Laser micro-machining of high density optical structures on large substrates

Transcription

1 Laser micro-machining of high density optical structures on large substrates Karl L. Boehlen*, Ines B. Stassen Boehlen Exitech Ltd, Oxford Industrial Park, Yarnton, Oxford, OX5 1QU, United Kingdom ABSTRACT A new laser mask projection technique, Synchronised Image Scanning (SIS), has been developed for the efficient fabrication of dense arrays of repeating microstructures on large area substrates. This paper details the technique and provides specific examples of the type of structures that can be produced. SIS allows for major improvements in the accuracy and speed with which 3D patterns can be created over large areas by laser ablation. An add-in for CAD software has been developed to build up linear feature arrays for the mask from 3D designs. Feature sizes down to a few microns can be produced with excellent surface quality. Large arrays of microstructures have wide ranging applications in many areas. One example is the machining of large polymer master panels that are then electroformed to produce a mould for replication for the manufacture of display enhancement films. Keywords: Laser processing, micro-optics, large area machining, 3D structures 1. INTRODUCTION The recent rise in digital communications and multi-media systems has led to increasingly complex technical demands on personal electronic products, interactive entertainment aids and commercial and domestic display devices. Some of these developments have partly been driven by the requirements of volume manufacturing, but other significant elements have had to be addressed purely due to the novel nature of modern microelectronic systems. To meet these demands, lasers are widely used in developmental and production environments since they provide a unique combination of flexibility, efficiency and the ability to produce a wide variety of microstructures. In many display applications, the use of micro-machined structures allows display properties such as the angle-of-view (AOV), feature definition and image brightness to be greatly improved [1]. In particular, liquid crystal display (LCD) devices, whether backlit or operating under ambient lighting conditions, have benefited from these developments. One way to increase lighting efficiencies for displays is to direct more backlight through the pixels. These applications use arrays of micro-elements, like lenses, strips of ramps, pyramidal and other shapes. Another display application where micro-lenses and particularly lenticular arrays play a key role is 3D displays. Such individual micro-optical elements may range in size from several hundred microns to less than 10 µm, and they are needed in sheets that match the size of the display. Manufacturing them can be a challenge. Even when the structures lend themselves to diamond machining or reflow technique, the process tends to be time-consuming and turnaround times, often of weeks, can be obstacles to rapid prototyping and the manufacture of masters for production. To address some of these issues, Exitech recently developed the MicrAblater PPM-2000 laser machining system for large-area micro-patterning of polymers. In the process, the desired mathematically defined surface topology is translated into a multi-aperture photo mask. Using a method called Synchronized Image Scanning hereafter called SIS, the laser etches through the photo mask the desired topology onto the polymer. This can then be used directly or to generate a master for replication. This paper presents an explanation how the SIS technique works and variety of structures and features that can be machined into large polymer substrates. *k.boehlen@exitech.co.uk; phone +44 (0) ; fax +44 (0) ; exitech.co.uk

