- Email: [email protected]
Excimer laser ablation for micro-machining: geometric effects
applied
surface science ELSEVIER
Applied Surface Science 96-98 (1996) 415-419
Excimer laser ablation for micro-machining: P.E. Dyer, D.M. Karnakis, P.H. Key Department
ofApplied
Physics,
Unioersify
of Hull,
geometric effects
*, P. Monk
Huli, HU6 7RX, UK
Received 22 May I995
Abstract KrF laser ablation of an epoxy material has been used to study the effects of surface geometry for micro-machining applications. Morphological features with surface height variations greater than the ablation depth per pulse and transverse dimensions on the scale of the irradiating spot are shown, during multi-pulse ablation, to evolve into surface profiles which may be predicted by considering the local variation in effective fluence. This simple model is seen to be in good agreement with ablation experiments carried out on convex and concave surfaces. Stepped features are more problematical and factors such as diffraction and local pressure inhibition of ablation must also be considered.
1. Introduction
The UV-excimer laser ablation characteristics of a wide variety of materials has been the subject of intense study in the last few years. For the most part, studies have been limited to the measurement of material removal rates and to the investigation of velocities, distributions and composition of products during ablation of planar surfaces by a range of diagnostic techniques. Some work has also been done on the effects of transverse irradiance geometry, at normal incidence, in order to assess patterning quality and capability. As laser ablation increases in application as a tool for micro-machining of components, multilayer materials and preshaped structures, it is important to obtain data applicable to non-normal incidence and non-planar ablation geometries. It is an interesting feature of, normal incidence,
* Corresponding author. Tel.: +44-1482
465203: fax: +44-
14x2 465606. 0169.4332/96/$15.00 0 SSDI 0169-4332(95)00496-3
1996 Elsevier Science B.V. All rights reserved
ablation of initially planar surfaces that small-scale surface asperities can be created by microscopic surface variations, impurities, shock waves, beam inhomogeneity or interference effects. That these asperities can be ‘amplified’ by subsequent ablation pulses is evidenced by the phenomena of hydro-dynamical sputtering [l] and conical structure formation [2]. The scale of the embryonic surface non-uniformity plays an important role in determining the final surface features produced. Post-ablation surface morphological features of this type are often chaotic because of the unpredictability of the original ‘seed’ asperity. For large scale morphological variations introduced deliberately, i.e. in micro-component machining, it is necessary to be able to predict the resultant surface profile after multi-pulse ablation. Large-scale features, in this regime, may be defined as having transverse dimensions similar to those of the laser spot size and surface height variations greater than the ablation depth per pulse. In this paper we present the results of a study of
416
P.E. Dyer et al./Applied Surface Science 96-98 (1996) 415-419
multi-pulse KrF laser ablation of large-scale, nonplanar, surface features using a simple irradiation geometry at normal incidence to the mean surface plane. High speed shadow photography (HSSP) was used to observe the ablation event at short delay times after the ablation laser pulse and the ablation sites examined after many ablation pulses by microscopy of sectioned samples. A simple model, following the scheme of Dyer et al. [2], was used to predict the final surface morphology.
2. Experimental A rectangular aperture was used to select an area of uniform intensity distribution from the KrF laser output beam. A single quartz lens was used to image the aperture onto a plane coincident with the mean height of target features. The lens was used at a conjugate ratio of 2 to produce the = 2 X 1 mm image and stopped down to give a depth of focus L f0.5 mm which is greater than the sum of the depths of the experimental surface features and ablation craters. Ablation was carried out at pulse repetition rates of 1-2 Hz to avoid cumulative heating effects. The HSSP results were obtained using 5 ns dye laser pulses for illumination in an experimental arrangement that has been described previously [3]. Target samples were prepared by casting rods and plates from a two-part epoxy. Initial experiments were carried out on planar samples, in which the mean ablation rate per pulse was measured as a function of laser fluence, in order to obtain the ablation characteristics of the material. From a fit of these data to Beer’s law of absorption, the effective absorption coefficient, (Y= 5 X lo4 cm-‘, and fluence threshold for significant ablation, FT = 40 ml cm-‘, were obtained. Subsequent ablation experiments were carried out at 4 X Fr, a regime of efficient ablation.
