Advanced Laser Micromachining Processes for

Invited Paper

Advanced Laser Micromachining Processes for MEMS and Optical Applications Andrew S. Holmes1, James E.A. Pedder1 and Karl L. Boehlen2 1

Department of Electrical & Electronic Engineering, Imperial College London Exhibition Road, London SW7 2AZ, UK 2 Exitech Ltd, Oxford Industrial Park, Yarnton, Oxford OX5 1QU, UK E-mail: [email protected] ABSTRACT

Laser micromachining has great potential as a MEMS (micro-electro-mechanical systems) fabrication technique because of its materials flexibility and 3D capabilities. The machining of deep polymer structures with complex, well-defined surface profiles is particularly relevant to microfluidics and micro-optics, and in this paper we review recent work on the use of projection ablation methods to fabricate structures and devices aimed at these application areas. In particular we focus on two excimer laser micromachining techniques that are capable of both 3D structuring and large-area machining: synchronous image scanning (SIS) and workpiece dragging with half-tone masks. The methods used in mask design are reviewed, and experimental results are presented for test structures fabricated in polycarbonate. Both techniques are shown to be capable of producing accurately dimensioned structures that are significantly deeper than the focal depth of the projection optics and virtually free from fabrication artifacts such as the steps normally associated with multiple-mask processes. Keywords: laser ablation, excimer laser, micromachining, SIS, half-tone, MEMS, microstructures

1. INTRODUCTION Laser micromachining by ablation has proved to be a useful technique in many fields of advanced manufacturing [1]. In addition to established industrial applications, such as drilling of ink-jet nozzles and printed circuit board via holes, new applications are emerging in the areas of optics and microelectromechanical systems (MEMS) that are ever more challenging in terms of geometry, machining accuracy, surface finish and/or processing speed. In the MEMS domain, these applications are mainly in areas where 3D processing of polymers and glasses are required, such as microfluidics [2] and RF MEMS [3]. The most important optical application currently is the production of plastic microlens arrays for large area displays [4]. This requires rapid processing of highly uniform, optical quality components over extremely large areas (up to 1.5 x 2 m2). Laser machining of complex surfaces may be carried out either in direct-write mode, where a focused laser spot is scanned over the workpiece, or in mask projection mode, where an imaging lens is used to project a mask pattern. Pulsed solid-state lasers, which tend to combine good beam quality and high pulse repetition rate with relatively low pulse energy, are normally used in direct-write mode. Mask projection tends to be reserved for excimer lasers, where the high peak power allows parallel processing over large areas. In this paper we consider two excimer laser micromachining techniques that allow both 3D structuring and large area processing: synchronized image scanning (SIS) and half-tone mask projection. The methods commonly used for process and mask design are briefly reviewed, and experimental results are presented for polycarbonate cylindrical microlens arrays produced by both techniques.

High-Power Laser Ablation VI, edited by Claude R. Phipps, Proc. of SPIE Vol. 6261, 62611E, (2006) · 0277-786X/06/$15 · doi: 10.1117/12.682929

