Femtosecond technology for precision manufacturing: Fundamental and technical aspects

Transcription

1 RIKEN Review No. 50 (January, 2003): Focused on Laser Precision Microfabrication (LPM 2002) Femtosecond technology for precision manufacturing: Fundamental and technical aspects Friedrich Dausinger # Inst. f. Strahlwerkzeuge (IFSW), University of Stuttgart, Germany During the past few years research groups demonstrated the potential of femtosecond pulses for ultra-precise machining. Initial investigations of industrial applications soon showed, however, several unexpected quality problems such as formation of recast, ripples and irregular hole shapes even in the femtosecond pulse regime. This contribution will present some progress in fundamental understanding of the ultrafast ablation process. On the basis of this knowledge technical means have been developed allowing to achieve an unprecedented level of accuracy at acceptable expenses. Acomparison between pico- and femtosecond regime will be presented. 1. Introduction Soon after the invention of the laser it was recognized that the focused beam can be used as a tool for material removal. The number of razor blades a laser pulse can penetrate was taken as a measure for its pulse energy in Gillette units. With industrial drilling lasers offering pulse energy values of some Joules metal plates thicker than one millimeter can be drilled in less than one millisecond. A lack of geometrical accuracy and other quality problems resulting from recast material strongly limited the number of industrial applications, however. In mechanical metal cutting techniques it is a common procedure to increase accuracy by reducing the volume of the chip. An analogous approach in laser material removal results in a breaking up the ablation process in a multitude of steps, see Fig. 1. This leads from single pulse drilling to the well known techniques percussion and trepanning, the latter being a percussion drilling process followed by a cutting procedure. A further step to increased accuracy is the so-called helical drilling method in which the ablation front penetrates the work piece on a helical path. 1) In the case of material removal with lasers the chip volume is determined by penetration depth and spot diameter. The penetration to which a laser pulse interacts with material is determined by optical and thermal penetration: l = l α + l th. (1) In dielectrics optical penetration depth dominates over the thermal one, generally, and is a strong function of the wave length. 1) In metals, on the other hand, the optical penetration depth turns out to be smaller than a tenth of the wavelength and can be neglected in most cases. The penetration depth is then determined by the thermal penetration depth which depends on thermal diffusivity and pulse duration τ H: l th =2 κ τ H for τ H > 10 ps. (2) From Eq. (2) it appears straightforward to reduce the pulse duration in order to increase accuracy. It has to be mentioned, however, that this relationship is based on classical heat conduction theory which is no longer valid for ultrashort time scale, see next paragraph. At this point it has to be made clear, that lα and lth are qualitative values, only. The quantitative values for the depth to which material is melted or vaporized depend additionally on the energy density transmitted to it. The same is true for the lateral dimension of the chip. In rare cases only the focal diameter df coincides accidentally with the diameter of the ablated zone which for a Gaussian beam is determined by the threshold for ablation Hs and the energy density in relationship to it: d Abl = d f r 1 2 ln H H S (3) Fig. 1. Process strategies to achieve high accuracy in laser drilling. 1) # address: dausinger@ifsw.uni-stuttgart.de The edges of the beam with energy density below the threshold for material removal may act on the work piece in an unwanted way by causing thermal damage. A steep limitation of the beam is desirable for a treatment with high quality, therefore. On the other hand, the sharpness of the laser beam may be affected by laser induced plasmas, be they a material vapor plasma near to the ablation front or a gas breakdown in front of the work piece or in the drilled capillary. Tönshoff et al. 2) discussed an optical effect on the laser beam occurring near to its focus in front of the work piece which they explained as filamentation. This resulted in a pronounced ripple structure on the wall of a boring which could be avoided 77

