Pulsed Laser Ablation of Polymers for Display Applications James E.A Pedder 1, Andrew S. Holmes 2, Heather J. Booth 1 1 Oerlikon Optics UK Ltd, Oxford Industrial Estate, Yarnton, Oxford, OX5 1QU, UK 2 Imperial College London, EEE Department, Exhibition Road, London, SW7 2AZ, UK ABSTRACT Laser micromachining by ablation is a well established technique used for the production of 2.5D and 3D features in a wide variety of materials. The fabrication of stepped, multi-level, structures can be achieved using a number of binary mask projection techniques using excimer lasers. Alternatively, direct-writing of complex 2.5D features can easily be achieved with solid-state lasers. Excimer laser ablation using half-tone masks allows almost continuous surface relief and the generation of features with low surface roughness. We have developed techniques to create large arrays of repeating micro-optical structures on polymer substrates. Here, we show our recent developments in laser structuring with the combination of half-tone and binary mask techniques. 1. INTRODUCTION Laser micromachining has become an increasingly important process for display manufacture over the past few years. With the advent of very high average power solid state lasers, laser ablation has become a viable technique for patterning thin film transparent conductive oxides on an industrial scale [1], and this approach is now being applied in volume production of flat panel displays. Laser micromachining is also emerging as the preferred technique for manufacturing large area plastic microlens arrays. Such arrays are used as diffusers to optimize the viewing angle and/or contrast in liquid crystal displays and rear projection televisions. They are also required for 3D television, where lenticular lens arrays are used to separate two interlaced images for the viewer in order to generate a stereoscopic effect. As a technique for manufacturing large area arrays of micro-optical elements, laser micromachining competes primarily with two other approaches: greyscale lithography and diamond turning. Greyscale lithography [2] produces a photoresist master which may be transferred either into an underlying substrate by anisotropic etching, or into a metal by electroforming. In either case a hard replica of the original master is produced which can be used as a hot embossing tool for mass replication. This technique works well at wafer scale, where lithographic exposure tools and plasma etchers are available with well defined and uniform characteristics, but is difficult to scale to larger areas. Display applications have unusually stringent requirements in terms of uniformity across the array, because the eye is highly sensitive even to slight irregularities, and it is difficult to meet these requirements if a large area replication tool has to be built up by stitching together smaller ones. Diamond turning [3] is scaleable to large areas, and can produce lenses with lateral dimensions from 50 μm to a few mm with surface roughness less then 15 nm rms. However, aspherical lenses are difficult to fabricate by diamond turning without complex multi-stage motion and grinding tips formed using focused-ion-beam milling. The ability of laser processing to produce almost arbitrary lens profiles is a significant advantage for display applications. For example, the field of view requirements are different in the horizontal and vertical directions, so spherical lenses are not the optimal choice for diffusers. A further advantage of laser processing is that it can yield close-packed arrays of lenses with a fill factor approaching 100% with outstanding uniformity and repeatability. Large area laser micromachining for displays is invariably carried out in mask projection mode, where a mask pattern is projected onto the workpiece using a reduction lens. This approach is ideally suited to excimer lasers which have high pulse power but relatively low beam quality and pulse repetition rate. The exposure field at the workpiece is generally quite small (typically up to 10 mm 2 ), with larger areas being covered by moving the workpiece on translation stages. For large panels this requires precision, high-speed stages that can achieve micron accuracy over displacements of up to several metres, together with dynamic focus control to compensate for variations in surface height. Machines that incorporate these facilities are now highly developed and available commercially.
When laser machining microlens arrays, control of the lens profile is typically achieved by exposing each lens site to a sequence of different mask apertures. In this way different parts of each lens receive different numbers of laser pulses, and a stepped surface that closely matches the desired lens shape is produced. The synchronized-image-scanning (SIS) technique developed by Exitech Ltd, which uses a linear array of apertures, is based on this principle [4]. An alternative approach is to use a variable transmission mask, implemented as a half-tone screen, in which case all regions of the workpiece receive the same number of pulses, but the etch depth per pulse varies with position due to the variation in transmitted laser fluence. In this paper we make a critical comparison of these two approaches as applied to the fabrication of large area polycarbonate microlens arrays. 