Proc. of SPIE Vol. 6261 62611E-1

2. LARGE AREA AND 3D EXCIMER LASER MICROMACHINING METHODS Projection ablation systems typically include beam forming optics designed to produce uniform illumination at the mask. With a standard binary mask (zero or 100% transmission), each laser pulse removes material to the same depth in all exposed regions of the workpiece. Several methods can be used to produce features with variable surface height in such systems. One approach is to vary the shape and/or size of the mask aperture during machining. This may be done using a motorized aperture or by indexed mask projection (IMP) which uses a sequence of different static mask apertures [5]. An alternative approach is to move or ‘drag’ the workpiece under a static mask while firing the laser. Dragging the workpiece beneath a single aperture will in general produce a channel with a cross-sectional profile that depends on the aperture shape. In particular, if the laser fluence and firing pitch (distance traveled by workpiece between laser pulses) are held constant, the local channel depth will be proportional to the length of the mask aperture in the direction of travel. Workpiece dragging is well suited to producing periodic structures over large areas, and has been widely used in the fabrication of structures such as gratings, microchannel arrays and anti-reflection surfaces [6]. Synchronised image scanning [4] is a variant of the above techniques in which the workpiece is dragged beneath a linear array of IMP mask apertures, all of which are illuminated by the laser beam. The laser firing pitch is set equal to the spacing between mask apertures so that each site on the workpiece is exposed to a single pulse with each of the apertures in turn. This concept is illustrated in Fig. 1. Note that for simplicity the projection optics is omitted from this figure, and the mask is shown in proximity to the substrate. SIS offers higher throughput than indexed mask projection by virtue of parallel processing. Also, because each site receives only one pulse with each SIS aperture, the stepping in the surface that arises from using a discrete set of masks can be minimized. The above techniques achieve variable depth by exposing different parts of the workpiece to different numbers of laser pulses. The fluence in the exposed regions is usually independent of position. An alternative option, also illustrated in Fig. 1, is to have the same number of pulses everywhere, but vary the local fluence using a half-tone mask. A half-tone mask comprises an array of pixels, each with a transmissive aperture of well-defined area on an opaque background. Provided the pixel structure is not resolved by the projection optics, such a mask will behave simply as an attenuator, with the local transmission being proportional to the square of the transmissive area in each pixel. In this way a conventional binary mask can be made to behave like a continuously variable or grey-scale mask.

Fig. 1. Fabrication of 3D structures by half-tone ablation (left) and SIS (right).


Half-tone masking has been used previously with projection ablation to produce diffractive optical elements [7], microfluidic channels [8] and other microstructures [9]. It is an attractive option when producing isolated 3D structures because it is less costly in terms of mask real estate than SIS. However this advantage is lost in large-area applications where parallel processing becomes imperative, requiring an array of half-tone apertures. Other factors also need to be taken into account when choosing between the different approaches, such as process efficiency and surface finish. These aspects are discussed in the following sections.

3. PROCESS AND MASK DESIGN FOR SIS AND HALF-TONE ABLATION For any given combination of laser source and material, the first step in designing a laser micromachining process is to decide on the operating fluence (pulse energy per unit area incident on the workpiece). In general the material removal rate, the morphology of the machined surface, and the nature and amount of ablation debris will all depend on the fluence, and these aspects can only be explored by experiment. For large area applications, the process efficiency in terms of volume of material removed per joule of incident laser energy is also an important consideration as this sets an upper limit on the throughput for a given laser source. Fig. 2 shows the measured variation with fluence of the ablation depth per pulse, and the corresponding process efficiency, for polycarbonate machined at 248 nm wavelength. The process efficiency peaks at around 250 mJ/cm2, corresponding to an ablation depth per pulse of around 120 nm. This defines the optimum fluence in terms of throughput for large-area applications, although in practice higher fluences are often used for small-area applications to increase the machining speed of isolated structures. 0.06 Normalised removal rate (mm /J)

0.3

3

Ablation depth per pulse (microns)

0.35

0.25 0.2 0.15 0.1 0.05 0 20

50

100 200 500 Fluence (mJ/sq.cm)

1000

2000

0.05 0.04 0.03 0.02 0.01 0 20

50

(a)

100 200 500 Fluence (mJ/sq.cm)

1000

2000

(b)

Fig. 2. (a) Ablation rate measurements for polycarbonate at 248 nm wavelength. Solid line is cubic polynomial fit to data; dashed line is best-fit “Beer’s law” ablation curve. (b) Process efficiency curve for polycarbonate at 248 nm wavelength, based on polynomial fit of Fig. 2a. 3.1 SIS mask design When designing an SIS mask, a CAD drawing of the target structure is divided into a number of horizontal slices, as shown in Fig. 1, with the outline of each slice being used to define an SIS mask aperture. This process is similar to that used in stereolithography, and CAD tools have been developed to automate the process [4]. Ideally, the thickness of each slice is chosen to match the ablation depth per pulse at the chosen operating fluence so that the machining process can be completed in a single pass. However, if the structures are particularly deep or on too large a pitch, it may not be possible to fit the required number of apertures within the area illuminated by the laser beam, in which case it is necessary to machine the structures in several passes. In general multiple-pass processes are undesirable because of issues with ablation debris. Most importantly, debris remaining from earlier passes can lead to micro-masking effects that degrade the final surface quality. For the same reason, SIS is usually carried out with a linear array of apertures rather than a 2D array. In a single-pass process with a linear array, ablation debris can be blown using an air-jet onto the part of the surface that has already been machined, and then cleaned away at the end of the machining process.


