On-Machine Measurement for Surface Flatness of Transparent and Thin Film in Laser Ablation Process

In printed electronics, laser ablation is used to repair defective patterns on transparent, flexible, and thin films, using high-power lasers. The distance between the film surface and laser focus is sensitive to changes as the narrow focus depth of the lens is the range of tens of microns. However, a film fixed on a conductive vacuum chuck (CVC) is always curved, owing to chucking bending; thus, laser focusing must be locally performed before ablation. Therefore, this study proposes a non-contact measurement method for the surface flatness of a transparent and thin film, to compensate for laser defocusing in a large area. The surface flatness was obtained using camera-focus points on the porous surface of the CVC. The focus points were interpolated to achieve a smooth and continuous surface flatness for chucking bending. A laser distance sensor was used to verify the surface flatness from the proposed method. The surface flatness was used to inspect the printed patterns, and to perform laser ablation on the film. The proposed method is advantageous for large-area laser ablation and is expected to become indispensable for repairing machines in printed electronics.


Introduction
The conventional concept of printed electronics (PEs) comprises the fabrication of circuits and patterns on a thin substrate by injecting conductive ink [1]. PEs have been applied in various devices such as solar cells, batteries, and organic light emitting diodes (OLEDs) [2,3]. Roll-to-roll (R2R) is a large-scale PE manufacturing process that continuously prints patterns on a thin film. R2R is an additive and large-area process; thus, R2R is low-waste, and more efficient than conventional processes for electronic devices [4,5]. In recent, 3D PE presents great potentials to build complex and multi-functional structures for stretchable electronics as well as conventional electronics [6].
Some faults and defects on a printed device can be repaired using laser ablation and electrohydrodymanics (EHD) [7,8]. EHD is an additive process whereas laser ablation is a subtractive process. These repair processes can regenerate a considerable portion of the defective printed devices [9]. Laser ablation comprises monitoring a targeting area using vision-microscopy, and blasting defective areas using a high-power laser. Thus, laser ablation for PEs usually comprises a process of removing defective patterns from a target area. Laser ablation (for a printed device) is applied to transparent and thin films, such as polyethylene terephthalate (PET) foils coated with indium-tin oxide for OLED fabrication [10]. Laser ablation can also remove selected areas of the transparent and thin films [11]. The physical conditions of laser ablation, such as the beam focus, exposure time, and intensity, are quite delicate, and high accuracy is required for the optics and mechanics. aimed for removal of the defective pattern on the film. While moving the ablation head using the stage, a high-power laser is emitted to the film surface. The ablation state can be investigated using the industrial camera. The coaxial optics in the ablation head share an optical axis of the high-power laser and industrial camera through a half-mirror, as shown in Figure 1. The focus depth of the lens for laser ablation is approximately 20-30 µm; thus, the relative distance between the film and the ablation head should be maintained.
Coatings 2020, 10, x FOR PEER REVIEW  3 of 15 stage, a high-power laser is emitted to the film surface. The ablation state can be investigated using the industrial camera. The coaxial optics in the ablation head share an optical axis of the high-power laser and industrial camera through a half-mirror, as shown in Figure 1. The focus depth of the lens for laser ablation is approximately 20-30 μm; thus, the relative distance between the film and the ablation head should be maintained. However, in this process, the film is always bent according to the chuck surface after chucking, owing to its flexibility and thinness [20,21]. The bending causes variations in the relative distance and defocusing of the laser ablation. The variation is insignificant in local processing because of the smoothness and continuity of the film surface, as shown in Figure 1. Nevertheless, it becomes crucial in large-area processing; thus, the variations and corresponding surface flatness values should be compensated at the ablation positions. The CVC is a porous metal plate made by sintering aluminum foam [22,23]. A vacuum supplied from a pump passes through micro-holes in the metal plate and adheres to the film. As shown in Figure 2, the surface of the metal plate is rough and uneven at the micro-scale owing to its porosity, but a contour of the surface flatness can be achieved at the macroscale. The target surface is a smooth face and occupies a larger area than that of the porosity. When measuring surface flatness, a laser spot can be placed on the porosity; thus, an area method, such as autofocus (AF) using a camera, is advantageous.  However, in this process, the film is always bent according to the chuck surface after chucking, owing to its flexibility and thinness [20,21]. The bending causes variations in the relative distance and defocusing of the laser ablation. The variation is insignificant in local processing because of the smoothness and continuity of the film surface, as shown in Figure 1. Nevertheless, it becomes crucial in large-area processing; thus, the variations and corresponding surface flatness values should be compensated at the ablation positions. The CVC is a porous metal plate made by sintering aluminum foam [22,23]. A vacuum supplied from a pump passes through micro-holes in the metal plate and adheres to the film. As shown in Figure 2, the surface of the metal plate is rough and uneven at the micro-scale owing to its porosity, but a contour of the surface flatness can be achieved at the macro-scale. The target surface is a smooth face and occupies a larger area than that of the porosity. When measuring surface flatness, a laser spot can be placed on the porosity; thus, an area method, such as autofocus (AF) using a camera, is advantageous.