2 2. MASK IMAGING 2.1. Overview of mask imaging techniques Various mask-imaging techniques for pulsed laser ablation are employed in many industrial applications. We give a brief overview over the state of the art. There are generally speaking three main elements when looking at mask projection: the mask, the lens and the work piece or target. Furthermore we have three modes of translation: static, move and stop (also known as step and repeat ) and continuous translation. Since it would be difficult and would not really make sense to move the lens we find that statistically we have 9 combinations for mask projection. In the following we take a closer look at these and see which are technically relevant and in industrial use. The simplest combination is to keep everything static. This gives a very straightforward setup with a fixed object and target and to expose the latter with enough laser shots until the right depth of whatever object is on the mask is reached. One could think of having a base plate with some sort of index pins to put a given work piece into the right position and this gives a simple but useful tool. Adding stages in the imaging plane makes it possible to introduce a step and repeat process where the same object pattern is ablated in different but well-defined locations on the work piece. Upgrading the stages by numeric control enables the use of work piece dragging [2]. By finding the right balance between the repetition rate of the laser and the velocity of the stages it is possible to vary the channel profiles, which represent the mask shape. Any mask object designed will be reproduced as an equivalent channel as long as there are no undercuts and the depth of the channel is within reason; e.g. a circular mask will give a groove with a cylindrical bottom. Applying different velocity profiles opens other routes to micro-structure materials. With a simple square aperture it is possible to machine steps, ramps and curves. If we add numeric controlled stages to the mask plane, then we find another set of possibilities. The first of them is mask or feature changing. By keeping the work piece static and exposing the same region with different apertures 3D structures can be machined; e.g. with four differently sized squares it is possible to produce a simple pyramid structure where the step size can vary depending on the number of shots per mask feature. The equivalent to the work piece dragging mentioned above is mask dragging, but to employ that it is better to introduce a second mask so that there is one mask that is static and defines the shape of the cut and a second mask moves to control the amount of energy per area. Having the mask and the work piece on the move in a synchronized manner opens the possibility to image mask features bigger than the area the laser beam can cover; this is called "Synchronised Mask Scanning". We now have looked at all the basic mask imaging techniques; to get more sophisticated we need to put more information into the mask. The possibilities are half tone[3], gray scale or object array techniques. We will concentrate on the latter in the following section, since this is used for the Synchronised Image Scanning technique The Synchronised Image Scanning technique Synchronised Image Scanning is a laser micro-machining technique where the information for the ablation of a specific 3D feature is stored as a linear array on a chrome-on-quartz mask. The feature is then written by synchronised motion and laser firing such that the firing frequency of the laser corresponds to the spatial pitch of the features. This requires highly accurate laser triggering with low-jitter signals and highly accurate stages and encoders. The technique of SIS is explained in depth in two previous papers [4, 5]; therefore we do not go into great technical details, but would like to describe the method by a practical example. The starting point is the micro-feature that has to be machined; there are virtually no limits to the design of the feature as long as there are no undercuts. Nevertheless to explain the technique we take a rather simple feature, a square pyramid, as an example. Let us assume a footprint of 30 µm x 30 µm and a depth of 30 µm and that the pyramids should be on a 30 µm pitch on the substrate. The next step is to slice this feature. This technique is well known in stereolithography [6], and we have written software in house to do this laborious task automatically (see Automated mask design, section 3). The thickness of the steps has to be chosen; since we are using an excimer laser and typically machine a polymer, we are looking at an ablation rate between ~ 40 nm and ~ 300 nm per shot [7]. Hence it is reasonable to choose a slice thickness of 150 nm, which will result in 200 cuts, every cut being a square. By bringing these squares into a linear array at a 30 µm pitch we obtain an object of 200 squares contained in an envelope of 6 mm x 30 µm. This object represents the heart of the SIS process. It can now be magnified to match the demagnification of the imaging lens. E.g. a 5 x lens gives a mask object of 30 mm x 150 µm. To make best use of the given laser energy we array 200 of those objects on a 150 µm pitch leaving us with a mask feature of 30 mm x 30 mm. Shaping the laser beam to a 31 mm x 31 mm square at the mask plane will cover the pattern without alignment issues. Moving the work piece linearly at a speed of 6 mm/s and triggering the laser to fire a pulse every 30 µm of travel will yield 40,000 copies of the designed pyramid in one second.

