Open Access
Add to BibSonomy
Abstract
Maskless lithography based on a digital micromirror device (DMD) has the advantages of high process flexibility and a low production cost. However, due to the trade-off relationship between the pixel size and exposure area, it is challenging to achieve high resolutions and high patterning speeds at the same time, which hinders the wider application of this technology in micro- and nano-fabrication processes. In addition, micromirrors in DMDs create pixelated edges that limit the pattern quality. In this paper, we propose a novel DMD maskless lithography method to improve the pattern quality during high-speed continuous patterning by means of pulse exposure and oblique scanning processes. A unique criterion, the pixel occupancy, was devised to determine the parameters related to the pulse exposure and oblique scanning optimally. We also studied how the duty cycle of the pulse exposure affects the pattern quality. As a result, we were able to increase the scanning speed up to the speed limit considering the damage threshold of the DMD and improve the pattern quality by resolving the pixelation problem. We anticipate that this method can be used in various microfabrication fields with short product life cycles or in those that require custom designs, such as the manufacturing of PCBs, MEMS devices, and micro-optics devices, among others.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Photolithography, a manufacturing process that utilizes patterned light and a photoresist (PR), is currently the de-facto standard in various fields of micro- and nano-fabrication, including printed circuit board (PCB) manufacturing and semiconductor manufacturing, as this method allows the creation of small patterns close to and even smaller than the diffraction limit of light [1,2]. Conventional lithography uses a photomask to transfer a geometric pattern to a PR on a substrate. However, this conventional method has several disadvantages, including considerable production costs and time-consuming steps when fabricating the photomasks as well as difficulty in managing contamination of the photomasks [3]. Research on maskless lithography, which generates a geometric pattern without using a photomask, aims to overcome these disadvantages of conventional lithography. Maskless lithography directly writes a pattern using a laser [4–6] or creates a pattern using active multi-element optical devices such as a grating light valve (GLV) [7,8] or a digital micromirror device (DMD) [9–12] instead of a photomask. One of the major advantages of maskless lithography is its flexibility, which allows users of this method to generate arbitrary pattern shape with a digital input. Because there is no need to fabricate a physical mask, generating various patterns becomes easy and inexpensive. In particular, DMD lithography has become a popular technique in maskless lithography given current improvements in DMD technology. However, due to the trade-off relationship between the pixel size and exposure area, the patterning speeds of high-resolution maskless lithography are slower than those used with mask lithography. Therefore, improving the patterning speed is one of the most important issues in the study of maskless lithography [13].
Early DMD lithography techniques were implemented in a step-and-repeat manner which exposes a pattern while the stage does not move and moves to the next pattern area after finishing the exposure [10]. The step-and-repeat method can be extremely slow to pattern a large area because many acceleration and deaccelerating steps of a motorized stage are needed. On the other hand, DMD scanning lithography, which generates a stationary pattern on the sample while continuously moving the stage with the high frame-rate of the DMD, makes it possible to achieve high throughput in maskless lithography [10,11]. Additionally, this scanning-based method is more robust against dead pixels because a specific location on the substrate is exposed by all pixels in the same column of DMD [14]. When using this scanning method, the maximum frame rate of the DMD is the limiting factor of the scanning speed if the exposure intensity and travel speed of the scanning stage are sufficient. During one frame duration of the DMD, if the travel distance of the stage exceeds the pixel size, a motion artifact will create linear blur along the direction of the stage scan, which causes a speed constraint due to the DMD frame rate. In order to increase the stage scanning speed beyond the speed limit set by the frame rate, pulse exposure can be used to reduce the motion blur [14].
Another problem associated with DMD scanning lithography is a pixelated pattern shape. Due to this pixelation problem, a small pixel size must be used to create a diagonal or curved pattern, which causes a decrease in the patterning speed. To address this issue, methods that enhance the pattern quality at relatively large pixel sizes have been proposed, such as gray-level dosing [15–17], sub-pixel shifting [18–21], and oblique scanning [11,12,22]. These methods can pattern a curved shape with a relatively large pixel size. However, these sub-pixel exposure methods can be even slower because many different DMD patterns are required.