Coatings 2020, 10, x FOR PEER REVIEW 3 of 15 stage, a high-power laser is emitted to the film surface. The ablation state can be investigated using the industrial camera. The coaxial optics in the ablation head share an optical axis of the high-power laser and industrial camera through a half-mirror, as shown in Figure 1. The focus depth of the lens for laser ablation is approximately 20-30 μm; thus, the relative distance between the film and the ablation head should be maintained. However, in this process, the film is always bent according to the chuck surface after chucking, owing to its flexibility and thinness [20,21]. The bending causes variations in the relative distance and defocusing of the laser ablation. The variation is insignificant in local processing because of the smoothness and continuity of the film surface, as shown in Figure 1. Nevertheless, it becomes crucial in large-area processing; thus, the variations and corresponding surface flatness values should be compensated at the ablation positions. The CVC is a porous metal plate made by sintering aluminum foam [22,23]. A vacuum supplied from a pump passes through micro-holes in the metal plate and adheres to the film. As shown in Figure 2, the surface of the metal plate is rough and uneven at the micro-scale owing to its porosity, but a contour of the surface flatness can be achieved at the macroscale. The target surface is a smooth face and occupies a larger area than that of the porosity. When measuring surface flatness, a laser spot can be placed on the porosity; thus, an area method, such as autofocus (AF) using a camera, is advantageous.   The surface flatness of the film forms a smooth curve as compared with that of the CVC, owing to surface conditions of the CVC and bending radius of the film. As shown in Figure 2, surface of the CVC in micro scale is coarse owing to scars from machining and porosity from sintering. However, a film for PE is hard enough to interpolate the surface conditions and follows general surface profiles of the CVC in macro scale. The radius of curvature on these micro-surface conditions is much smaller than that of the general surface profile. Moreover, a film is flattened before being fixed on the CVC because the radius of curvature affects to the characteristics of printed devices [24]. PET film is one of the popular substrates in PE and the radius of the curvature is not less than 4 mm for 100 µm thickness [25]. The radius of curvature of the general surface profile is normally much larger than the minimum, therefore, the surface flatness can be obtained by interpolating the AF position on the CVC.

Surface Flatness of Conductive Vacuum Chuck
In PEs, non-contact and on-machine measurements are preferred; thus, an optical distance sensor (ODS) is conventionally used to measure the surface flatness in large areas [26]. It is desirable to measure the surface flatness from the upper face of the film, but this is generally not possible owing to optical discontinuities from printed patterns and film transparency. Commercial ODSs can measure these transparent films, but the refractive index of the printed patterns is usually unknown [27]. Thus, after measuring the surface flatness of the CVC, that of the film surface can be obtained based on interpolation.