2 Fig. 2. Material removal calculated for aluminum with piston model. 3,4) or at least weakened by drilling in vacuum or helium atmosphere. In any case a radial widening of the interaction zone increasing with laser intensity is to be expected. So far it can be summarized that from a reduction of the laser parameters pulse duration, focal diameter and pulse energy a decrease of the chip volume can be expected which can be used to increase the accuracy of a removal process, see Fig. 1. On the other hand, the phase composition of the removed material strongly depends on energy density. Figure 2 showing the results of a calculation 3) based on von Allmen s piston model 4) makes clear, that with increasing intensity the share of vaporized material strongly increases at the expense of molten material. From that point of view higher energy density appears favorable for achieving high accuracy, a clear contradiction to the previous finding. 2. Fundamental understanding 2.1 Thermal effects Often the advantage of ultrashort laser pulses is asserted by comparing drilling results in comparison with those achieved with nanosecond pulses. The nanosecond holes exhibit a large amount of recast material which is not visible in the femtosecond case. 2) On the contrary, initial trials in our group revealed an unexpected large amount of recast material in drilling and surface structuring of metals, which decreases with pulse duration but did not reach a zero value even at a pulse duration of 120 fs, see Fig. 3. This experimental finding which was checked for possible errors like long underground pulses is supported by calculations with an analytical model. 5) Based on von Allmen s stationary piston model 4) the new development now allows to investigate transient processes even at ultrashort time scale. At a pulse duration shorter than about 10 ps the classical heat conduction theory based on the assumption that a material can be characterized by one temperature, only, is no longer valid. In that case the time for electron-lattice interaction which is in the order of several picoseconds has to be taken into account. During laser pulses shorter than this relaxation time the laser energy deposited primarily in the electronic system can not be transferred to the lattice. The consequence of this is, that the melt depth produced by a laser pulse approaches a min- Fig. 3. Burr height of shallow structures in steel. The number of scans was selected to obtain an average ablation depth of 40 μm. imum value of several tenths of a micrometer, see Fig. 4, 5) instead of vanishing completely when the pulse duration is decreased. It is worth to note that the maximum melt depth is achieved long after the end of the laser pulse. The time during which the material remains molten is in the order of several tens of nanosecond, the evaporation stops after a few nanoseconds, see Fig. 4. That means that for ultrashort pulses below typically 10 ps the characteristic process times are nearly independent of the pulse duration and still in the nanosecond regime! The pressure built up by vaporization accelerates the melt layer to velocity values in the range of 100 m/s. During a typical time span in which a material stays molten of 10 ns it moves about 1 µm, only. For a micro-machining process one should attempt to avoid accumulation of these submicron layers. In the model used for the above considerations ablation takes place by vaporization. In the context of ultrashort pulses one may speculate that this is no longer true. The intense energy transfer to the electrons might cause their separation from the bulk material leaving behind ions which repel each other heavily and leave the bulk material. This so-called Coulomb explosion is a non-thermal ablation mechanism, from which melt-free material removal could be expected. Stoian et al. 6) Fig. 4. Vaporization and solidification times as well as maximum melt thickness for the ablation of aluminum with different pulse durations (H =10J/cm 2 ). 78

3 Fig. 6. Distortion of original beam (dotted circle) 60 cm behind focus in air at ultrashort pulse duration (120 fs). Left: 80 J/cm 2, right: 260 J/cm 2. Fig. 5. Temporal evolution of the vapor plume after irradiation of an aluminum target with a 500 fs pulse ( = 800 nm, Q = 500 μj, d f 18 μm, H 200 J/cm 2 ). investigated femtosecond ablation of dielectrics and metals with time of flight measurements. They pointed out that in dielectrics the material was partly removed by Coulomb explosion, partly thermally. In metals no hint for this effect was detected, on the other hand. From their finding one can conclude that in metals ablation even with femtosecond pulses is essentially of thermal nature. Figure 5 shows the effect of a single femtosecond pulse long after its end. The pictures were obtained with shadowgraphy by illuminating the atmosphere in front of the work piece with a probe laser wave length which is resonantly absorbed in aluminum vapor. Obviously the metal vapor plume develops at a time scale six orders of magnitude longer than the laser pulse. This means, that even if the ablation process would be non-thermal, it is followed by a long lasting expansion of hot material which certainly affects the work piece thermally. 2.2 Optical effects One of the advantages expected from the use of ultra-short pulses is the absence of a laser induced plasma affecting beam propagation which assumably can not develop during the ultra-short pulse duration. Figure 6 makes clear, however, that also with femtoseconds a strong deformation of the laser beam can be observed. The picture was taken from a screen positioned 60 cm behind the focus in air. Around the original beam marked in the left picture a colored ring structure is observed which increases in intensity and diameter when the laser energy density is raised. A more detailed description of this phenomenon which is assumed to be caused by cold atmospheric plasma is given in; 7) in this context only ist influence of it on the energy distribution will be considered. Figure 7 reveals, that a quite large amount of the beam energy is scattered into the ring structure. At 125 fs less than 50% are left in the diameter of the original beam. At an Fig. 7. Energy distribution atmospheric plasma. Shown is the share of scattered light ( = 800 nm, f p =1kHz,d f 18 µm, in air). energy density of 65 J/cm 2, the percentage of scattered energy decreases with the pulse duration reaching zero slightly above 5 ps. At 260 J/cm 2 the share of scattered energy remains above 50% in the whole investigated range of pulse duration. Figure 8 shows the total transmission through the area near the focus where the scattering effect occurs. At the shortest pulse duration 125 fs no loss could be detected. In contrast to this more than 20% of energy loss are observed near 5 ps. It can not be expected that such a strong deformation of the beam remains without consequences for the shape of a hole to be drilled. The effect on a steel plate treated with a series Fig. 8. Transmission through atmospheric plasma ( = 800 nm, f p =1kHz,d f 18 μm, in air). 79