3. Results and discussion HSSP results for the reference planar surface are shown in Fig. 1. The HSSP data, at time delays of 0.4 and 4.0 ps after the start of the ablation laser pulse, show in Fig. l(a); shock-wave formation above
the ablation site and in Fig. l(b); the expanded shock-front and ablation products which have not yet detached from the surface. The forward expansion velocity of the shock-front is, at 500 ms-‘, = 40% higher than the transverse expansion velocity. Since the driving pressure is proportional to the square of velocity, we can deduce that the initial pressure in the centre of the ablation site is approximately twice that at the edges. The product expansion velocity is much smaller at 65 ms-’ and is no longer visible after 20 l.~s, due either to the density being too low to form the shadowgraph or to decomposition into gaseous products. Re-deposition of ablated material extends to a distance of 0.5 mm at the edges of the ablation site. HSSP results at time delays of 1.0 and 2.0 p,s for 2 mm diameter convex and concave surfaces are shown in Fig. 2. For the convex surface, Fig. 2(a) and (b), the expansion of the shock-front is radial, being unconfined by edge effects. It can be seen, however, that there is a slight forward bias in the early stages, indicating a higher ablation rate over that part of the target surface which is irradiated at normal incidence. The concave surface, shown in Fig. 2(c) and (d), exhibits an increased forward shock-front velocity during the early stages of expansion which rapidly becomes similar to the convex case when free of the sample surface. The early stages of the shock-front and product expansion for these surface geometries may contribute to inhibition of the ablation-rate and is the subject of continuing investigations. A first order model, in which dynamic effects are neglected, for the influence of surface geometry on spatial variations in ablation-rate may be used by considering the local variation in effective laser fluence by analogy with the scheme for cone formation proposed by Dyer [2]. In the present case we consider a reference plane, x-y. normal to the direction, z, of the incident laser radiation. The local variation in height of the surface above the reference plane is given by the coordinates z(x) and z(y). The gradients in the n and y directions at any point are given by d z/dx and d z/d y. In the current geometry we observe in the y direction for which d z/d y = 0 and, therefore, only consider a two-dimensional surface profile. The surface gradient at any point, x, determines the local effective fluence through the cosine
P.E. Dyer cr al. /Applied
417
Surface Science 96-98 (1996) 415-419
Fig. 3. Multi-pular ablation models for cylindrical convex and concave surfaces. Each broken line represents the new surface after 150 laser oulses.
incident, write:
Fig. I. High speed shadow photographs of ablation of planar surface at time delays of (a) 0.4 IJ.S and (b) 4.0 ks after the ablation pulse.
of the incidence angle. Using Beer’s law to define the depth of material removed per laser pulse, A z( x>, in terms of the effective ahsorption index, LY,and the
F. and threshold,
FT, fluenccs
we can
the new surface then being described by: z(x)A z.(x). The model is applied iteratively to determine the final surface profile after the required number of laser pulses. Predictions of the final surface profile for the convex and concave surfaces after 900 laser pulses at 160 mJ cm-’ are shown in Fig. 3. From the sections, shown in Fig. 4, it can he seen that the model is in good agreement with the experimental results.
Fig. 2. High speed shadow photographs of ahlation of convex and concave surfaces at time delays of t .O p.s, (a) and (c), and 2 0 p,s, (b) and Cd), after the ablation
pulse.