The ideal position for the laser ablation, z r , must be corrected according to the surface flatness of the CVC, i.e., ∆z(x,y). Figure 2 shows the relationships among the ablation head, ODS and CVC. The ODS is attached near the ablation head. The distance between the coaxial optics and ODS, ∆z d , forms kinematic offsets (x o , y o , z o ). Thus, the coordinates of the ODS, (x , y , z ), are calculated as follows: Measuring a surface flatness by scanning the ODS is fast and convenient. However, the ODS can cause measurement errors owing to the porosity of the CVC. The narrow spot of the ODS provides the distance from a local area; thus, a measurement error can occur owing to the porosity, as shown in Figure 2. The degree of focus of an image represents a larger area; thus, the focus position is more reliable for determining the surface flatness. The focus position is found using the AF, which scans the variations of contrast along an optical axis. The focus positions on a plane of the CVC are applicable for determining the surface flatness [28].
The AF requires a long scanning time, as the focus depth of the coaxial optics is extremely narrow. For instance, the scanning interval for the AF should be less than the focus depth; it was 10 µm in this study. Therefore, the AF should be joined with the ODS, i.e., the AF should be performed based on the ODS results. The AF determines a focal position at the optimum of contrast, σ, according to movement along the optical axis near z d [29]. The range of the movement near z d , δ, is a small value. The focus position of the AF, z a , can be determined at a local maximum of the DFIs as follows: The contrast for the AF can be evaluated using the DFIs. The DFIs are calculated from high-frequency components and pixel-based operations, as shown in Appendix A [30]. The surface flatness using the AF can be defined according to the difference between the maximum and minimum values of z a .

Interplation of Surface Flatness of Thin Film
Commercial coordinate measuring machines usually obtain surface flatness values using contact methods and sampling measurement data [31]. Owing to the AF scanning time, the number of AF points in the planar directions should also be limited. The curvature of the film in the fixed state forms smoother curve than that of the CVC. Thus, the surface flatness of the film can be estimated based on a 2D interpolation of the sampling points. Bilinear interpolation is a popular interpolation approach, and normalized positions for the interpolation can be defined as follows [32]: Then, the interpolated position in the planar direction by the AF can be written using the following equation [33].
Surface flatness can be also interpolated using the above concept. (7) Then, the interpolated position for the laser ablation on the film surface, z f , can be calculated, while considering the film thickness and air buffer from the CVC. The surface flatness is used for ablation and inspection just by adding constant values in practice.

Laser Ablation System
The laser ablation system is composed of a linear stage, ablation head, high-power laser source, light source, CVC, controller and ODS. The linear stage transfers the ablation head above the CVC based on Cartesian coordinates. In this study, the xy axes comprised the gantry structure, and were driven using high-speed linear motors. The z-axis was constructed using a high-accuracy ball screw and was installed on the gantry. The ablation head was attached on the z-axis, and the ODS was equipped near the ablation head, as shown in Figure 3a.
The light source supplied variable color light to the ablation sample, so as to monitor the micro-patterns using the industrial camera. For optimization of the light color and intensity, the original light colors of RGBW LEDs were used. The initial intensity of a selected color for the AF was set to 30% of the maximum light power. Then, the focus was scanned on the CVC by moving the focal axis and was determined at the maximum of the DFIs. The light intensity was also determined at the maximum of the DFIs by adjusting the light power in the focus position. The optimal light color and intensity values for the AF were also achieved using the maximum values of the DFIs. After the focus of the industrial camera was determined under RGBW colors, the light color and intensity were also determined for the AF at the maximum of the DFIs.  The ablation sample was produced through the R2R process, i.e., printing micro-patterns on a transparent and thin film. The patterns (shown in Figure 3b) consisted of micro-lines ranging from 10 µm to 50 µm, and were printed using silver nanoparticle ink. The film was polyethylene terephthalate (PET) with a thickness of 100 µm. The ablation sample was fastened to the CVC using a vacuum. The optimal color light reflected on the ablation sample passed the object lens and half-mirror. The industrial camera acquired images for inspecting the patterns and aligning the ablation direction. Then, a laser beam from the high-power laser source was injected into the side port of the ablation head, reflected on the half-mirror into a right angle, and focused on the ablation sample through an object lens. The laser beam blasted a defective pattern on the film, and the laser ablation was observed using the industrial camera.
