Microlens Laser Beam Homogenizer From Theory to Application

Transcription

1 Invited Paper Microlens Laser Beam Homogenizer From Theory to Application Maik Zimmermann* a, Norbert Lindlein b, Reinhard Voelkel c, Kenneth J.Weible c a Bayerisches Laserzentrum GmbH, Konrad-Zuse-Str. 2-6, Erlangen, Germany b Institute o Optics, Inormation and Photonics (Max Planck Research Group), University o Erlangen, Staudtstr. 7/B2, Erlangen, Germany c SUSS MicroOptics SA, Jaquet-Droz 7, CH-2000 Neuchâtel, Switzerland ABSTRACT Many applications in laser manuacturing like semiconductor lithography, micro-machining, micro-structuring or material-analysis require a homogeneous intensity distribution o the laser beam over its complete proile. Reractive and diractive beam homogenizer solutions have been developed or this challenge, but their applicability strongly depends on the physics o the individual laser beam. This paper investigates the inluence o laser beam properties like spatial coherence or microlens beam homogenizers. Diraction at the small lens apertures and intererence eects o periodic arrays are explained by using diraction theory. Dierent microlens beam homogenizer conigurations are presented. Design considerations that might be helpul or the layout o a speciic microlens beam homogenizer system are discussed. It is shown that, among other actors, the Fresnel number is the most important quantity to characterize the inluence o diraction eects on microlens laser beam homogenizers. The inluence o the spatial partial coherence will be explained or the example o a Fly s eye condenser. For cw laser sources, the inluence o a rotating diuser plate on grating intererence and speckles eects is investigated. Finally, the theory will be compared to some practical examples in planar laser measurement techniques, in combustion diagnostics and micromachining with Excimer lasers. Keywords: homogenizer, lat-top, microlens, shaping, laser, coherence, diraction, diuser 1. INTRODUCTION The homogenization o laser beams is an important issue not only in many ields o laser material processing, but also in laser measuring techniques and laser analysis. Most o all, those laser applications, which image a mask pattern onto a work piece, require a homogenous distribution o radiation intensity over the whole mask area and consequently over the whole machining plane. Examples o applications are photolithographic methods or the light exposure on semiconductors [1] or the selective removal o multi layer systems. To produce multi-hole arrays, e. g. or ink-jet printer or microcavities or cupping tools [2], the diameter, the quality o the curb and the depth o the holes within the arrays must be very uniorm. A beam uniormity in the range o ± 5 % (rms) is standard or laser machining applications and ± 2 % or photolithography. Similar beam quality requirements also exist or the direct radiation o a specimen with the laser, e. g. or luorescence detection in bio-sensing. Here both beam directions are homogenized and the beam proile is mostly projected as a rectangular or squared proile on the sample. Other applications require a homogenous thin laser line, only one beam direction is homogenized. The most important example or this application is lat-panel-display annealing. Here a thin layer o amorphous silicon is ormed on a glass substrate in a vapor-deposition process. This layer must be transormed into poly-crystalline silicon. In the past, this phase transormation was perormed in high-temperature ovens requiring the use o expensive, thermally resistant glass. Today, most lat-panel silicon annealing is carried out by using high-power Excimer lasers in combination with lowtemperature ovens, allowing the use o low-cost glass panels. The key requirements in this application are uniorm irradiation across the entire panel and a ast throughput. This is accomplished by sweeping a long, thin, highly uniorm line across the panel. Further examples or thin line homogenizers are laser light sheets, e. g. or the analysis o streams and particle low. For methods like the Particle Image Velocimetry (PIV) or the Planar Laser Induced Fluorescence (PLIV) a thin laser sheet is produced in the medium that is to be tested [3]. A quantitative precise analysis requires a homogenous distribution o intensity over the whole light-section during measuring. *m.zimmermann@blz.org; phone ; ax ; Laser Beam Shaping VIII, edited by Fred M. Dickey, David L. Shealy, Proc. o SPIE Vol. 6663, , (2007) X/07/$18 doi: / Proc. o SPIE Vol