418
P.E. Dyer et al./Applied
Surface Science 96-98
(1996) 415-419
Multipulse ablation across a step-edge between two planar surfaces at different heights is, clearly, a problem in that both surfaces cannot readily form the image plane simultaneously. For large depth of focus optics and a small step height the problem may not be too severe, providing the ablation depth is not too great. Otherwise the edge resolution of the ablation site can rapidly deteriorate. It has been previously proposed [3] that the shock-front expansion velocity mirrors the local pressure profile at the early stages of ablation leading to an inhibition of local, overall, ablation rate. If this be the case, an examination of HSSP results will allow a qualitative prediction of the surface profile after multipulse ablation. Fig. 5 shows the shock-front positions after 0.4 and 1.0 us. Considering first the upper surface of the step; it can be seen that after 0.4 Fig. 5. High speed shadow photographs of ablation of a stepped feature at time delays of 0.4 p.s (a) and 1.0 ks (b) after the ablation pulse.
Fig. 4. Photo-micrographs of sectioned convex and concave surfaces after ablation with 1400 laser pulses. The laser spot was smaller than the full site width for the concave sample and off-set to show the original profile.
ps the forward expansion of the shock-front is greater than the transverse expansion and this continues to be the case at 1.0 ps. This may once again be understood in terms of the driving pressure being more readily relieved at the free edges of the ablation site. In the case of the lower surface of the step; there is additional confinement at the step wall leading to a slight increase in the forward expansion velocity, the transverse velocity at the free edge, however, is similar to that at the upper surface. Tt is interesting to note that at 1.0 ps, Fig. 5(b) shows the shock-fronts from the upper and lower surfaces have started to coalesce and longer time-scales reveal a single shock-front. Qualitatively, from consideration of the early pressure inhibition of ablation, we might predict that multipulse ablation of the step will result in the maximum material removal at the three free edges and minimum material removal at the foot of the step wall. Mitigating against this scenario, however, is the loss of a pressure relief mechanism at the outer edges as ablation proceeds to create a deep crater and the possibility of Fresnel diffraction effects at the base of the walls. Fig. 6 shows a section through the ablation site with the original step profile super-
P.E. Dyer et al./Applied
Surface Science 96-98
Fig. 6. Photo-mmograph of sectioned stepped feature after ablation with 700 laser pulses.
imposed. It can be seen that the ablation profile for the upper step is very uniform with extensive rounding of the step edge, suggesting that the two free edges of this part of the site prevent the formation of a large pressure gradient. The ablation profile of the lower step, however, is also very uniform, suggesting that a high local pressure inhibition of ablation at the base of the step wall does not in fact occur. The undercutting observed at the base of the walls would suggest that diffraction effects have contributed to a local high fluence region.
(19%) 415-419
419
work suggest that initially convex surfaces are planarised whilst initially concave surfaces develop nearly straight walls with angles of inclination equal to those found during the formation of conical structures. Where vertical walls exist, Fresnel diffraction may lead to trench formation. Since all of the phenomena observed are a strong function of the ratio of initial fluence to threshold the degree to which non-uniform fluence, F/F,, ablation occurs is largely influenced by the ablation regime chosen. The spatially non-uniform ablation rate effects seen under conditions of uniform irradiation of non-planar surfaces may be mirrored by non-uniform irradiation of planar surfaces. It should, therefore, be possible to maintain a desired surface profile during multipulse ablation by tailoring the laser beam energy distribution to the initial surface morphology.
Acknowledgements The authors wish to thank G. Robinson for the microphotography and P.H.K. gratefully acknowledges the continuing support of Exitech Ltd, Oxford, UK.