3 This simple example shows that SIS is a highly efficient technique to produce micro-features. One area where this plays an important part is in display applications, where there are high density micro-imaging elements on large areas. We look into that in the next section. In Figure 1, the basic principle of SIS is demonstrated in three steps, the first laser pulse imprints the three squares on the mask into the work piece, the next pulse will be triggered as soon as the work piece has traveled exactly the pitch distance. Positions 2 and 3 on the work piece have now received two exposures. By moving the work piece further and triggering the laser again at pitch distance, the third laser pulse will finish off the feature on position 3. Continuing this process will produce a three step pyramid every laser pulse. To make the whole process efficient the work piece has to move continually, the laser being triggered on the fly. This requires highly accurate laser triggering with low-jitter signals, and highly accurate stages and encoders. Mask static 1. Shot 2. Shot 3. Shot Work piece on the move Figure 1: Principle of SIS showing the first three shots with moving work piece and static mask. The laser beam illuminates all apertures Scan direction There are two scan direction to machine a 3D feature layer by layer; one is starting with the smallest aperture as shown in Figure 2 a) and the other starts with the biggest aperture at the mask b). As long as the total depth of the machined feature does not exceed the depth of focus of the imaging lens it will reach the same depth regardless of the scan direction. Nevertheless the scan direction has an effect on the surface quality. The ever-increasing aperture of the scanning procedure starting with the smallest one will smooth the edges of the ablation steps and thus decrease the roughness of the surface significantly. When machining into polymers, it is sometimes the case that the very top layer of the material has a slightly higher ablation threshold then the rest; by using scan mode a) this can then cause some distortion of the ablation curve since every single aperture has to remove a small part of the top layer, while in scan mode b) the very first aperture will remove all of the top layer where the feature will be created and for the subsequent shots the material will be homogenous from a ablation point of view. a) b) Figure 2: Schematic build up by applying different scan directions: a) starting with the smallest and b) starting with the biggest aperture

4 In Figure 3 the fist picture shows that the steps are visible in the pyramids machined in mode b) in Figure 2, while the second picture illustrates that the scan mode a) in Figure 2 results in a smoother surface. Figure 3: Results of different scan directions: picture on the left scan starting with the biggest aperture, picture on the right scan starting with smallest aperture (deeper structure). 3. AUTOMATED MASK DESIGN To achieve a reasonable resolution in z, the ablation depth per shot should not exceed 200 nm by far. Taking this into consideration and assuming a depth of field of ~ 15 µm, the arrays on the mask typically consist of 150 or more individual features. Drawing these features one by one would be a very tedious and time-consuming job. Hence we have looked into ways of automating this process. Using a CAD program with an Application Programming Interface (API), we have taken the following approaches: For simple objects like cones and pyramids, we have written a program that takes the 2D outline of the largest feature, copies it the required number of times and shrinks each subsequent copy as defined by the user. In the case of a more complicated structure, we start with a 3D CAD model, slice it horizontally in such a way that the thickness of the slice reflects the depth of ablation of the corresponding laser shot. The sections of the 3D model form the features on the mask. When we are dealing with micro-lenses, however, the specification is often given by a fairly complex mathematical equation in x, y and z. In the simplest case, namely for a spherical lens, this equation looks like this: x 2 + y 2 + (z + R H) 2 = R 2 for 0 z H where H is the thickness of the lens; obviously it will be much more complicated for other shapes of lenses. In order to avoid the detour of first designing the corresponding CAD model and then using the slicing algorithm described above, we have written a software package that draws the feature array for the mask directly from the defining equation. As we need again horizontal sections, for each feature z equals some constant z n. The section we want to draw is therefore a curve described by an equation in x and y. In many cases, this equation will be such that y is not a simple analytical function of x. But this does not really matter: Because the resolution of the imaging system is finite, we do not loose anything by approximating the curve by a short-edged polygon. This approach is simpler and allows us to have a single program that deals with all the different types of curves; it is also more appropriate for the CAD environment. Once the features have been generated in one of the three ways described above, they are arranged in an array; the pitch between them corresponds to the pitch between the structures on the substrate. One more problem needs to be solved at