2 The examples above show the wide ield o applications and the demand or beam homogenization. Various elements and optical systems have been developed or laser beam shaping. Honagle et al. [4] described a reractive beam shaper which can be used to sort the light into a lat-top distribution using two specially designed aspherical lenses. The disadvantages o such systems are the strict dependence on the entrance proile and the proper alignment. Alignment errors and luctuations o the laser beam have a strong inluence on the achieved uniormity. Beam shaping with diractive optical elements represents a very elegant and powerul method or the generation o arbitrary irradiation patterns [5]. These elements are usually designed or a speciic wavelength and phase unction. To achieve high perormance, i. e. beam uniormity and eiciency, expensive multi-level elements are necessary. Another concept or lat-top generation uses multi-aperture elements, which divide the incoming beam into a number o beamlets. The beamlets are overlapped with the help o an additional lens. The advantages o these shapers are the independence rom entrance intensity proile and wide spectrum o wavelengths. However, the periodic structure and the overlapping o beamlets produce intererence eects especially with the usage o highly coherent light. Nevertheless a successul homogenization with these elements can be achieved with the consideration o physical optics [6] and in certain cases with the usage o additional elements like random diusers. 2. GEOMETRICAL DESCRIPTION OF MICROLENS BEAM HOMOGENIZERS The principle o light homogenization with arrayed elements was ound more than 100 years ago. A publication o 1940 suggests the application o a multi-aperture system or the illumination in ilm projectors [7]. Classical homogenizers consist o lens array elements, which are produced by compression molding or manuactured through assembly o single cylindrical lenses. Classical array homogenizers have many disadvantages: manuacturing tolerances or the individual lenslets; misalignment in the array; scattering or losses at the transition and relatively high costs or mounting. Microlens arrays are a cheaper alternative which are replicated into a monolithic material o high optical quality by the use o waer-based manuacturing processes like photolithography and reactive-ion-etching. The challenge o these processes is the optimization o the lens proile, which is essential or the quality o the homogenization [8]. However, the manuacturing costs per square millimeter are much lower compared to conventional methods. There are two main types o microlens beam homogenizers: the non-imaging and the imaging homogenizer [9]. Both types use lens arrays to split the incident beam into beamlets. These beamlets are then passed through a spherical lens and overlap at the homogenization plane located in the back ocal plane o the spherical lens. The spherical lens causes parallel bundles o rays to converge in the homogenization plane and is thereore called a Fourier lens. It carries out a two-dimensional Fourier transormation. The intensity pattern in the homogenization plane is related to the spatial requency spectrum generated by the microlens array or arrays prior to entering the Fourier lens. The non-imaging homogenizer consists o a single lens array and a spherical lens. The imaging homogenizer with two lens arrays shown in Figure 1 usually provides much better lat-top uniormity [9]. 1 2 FP a, Fig. 1. Imaging Homogenizer: Two microlens arrays 1 and 2, one spherical Fourier lens. The beam propagation in an imaging homogenizer is based on the Kohler illumination system which is well known rom microscopy. In a ly s eye condenser, the lens arrays orm many parallel Kohler illumination systems side-by-side. Multiple light sources are generated by dividing the entrance beam into multiple beamlets. The second microlens array 2, in combination with the spherical Fourier lens, acts as an array o objective lenses that superimposes the images DFT Proc. o SPIE Vol