The operating software for laser ablation was constructed using Visual Studio and OpenCV. The software maneuvered the stage, industrial camera, color light source, high-power laser, and ODS. The source code for the DFIs was implemented into a dynamic-link library based on OpenCV. The procedure of experiment is shown in Figure 4, and the specifications of the ablation head are summarized in Table 1. The light source supplied variable color light to the ablation sample, so as to monitor the micropatterns using the industrial camera. For optimization of the light color and intensity, the original light colors of RGBW LEDs were used. The initial intensity of a selected color for the AF was set to 30% of the maximum light power. Then, the focus was scanned on the CVC by moving the focal axis and was determined at the maximum of the DFIs. The light intensity was also determined at the maximum of the DFIs by adjusting the light power in the focus position. The optimal light color and intensity values for the AF were also achieved using the maximum values of the DFIs. After the focus of the industrial camera was determined under RGBW colors, the light color and intensity were also determined for the AF at the maximum of the DFIs.
The ablation sample was produced through the R2R process, i.e., printing micro-patterns on a transparent and thin film. The patterns (shown in Figure 3b) consisted of micro-lines ranging from 10 μm to 50 μm, and were printed using silver nanoparticle ink. The film was polyethylene terephthalate (PET) with a thickness of 100 μm. The ablation sample was fastened to the CVC using a vacuum. The optimal color light reflected on the ablation sample passed the object lens and halfmirror. The industrial camera acquired images for inspecting the patterns and aligning the ablation direction. Then, a laser beam from the high-power laser source was injected into the side port of the ablation head, reflected on the half-mirror into a right angle, and focused on the ablation sample through an object lens. The laser beam blasted a defective pattern on the film, and the laser ablation was observed using the industrial camera.
The operating software for laser ablation was constructed using Visual Studio and OpenCV. The software maneuvered the stage, industrial camera, color light source, high-power laser, and ODS. The source code for the DFIs was implemented into a dynamic-link library based on OpenCV. The procedure of experiment is shown in Figure 4, and the specifications of the ablation head are summarized in Table 1.

Interpolation and Surface Flatness
The AF points in the planar directions, (x i , y j ), were determined by dividing the CVC surface into a rectangular grid. The values of z d (x i + x 0 , y j + y 0 ) were measured using the ODS, and then, AF was performed to correct z d under the optimum light color and intensity values. z a (x i , y j ) was obtained from the corrected value, and a contour map of the surface flatness was constructed by bilinear interpolation. Signals of thickness measurement mode in the ODS were used to verify the interpolation. After entire area of the CVC was covered with a clean film, a vacuum was applied to fix the film. Then, the ODS was tilted and sensor parameters were varied to switch into the thickness measurement mode. Signals of the reflection from the upper surface were used to obtain surface flatness.
After the ablation sample was mounted on the CVC, and the patterns on the sample were inspected at the interpolated position, z f (x i + α, y j + β), to verify the proposed method. The distinctness of the images acquired at the interpolated position was examined to determine the usability for laser ablation. Laser ablation was performed to remove line patterns on the ablation sample.

Results and Discussion
The optimum light conditions for the AF were determined at the maximum of the DFIs. Figure 5 shows that the curves of the DFIs usually form a single peak during focusing and lighting. Many DFIs showed maxima at the same focal position, although some DFIs presented different foci or had diverged. After determining the focus position, the light intensity was determined at the maximum values of the DFIs by increasing the light power. These focusing and lighting procedures were repeated for the other original colors. Figure 5a shows the variations of the DFIs according to the intensity of blue light. A threshold pixel became saturated, and a derivative-based correlation showed different focus positions; thus, these DFIs were excluded from determining the light conditions. The energy Laplacian and Tenenbaum gradient showed larger variations than that for the other DFIs. However, the energy Laplacian showed noise patterns and coarse curvature in the graph; the Tenenbaum gradient was finally selected for the AF. Figure 5b shows the responses of the Tenenbaum gradient according to the original color lights. From comparing the color responses of the Tenenbaum gradient, the highest maximum was formed when using the red color; thus, red light was selected for the AF. In addition, the color of the high-power laser was green (532 nm), i.e., a complementary color of the AF color; thus, the determined light color was advantageous for monitoring the laser ablation.