4 of drilling procedures with increasing number of pulses can be clearly seen in Fig. 9. Around a central hole with largest depth, side holes are drilled by the scattered part of the beam. At high enough pulse numbers ablation by scattered light leads to full penetration, as well, causing an overall widening of the hole. From Figs. 7 and 8 one may conclude, that the scattering effect can be weakened or even avoided by choosing a large enough pulse duration of some picoseconds and by decreasing the energy density. The latter measure has an unwanted side effect, however, as can be seen in Fig. 10 the volume ablation rate strongly increases with energy density. The choice of a low energy density may slow down the drilling process such that it is no longer economically feasible. Another way to avoid scattering is to reduce pressure, see Fig. 11. While at atmospheric pressure a strong widening is observed above 10 J/cm 2, the hole diameter at reduced pressure increases only slightly following Eq. (3) for a Gaussian beam. As soon as the hole widening is avoided, the linear drilling rate steadily increases with the energy density. On the contrary, a saturation is observed when widening happens at atmospheric pressure. As can be observed from Fig. 10, pressure does not affect at the volume ablation rate at the used short pulse duration of 125 fs. Obviously the scattering effect is not energy absorbing at this pulse duration, an observation which is in agreement with the finding described in Fig. 8. Fig. 11. Effect of pressure on drilling rate and diameter depending on energy density (Steel, 30 μm, fi H =110fs, =800nm,f p = 1kHz, d f 18 μm, in air). Besides widening a further optical impact on hole quality is Fig. 12. Ripple formation in helical drilling with linearly polarized ultrashort pulsed laser (fi H = 120 fs, =800nm,f p =1kHz, d f 18 μm, H =310J/cm 2,inair,steel,0.5mm). Fig. 9. Holes drilled in steel with varying pulses of distorted beam (fi H =125fs, =800nm,f p =1kHz,d f 18 μm, H 330 J/cm 2, in air). displayed in Fig. 12, namely the formation of ripples. The start of the ripples at the entrance of the hole is probably caused by the scattering effect discussed above. Their propagation through the hole depends on the orientation of the beam polarization relative to the wall. At p-polarization, the ripples disappear after a short distance. At s-polarization, on the contrary, they propagate through the whole boring. A recipe to suppress ripples is to maintain p-polarization around the whole circumference, see next chapter. 3. Technical consequences Based on the fundamental understanding of material removal with ultra-short pulses technical means have been developed in order to increase efficiency and quality. Fig. 10. Volume ablation rate (fi H = 110 fs, = 800 nm, d f = 18 μm, steel). 3.1 Avoiding of burr formation From the discussion of the previous chapter one can conclude that the burr observed at the edges of ablated structures, see Fig. 3, is produced by accumulation of thin layers of melt accelerated by vapor pressure to the side. A simple means following from this is to reduce acceleration by reducing the energy density to a value slightly above the threshold for ablation. A confirmation of this conclusion is given in. 8) The disadvantage is a very low ablation volume per pulse which 80