In this paper, we propose a maskless lithography method based on a DMD that simultaneously improves the pattern quality and patterning speed while utilizing both pulse exposure and oblique scanning strategies. To ensure the effective and simultaneous implementation of pulse exposure and oblique scanning, we devised a unique criterion, referred to as pixel occupancy, a unitless measure that represents the uniformity of the DMD pixel coverage. Using this criterion, we were able to optimize parameters related to pulse exposure and oblique scanning to ensure high speeds and high quality levels. We also studied how the duty cycle of pulse exposure affects the pattern quality, including the edge response of the pattern. Finally, we experimentally demonstrate that the proposed maskless lithography method can achieve higher speeds compared to those with conventional DMD scanning lithography with high-quality patterns by resolving the pixelation problem at the same time.
2. Method
2.1 Pulse illumination method
Figure 1 shows a schematic of the DMD scanning lithography system and the DMD projector. The prototype system is designed as a gantry-type system that moves the DMD projector and not the substrate for efficient large-area patterning. By illuminating all pixels of the DMD with a pulsed LED illuminator, only the light reflected from the bright pixel passes through the aperture of the projection lens, with the desired pattern imaged on the substrate. Because the DMD projector moves continuously via the stage during the exposure step, the DMD pattern must be updated frequently and synchronized with the stage to form a static image on the substrate. In general, as shown in Fig. 2(a), the pattern is scrolled by one row every time the stage moves a distance of one pixel with continuous illumination. Therefore, the resultant exposure pattern on the substrate is stationary. If the pitch, the travel distance of the stage during a frame duration of the DMD (${\tau _0}$), increases without pulsed behavior, the exposed pattern becomes blurred along the stage scanning direction, as shown in Fig. 2(b). On the other hand, as shown in Fig. 2(c), pulsed operation of the DMD can create a clear pattern without blurring even at a large pitch, meaning that high-speed scanning is possible.
Fig. 1. Schematic diagram of the DMD scanning lithography system. A gantry-type stage contains the linear encoder. The encoder signals in the scanning direction are sent to the DMD controller at the desired interval, a multiple of the encoder resolution (5 nm). The DMD controller triggers the DMD frame update and the on/off operation of the pulsed LED illuminator simultaneously for every encoder signal. The pattern data are a series of raster images generated by a pattern generation algorithm devised in this paper.
Fig. 2. Schematic images showing the exposure process in DMD scanning lithography. Frames #1-4 present the DMD patterns used to create a rightward arrow pattern on a substrate. As the stage moves according to the distance of the pitch, the DMD pattern updates to the next frame. (a) A typical exposure process implemented at a one-pixel pitch without pulsed illumination. For a higher scanning speed, the pitch can be increased. The pitch is doubled (b) without or (c) with pulsed illumination. The result of (c) is identical to that of (a) if the duty cycle of illumination is 50%.
In addition, because damage to the DMD is mostly governed by the average power of the incident light when the pulse repetition rate is high enough [23], pulsed illumination does not increase the risk of damaging the DMD compared to continuous illumination. As a result, the combination of pulsed illumination and multiple pixel shifting enables an increase in the scanning speed beyond the typical speed limit due to the DMD frame rate. In fact, except for the mechanical speed limit of the moving stage, the maximum scanning speed of DMD scanning lithography is limited by either the frame rate or the damage threshold of the DMD. In particular, the speed limit established by the damage threshold is high enough in a high-magnification system owing to the high energy density on the substrate. In addition, a high-sensitivity PR and image-amplification system [24] can improve this type of speed limit. Thus, the combination of pulsed illumination and multiple pixel shifting will be practically effective in various DMD scanning lithography systems to enhance the patterning speed.
2.2 Oblique scanning and pixel occupancy
Oblique scanning in DMD scanning lithography refers to the strategy of positioning the DMD image field at an angle relative to the scanning direction, as shown in Fig. 3. This angle serves to enhance the positional accuracy of the pattern by allowing the projected pixel positions to be spread across various positions other than the DMD pixel grid [25]. To analyze the pitch regardless of the pixel size, we define the dimensionless pitch d by
where, as shown in Fig. 3, p is the pitch, W is the width of the projected micromirror, and $\theta $ is the rotational angle between the scanning direction and the projected DMD image.Fig. 3. Schematic image of projected micromirrors of several consecutive frames of a DMD when the oblique scanning is used.