the Tenenbaum gradient, the highest maximum was formed when using the red color; thus, red light was selected for the AF. In addition, the color of the high-power laser was green (532 nm), i.e., a complementary color of the AF color; thus, the determined light color was advantageous for monitoring the laser ablation. The variation in the grid density of the AF points and surface flatness values when using the ODS is shown as contour maps in Figure 6. A longer distance implies a farther distance from the CVC surface, and is lower on the CVS surface. Thus, the color in the contour maps changes from blue to red according to the −Δzd value. The red area indicates a local high surface with a long distance from the ODS, and the blue area indicates the inversion. The values of the indicators on the right of the contour maps show the minimum and maximum values of −Δzd. The contour maps show that the top-left of the CVC has lower values and the bottom-left has higher values. Additionally, the contour maps show that the top-right has slightly higher values and the bottom-right represents the average. The overall tendency of the contour maps in Figure 6 is similar to the variation of the grid density. However, the contour maps became coarse as the grid density increased. The contour map at a low grid density was smooth, but discontinuities such as dots and speckles increased at high grid density. The blue dots in the 50 × 50 map were caused by local cracks and porosity on the CVC. This indicates that a large number of sampling positions does not guarantee high-accuracy of surface flatness values. As shown in Figure 2, the concave curvature of the porosity is very small whereas the curvature of the CVC is very large. Thus, the blue dots should be excluded when approximating the surface flatness and positions of the ODS. The curvature of a thin film fixed on the CVC will have a smooth surface; thus, the effect of these local flaws on the surface flatness of the film decreases. In the The variation in the grid density of the AF points and surface flatness values when using the ODS is shown as contour maps in Figure 6. A longer distance implies a farther distance from the CVC surface, and is lower on the CVS surface. Thus, the color in the contour maps changes from blue to red according to the −∆z d value. The red area indicates a local high surface with a long distance from the ODS, and the blue area indicates the inversion. The values of the indicators on the right of the contour maps show the minimum and maximum values of −∆z d . The contour maps show that the top-left of the CVC has lower values and the bottom-left has higher values. Additionally, the contour maps show that the top-right has slightly higher values and the bottom-right represents the average. The overall tendency of the contour maps in Figure 6 is similar to the variation of the grid density. However, the contour maps became coarse as the grid density increased. The contour map at a low grid density was smooth, but discontinuities such as dots and speckles increased at high grid density. The blue dots in the 50 × 50 map were caused by local cracks and porosity on the CVC. This indicates that a large number of sampling positions does not guarantee high-accuracy of surface flatness values. As shown in Figure 2, the concave curvature of the porosity is very small whereas the curvature of the CVC is very large. Thus, the blue dots should be excluded when approximating the surface flatness and positions of the ODS. The curvature of a thin film fixed on the CVC will have a smooth surface; thus, the effect of these local flaws on the surface flatness of the film decreases. In the contour maps, an 8 × 8 grid density was considered optimal when considering the smoothness, local variations, and focusing time.
Coatings 2020, 10, x FOR PEER REVIEW 9 of 15 contour maps, an 8 × 8 grid density was considered optimal when considering the smoothness, local variations, and focusing time.  Figure 7a shows contour maps of the surface flatness after the AF and −Δza according to the grid density. The overall tendencies of the contour maps were similar as shown in Figure 6, but the maps varied locally as the grid density increased. The difference between the contour maps from the ODS and AF was small at a lower grid density, but the difference became distinct at higher grid densities. Figure 7b shows a contour map of the surface flatness using the thickness measurement mode on a clean film. Overall shapes of Figure 7a,b are similar and some local discontinuities were removed compared with Figure 6. The error between the reflection method and our interpolation was ±3.0 μm but the error between the reflection method and the CVC was ±4.5 μm.  Figure 7a shows contour maps of the surface flatness after the AF and −∆z a according to the grid density. The overall tendencies of the contour maps were similar as shown in Figure 6, but the maps varied locally as the grid density increased. The difference between the contour maps from the ODS and AF was small at a lower grid density, but the difference became distinct at higher grid densities. Figure 7b shows a contour map of the surface flatness using the thickness measurement mode on a clean film. Overall shapes of Figure 7a,b are similar and some local discontinuities were removed compared with Figure 6. The error between the reflection method and our interpolation was ±3.0 µm but the error between the reflection method and the CVC was ±4.5 µm.