3 o each o the beamlets in the irst array onto the homogenization plane FP. Square-type lens apertures o the irst microlens array 1 generate a square lat-top intensity distribution in the Fourier plane. Circular or hexagonal microlenses will generate a circular or hexagonal lat-top, respectively. Usually square-lens or crossed cylindrical-lens arrays are used to ensure a high illing actor that is typical better than 98 %. The lens array is characterized by the pitch p, i. e. the vertex clearance between two neighboring lenses o the array. To show the correlation between the element-properties within the homogenizer we used paraxial matrix method or a irst approximation o the geometrical optic. For the description given here, we assume a point light source at an ininite distance in ront o the irst lens array. The size o the lat-top D FT depends on the ocal lengths o the lenses within the array 1, 2 and the Fourier lens and is given by D FT ( a12 ) = p. (1) 1 2 To ulil the imaging conditions as mentioned above, the separation a 12 between 1 and 2 has to be equal to the ocal length o the second lenses 2. Through this and together with the optical power o the Fourier lens, the apertures are imaged to the plane FP. With the determination that a 12 = 2, equation (1) is simpliied to D FT = p. (2) 2 For imaging homogenizers the divergence θ (hal angle) ater the homogenized plane is given by 1 d IN 2 p + DFT 2 p tanθ = +, with a 12 = 2 and s = 0, (3) 2 1 where d IN is the diameter o the incident beam and s the distance between the second lens array and the Fourier lens. In this equation we assume that the separation s between the second lens array and the Fourier lens is zero. Normally the divergence increases by increasing this separation. Usually, imaging homogenizers consist o two similar lens arrays with identical lens pitch p and ocal length ( 1 = 2 = ). This coniguration is the classical ly s eye condenser. For collimated laser beams, the beam is then ocused into the plane o the second lens array. Thus, care must be taken not to damage the second microlens array by ocusing high-power laser beams into the lens material. In this case, the ocal length o the second array and according to imaging conditions the separation o 1 and 2 has to be increased. For extended laser sources like a multimode laser or a collimated iber-coupled laser, an image o the light source is ound at the plane o the second microlens array. For imaging homogenizers the diameter o the individual beamlets at the second microlens array 2 must be smaller than the lens pitch to avoid overilling o the lens aperture and the loss o light. An overilling o the second lens array results in unwanted multiple-images in the plane FP. I an extended light source with the diameter D source is collimated with a positive spherical lens with a ocal length CL, the image size D image at the second lens array is D image = D Source 1 CL p, i a 12 = 1 = 2. (4) To avoid overilling o lens apertures the pitch p has to be larger than D image. For laser beams with a signiicant beam divergence the diameter o the beamlets at the second microlens array scales with the beam divergence. The allowed p maximum beam divergence σ or a given microlens homogenizer is tanσ or a 12 = 1. 2 The number o lenses N across the laser beam diameter d IN is N = d IN / p. A simple demonstration o the intensity distribution ater the homogenizer with a Gaussian entrance proile illustrates the dependence upon the number o lenses in Figure 2. For standard laser beams, an overlay o some 8-10 microlenses is usually suicient to achieve a good lat top uniormity. Larger numbers o microlenses do not have negative eects. For large Excimer laser beams, homogenizers with many thousands o microlenses provide excellent lat-top proiles. 1 Proc. o SPIE Vol