References 4. Conclusions In laser micro-machining using a projection image technique on non-planar surfaces a number of factors must be taken into account in order to predict the final surface profile. The trends observed in this
[I] R. Kelly and J.E. Rothenberg, Nucl. Instr. Meth. B 7/8 ( 1985) 7%. [2] P.E. Dyer, SD. Jenkins and J. Sidhu, Appl. Phys. Lett. 49 (1986) 453. [31 P.E. Dyer, P.H. Key, D. Sands, H.V. Snelling and F.X. Wagner, Appl. Surf Sci. 86 (1995) IX.
surface science ELSEVIER
Applied Surface Science 96-98 (1996) 415-419
Excimer laser ablation for micro-machining: P.E. Dyer, D.M. Karnakis, P.H. Key Department
ofApplied
Physics,
Unioersify
of Hull,
geometric effects
*, P. Monk
Huli, HU6 7RX, UK
Received 22 May I995
Abstract KrF laser ablation of an epoxy material has been used to study the effects of surface geometry for micro-machining applications. Morphological features with surface height variations greater than the ablation depth per pulse and transverse dimensions on the scale of the irradiating spot are shown, during multi-pulse ablation, to evolve into surface profiles which may be predicted by considering the local variation in effective fluence. This simple model is seen to be in good agreement with ablation experiments carried out on convex and concave surfaces. Stepped features are more problematical and factors such as diffraction and local pressure inhibition of ablation must also be considered.
1. Introduction
The UV-excimer laser ablation characteristics of a wide variety of materials has been the subject of intense study in the last few years. For the most part, studies have been limited to the measurement of material removal rates and to the investigation of velocities, distributions and composition of products during ablation of planar surfaces by a range of diagnostic techniques. Some work has also been done on the effects of transverse irradiance geometry, at normal incidence, in order to assess patterning quality and capability. As laser ablation increases in application as a tool for micro-machining of components, multilayer materials and preshaped structures, it is important to obtain data applicable to non-normal incidence and non-planar ablation geometries. It is an interesting feature of, normal incidence,
* Corresponding author. Tel.: +44-1482
465203: fax: +44-
14x2 465606. 0169.4332/96/$15.00 0 SSDI 0169-4332(95)00496-3
1996 Elsevier Science B.V. All rights reserved
ablation of initially planar surfaces that small-scale surface asperities can be created by microscopic surface variations, impurities, shock waves, beam inhomogeneity or interference effects. That these asperities can be ‘amplified’ by subsequent ablation pulses is evidenced by the phenomena of hydro-dynamical sputtering [l] and conical structure formation [2]. The scale of the embryonic surface non-uniformity plays an important role in determining the final surface features produced. Post-ablation surface morphological features of this type are often chaotic because of the unpredictability of the original ‘seed’ asperity. For large scale morphological variations introduced deliberately, i.e. in micro-component machining, it is necessary to be able to predict the resultant surface profile after multi-pulse ablation. Large-scale features, in this regime, may be defined as having transverse dimensions similar to those of the laser spot size and surface height variations greater than the ablation depth per pulse. In this paper we present the results of a study of
416
P.E. Dyer et al./Applied Surface Science 96-98 (1996) 415-419
multi-pulse KrF laser ablation of large-scale, nonplanar, surface features using a simple irradiation geometry at normal incidence to the mean surface plane. High speed shadow photography (HSSP) was used to observe the ablation event at short delay times after the ablation laser pulse and the ablation sites examined after many ablation pulses by microscopy of sectioned samples. A simple model, following the scheme of Dyer et al. [2], was used to predict the final surface morphology.
2. Experimental A rectangular aperture was used to select an area of uniform intensity distribution from the KrF laser output beam. A single quartz lens was used to image the aperture onto a plane coincident with the mean height of target features. The lens was used at a conjugate ratio of 2 to produce the = 2 X 1 mm image and stopped down to give a depth of focus L f0.5 mm which is greater than the sum of the depths of the experimental surface features and ablation craters. Ablation was carried out at pulse repetition rates of 1-2 Hz to avoid cumulative heating effects. The HSSP results were obtained using 5 ns dye laser pulses for illumination in an experimental arrangement that has been described previously [3]. Target samples were prepared by casting rods and plates from a two-part epoxy. Initial experiments were carried out on planar samples, in which the mean ablation rate per pulse was measured as a function of laser fluence, in order to obtain the ablation characteristics of the material. From a fit of these data to Beer’s law of absorption, the effective absorption coefficient, (Y= 5 X lo4 cm-‘, and fluence threshold for significant ablation, FT = 40 ml cm-‘, were obtained. Subsequent ablation experiments were carried out at 4 X Fr, a regime of efficient ablation.