2. LARGE AREA LASER MICROMACHINING The principle of SIS is shown in Figure 1, where the projection lens has been omitted to simplify the diagram. The mask carries a series of apertures which are constructed simply by taking slices through a CAD model of the basic repeating structure. The different apertures are arranged in a linear array on the mask, with the spacing between apertures being equal to the desired pitch of the repeating structures on the workpiece (multiplied by the reduction ratio of the projection optics). The workpiece is moved continuously, and firing of the laser is synchronized to the stage motion so that each lens site on the workpiece is exposed to each aperture in turn. Fig. 1. Laser machining of 2.5D repeating structures by SIS. SIS is ideally implemented as a single pass process, with each region of the workpiece passing beneath the mask only once. This avoids problems with ablation debris that can occur in multiple-pass processes, whereby debris from earlier passes can lead to micro-masking and hence increased surface roughness. In a single-pass process with a linear array of apertures, ablation debris can be blown using an air-jet onto the part of the surface that has already been machined, and then removed at the end of the process by, for example, solvent cleaning. Figure 2 shows how a similar optical set-up can be used to form large area microlens array with a half-tone mask. In this case all the apertures have the same intensity transmission profile, chosen such that the desired lens shape is produced after N HT pulses where N HT is the total number of apertures. Note that the same result could, in principle, be achieved by a multiple-pass process with a single half-tone aperture. However, such a process would be expected to suffer from severe ablation debris issues, and would be inefficient in its use of the available laser power if multiple apertures could be accommodated within the beam cross-section. Half-tone mask design for processes of the kind shown in Fig. 2 is a relatively complex process in the general case. However, in the simplest analysis, where secondary effects such as the reduction of ablation rate on inclined surfaces
are ignored, the mask transmission profile for a given operating fluence may be derived directly from the CAD model and the material ablation curve at normal incidence. This approach, which was used in the present work, has been reported previously and will not be described in detail here [5,6]. Fig. 2. Alternative SIS process using an array of similar half-tone mask apertures. 2.1 Operating fluence and process efficiency Choosing the correct operating fluence is a key step in the design of any laser micromachining process. The fluence always has to be in a range where the surface finish and debris levels are acceptable for the intended application, and for prototyping or low-volume production these are often the only considerations. For large area applications, however, the achievable throughput and the process efficiency in terms of ablated volume per Joule of incident laser energy also need to be taken into account. For polymer materials under nanosecond ablation, the shape of the ablation curve is invariably such that there is a peak in the process efficiency some way above the ablation threshold. For polycarbonate machined at 248 nm wavelength, this occurs at a fluence of around 250 mj/cm 2 where the ablation depth per pulse is around 120 nm. For a given laser system, using this fluence level will give the maximum throughput provided all the available laser energy is used. In practice, however, other constraints may determine the fluence level. For example, if the machining tool has not been designed specifically for the application in question, it may not accommodate enough SIS apertures to achieve the required machining depth in a single pass at the optimal fluence, in which case a higher fluence will need to be used. In a different application, a lower than optimal fluence might be used, with a correspondingly higher number of SIS apertures, to reduce the stepping in the machined surfaces. The process efficiency when using half-tone apertures is inevitably lower than in conventional SIS, implying either a higher operating fluence or a larger number of apertures at the same fluence. The reasons for this are twofold. Firstly, the maximum transmission of the half-tone screen is always below 100% due to limitations in the mask manufacturing process. Secondly, there is always a minimum fluence below which ablation debris is not ejected effectively and hence accumulates on the machined surface. It is essential to keep the transmitted fluence above this minimum level, and consequently it is not possible to produce structures where the depth varies continuously down to zero. This minimum fluence requirement generally results in additional material removal beyond the minimum required to form the desired structure, and hence wasted laser energy. This effect can be quite significant at lower operating fluences. For example, a safe minimum fluence for polycarbonate at 248 nm wavelength is around 100 mj/cm 2. At an operating fluence of 250 mj/cm 2, and assuming a maximum half-tone mask transmission of 92%, the process efficiency is only about 60% of that of conventional SIS.
3. EXPERIMENTAL TECHNIQUE In order to compare the performances of conventional and half-tone SIS processes, arrays of close-packed convex microlenses were machined in polycarbonate. Each lens was designed to have a spherical surface of radius 41 μm, and the centre-to-centre separation of nearest neighbors was 50 μm, implying a maximum machined depth of about 12 μm. Experiments were carried out using an Exitech M8000 laser micromachining workstation designed specifically for SIS. The workstation was equipped with a KrF excimer laser (248 nm wavelength), a 5X 0.13NA projection lens, and beamforming optics configured to produce a 70 x 6 mm 2 exposure field at the mask plane. Exposures were carried out at a number of different operating fluences between 200 mj/cm 2 and 1000 mj/cm 2. A different, specially designed mask was used at each fluence so that all the lenses produced would have the same nominal shape. Table 1 shows the numbers of apertures and the half-tone transmission ranges used in each case. Figure 3 shows a representative lens array fabricated by half-tone exposure at 200 mj/cm 2. Table 1. Aperture numbers and half-tone transmission ranges used at different operating fluences. Fluence mj/cm 2 N SIS N HT Min HT transmission % Max HT transmission % 200 134 247 50 92 400 69 94 25 91 600 54 66 17 91 800 45 55 13 90 1000 39 50 13 90 Fig. 3. Close-packed microlens array formed by SIS with a half-tone mask at 200 mj/cm 2. 4. RESULTS The lens arrays produced were compared in terms of both profile accuracy and surface roughness. Data for both comparisons was acquired using an AFM (Veeco Multimode SPM). The maximum vertical travel of the AFM tip was about 5 μm, meaning that only a 30 x 30 μm 2 region at the centre of each lens could be explored.