Previous works on oblique scanning determined the two parameters d and $\theta $ based on analytical equations [11,26–28]. Since this equation-based method was designed for the pitch smaller than or adjacent to the pixel width W, we needed another method that can be used for the larger pitch with the pulse illumination method. Therefore, we propose a novel parameter selection strategy in a numerical way based on the distribution of projected pixel positions.
Figure 4 shows the possible pixel positions projected onto the substrate and an exemplary dose distribution having a curve and diagonal line according to the dimensionless pitch d and rotational angle $\theta $. Figure 4 indicates that the patterning result varies depending on the parameters d and $\theta $ owing to the different pixel distributions. Without oblique scanning, the pixilation problem clearly arises, as shown in Fig. 4(a). Oblique scanning with inappropriate parameters produces the distorted pattern shape shown in Fig. 4(b). To maximize the effect of oblique scanning to enhance the positional accuracy, these two parameters must be assigned appropriate values. The parameters used in Fig. 4(c) were derived by the equation-based method in the literature [11]. Figure 4(d) with other optimized parameters also can produce a smooth edge shape of the curves and oblique lines. Here, we introduced a new metric, called the pixel occupancy, to accomplish this optimization by quantifying the positional accuracy of the pattern according to the parameters. Figure 5 shows the procedure used to calculate the new metric. Each center point of a projected pixel is converted to a square of a certain size. Then, the area that the squares occupy within a region of interest (ROI) is calculated, with the ratio of the area becoming the value of the new metric, i.e., the pixel occupancy. Since the distribution of the center points is periodic with respect to W, the ROI can be set as $W \times W$ rectangle regardless of its location. If the calculated pixel occupancy equals one, positional accuracy is guaranteed at the level of the converted square size because there is no gap larger than the size of the square. For example, Fig. 5(a), where the pixel occupancy is less than one, cannot achieve accuracy of one-tenth of the pixel size, whereas Fig. 5(b), where the pixel occupancy is one, can do so.
Fig. 4. Simulation results of the projected pixel positions (left) and accumulated dose distribution (right) according to the parameters d and θ: (a) d = 1 and θ=0; (b) d = 0.947 and θ=0.056 rad; (c) d = 0.968 and θ=0.030 rad; (d) d = 0.966 and θ=0.056 rad. White squares represent the pixel sizes.
Fig. 5. The procedure of calculating the pixel occupancy (left): Each pixel converts to a square of a certain width ($W$/10), and a $W \times W$ rectangle is established as the Region of Interest (ROI); then, the ratio of the total area of squares (white area) to the entire ROI (red box), i.e., the pixel occupancy, is calculated. When (a) d = 2.9 and θ=0.034 rad, because the calculated pixel occupancy (PO) is not 1.0 but 0.437, the patterning accuracy is worse than $W$/10. When (b) d = 2.99 and θ=0.033 rad, as the calculated pixel occupancy is 1.0, the patterning accuracy is guaranteed at the level of $W$/10. Corresponding dose patterns are illustrated on the right.
The pixel occupancy should be computed in the two-dimensional space for d and $\theta $ to select parameters that facilitate the desired level of patterning accuracy. Figure 6 shows examples of a pixel occupancy map computed with squares one tenth of the pixel size. By simply avoiding points with a value lower than one on the map, we can select the parameters d and $\theta $ such that one-tenth pixel-size positional accuracy can be achieved. Unlike the conventional oblique scanning methods, this parameter-selection method can be applied regardless of the range of d and can also predict the patterning accuracy. Figure 6(b) has smaller areas, indicating a pixel occupancy value of 1.0 as compared to Fig. 6(a), indicating that achieving high accuracy is more difficult with a large d of a sparse pixel distribution. When using this method in actual patterning, the pixel occupancy map may be computed near the maximum d determined by the system specs to maximize the scanning speed.