5 this stage: In many cases, the structures on the substrate and hence the features on the mask touch each other and thus some of the lines are on top of each other. This happens for instance when we limit the illuminated area around a convex structure (cf. Figure 4). As it is the requirement of the mask manufacturers that connected features on the mask are defined by one continuous outline, these overlaps have to be identified and eliminated as indicated in Figure 4. Doing this by hand is a laborious process, and so we have written a piece of software to do this. Figure 4: Basic mask pattern for a convex structure showing the elimination of overlaps 4. DEMONSTRATORS OF SIS 4.1. Micro-lenses As mentioned above, virtually any repeating feature can be generated using the SIS technique. Micro-lenses are of particular high interest in industry, and the capability of machining micro-lenses rapidly and cost effectively is very attractive. The process to produce micro-lenses is the same as the one described earlier for the pyramids. Nevertheless, we had special software written to produce the mask pattern, as explained in section 3. This software enables us to generate any sort of lens defined by a mathematical function. In contrary to the reflow technique for making microlenses, there is no restriction to the shape nor the curvature of the lens. Figure 5 shows on the left a negative spherical lens in a hexagonal array. The lens design has an optical radius of 100 µm and the outer circle containing the hexagon has a diameter of 160 µm. The scan direction was such that the lenses are built up from the Figure 5: Two different negative lens designs machined by SIS: On the left a spherical lens with a optical radius of 100 µm on a hexagonal array and on the right a ellipsoidal lens with a small diameter of 100 µm and a big diameter of 150 µm. fully open to the closed aperture in the object. Figure 5 on the right shows an negative ellipsoidal lens with a small diameter of 100 µm and a big diameter of 150 µm. There are no restrictions to the form and shape, as long as the smallest feature is bigger than the resolution of the lens and all apertures are filled with the laser beam. A state of the art micro-machining projection lens with a reasonable field of view has a resolution limit of 2 µm, therefore details smaller than that will not be resolved. The requirement that all the apertures have to be filled with the laser beam at ones is not essential. It is possible to cut one long array into shorter ones and then move one array after the other into the beam and make multiple scans on the work piece superimposed on each other. This is a good solution for big and deep features.

6 Figure 6 demonstrates the vast variety of shapes and forms that can be directly laser machined. The top row shows the design and the bottom row the result. In the first column are spherical lenses with an optical radius of 80 µm and on a 80 µm pitch with a rod-like separator in one axis. In the second column are Fresnel lenses with the same optical radius of 80 µm and again on a pitch of 80 µm. The rod-like separator appears here only every second row of lenses. These two lens designs show how powerful SIS can be for machining highly complex structures in an efficient manner. Both lens designs have the same optical behavior, i.e. focal length. The Fresnel design has the big advantage of being only half as deep and is therefore twice as efficient to machine. This is maybe not that important in this example, but can make a big difference for larger lenses. Think of a spherical lens with a sag of 60 µm, which means that the total depth is more than the depth of field of a typical projection lens with an NA of By converting the spherical lens into a Fresnel lens with three steps, the total depth will only be 20 µm an easy task for SIS. 80 um 80 um R 80 um 10.5 um R 80 um 5.3 um Figure 6: Left hand side: sketch and machined positive spherical lens with an optical radius of 80 µm and separation border. Right hand side: sketch and machined positive Fresnel lens with an optical radius of 80 µm and separation border every second row Moiré magnifier A interesting phenomenon can be seen by putting the two machined micro-lens arrays pictured in Figure 6 face to face together so that the second one is in the focus of the first. We then see the so called moiré magnifier [8]; the moiré magnification can be seen whenever an array of lenses is used to view an array of identical objects placed at the focal plane of the top lens array. As the lens array is aligned with the object array, a moiré pattern is observed in which each moiré fringe consists of a magnified image of the repeat element of the object array. As the arrays are rotated with respect to each other, the magnifications and orientation of the image changes. [8] This is exactly what can be seen in Figure 7 where we use the spherical lens array in Figure 6 on the left as an imaging array to look at the Fresnel lens array in Figure 6 on the right. Figure 7 demonstrates what happens by a small increase of the angle the samples have in respect to each other. It is possible to image any kind of object as long as it has the same pitch as the imaging lens array, in this case 80 µm.