3. Results and discussion HSSP results for the reference planar surface are shown in Fig. 1. The HSSP data, at time delays of 0.4 and 4.0 ps after the start of the ablation laser pulse, show in Fig. l(a); shock-wave formation above
the ablation site and in Fig. l(b); the expanded shock-front and ablation products which have not yet detached from the surface. The forward expansion velocity of the shock-front is, at 500 ms-‘, = 40% higher than the transverse expansion velocity. Since the driving pressure is proportional to the square of velocity, we can deduce that the initial pressure in the centre of the ablation site is approximately twice that at the edges. The product expansion velocity is much smaller at 65 ms-’ and is no longer visible after 20 l.~s, due either to the density being too low to form the shadowgraph or to decomposition into gaseous products. Re-deposition of ablated material extends to a distance of 0.5 mm at the edges of the ablation site. HSSP results at time delays of 1.0 and 2.0 p,s for 2 mm diameter convex and concave surfaces are shown in Fig. 2. For the convex surface, Fig. 2(a) and (b), the expansion of the shock-front is radial, being unconfined by edge effects. It can be seen, however, that there is a slight forward bias in the early stages, indicating a higher ablation rate over that part of the target surface which is irradiated at normal incidence. The concave surface, shown in Fig. 2(c) and (d), exhibits an increased forward shock-front velocity during the early stages of expansion which rapidly becomes similar to the convex case when free of the sample surface. The early stages of the shock-front and product expansion for these surface geometries may contribute to inhibition of the ablation-rate and is the subject of continuing investigations. A first order model, in which dynamic effects are neglected, for the influence of surface geometry on spatial variations in ablation-rate may be used by considering the local variation in effective laser fluence by analogy with the scheme for cone formation proposed by Dyer [2]. In the present case we consider a reference plane, x-y. normal to the direction, z, of the incident laser radiation. The local variation in height of the surface above the reference plane is given by the coordinates z(x) and z(y). The gradients in the n and y directions at any point are given by d z/dx and d z/d y. In the current geometry we observe in the y direction for which d z/d y = 0 and, therefore, only consider a two-dimensional surface profile. The surface gradient at any point, x, determines the local effective fluence through the cosine
P.E. Dyer cr al. /Applied
417
Surface Science 96-98 (1996) 415-419
Fig. 3. Multi-pular ablation models for cylindrical convex and concave surfaces. Each broken line represents the new surface after 150 laser oulses.
incident, write:
Fig. I. High speed shadow photographs of ablation of planar surface at time delays of (a) 0.4 IJ.S and (b) 4.0 ks after the ablation pulse.
of the incidence angle. Using Beer’s law to define the depth of material removed per laser pulse, A z( x>, in terms of the effective ahsorption index, LY,and the
F. and threshold,
FT, fluenccs
we can
the new surface then being described by: z(x)A z.(x). The model is applied iteratively to determine the final surface profile after the required number of laser pulses. Predictions of the final surface profile for the convex and concave surfaces after 900 laser pulses at 160 mJ cm-’ are shown in Fig. 3. From the sections, shown in Fig. 4, it can he seen that the model is in good agreement with the experimental results.
Fig. 2. High speed shadow photographs of ahlation of convex and concave surfaces at time delays of t .O p.s, (a) and (c), and 2 0 p,s, (b) and Cd), after the ablation
pulse.