4 Microlens Array, IuI4J N=3 N=9 FP: Overlay o magniied lens apertures Simulated Proile in FP Measured proile in FP Fig. 2. Illustration o the intensity redistribution in Fourier-plane in dependence on the number o lenses. As shown above an increasing o the number o lenslets across the pupil diameter will improve the quality o the homogeneous intensity distribution in the case o geometrical optics. Normally, the total size o the elements is limited by cost and manuacturing issues. The consequence is a miniaturization o the lenses within the array, while retaining the diameter o the incident beam. The result seen rom physical optics is an increase o diraction eects because o the smaller apertures o the lenses. For a simpliied quantiication o the diraction inluence, the calculation o the Fresnel number can be very helpul. 3. DIFFRACTION AND GRATING EFFECTS A regular microlens array is a periodic structure with a period p showing eects like grating inerence and Talbot selimaging [6, 10]. Light interacting with a periodic structure will always keep traces o this periodicity in its urther propagation. This is well explained by Fourier optics: The Fourier transormation o a comb unction is again a comb unction. This remaining periodicity usually generates an unwanted modulation in the homogenization plane and limits the degree o uniormity that can be achieved in the lat-top. The modulation is undamental or all reractive and diractive laser beam homogenizers. The homogenizing process requests that the incident beam is divided into individual beamlets and that these beamlets are reorganized and overlap in the homogenization plane. Thus, the splitting o the beam itsel introduces the modulation o the lat-top intensity. Obviously, the inluence o these eects depends strongly on the coherence o the light source. Using an imaging microlens beam homogenizer as shown in Figure 1 and a coherent and well collimated laser beam, the lat-top intensity proile is subdivided into sharp peaks. Due to the inite extension o the microlens array the peaks will o course not be exact delta unctions, but the comb unction will be convoluted with the Fourier transorm o the aperture unction o the illuminated part o the microlens array. The single peaks will be Airy discs i the illuminated part is circular. In the case o a Fly s eye condenser, a Fourier lens will transorm the ar ield behind the microlens arrays into the ocal plane o this lens. So, it is quite clear that in the case o coherent illumination there will be sharp peaks in the Fourier plane o the Fly s eye condenser. These peaks can only be avoided i the illumination is partly coherent or i the periodicity is avoided. Wippermann et al. [11] described a solution or avoiding the periodic structure with the help o a so called chirped microlens array. Another way to temporally smooth the spiked proile is to use rotating random diusers. 3.1 Fresnel Number A simple illustration and quantiication o the eect o diraction can be given by calculating the Fresnel number FN o the Fly s eye condenser (Fig. 3). The Fresnel number o a lens is an important quantity to characterize the inluence o diraction eects onto the lens. The Fresnel number describes the number o Fresnel zones o a spherical wave which is ormed by a lens with a diameter D lens (we assume that p = D lens or a illing actor o 100%) or an incident plane wave with wavelength λ. Proc. o SPIE Vol

5 This can be seen easily by calculating the optical path length dierence ( OPD) between a ray at the rim o the lens and at the centre o the lens: ( D / 2) ( D / 2) lens λ lens + = FN 2 2 OPD =, (5) where is the ocal length o a microlens within an array. Here, the square root was expressed by the irst two terms o its Taylor series or the case D lens /2 <<. When this condition is ulilled we can deine the Fresnel number o a lens with a ocal length and a lateral diameter D lens o the lens, the Fresnel number FN to be: FN ( D / 2) 2 lens = λ. (6) I, '"7,' 'dl,, 'liii F ::s-: --- I' it,,! 1 Ill,' I,,,, S I S F F 4- sue-i ainpade 'a d Fig. 3. Illustration o the Fresnel number. With consideration o the lat-top dimension beam homogenizer [9] is deined as FN p D 4 λ FT uoqeasqo lutod 12 Z = Vi D p iausaij sauo FT FT = and D lens p the Fresnel number FN o a microlens, (7) whereas p is the pitch o the microlens array, D FT is the dimension o the lat-top intensity proile in the homogenization plane FP, is the ocal length o the Fourier lens and λ is the wavelength. Non-imaging homogenizers oten show dominant diraction eects due to Fresnel diraction at the microlens array. The Fresnel diraction is related to the Fresnel number FN. Higher Fresnel numbers give sharper edges and smaller variations o the lat-top proile. In practice, non-imaging homogenizers should have Fresnel numbers FN > 10, better FN > 100, to obtain a good uniormity. Non-imaging homogenizers are well suited or large area illumination, as the lattop dimension D FT is proportional to the Fresnel number FN. For small Fresnel numbers FN < 10 or high uniormity lattop requirements, the imaging homogenizer is the preerred solution. 3.2 Spatial coherence in the case o a Fly s eye condenser Fulilling the image condition or imaging beam homogenizers holds a severe disadvantage or coherent laser beams. Microlens arrays are periodic structures, where the pitch p is the grating period. Each microlens array will behave like Proc. o SPIE Vol