3.2 Half-tone mask design Mask design for half-tone ablation involves deriving a mask transmission function T(x,y) from a depth map h(x,y) of the target structure. In the simplest analysis, where perfect geometrical imaging is assumed and secondary effects such as the reduction of ablation rate on inclined surfaces are neglected, the local ablation depth after N pulses will be Nf (F0 T (mx , my) ) where f(.) is the material ablation curve at normal incidence, F0 is the fluence at 100% transmission (the operating fluence), and m is the reduction ratio of the imaging lens. In this approximation the mask transmission function is obtained simply as: 1 −1 ⎛ h( x , y ) ⎞ T (mx , my ) = f ⎜ ⎟ F0 ⎝ N ⎠ A number of factors need to be taken into account when choosing the values of F0 and N. Firstly, there will be minimum and maximum achievable transmission levels that depend on the resolution of the laser system and the technology used to manufacture the mask. In order for the pixel structure of the half-tone pattern not to be resolved, the pitch p of the half-tone pixels must be less than mλ/(NA + NAi), where λ is the laser wavelength, NA is the numerical aperture of the projection lens, and NAi is the numerical aperture of the illumination [10]. For λ = 248 nm, and with typical values of m = 5, NA = 0.15 and NAi = 0.07, this gives p < 5.6 µm. At the same time, the mask manufacturing process will have a critical dimension (CD) or smallest possible feature, opaque or transparent, that can be produced. The CD and the pitch together define the range of transmission values that can be realized using a particular pixel design. For example, for a mask as in Fig. 3a , with p = 5 µm and CD = 1 µm, the highest transmission is Tmax = 0.9216 (corresponding to w = 1 µm), while the lowest transmission is Tmin = 0.1296 (w = 4 µm). Lower transmission levels could be achieved by allowing other pixel geometries, for example square windows on an opaque background. In practice the value of Tmin may be set by the material ablation characteristics rather than the mask technology. In general there is a minimum useable fluence, Fmin, below which ablation debris is not ejected effectively and hence accumulates on the machined surface. This minimum fluence requirement translates to a minimum machined depth in the final structure, so that in general it is not possible to produce structures where the depth varies continuously down to zero. This is an important limitation on the half-tone ablation process. 1

w = CD + ng CD = critical dimension g = mask-writing grid n = half-tone index 0 ≤ n ≤ (p – 2CD)/g

Normalised ablation depth

T = [1 – (w/p)2]2 with

0.9 0.8 0.7

10 nm grid

0.6 0.5 25 nm grid 0.4

0.2

400

0.1 0

200

F0 =200 mJ/sq.cm

0.3

0

(a)

50

600

100 150 200 Half-tone index

400

250

600

300

(b)

Fig. 3. (a) Assumed geometry for half-tone pixels, and method for calculating transmission as a function of half-tone index. (b) Variation of normalized ablation depth in polycarbonate with half-tone index for different operating fluences and mask writing grids, assuming p = 5 µm, CD = 1 µm and Fmin = 100 mJ/cm2. The number of discrete transmission levels available with a half-tone mask is determined by the snap-grid g used in the mask manufacturing process. For example, with p = 5 µm, CD = 1 µm and g = 25 nm, the index n in Fig. 3a can take any value from 0 to 120, giving 121 discrete transmission levels between T = 0.1296 and T = 0.9216. However, because of the minimum fluence requirement, not all of these levels will be available at lower F0 values. Fig. 3b illustrates this point for the case Fmin = 100 mJ/cm2. With a writing grid of 25 nm, 114 levels are available at an operating fluence of