Fig. 6. Two-dimensional pixel occupancy plot, where the range of d is (a) 0.8-1.0 and (b) 2.5-3.0. Each pixel occupancy value is calculated using squares one tenth of the pixel size. Specific locations indicating the parameter set used in Figs. 4, 5 and 11 are marked as arrows and with characters. The pixel occupancy outcomes of (4c), (4d), (5b), (11b), and (11c, d) are all 1.0.
2.3 Edge response analysis
In DMD scanning lithography, the edge response of a pattern, i.e., the edge sharpness, is affected by the edge direction, duty cycle, PR contrast, and dimensionless pitch. The PR contrast $\mathrm{\gamma }$ is defined as shown below.
Here, ${D_0}$ is the minimum exposure dose required for the PR to react to light, and ${D_f}$ is the exposure dose that completely removes the PR (in the case of a positive PR). The simulated dose distribution can be converted to the developed PR height using Eq. (2). When calculating the dose distribution, convolution of an airy disk with a radius of $0.09W$ was performed to consider diffraction similar to the experimental conditions in this study. Figure 7 shows the simulation results of the edge responses of the developed PR patterns. In DMD scanning lithography, the edge responses along the x- and y-axes (parallel and perpendicular to the scanning direction, respectively) are different, as shown in Fig. 7. The edge along the y-axis is steep because there is little motion along the y-axis. However, the edge can be blurred along the x-scan axis because the pattern on the substrate is translated by the pitch distance ($p$) while the DMD pattern is maintained and illuminated by the light source. The amount of blur is affected by the distance traveled while being illuminated. Therefore, the edge response is highly dependent on the duty cycle of the illumination and the pitch of the scanning stage. We simulated the 10-to-90% edge response according to the duty cycle with different PR contrasts and pitches. Figure 7(b) shows the edge response along the y-axis, where the result is only affected by the PR contrast and not the duty cycle or pitch. The conventional DMD scanning lithography results indicated by the black stars in Fig. 7(c) show a small amount of widening of the edge response due to the scan motion during a single frame. Figure 7(d) shows the edge response when the scan pitch is 3 with faster scanning by three times. Without pulse illumination, the edge response becomes almost three times greater compared to the ideal stiff edge. A duty cycle of 33.3% with a pitch of 3 means that the exposure on the substrate is identical to that with conventional DMD scanning lithography with a pitch of 1 and duty cycle of 100%, producing the same small amount of blurring along the scanning direction. As the duty cycle is reduced further, meaning a shorter pulse illumination time, the edge response along the scan axis becomes stiffer. However, this enhancement of the edge response according to the duty cycle becomes saturated, as shown in Figs. 7(c), (d). Because the edge response along the scan axis cannot be stiffer than that in the perpendicular direction, there is no need to reduce the duty cycle to less than a specific value, as indicated by the red stars in Figs. 7(c), (d). A low duty cycle and the resulting high peak power to achieve a certain energy to cure the PR can potentially cause thermal damage to the DMD, although the damage is not critical at a high frame rate [23]. We can select the optimal duty cycle as the value denoted by the red stars in Fig. 7(c), (d) to achieve a stiff edge response.
Fig. 7. Simulation result of the edge response: (a) 3-D pattern image including vertical and horizontal edges. The white arrow represents the scan direction. (b) Edge response along the Y-axis. (c), (d) Edge response along the X-axis (scan axis). The edge response is acquired at the cross-sectional plane in (a) by calculating the thickness from 10% to 90% of the maximum height. The edge response is plotted as a multiple of the pixel width W.
2.4 Pattern generation for implementation
To implement oblique scanning in DMD lithography, datasets of DMD images must be prepared. We referred to the pattern generation algorithm in earlier work [12] to create a dataset. The algorithm generates a series of binary (on/off) raster images using a vector image representing the desired pattern geometry. To determine the on/off status of each pixel, the algorithm must determine whether or not the points inside the pixel are within the pattern boundary; to implement this with low computational complexity, we use the Hormann-Agathos method [29], where the code was modified to operate not only for polygons but also for an arc and circle. Figure 8 shows the process of determining the on/off status, where each pixel is subsampled by the half-pixel width and the subsampled points are classified into interior (red, 1) and exterior (black, 0) points of the pattern boundary. The on/off status is determined by comparing the final calculation value, which is acquired through convolution followed by average pooling, with the threshold value referred to as the occupancy limit, as defined in the literature [12].
