry from in After receiving his doctorate, he spent a year as a postdoctoral fellow with the Department of Chemistry, Iowa State University, J. Microlith., Microfab., Microsyst., Vol. 2 No. 4, October