600 mJ/cm2, while only 69 are available at F0 = 200 mJ/cm2. The vertical axis in Fig. 3b shows the ablated depth f (F0 T ) in polycarbonate as a function of the half-tone index, normalised to the depth f (F0 ) that would be achieved at 100% mask transmission. It can be seen that raising the operating fluence increases the ratio of the maximum to minimum machined depths, while reducing the mask-writing grid leads to an increase in the number of transmission levels. An important consequence of the minimum fluence requirement is that it can significantly reduce the process efficiency of half-tone machining. In order to produce a structure in which the difference between the maximum and minimum machined depths is D, NHT pulses are required where: D N HT = f ( F0 T max) − f ( F0 T min) whereas the number of pulses required for the corresponding SIS process would be simply N SIS = D / f ( F0 ) . The efficiency of the half-tone process is therefore lower by a factor:

η rel =

N SIS f ( F0 T max) − f ( F0 Tmin ) = N HT f ( F0 )

For any given operating fluence, ηrel is simply the difference between the maximum and minimum normalized ablation depths in Fig. 3b. The variation of this parameter with fluence for polycarbonate is shown in Fig. 4, assuming the same mask parameters and Fmin value as in Fig. 3b. The half-tone process is less efficient for two reasons: firstly the maximum half-tone transmission is less than 100%, and secondly extra material is removed that does not form part of the final structure. The latter effect becomes less pronounced as the fluence increases, so the half-tone process becomes relatively more efficient at high fluences. Note that the abrupt change of gradient at around 770 mJ/cm2 arises because Tmin is determined by the minimum fluence requirement at lower fluences and by the mask manufacture process at higher fluence. Efficiency of half-tone relative to SIS

0.9 0.85 0.8 0.75 0.7 0.65 0.6 0.55 0.5 200

400

600 Fluence (mJ/sq.cm)

800

1000

Fig. 4. Efficiency of half-tone process relative to SIS as a function of operating fluence for polycarbonate, assuming the same mask parameters and minimum fluence as in Fig. 3b. 3.3 Additional considerations The mask design methods outlined above tend to become less accurate in deeper structures as a result of diffraction, illumination-related effects (in particular separation of the beams from the different homogenizer sourcelets), and the reduction of the ablation rate on inclined surfaces. For structures such as microlenses where the surface height is relatively slowly varying, the latter effect appears to be most important. The issue here is that the ablation depth parallel to the optic axis is assumed independent of surface gradient, whereas in fact it decreases on steeply inclined surfaces. For example, Fig. 5 shows the measured angular variation of the ablation rate in polycarbonate at 248 nm wavelength. It


Ablation depth in direction of beam (microns)

can be seen that the angle-independent approximation (indicated by dashed lines) remains good up to a certain limiting slope, beyond which the ablation rate falls off. 0.35 F = 925 mJ/sq.cm

0.3 0.25 0.2

F = 375

0.15 0.1 0.05 0

F = 75

0

20

40 60 Surface slope (degrees)

80

Fig. 5. Measured variation of the ablation depth δz parallel to the optic axis as a function of surface slope θ for polycarbonate at 248 nm wavelength. If neglected during the mask design stage, the angular variation of the ablation rate can lead to significant errors in the depth profiles of deeper structures. Furthermore, for any given operating fluence there will be a limiting surface gradient that cannot be exceeded, and this can lead to unwanted artifacts such as seams of material at the interfaces between adjacent regions that have been machined at different times. Such effects can only be eliminated reliably by taking account of the angular dependence of the ablation rate during mask design, which generally requires numerical modelling. We have been developing simulation tools that can address this problem [11], and have demonstrated a significant improvement in profile control for structures with depths of up to several hundred microns [12]. In addition to the above issues which relate to the overall profile accuracy, there are other features of the SIS and halftone processes relating to surface morphology that cannot currently be modeled. For example, the running order of the SIS apertures can have a marked effect on the roughness of the final surface. In particular, the stepping is reduced if the boundary of each aperture lies inside the one that follows it. In qualitative terms this is because the steps caused by the hard edges on each aperture become rounded by subsequent exposures. The net effect is that the surface steps in the final structure can be significantly smaller than might be expected based on the ablation depth per pulse.