Fig. 8. Schematic of the method used to determine on/off status of each DMD pixel. Red and black dots, subsampled by the half-pixel width, are inside and outside of the pattern boundary, respectively. The green rectangle represents a single pixel of the DMD. Pixels of the final calculation result higher than the threshold (e.g., 0.5) are determined to be an ‘on’ pixel and the remaining pixels are ‘off’ pixels. The axes of i and j are identical to those in Fig. 3.
3. Experiments
3.1 Experimental setup
We conducted experiments that involved the fabrication of PR patterns on wafers to demonstrate the advantages of the proposed patterning method, achieving higher patterning speed and quality levels compared to those by the conventional DMD scanning lithography method at an identical DMD frame rate and pixel size. Figure 9 shows a picture of the DMD lithography prototype system with a configuration identical to that in Fig. 1. The experimental system includes a scanning stage, a DMD projector, and a PC. The scanning stage consists of an encoder, a controller, and XYZ linear motors. The encoder generates signals for DMD triggering, while the controller and motors are responsible for the X (scanning), Y (stepping for large-area patterning), and Z (adjusting focus) movements of the projector. The projector consists of a pulsed LED illuminator (3300B-651, Innovations in Optics Inc., USA), a DMD module including a chip (DLP7000, Texas Instruments Inc., USA) and a controller (V-7001, Digital Light innovations Inc., USA), a TIR prism (K9 glass custom product, Henan Kingopt Import & Export Co. Ltd., China), and a bi-telecentric projection lens (S5LPJ2645, Sill Optics GmbH & Co. KG, Germany). The pulsed LED illuminator with a wavelength of 405 nm uniformly illuminates the entire area of the micromirrors (1024${\times} $768 pixels with a 13.68µm pitch) of the DMD. The light reflected by the micromirrors is projected onto the substrate through the projection lens with magnification of 0.7, which enlarges the pixel pitch of 13.68µm to 19.54µm. The TIR prism separates the illumination and projection light paths. The scanning stage (Custom, C&G Microwave Co. Ltd., Republic of Korea) is gantry-type stage; thus, the DMD projector moves for scanning while the substrate remains stationary. The maximum travel speed, repeatability, and encoder resolution of the gantry stage are 50 mm/s, ${\pm} $1um, and 5 nm, respectively.
Fig. 9. Picture of lithography equipment for experiments where the configuration is identical to that in Fig. 1.
All patterning experiments follow the same process, as summarized in Fig. 10. The first step is a hexamethyldisilazane (HMDS) treatment of the silicon wafer, which increases the adhesion between the wafer and PR [30]. The next step is PR (AZ-4533 PR) coating onto the wafer by a spin coating at 2000rpm for 20 seconds. Then, the PR-coated substrate is soft-baked at 100°C for 50 seconds in order to remove the solvent of PR. After the soft-baking step, the exposure process is conducted with DMD scanning. Subsequently, the PR is developed by the AZ300 MIF developer for 80 seconds. Then, a post-baking step is conducted at 115°C for 50 seconds. At the end of these processes, the patterned PR-coated wafer is completed and the images of the result are taken using an optical microscope (BX51M, Olympus Corporation, Japan) and a confocal scanning microscope (NS-3600, Nanoscope System Inc., Republic of Korea).