4.1 Profile accuracy Figure 4 shows an AFM image of a half-tone lens, together with extracted surface profiles for lenses machined at all fluence levels. The measured profiles for lenses machined at fluences of 600, 800 and 1000 mj/cm 2 overlay the design profile almost perfectly. The profile accuracy is less good at lower fluences, but the nature of the error is such that it could be virtually eliminated by slight adjustment of the operating fluence. Fig. 4. AFM plot for half-tone lens machined at 200 mj/cm 2 (left), and extracted surface profiles for half-tone lenses machined at different fluences with design profile for comparison (right). The profile accuracy achieved with conventional SIS was generally found to be less good, and also dependent on the direction of motion of the workpiece. Figure 5 shows extracted profiles for lenses machined at all four fluences levels and both scan directions. In Figure 5a where the workpiece was moving in the direction of the arrow in Figure 1, the lens profiles are all severely distorted by a flat region at the top. This flattening is thought to arise from ablation debris being deposited in the central region of the lens during the early stages of machining, and then inhibiting ablation of the central region towards the end of the process. This effect cannot occur when the workpiece is moving in the opposite direction, because in that case the light transmitted by each aperture always falls on a region that has been ablated by all previous pulses, and debris does not have the opportunity to accumulate in such a region. (a) (b) Fig. 5. Extracted surface profiles for conventional SIS lenses formed at different fluences and with the workpiece moving either (a) Backward (along the direction of the arrow in Figure 1), or (b) Forward (in the opposite direction).
The dependence of the lens profile on the scanning direction is an undesirable feature of conventional SIS, since for large area applications the most efficient way to move the workpiece is in a raster pattern, with adjacent rows of elements being produced by opposing scan directions. This directional dependence does not arise with half-tone SIS because the mask apertures are all similar. 4.1 Surface roughness Figure 6 shows close-up SEM images of microlenses machined by conventional SIS with forward scanning, and by half-tone SIS, the operating fluence being 200 mj/cm 2 in both cases. The terracing in the surface of the conventional SIS lens is typical of structures fabricated by this method, and results from the sharp edges of the SIS apertures. This effect, which becomes more pronounced at higher fluences, is also evident in the surface profiles of Figure 5b. Interestingly it cannot be seen in the profiles in Figure 5a, and this is because with backward scanning the steps in the surface produced by each aperture are smoothed by successive laser pulses, in what is effectively a self-polishing process. In a similar fashion, the surface structure on the lens in Figure 6a could be reduced by a post-process involving flood exposure. No roughness is discernable in the SEM image of the half-tone SIS lens, except close to the boundaries between lenses. Figure 7 shows the variations in surface roughness with operating fluence for both processes, based on Ra values extracted from AFM data. (a) Fig. 6. Close-up SEM images of microlenses formed by (a) conventional SIS with forward scanning, and (b) half-tone SIS. The fluence was 200 mj/cm 2 in both cases. (b) (a) (b) Fig. 7. Variations of roughness parameters with fluence for (a) conventional SIS with forward scanning, and (b) halftone SIS. Roughness parameters are extracted from line scans of the AFM data.
The roughness with conventional SIS appears to rise steeply with fluence at low fluence levels, but then level out at around 90 nm. The best achievable roughness, at a fluence of 200 mj/cm 2, corresponds to an Ra of around 15 nm. It must be noted that is for a forward scanning process with no polishing. The Ra value with half-tone SIS is consistently below 10 nm, with values of 6 nm or below being achieved over most of the fluence range tested. 5. CONCLUSION We have compared SIS laser micromachining processes employing binary and half-tone masks as techniques for fabricating large area microlens arrays in polycarbonate. The results obtained with conventional binary masks indicate that the profile accuracy depends to some extent on the scan direction, and that there is an apparent trade-off between profile accuracy and surface finish. In the scan direction that gives the best profile accuracy, the minimum Ra value achieved was around 15 nm. The half-tone process was found to give good profile accuracy and low surface roughness over a wide fluence range. The Ra was consistently below 10 nm, with a typical value of around 6 nm. 6. REFERENCES 1. M. Henry, Laser Direct Write of Active Thin-Films on Glass for Industrial Flat Panel Display Manufacture, Proc. 4 th Int. Congress. Laser Adv. Mat. Processing, Kyoto, 2006, Paper No. 06-16. 2. W. Daschner et al, General aspheric refractive micro-optics fabricated by optical lithography using a high energy beam sensitive glass gray-level mask, J. Vac. Sci. Tech. B, Vol. 14, Nov. 1996, pp 3730-3733. 3. A.Y. Yi, L. Li, Design and fabrication of a microlens array by use of a slow tool servo, Optics Letters, Vol. 30, July 2005, pp 1707-1709. 4. K.L. Boehlen et al, Advanced Laser Micro-Structuring of Super-Large Area Optical Films, Proc. SPIE, Vol. 5720, Jan. 2005, pp 204-211. 5. A.S. Holmes, Excimer laser micromachining with half-tone masks for the fabrication of 3D microstructures, IEE Proc. Sci. Meas. & Technol., Vol. 151, No. 2, 2004, pp. 85-92. 6. A.S. Holmes, J.E.A. Pedder, K. Boehlen, Advanced laser micromachining processes for MEMS and optical applications, SPIE Proceedings Vol. 6261, Proc. High Power Laser Ablation Conference 2006, Taos, New Mexico, 7-12 May 2006, pp. 1E1-1E9.