Coatings 2020, 10, x FOR PEER REVIEW 9 of 15 contour maps, an 8 × 8 grid density was considered optimal when considering the smoothness, local variations, and focusing time.  Figure 7a shows contour maps of the surface flatness after the AF and −Δza according to the grid density. The overall tendencies of the contour maps were similar as shown in Figure 6, but the maps varied locally as the grid density increased. The difference between the contour maps from the ODS and AF was small at a lower grid density, but the difference became distinct at higher grid densities. Figure 7b shows a contour map of the surface flatness using the thickness measurement mode on a clean film. Overall shapes of Figure 7a,b are similar and some local discontinuities were removed compared with Figure 6. The error between the reflection method and our interpolation was ±3.0 μm but the error between the reflection method and the CVC was ±4.5 μm. After mounting a transparent and thin film fabricated using the R2R, images were acquired at the interpolated position for the laser ablation, z f . The (x, y) points of the interpolated positions were achieved by slicing the contour maps of the 8 × 8 grid (shown in Figure 6) into a 24 × 24 grid. Then, the focus quality was investigated at the interpolated position. In the 24 × 24 grid of the CVC, 18 × 20 points were placed on the ablation sample. Figure 8 shows the sample images acquired at the interpolated positions. Three images showed focused patterns; another (at the bottom-right) was slightly defocused.
Coatings 2020, 10, x FOR PEER REVIEW 10 of 15 After mounting a transparent and thin film fabricated using the R2R, images were acquired at the interpolated position for the laser ablation, zf. The (x, y) points of the interpolated positions were achieved by slicing the contour maps of the 8 × 8 grid (shown in Figure 6) into a 24 × 24 grid. Then, the focus quality was investigated at the interpolated position. In the 24 × 24 grid of the CVC, 18 × 20 points were placed on the ablation sample. Figure 8 shows the sample images acquired at the interpolated positions. Three images showed focused patterns; another (at the bottom-right) was slightly defocused. The distribution of the focus quality is mapped in Figure 9. O denotes focused, Δ denotes slightly defocused but ablatable, and X denotes completely defocused. The corresponding counts of points were 322, 10, and 28. Thus, 92.2% of the points in the ablation sample were available for laser ablation. However, 96.4% of the defocused points were placed at the edges of the ablation sample, and one point was located inside. This indicates that the chucking force using the vacuum is weak at the edges owing to leak; hence, the edge area floats relative to the inside and margin at the edge area is required when designing PE patterns. The defocused point inside was caused by large scratches on the CVC, which formed wrinkles in the film. These facts imply that scratches and bruises must be removed constructing the CVC for laser ablation of PEs. The distribution of the focus quality is mapped in Figure 9. O denotes focused, ∆ denotes slightly defocused but ablatable, and X denotes completely defocused. The corresponding counts of points were 322, 10, and 28. Thus, 92.2% of the points in the ablation sample were available for laser ablation. However, 96.4% of the defocused points were placed at the edges of the ablation sample, and one point was located inside. This indicates that the chucking force using the vacuum is weak at the edges owing to leak; hence, the edge area floats relative to the inside and margin at the edge area is required when designing PE patterns. The defocused point inside was caused by large scratches on the CVC, which formed wrinkles in the film. These facts imply that scratches and bruises must be removed constructing the CVC for laser ablation of PEs.  The right image shows laser ablation cleared line-patterns on horizontal direction using 7 μm spot diameter. Figure 10 shows that the micro lines printed on PET was removed after laser ablation. The results show that these surface flatness values should be applied to calibrate laser ablation systems for large-area processes and that their compensation should be an essential function of manufacturing systems for PEs. Laser ablation is useful for recovering faulty devices; thus, it is advantageous to increase the production yield. Measurement time of a laser interferometer is shorter   Figure 10 shows the laser ablation on the line patterns on the compensated position using alignment and surface flatness. Automatic alignment found positional deviation on XYθ directions and the deviation was applied to XY ablation position. Positions for laser ablation calculated using end edge of a pattern group, line pitch and line length. Z ablation positions were compensated using surface flatness according to XY machine coordinates. The left image shows that the ablation laser was focused with a 5 µm spot diameter and removed 10 µm-width line patterns on vertical direction. The right image shows laser ablation cleared line-patterns on horizontal direction using 7 µm spot diameter. Figure 10 shows that the micro lines printed on PET was removed after laser ablation.  