3.2 Experimental results
The fabricated PR patterns are shown in Fig. 11. The stage scanning speed was 7 mm/s or 21 mm/s in Figs. 11(a), (b) or (c), (d), respectively. The frame rate of the DMD was 360 Hz for all cases. Therefore, the repetition frequency of the pulsed LED was also 360 Hz because it was synchronized by the trigger signal of the DMD driver. The average power of the illumination was 74 mW/cm2 in Figs. 11(a), (b); it was three times higher at 220 mW/cm2 in (c), (d) as the exposure time was one third as long. The pulsed illumination was used in Fig. 11(d) with a duty cycle of 33% and a pulse width of 917µs. In Figs. 11(a)-(c), the continuous wave illumination was used, meaning that the duty cycle was 100%. Oblique scanning produced an extremely smooth edge, as shown in Figs. 11(b), (d) compared to the digitized pattern without oblique scanning shown in Fig. 11(a), verifying the effectiveness of our parameter-selection method for oblique scanning. The sidewall thickness during high-speed patterning with oblique scanning and pulse illumination (Fig. 11(d)) is similar to that of slow-speed patterning (Fig. 11(b)) in spite of the scanning speed being three times faster, whereas the edges without pulse illumination (Fig. 11(c)) are blurred along the scanning direction as each DMD frame is fully illuminated while the stage travels a three-pixel distance. This result demonstrates that patterning at higher speeds during DMD scanning lithography with a restricted DMD frame rate is possible using pulse illumination.
Fig. 11. Images of fabricated PR patterns exposed by DMD scanning lithography taken by 20x and 50x (insets) with an optical microscope. The target pattern shape has concentric circles, each with a thickness of 50 µm. The scanning direction is horizontal. Black squares represent the pixel sizes. (a) Exposed pattern without oblique scanning ($\mathrm{\theta }$=0), where the pitch equals one pixel (d = 1) and illumination is realized by a continuous wave (duty cycle = 100%). The remaining three patterns are exposed using oblique scanning ($\mathrm{\theta }$=7°) under the following conditions: (b) d≈1 and CW illumination (duty cycle = 100%), (c) d≈3 and CW illumination (duty cycle = 100%), and (d) d≈3 and pulsed illumination (duty cycle = 33%)
Figure 12 shows the three-dimensional surface profile around the edge of the fabricated PR pattern taken by a confocal scanning microscope. The edge responses for 10 to 90% of the width were calculated using the cross-section profiles, as shown in Fig. 12(b). We also measured the edge response according to the pulse illumination duty cycle. The edge response improved as the duty cycle was reduced (Fig. 13). However, this improvement became saturated at a duty cycle of around 20%. Because the pixel size is 19.54µm, the minimum edge response of 16µm in Fig. 13 is around 82% of the pixel size. The discrepancy between the experimental results and the simulation results may stem from additional factors such as optical aberrations, motion errors, and thermal/chemical effects of the PR.
Fig. 12. Area around the edge of the fabricated PR pattern, as measured by 100x confocal microscopy: (a) 3-D reconstructed image and (b) corresponding cross-section profile. The exposure condition is equal to that in Fig. 11(d). $\mathrm{\Delta }X$ is the thickness from 10% to 90% of the maximum height.
Fig. 13. Measurement result of the edge response according to the duty cycle. The exposure condition is equal to that in Fig. 11(d) except for duty cycle.
4. Conclusion
In this study, we simultaneously improved the pattern quality and patterning speed of DMD maskless lithography through pulse exposure and oblique scanning. We proposed a novel method that determines the oblique angle regardless of the pitch using a new metric called the pixel occupancy so that oblique scanning can accompany pulse exposure for high-speed patterning, unlike the conventional equation-based oblique scanning method that cannot determine the oblique angle with a multiple-pixel pitch.
Through the edge response analysis in this study, we noted the characteristics of oblique scanning: the edge response of the dose distribution cannot improve beyond certain limits even when the duty cycle of the pulse exposure is reduced further. Therefore, we could optimally determine the duty cycle for the thinnest edge thickness. Theoretically, to improve the edge response further, a high-contrast PR must be used.
The method involving the use of a microlens arrays in DMD lithography [22] can improve the minimum feature size and positional accuracy while maintaining the patterning speed, but complicated fabrication and alignment of components are required. However, our method can have the same effect by simply reducing the projected pixel size using a high-magnification lens. In the high-magnification system with a reduced pixel size and exposure field, the scanning speed is further limited by the frame rate of the DMD rather than the damage threshold. Thus, the speed improvement by our method is more powerful in a high-magnification system. When using high-magnification lenses, techniques that take into account diffraction effects, such as optical proximity correction [31] and inverse lithography technology [32] can be employed because our method can create high-quality, arbitrary pattern shapes.