4. PROCESSING EXAMPLE – MICROLENS ARRAYS To illustrate the basic capabilities of half-tone ablation and SIS in terms of profile accuracy and surface finish, we have fabricated test structures in the form of polycarbonate microlens arrays. Convex cylindrical lenses were fabricated with a width of 650 µm and a radius of curvature of 1 mm, corresponding to a maximum machined depth of around 50 µm and a maximum surface slope of 18°. Experiments were carried out with an Exitech M8000 laser micromachining workstation designed specifically for SIS. The workstation was equipped with a KrF excimer laser, a 5X 0.13NA projection lens, and beamforming optics configured to produce a 70 x 6 mm2 rectangular illuminated region at the mask plane. An operating fluence of 250 mJ/cm2 was used for the SIS process, corresponding to the peak of the process efficiency curve. At this fluence a total of 544 pulses was required to achieve the required maximum depth. These were delivered in two passes, using an SIS mask with 272 apertures on a pitch of 250 µm (corresponding to a laser firing pitch of 50 µm). A higher fluence of 890 mJ/cm2 was used for the half-tone process to minimize the effects of the minimum fluence requirement. In this case the desired cylindrical lens structure was produced by dragging a 4.75 mm-long rectangular half-tone mask aperture with the required transverse transmission profile. A total of 288 pulses was


delivered in a single pass, requiring a laser firing pitch of 3.3 µm. Fig. 6 shows an SEM image of the structure produced by the half-tone process.

Fig. 6. SEM image of a microlens array formed by workpiece dragging with half-tone mask. Fig. 7 compares stylus profilometer measurements made on individual half-tone and SIS lenses (solid lines in left and right graphs respectively) with the expected profiles based on the mask designs (dashed lines). Good overall profile accuracy is achieved by both techniques, although in both cases the agreement is closer on one side of the lens than on the other due to a slight asymmetry in the actual profiles. The exact cause of this asymmetry is unknown, although it is thought to have arisen from non-uniformity in the illumination. Note that the lower fluence used in the SIS process has resulted in a seam wall on either side of the SIS lens. 70

60

60

50 Depth (microns)

Depth (microns)

50 40 30 20 10

30 20 10 0

0 -10 -400

40

-200

0 Position (microns)

200

400

-10 -400

-200

0 Position (microns)

200

400

Fig. 7. Transverse surface profiles of lenses formed by workpiece dragging with a half-tone mask (left) and by SIS (right). Solid lines are measured profiles; dashed lines are expected profiles based on mask designs. Surface analysis of the lenses was carried out by digital holographic microscopy (DHM), an interferometric technique that can provide surface profile information with micron and nanometer resolutions in the lateral and vertical directions respectively [13]. Fig. 8a shows a linescan taken along the long axis of the half-tone lens near the point of minimum machined depth. A strong periodic component can be seen with a period equal to the laser firing pitch of 3.3 microns. It was found that this could be eliminated by flood exposing the surface with a small umber of laser pulses at a much lower fluence. This ‘polishing’ process was carried out by workpiece dragging using a simple rectangular aperture with graded half-tone transmission profiles on its leading and trailing edges. Fig. 8b shows the surface profile after polishing, which has an Ra of only 8.4 nm. The improvement in surface finish afforded by polishing is also evident in the phase plots on the right of Fig. 8, which show a significant reduction in the noise on the fringes after polishing.


Surface height (microns)

Fig. 9 shows DHM measurements at the equivalent position on a lens machined by SIS. Again there is a periodic component in the linescan which corresponds to the laser firing pitch, but even without polishing the Ra is comparable to that of the polished half-tone lens. This comparison is somewhat unfair to the half-tone process because the crown of the SIS lens receives only a very small number of laser pulses, and indeed Ra values as high as 30 nm were measured elsewhere on the surface, compared to a maximum Ra of around 20 nm on the as-machined half-tone lens. More detailed surface roughness measurements are in progress and will be reported in a later paper.