Figure 10 shows the laser ablation on the line patterns on the compensated position using alignment and surface flatness. Automatic alignment found positional deviation on XYθ directions and the deviation was applied to XY ablation position. Positions for laser ablation calculated using end edge of a pattern group, line pitch and line length. Z ablation positions were compensated using surface flatness according to XY machine coordinates. The left image shows that the ablation laser was focused with a 5 μm spot diameter and removed 10 μm-width line patterns on vertical direction. The right image shows laser ablation cleared line-patterns on horizontal direction using 7 μm spot diameter. Figure 10 shows that the micro lines printed on PET was removed after laser ablation. The results show that these surface flatness values should be applied to calibrate laser ablation systems for large-area processes and that their compensation should be an essential function of manufacturing systems for PEs. Laser ablation is useful for recovering faulty devices; thus, it is advantageous to increase the production yield. Measurement time of a laser interferometer is shorter The results show that these surface flatness values should be applied to calibrate laser ablation systems for large-area processes and that their compensation should be an essential function of manufacturing systems for PEs. Laser ablation is useful for recovering faulty devices; thus, it is advantageous to increase the production yield. Measurement time of a laser interferometer is shorter and measurement accuracy is higher than those of the proposed method. However, the laser interferometer requires installation time and coordinate transformation which cause kinematic errors. The proposed measurement uses on-machine devices under a single coordinate, thus additional installation and coordinate transform are not required. Automatic measurement process provides operational convenience and efficient working time.
On the other hand, the ablation head was used for simple monitoring in this study, but various functions can be achieved using artificial intelligence (AI) in the future. AI is currently introduced in recent researches of laser ablation [34] and the laser ablation system in this study is advantageous for data mining. The ablation head can collect image-based data which can be linked with the mechanical data of high accuracy. These high-quality data will be appropriate for training AI functions, such as kinematic alignment, light optimization, determination of ablation quality as well as acquisition of surface flatness.

Conclusions
In this study, a method for measuring the surface flatness of a transparent and thin film was proposed, based on the AF for the laser ablation of PEs. The defocusing of laser ablation owing to the chucking bending of the flexible and thin film can be reduced by the surface flatness. The surface flatness of the film was estimated from that of the CVC based on a non-contact and optical method considering the transparency, thinness, and printed patterns of the film. The surface flatness of the CVC was measured using an optical distance sensor according to the variable grid density. The proper grid density of the measurement points was determined by considering the porosity of the CVC and coarseness of the contour maps. AF was performed to correct the measured position by the ODS, as the errors caused by the ODS were caused by localities. An interpolated position was obtained from the surface flatness as corrected by the AF. The results showed that 92.2% of the test points were usable for laser ablation and most of defocus was caused by the vacuum leak at the edge. The proposed method is available in the normal areas of laser ablation and makes it possible to provide large-area ablation; thus, it should be an indispensable function in a laser ablation system for PEs. The automatic on-machine measurement provides operational convenience and efficient working time without additional device installation. The proposed method also contributes to the repair of defective devices, thus improving the production yield of PEs. Funding: This study has been conducted with the support of the Korea Institute of Industrial Technology as "Development of root technology for multi-product flexible production (KITECH EO-20-0015).

Conflicts of Interest:
The authors declare no conflict of interest.