5 raises the call for lasers with very high repetition rate. 3.2 Helical drilling An energy density slightly above the ablation threshold is effective only for very shallow structures. The minimal required energy density strongly increases with hole depth. 1) For high aspect ratio drilling the helical drilling procedure has turned out to be useful also in the ultra-short regime, already. 9) In the light of the theoretical findings described above an explanation for the reduction of recast observed when using this method can be found: In helical drilling the melt can move laterally (and is then ablated by following pulses) and does not need to be moved axially on the long way to the hole entrance as would be the case in percussion drilling. An appropriate optical setup, e.g. the one described in 9) allows not only to choose the helix diameter but also the inclination angle of the beam. At perpendicular incidence (inclination angle γ = 0) more than 10 s are needed to achieve a cylindrical hole with the used set of parameters, see Fig. 13. For the shortest pulse duration 120 fs, cylindricity can not be reached under these conditions. If an inclination angle of 4 is chosen, on the other hand, cylindricity is reached for all pulse duration values at about half the time. Additionally a negative conicity which is desirable for certain applications can be achieved easily, see Fig. 13. The addition of a quarter wave plate to the optical setup allows to maintain the polarization in the p-orientation which turned out to be beneficial for avoiding ripples and non-circular hole exit shape. An additional benefit is to minimize reflections at the hole wall which may lead to a deflection of the hole propagation and finally to an unwanted deformation of the hole exit. Fig. 14. Entrance (above) and exit (below) of a hole drilled through 0.5 mm steel plate with an experimental picosecond laser (fi H = 5ps, =1030nm,f p =10kHz,Q = 600 μj). Figure 14 evidences what can be achieved with helical drilling and p-polarization. The exit appears perfectly circular with sharp edges. Around the entrance a surface structure is observed affecting the sharpness of the edges. This can be attributed to the high pulse energy used in the experiment causing scattering effects even at the pulse duration of 5 ps what could be avoided when applying vacuum. It is expected that in further investigations with reduced energy and a more stable laser system these flaws can be avoided. 3.3 Optimal pulse duration For applications in machining laser systems have to be reliable, robust and affordable. Since technical effort increases with shortening the pulse duration, the latter should be as short as necessary, only, for achieving a satisfying result. The theoretical findings of section 2 point out that below 10 ps a further reduction of pulse duration is expected to be without significant effect on thermal behavior, at least as far as metals are considered. On the other hand, the observed scattering effect increases when the pulse duration is lowered beneath 5 ps, see Fig. 7. From all these considerations a pulse duration near to 5 to 10 ps appears to be optimal for micromachining of metals. 4. Conclusions Fig. 13. Influence of drilling time and pulse duration on hole cylindricity. Above: with perpendicular incidence; below: with an inclination of 4. The recipe to increase the quality of material removal processes by reducing the pulse duration to the femtosecond range has turned out to be less curative than expected. Fundamental investigations revealed three sources for remaining flaws 81

6 strong deformation of laser beam near focus, deflection of s-polarized radiation at hole wall, essentially thermal nature of ablation process even in femtosecond range. Technical measures have been developed to overcome the problems. Like in drilling with nanosecond pulses the helical technique turned out to be efficient for strongly reducing formation of recast. An optical setup has been developed offering the possibilities to vary hole diameter and conicity in a wide range and to maintain p-polarization. A pulse duration near 5 to 10 ps is regarded as optimal. For shallow structures energy density values of about 5 J/cm 2 offer the possibility to avoid dross formation. For high aspect ratio drilling a vacuum atmosphere allows to apply higher energy density values and still to avoid flaws. The author wishes to thank his coworkers Detlef Breitling, Christian Föhl, Andreas Ruf and Michael Weikert as well as the colleagues of the group of Vitali Konov from General Physics Institute in Moscow, Serge Garnov, Sergei Klimentov and Taras Kononenko for their contributions and the German Ministry for Education and Research (BMBF) for funding within the project PRIMUS (13N771016). The author is responsible for the content. References 1) T. Abeln, J. Radtke, and F. Dausinger: Proc. Laser Microfabrication Conf. ICALEO 99, San Diego, USA, (Laser Institute of America, 2000), p ) H. Tönshoff, C. Momma, A. Ostendorf, S. Nolte, and G. Kamlage: Proc. Laser Materials Processing Conf. ICA- LEO 98, Orlando, USA, (Laser Institute of America,1998), pp. A 28. 3) A. Ruf, P. Berger, F. Dausinger, and H. Hügel: Frühjahrstagung 2001 der Deutschen Physikalischen Gesellschaft, Berlin, Germany, , Verhandl. DPG (VI) 36, S ) M.vonAllmen:J.Appl.Phys.47, S (1976). 5) A. Ruf, D. Breitling, P. Berger, F. Dausinger, and H. Hügel: Proc. SPIE 4830 (in press). 6) R. Stoian, M. Boyle, A. Thoss, A. Rosenfeld, G. Korn, and I. Hertel: Proc. SPIE 4830 (in press). 7) V. Konov, S. Klimentov, T. Kononenko, P Pivovarov, and S. Garnov: Proc. SPIE 4830 (in press). 8) M. Weikert, C. Föhl, and F. Dausinger: Proc. SPIE 4830 (in press). 9) C. Föhl, D. Breitling, K. Jasper, J. Radtke, and F. Dausinger: Proc. SPIE 4426, 104 (2002). 82