418
P.E. Dyer et al./Applied
Surface Science 96-98
(1996) 415-419
Multipulse ablation across a step-edge between two planar surfaces at different heights is, clearly, a problem in that both surfaces cannot readily form the image plane simultaneously. For large depth of focus optics and a small step height the problem may not be too severe, providing the ablation depth is not too great. Otherwise the edge resolution of the ablation site can rapidly deteriorate. It has been previously proposed [3] that the shock-front expansion velocity mirrors the local pressure profile at the early stages of ablation leading to an inhibition of local, overall, ablation rate. If this be the case, an examination of HSSP results will allow a qualitative prediction of the surface profile after multipulse ablation. Fig. 5 shows the shock-front positions after 0.4 and 1.0 us. Considering first the upper surface of the step; it can be seen that after 0.4 Fig. 5. High speed shadow photographs of ablation of a stepped feature at time delays of 0.4 p.s (a) and 1.0 ks (b) after the ablation pulse.
Fig. 4. Photo-micrographs of sectioned convex and concave surfaces after ablation with 1400 laser pulses. The laser spot was smaller than the full site width for the concave sample and off-set to show the original profile.
ps the forward expansion of the shock-front is greater than the transverse expansion and this continues to be the case at 1.0 ps. This may once again be understood in terms of the driving pressure being more readily relieved at the free edges of the ablation site. In the case of the lower surface of the step; there is additional confinement at the step wall leading to a slight increase in the forward expansion velocity, the transverse velocity at the free edge, however, is similar to that at the upper surface. Tt is interesting to note that at 1.0 ps, Fig. 5(b) shows the shock-fronts from the upper and lower surfaces have started to coalesce and longer time-scales reveal a single shock-front. Qualitatively, from consideration of the early pressure inhibition of ablation, we might predict that multipulse ablation of the step will result in the maximum material removal at the three free edges and minimum material removal at the foot of the step wall. Mitigating against this scenario, however, is the loss of a pressure relief mechanism at the outer edges as ablation proceeds to create a deep crater and the possibility of Fresnel diffraction effects at the base of the walls. Fig. 6 shows a section through the ablation site with the original step profile super-
P.E. Dyer et al./Applied
Surface Science 96-98
Fig. 6. Photo-mmograph of sectioned stepped feature after ablation with 700 laser pulses.
imposed. It can be seen that the ablation profile for the upper step is very uniform with extensive rounding of the step edge, suggesting that the two free edges of this part of the site prevent the formation of a large pressure gradient. The ablation profile of the lower step, however, is also very uniform, suggesting that a high local pressure inhibition of ablation at the base of the step wall does not in fact occur. The undercutting observed at the base of the walls would suggest that diffraction effects have contributed to a local high fluence region.
(19%) 415-419
419
work suggest that initially convex surfaces are planarised whilst initially concave surfaces develop nearly straight walls with angles of inclination equal to those found during the formation of conical structures. Where vertical walls exist, Fresnel diffraction may lead to trench formation. Since all of the phenomena observed are a strong function of the ratio of initial fluence to threshold the degree to which non-uniform fluence, F/F,, ablation occurs is largely influenced by the ablation regime chosen. The spatially non-uniform ablation rate effects seen under conditions of uniform irradiation of non-planar surfaces may be mirrored by non-uniform irradiation of planar surfaces. It should, therefore, be possible to maintain a desired surface profile during multipulse ablation by tailoring the laser beam energy distribution to the initial surface morphology.
Acknowledgements The authors wish to thank G. Robinson for the microphotography and P.H.K. gratefully acknowledges the continuing support of Exitech Ltd, Oxford, UK.
References 4. Conclusions In laser micro-machining using a projection image technique on non-planar surfaces a number of factors must be taken into account in order to predict the final surface profile. The trends observed in this
[I] R. Kelly and J.E. Rothenberg, Nucl. Instr. Meth. B 7/8 ( 1985) 7%. [2] P.E. Dyer, SD. Jenkins and J. Sidhu, Appl. Phys. Lett. 49 (1986) 453. [31 P.E. Dyer, P.H. Key, D. Sands, H.V. Snelling and F.X. Wagner, Appl. Surf Sci. 86 (1995) IX.