6 a diraction grating and will generate diraction orders with a period Λ FP = λ / p in the homogenization plane FP. The inluence o the partial spatial coherence onto the light distribution will be explained or the case o the Fly s eye condenser shown in Figure 4. CL, CCD 4 l = 2 Fig. 4. Scheme o a Fly s eye condenser with identical microlens arrays. Each point o the monochromatic, but spatially incoherent light source orms a plane wave with a certain tilt ϑ in = x source / CL (x source is the distance o the light source point rom the optical axis) behind the collimator lens CL. Each o these plane waves is incoherent to each other. In the wave-optical model each plane wave is split behind the irst lens array 1 into several diraction orders m 1. The period o the lens arrays 1 and 2 is p and the wavelength o the light is λ. Then, the diraction orders behind the irst lens array propagate with the ar ield angle ϑ 1 : = λ. p ϑ1 ϑin + m1 Here, only small angles are used (p >> λ or a typical microlens array) so that we can approximatesin ϑ ϑ. Behind the second microlens array 2 with the same period p the diraction orders m 2 have the same angular dierence thus the ar ield angles ϑ 2 o the dierent diraction orders behind the second microlens array are: λ λ λ ϑ 2 ϑ m ϑ ( m + m ) = ϑin + m. (9) p p = = in p Finally, the Fourier lens ocuses each plane wave o the dierent diraction orders into a spot. This spot is in the ocal plane at the lateral position: λ x = ϑ 2 = ϑin + m. p A wave-optical simulation o the intensity in the ocal plane o Fourier lens or a single light source point is shown in Figure 5. For a circular diameter Ø o the illuminated aperture o Fourier lens each spot is o course an Airy disc (Figure 6) with a radius r Airy between the central peak and the irst minimum o: r Airy λ =. (11) 1.22 Ø Here, again the small angle approximation is used which is valid or Ø <<. (8) (10) Proc. o SPIE Vol

7 Fig. 5. Wave-optical simulation o the intensity in the ocal plane o Fourier lens or a single light source point: total ield in the ocal plane. JIA LJ L Fig. 6. Zoomed central part o Fig. 5 showing the single Airy discs o the dierent diraction orders. The intensities o the spot arrays in the ocal plane o Fourier lens which are ormed by dierent light source points are then added. Each light source point with axial distance x source orms a laterally shited spot array with a shit o x = ϑ in = xsource / CL. I the diameter D source o the light source is such that the lateral shit x o all light source points is just one period Λ FP o the spot array with Λ FP = λ / p, a smooth homogeneous intensity will be obtained in the ocal plane o lens, i all light source points emit identically. However, i the diameter o the light source increases urther there will be again high luctuations o the intensity pattern. The simulated intensity distributions in the ocal plane o lens or an extended light source or is shown in Figs. 7 and 8. Fig. 7. Simulated intensity distributions in the ocal plane or an extended light source = p. D / λ / source CL Fig. 8. Simulated intensity distributions in the ocal plane or an extended light source = p. D / 1.5λ / source CL Proc. o SPIE Vol