We anticipate that our method can be used for various microfabrication tasks, including manufacturing PCBs, MEMS devices, and micro-optics, among others. Specifically, because our method is capable of generating extremely smooth pattern edges, the manufacturing of PCB and inverse lithography technology masks, which include many diagonal lines and curved shapes, is possible when using it. In addition, high positional accuracy requirements compared to the relatively large feature sizes of these applications are also feasible when using the method proposed here.
Funding
Basic Research Fund of the Korea Institute of Machinery and Materials (NK236E); National Research Foundation of Korea (NRF-2020R1A2C3006745).
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. C. Mack, Fundamental Principles of Optical Lithography: The Science of Microfabrication, 1st ed. (John Wiley & Sons, 2008).
2. B. W. Smith and K. Suzuki, Microlithography: Science and Technology, 3rd ed. (CRC Press, 2020).
3. J. Lee, H. Lee, and J. Yang, “A rasterization method for generating exposure pattern images with optical maskless lithography,” J. Mech. Sci. Technol. 32(5), 2209–2218 (2018). [CrossRef]
4. H. Ulrich, R. Wijnaendts-van-Resandt, C. Rensch, and W. Ehrensperger, “Direct writing laser lithography for production of microstructures,” Microelectron. Eng. 6(1-4), 77–84 (1987). [CrossRef]
5. C. Rensch, S. Hell, M. V. Schickfus, and S. Hunklinger, “Laser scanner for direct writing lithography,” Appl. Opt. 28(17), 3754–3758 (1989). [CrossRef]
6. Y. Kito, M. Hori, Y. Hayashida, T. Suzuki, H. Komiyama, T. Watanabe, T. Kurashige, M. Kato, and K. Nara, “Direct imaging exposure equipment with high overlay accuracy for flexible substrate in roll-to-roll method,” J. Soc. Inf. Disp. 25(7), 411–420 (2017). [CrossRef]
7. R. Menon, A. Patel, E. E. Moon, and H. I. Smith, “Alpha-prototype system for zone-plate-array lithography,” J. Vac. Sci. Technol. B 22, 3032–3037 (2004). [CrossRef]
8. J. I. Trisnadi, C. B. Carlisle, and R. Monteverde, “Overview and applications of grating-light-valve-based optical write engines for high-speed digital imaging,” Proc. SPIE 5348, 52–64 (2004). [CrossRef]
9. W. C. Sweatt and R. H. Stulen, “Maskless lithography,” U.S. patent 5,870,176 (9 February 1999).
10. K. Jain, T. J. Dunn, and J. M. Hoffman, “Seamless, maskless lithography system using spatial light modulator,” U.S. patent 6,312,134B1 (6 November 2001).