Ra = 17.8 nm

0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08

0

10

20

30

40 50 60 Position (microns)

70

80

90

80

90

(a)


0.06 Ra = 8.4 nm

0.04 0.02 0 -0.02 -0.04 -0.06

0

10

20

30


70

(b)


Fig. 8. DHM measurements showing longitudinal surface profiles and wrapped phase plots for a microlens formed by workpiece dragging with a half-tone mask: (a) as machined and (b) after laser polishing.

Ra = 8.5 nm

0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08

0

10

20

30


70

80

90

Fig. 9. DHM measurements for a microlens formed by SIS.


5. CONCLUSIONS Several processes are available for large-area excimer laser micromachining of complex surfaces by projection ablation. In this paper we have focused on two techniques: SIS and workpiece dragging with half-tone masks. Based on results to date it would appear that SIS is the most appropriate technique for large area machining because of its higher process efficiency. Excellent surface finish has been achieved in polycarbonate, and it is expected that further improvements could be made by implementing SIS masks with graded transmission profiles at the edges. Half-tone masking remains an attractive option for small area applications because of its more efficient use of real estate on the mask.

6. ACKNOWLEDGEMENTS This work was funded by the UK Engineering and Physical Sciences Research Council (EPSRC). Experiments were carried out using laser facilities at Exitech Ltd. One of the authors, J.E.A. Pedder, is supported jointly by the EPSRC and Exitech Ltd. The DHM measurements were carried out at favourable rates by Lyncée Tec SA, Lausanne, Switzerland.

7. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

M.C. Gower, “Industrial applications of laser micro-machining”, Optics Express, vol. 7, no. 2, pp. 56-67, 2000. J.P.H. Burt, A.D. Goater, C.J. Hayden, J.A. Tame, “Laser micromachining of biofactory-on-a-chip devices”, Proc. SPIE vol. 4637, pp. 305-317, 2002. H.J. Booth, C.E. Abbott, R.M. Allott, K.L. Boehlen, J. Fieret, J. Greuters, P. Trimble, J.E.A. Pedder, “Laser micromachining techniques for industrial MEMS applications”, Proc. SPIE vol. 5713, pp. 190-199, 2005. K.L. Boehlen, I.B. Stassen-Boehlen, “Laser micromachining of high-density optical structures on large substrates”, Proc. SPIE vol. 5339, pp. 118-126, 2004. N.H. Rizvi, P.T. Rumsby, M.C. Gower, “New developments and applications in the production of 3D microstructures by laser micro-machining”, Proc. SPIE vol. 3898, pp. 240-249, 1999. N.H. Rizvi, D. Milne, P.T. Rumsby, M.C. Gower, “Laser micromachining: new developments and applications”, Proc. SPIE vol. 3933, pp. 261-271, 2000. F. Quentel, J. Fieret, A.S. Holmes, S. Paineau, “Multilevel diffractive optical element manufacture by excimer laser ablation using halftone masks”, Proc. SPIE vol. 4274, pp. 420-431, 2001. C.J. Hayden, “Three-dimensional excimer laser micromachining using greyscale masks”, J. Micromech. Microeng., vol. 13, pp. 599-603, 2003. A. Braun, K. Zimmer, “Diffractive gray scale masks for excimer laser ablation”, Appl. Surf. Sci., vol. 186 (1-4), pp. 200-205, 2002. A.S. Holmes, “Excimer laser micromachining with half-tone masks for the fabrication of 3-D microstructures”, IEE Proc. Sci. Meas. Technol., vol. 151, no. 2, pp. 85-92, 2004. A.S. Holmes, A.I. Onischenko, D.S. George, J.E.A. Pedder, “Modelling of solid-state and excimer laser processes for 3D micromachining”, Proc. SPIE vol. 5713, pp. 180-189, 2005. J.E.A. Pedder, A.S. Holmes, “A study of angular dependence in the ablation rate of polymers by nanosecond pulses”, Proc. SPIE vol. 6106, pp. 286-294, 2006. F. Charrière, J. Kühn, T. Colomb, F. Montfort, E. Cuche, Y. Emery, K. Weible, P. Marquet, C. Depeursinge, “Characterization of microlenses by digital holographic microscopy”, Appl. Opt., vol. 45, no. 5, pp. 829-835, 2006.