8 Only in the case that the light source diameter is so large that the spot arrays o the dierent light source points are shited by many periods there will be a nearly homogeneous intensity even i the light source diameter changes by a ew percent or i the light source emits dierently or dierent points. Figs. 9 and 10 show clearly this eect or a large light source with a ratio R o the angular extension o the light source D source / CL and o the angular extension o the spot array λ / p o R = 82 or R = There is nearly no dierence between both intensity distributions. Only, by looking exactly on both Figures the right one shows a little bit higher luctuations. Fig. 9. Simulated intensity distributions in the ocal plane o Fourier lens or a larger extended light extended light source = p. D / 82λ / source CL Fig. 10. Simulated intensity distributions in the ocal plane o Fourier lens or a larger extended light extended light source = p. D / 82.5λ / O course, there is an upper limit or the light source diameter depending on the lenses o the microlens arrays because the light source images which are ormed by the collimator lens CL and the lens array 1 on the apertures o the lens array 2 are not allowed to be larger than one lens aperture o array 2. Otherwise, there will be cross talking between the dierent channels and the intensity distribution in the ocal plane o Fourier lens will have undesired side bands. The maximum diameter o the light source which is allowed is: Dsource / p D 1 CL source p R = 4FN λ / p λ =. (12) CL 2 1 So, the ratio R which has to be large in order to have a good homogenization is limited by the Fresnel number FN o the lenses o the microlens arrays. Here, it is assumed that the lenses o the arrays have a diameter o one period. Consequently it is important that the Fresnel number o the microlenses is not too small, otherwise the homogenization cannot be uniorm. O course, the ratio R between the angular extension (ield divergence o the light source) D source / CL o the light source and the diraction angle λ/p o the lens array is additionally limiting the homogenization, even i the Fresnel number o the lenses is larger. A spatially coherent or nearly spatially coherent light source cannot be homogenized suiciently with a Fly s eye condenser using periodic lens arrays. There is only one way to help in this case. By misadjusting the collimator lens CL in such a way that the extended light source is no longer exactly in its ocal plane, a spherical wave will be ormed by each point o the light source. Then, the intererence pattern in the ocal plane o Fourier lens or the dierent light source points will be o a higher requency and more complex so that a slightly better homogenization can be achieved. The ollowing Figures 11 and 12 show such a simulation which has to be compared with the upper Figures. O course, this trick is only allowed to some extent because it smears out the lat top proile. source CL Proc. o SPIE Vol

9 Fig. 11. Intensity distribution o one light source point i the collimator lens CL is axially misaligned so that spherical waves enter the lens arrays. Fig. 12. Intensity distribution or an extended light source or the case D source / CL = 1.5λ / p i the collimator lens CL is axially misaligned so that spherical waves enter the lens arrays. The spot patterns generated by grating diraction and intererence could be reduced by dierent measures. A larger microlens pitch p leads to a iner period Λ FP in the Fourier plane. A detuning o beam homogenizer components, e. g. a shit o the microlens arrays or the working plane out o its correct positions allows a reduction o the modulation o the lat top proile. These grating intererence eects can be demonstrated with a ly s eye condenser and an illumination with a collimated diode-laser beam at a wavelength o 670 nm. The results are shown in Fig. 13. The pitch p o the used microlens array is 0.3 mm. The results show a good agreement with theoretical calculations. The period o the spots is approximately 677 µm or a Fourier lens ocal length o 300 mm. Fig. 13. Measured intensity proile at ocal plane o the Fourier lens. 3.3 Diusers As discussed above, eects like grating intererence and speckles are well correlated with the coherence o a laser beam. A laser with many incoherent modes (larger M 2 ) will have lower contrast speckle than a single-mode coherent laser. A reduction o coherence related eects by using rotating or static diuser plates is possible or applications where light or a cw laser, or multiple pulses rom a pulsed laser, are integrated at the detector or object. The diusers can be abricated in used silica or wavelengths between ultraviolet and near inrared or silicon or the middle inrared. The scattering angle can be inluenced by processing parameters and is typical between 1 and 20. Figure 14 shows the angular light distribution o a structured used silica diuser in comparison to a typical ground glass diuser. Proc. o SPIE Vol