11. W. Mei, “Point array maskless lithography,” U.S. patent 6,473,237B2 (29 October 2002).
12. M. Seo and H. Kim, “Lithography upon micromirrors,” Comput.-Aided Des. 39(3), 202–217 (2007). [CrossRef]
13. K.-R. Kim, J. Yi, S.-H. Cho, N.-H. Kang, M.-W. Cho, B.-S. Shin, and B. Choi, “SLM-based maskless lithography for TFT-LCD,” Appl. Surf. Sci. 255(18), 7835–7840 (2009). [CrossRef]
14. Y. Watanabe, H. Kono, Y. Kanaya, Y. Saito, T. Sakamoto, S. Owa, N. Hirayanagi, T. Koo, C. Poppe, D. Tseng, C. Sorensen, H. Lee, S. Renwick, and B. Yuan, “Digital scanner, optical maskless exposure tool with DUV solid state laser,” Proc. SPIE 11613, 1161309 (2021). [CrossRef]
15. E. J. Hansotte, E. C. Carignan, and W. D. Meisburger, “High speed maskless lithography of printed circuit boards using digital micromirrors,” Proc. SPIE 7932, 793207 (2011). [CrossRef]
16. M. Kang, C. Han, and H. Jeon, “Submicrometer-scale pattern generation via maskless digital photolithography,” Optica 7(12), 1788–1795 (2020). [CrossRef]
17. Z. Zhang, Q. Meng, and N. Luo, “A DMD based UV lithography method with improved dynamical modulation range for the fabrication of curved microstructures,” AIP Adv. 11(4), 045008 (2021). [CrossRef]
18. R. Chen, H. Liu, H. Zhang, W. Zhang, J. Xu, W. Xu, and J. Li, “Edge smoothness enhancement in DMD scanning lithography system based on a wobulation technique,” Opt. Express 25(18), 21958–21968 (2017). [CrossRef]
19. K. Kim, S. Han, J. Yoon, S. Kwon, H.-K. Park, and W. Park, “Lithographic resolution enhancement of a maskless lithography system based on a wobulation technique for flow lithography,” Appl. Phys. Lett. 109(23), 234101 (2016). [CrossRef]
20. S. Guo, Z. Lu, Z. Xiong, L. Huang, H. Liu, and J. Li, “Lithographic pattern quality enhancement of DMD lithography with spatiotemporal modulated technology,” Opt. Lett. 46(6), 1377–1380 (2021). [CrossRef]
21. Y. Zhang, J. Luo, Z. Xiong, H. Liu, L. Wang, Y. Gu, Z. Lu, J. Li, and J. Huang, “User-defined microstructures array fabricated by DMD based multistep lithography with dose modulation,” Opt. Express 27(22), 31956–31966 (2019). [CrossRef]
22. D.-H. Dinh, H.-L. Chien, and Y.-C. Lee, “Maskless lithography based on digital micromirror device (DMD) and double sided microlens and spatial filter array,” Opt. Laser Technol. 113, 407–415 (2019). [CrossRef]
23. Texas Instruments, “Laser power handling for DMDs,” White Paper DLPA027 (January 2012).
24. F. N. Ghauri, R. Jones, R. Feeler, and M. E. Kushina, “Replacing scanning based laser processing applications with variable mask image amplification systems using TI DMD, pulsed Nd: YAG lasers, and amplifiers,” Proc. SPIE 9376, 93760J (2015). [CrossRef]
25. M. Seo and H. Kim, “Influence of dynamic sub-pixelation on exposure intensity distribution under diffraction effects in spatial light modulation based lithography,” Microelectron. Eng. 98, 125–129 (2012). [CrossRef]
26. S. Huang, M. Li, L. Wang, Y. Su, and Y. Liang, “Precise fabrication of large-area microstructures by digital oblique scanning lithography strategy and stage self-calibration technique,” Appl. Phys. Express 12(9), 096501 (2019). [CrossRef]
27. J.-G. Hur, “Maskless fabrication of three-dimensional microstructures with high isotropic resolution: practical and theoretical considerations,” Appl. Opt. 50(16), 2383–2390 (2011). [CrossRef]
28. K. F. Chan, Z. Feng, R. Yang, A. Ishikawa, and W. Mei, “High-resolution maskless lithography,” J. Micro/Nanolithogr., MEMS, MOEMS 2(4), 331–339 (2003). [CrossRef]
29. K. Hormann and A. Agathos, “The point in polygon problem for arbitrary polygons,” Comput. Geom. 20(3), 131–144 (2001). [CrossRef]
30. M. Hair and W. Hertl, “Reaction of hexamethyldisilazane with silica,” J. Phys. Chem. 75(14), 2181–2185 (1971). [CrossRef]
31. O. W. Otto, J. G. Garofalo, K. Low, C.-M. Yuan, R. C. Henderson, C. Pierrat, R. L. Kostelak, S. Vaidya, and P. Vasudev, “Automated optical proximity correction: a rules-based approach,” Proc. SPIE 2197, 278–293 (1994). [CrossRef]
32. L. Pang, “Inverse lithography technology: 30 years from concept to practical, full-chip reality,” J. Micro/Nanopattern. Mats. Metro. 20, 030901 (2021). [CrossRef]