7 Figure 7: Moiré magnified images produced by two different angles between the object and the imaging lens arrays. 5. LARGE SUBSTRATE MACHINING The big advantage of SIS is that there is no restriction in principle to the substrate size. There are limits to the substrate size by diamond turning or other micro-structuring techniques it would be simple to integrate micro-feature production on a real-to-real system and to generate a quasi-infinitely long array. This is not implemented yet, but Exitech developed the MicrAblater PPM 2000, which is capable of patterning thin substrates of a size up to 2 m x 1.4 m. A key breakthrough in achieving high pattern reproducibility was made by developing a novel FocaFloat optics head that ensures the film front surface is positioned to within ±10µm of the lens image plane at all times even on sheets with significant variations in thickness (±100µm) and with significant sag (1mm). This head incorporates debris extraction close to the material is ablated and therefore protects the surface from re-deposition, a very important point in a process that removes about 0.5 mm 3 /s. It is also possible to have a local inert gas environment where the ablation takes place. A picture of a PPM1600 that Exitech has built and installed at a customer site can be seen in Figure 8. This machine is designed to process 1.4 m x 1.2 m substrates and is fitted with a automatic loader visible at the front. Figure 8: Exitech PPM1600 with automatic work piece loader.

8 Not only the size of the sample but also the number of lenses machined into it is impressive: We are looking at more than 400 million lenses. 400 million lenses where the reproducibility is close to 100 %. Figure 9 gives an indication of the reproducibility achieved by SIS. Figure 9: Demonstration of reproducibility of the SIS process One very important point when machining large areas is to obtain a homogenized appearance. Even though the microfeatures are identical, the human eye is amazingly good at catching repeating seams or pattern that should not be there. The in house developed multi-scan method enables us to machine big areas macro-pattern free. Figure 10 illustrates the differences with and without multi-scan. Although the same micro-features are machined, the macro-appearance of the processed area is very different. In the left hand picture are five vertical bands clearly visible plus some horizontal inhomogeneity. The right hand picture on the contrary shows an even appearance, and no banding or seam at all is visible. Figure 10: Comparison of processed areas with multi-scan method on the right and without on the left. On the left five vertical bands are clearly visible plus some horizontal inhomogeneity while on the right the appearance is very smooth. 6. CONCLUSIONS We have demonstrated the possibility to produce low-cost high-quality optical micro-features into large-area substrates by direct laser ablation. We have described the main mask imaging techniques and focussed particularly on SIS, which opens a new route to generate complex microstructures on a big scale in an efficient way. We have shown that we are

9 able to produce high-quality, high-resolution features like spherical and Fresnel lenses and pyramids into flat substrates. One area of industrial application are micro-lens arrays, but we are convinced that there is a great potential in all kinds of micro-structuring. SIS is well suited to machine features with an outstanding repeatability from a few microns up to a few hundreds of microns regardless of the complexity of the profile. The unique advantage of SIS is that the method itself does not limit the area that can be machined unlike diamond turning or reflow technique, and Exitech has developed a machine that can process substrates of up to 2.8 m ACKNOWLEDGEMENTS It is my pleasure to thank my colleagues at Exitech for their help, support and contributions. REFERENCES 1. R. J. T. Clabburn & A. M. Fairhurst, "Photopolymer materials to improve LCD image quality" Proc. SID International, Symposium, San Jose (USA) May N. Rizvi et al. "Laser micromachining - new techniques and developments for display applications", Proc. of SPIE Vol. 4274, p , F. Quentel et al. "Multilevel diffractive optical element manufacture by Excimer laser ablation and halftone masks", Proc. of SPIE, Vol. 4274, p , H.J. Booth "Recent Applications of pulsed lasers in advanced materials processing", Thin Solid Films 2004, to be published 5. C. Abbott et al. "New techniques for laser micromachining MEMS devices", Proc. of SPIE 2002, Vol. 4760, p.281, C. W. Hull, "Apparatus for production of three dimensional objects by stereolithography", US Patent, R.C.Crafer & P.J. Oakley, "Laser Processing in Manufacturing", Chapter 8, Chapman & Hall, London, M. C. Hutley et al., "Pure Appl. Opt. 3, page , UK, 1994