10 100,0 0, AngIe [ VSpMg D 4WDEpI Fig. 14. Angular distribution o ground glass and structured used silica, SEM pictures o diuser surace. The rotating diuser plate is usually placed in a separate telescope near the common ocus plane (Fig. 15). The diuser plate is rotating, thus the speckle pattern is changing temporally. By shiting the diuser position or detuning the telescope, the source divergence is changed. Time-averaged, the light emitting rom the rotating diuser is similar to an incoherent extended light source. Placing the diuser near the ocus o the beam, e. g. by using a beam expander in ront o the homogenizer, reduces the coherence by creating a new extended light source, whereas the size o the source is equal to the spot diameter on the rotating diuser. Unortunately rotating diusers do not work or pulsed lasers like Nd:YAG with nano- or picosecond pulses. For single pulse exposure dierent measures, such as stair case beam splitters and pulse stretchers are applied. Telescope 1 2 FP DFT Rotating Diusor Fig. 15. Schematic setup o a rotating diuser used with a microlens homogenizer. Experimental results o the application o a diuser within a microlens homogenizer in a ly s eye setup are shown in Figures 16 and 17. A diode laser with a wavelength o 670 nm is used. The pitch o the micro lens array is 250 µm. A Fourier lens with a ocal length o 40 mm generates a lat-top with dimensions o approx. 6.4 x 6.4 mm 2. For this coniguration the Fresnel number is approx. 15. It can be demonstrated that the grating eects can be smoothed with the application o a rotating diuser. With a good alignment the beam uniormity is in the range o ± 5 % and is better than ± 1% with an incoherent illumination. Proc. o SPIE Vol

11 Fig. 16. Intensity distribution distributions in the ocal plane o Fourier lens without the application o a rotating diuser. Fig. 17. Intensity distribution distributions in the ocal plane o Fourier lens : smoothing o intererence spikes by application o a rotating diuser. The 2-D diusers shown in Figure 14 are o very limited use or thin-line homogenizers. 1D random diusers consisting o a random pattern o arbitrary and statistically placed diusing elements are under investigation. These diusers are used to improve the uniormity o line homogenizers and laser light sheets. top view 4. MICROLENS BEAM HOMOGENIZER APPLICATIONS 4.1 Light Sheet Homogenizer Planar laser diagnostics means a spatial high-resolution measurement o the 2 dimensional concentration distribution o gases or luids, which are excited by pulsed lasers, or example Excimer lasers [3]. Thereore a thin laser light sheet is established within the media which has to be investigated. The luorescence signal o the tracing substance, which is excited by laser irradiation, is detected with a high resolution ccd-camera. A schematic setup is shown in Figure 18. homogenizer IlaserbeamI- IQ _ cylindrical cylindrical cylindrical image plane microlens array Fourier-lens ocusing lens side view I1a5&bmi - El Fig. 18. Setup o the light sheet homogenizer. The detected signal is directly proportional to the absolute concentration o the tracer substance and thereore to the concentration o the carrying luid. To investigate the absolute concentration out o single pictures, a homogeneous intensity distribution in the image plane is necessary. Indeed spatial inhomogeneity can be corrected ater measurement but this procedure is very time consuming. Temporal luctuations and hot-spots cannot be corrected. The homogenizer consists o two cylindrical microlens arrays with a pitch o 0.5 mm. The plano-convex Fourier lens generates a lat-top with a size o approximately 50 mm. The second beam direction is ocussed by a lens. Proc. o SPIE Vol

12 çtit (Arb. Units) (Arb. Units) Fig. 19. RMS image o the normalized luorescence signal: let: without the application o the beam-homogenizer. middle: with the application o the beam-homogenizer. Right: tracer-lif measurement o the mixing ield o two dierent turbulent lows (by courtesy o the research group o Lehrstuhl ür Technische Thermodynamik Friedrich-Alexander Universität Erlangen-Nürnberg, Germany). Figure 19 shows examples o camera pictures o a planar laser induced luorescence with and without a homogenizer. The intensity distribution without homogenizer shows stripes, which can be traced back to the instable resonator type o the Excimer laser (Excimer laser wavelength 248 nm, pulse energy 250 mj). The tracer-lif measurement o the mixing ield o two dierent turbulent lows is shown in the right camera picture o Figure Laser Micro Machining A homogenizer or illumination o a mask is used within an ArF-Excimer laser working station. Due to the short operation wavelength o 193 nm, the maximum pulse energy o 120 mj and the peak power o 7 MW, the homogenizer material is a high purity grade used silica with low absorption and degradation at this wavelength. Furthermore the telescope homogenizer setup with two dierent microlens arrays has to be used to reduce the energy density at the second microlens array. The typical raw beam proile o the Excimer laser with a Gaussian intensity distribution in the x- direction and super-gaussian in y-direction is shown in Figure 20. The dimension o the laser beam is approx. 5 x 14 mm with a beam divergence o 1 x 2 mrad. For conversion o the ultraviolet radiation into visible and detectable light a special luorescence plate is used. Fig. 20. Measured intensity distribution o Excimer laser raw beam. Fig. 21. Measured intensity distribution o homogenized beam. Fig. 22. SEM picture o laser ablation in borosilicate glass. We used two square microlens arrays with a pitch o 0.3 mm and a ocal length o 3.9 mm and 7 mm. The ocal length o the spherical Fourier lens is 175 mm. A square lat-top with dimensions o approx. 7.5 x 7.5 mm² and beam uniormity (measured with the norm ISO 13694) o better than ± 5% is realized (Fig. 21). The illuminated mask with a drilling diameter o 600 µm is projected with a demagniication actor o approx. 4 onto the glass substrate (borosilicate). The ablation which is shown in Figure 22 shows a very smooth ablation ground and sharp edges. Proc. o SPIE Vol

13 5. CONCLUSION We showed capabilities and limits o microlens arrays or shaping laser beams into homogeneous patterns like light sheets or lat-tops. Design rules which consider geometrical and physical optics were presented. We demonstrated a simple estimation o diraction inluence with calculating the Fresnel number o the system. Furthermore we presented a useul method or reducing intererence patterns with the application o a rotating random diuser. Examples o the usage o microlens beam homogenizer in planar laser diagnostic and Excimer laser micromachining were shown. REFERENCES 1. R. Völkel, H. P. Herzig, P. Nussbaum and R. Dändliker, Microlensarray imaging system or photolithography, Optical Engineering 45 (11), (1996). 2. U. Popp, Th. Neudecker, U. Engel, M. Geiger, Excimer Laser Texturing o Hard Coated Cold Forging Tools - Investigations on Tool Lie, Advanced Technology o Plasticity 2002, Vol. II., (2002). 3. S. Padler, F. Beyrau, M. Löler, and A. Leipertz, Application o a beam homogenizer to planar laser diagnostics, Optics Express 14, (2006). 4. J. A. Honagle and C. M. Jeerson, Design and Perormance o a Reractive Optical System that Converts a Gaussian to a Flattop Beam, Appl. Opt. 39, (2000). 5. T. Kajava, M. Kaivola, J. Turunen, P. Paakkoned, M. Kuittinen, P. Laakkonen and J. Simonen, Excimer laser beam shaping using diractive optics, Lasers and Electro-Optics Europe, Conerence Digest Conerence on Volume, Issue, 2000 Page(s): 1 pp. (2000). 6. I. Harder, M. Lano, N. Lindlein and J. Schwider, Homogenization and beam shaping with micro lens arrays, Photon n Management, Proceedings o SPIE, 5456, (2004). 7. K. Räntsch, L. Bertele, H. Sauer and A. Merz, Illuminating System, US-Patent , United States Patent Oice, A. Büttner and U. Zeitner, Wave optical analysis o light-emitting diode beam shaping using microlens arrays, Optical Engineering 41 (10), (2002). 9. Fred M. Dickey and Scott C. Holswade, Laser Beam Shaping: Theory and Techniques, Publisher: Marcel Dekker, (2000). 10. B. Besold and N. Lindlein, Fractional Talbot eect or periodic microlens arrays, Optical Engineering 36 (4), (1997). 11. F. Wippermann, U.D. Zeitner, P. Dannberg, A. Bräuer and S. Sinzinger, Beam homogenizers based on chirped microlens arrays, Optics Express 15, (2007). Proc. o SPIE Vol