Drilling of Glass by Excimer Laser Mask Projection Technique Bernd Keiper, Horst Exner, Udo Löschner, Thomas Kuntze Laserinstitut Mittelsachsen e.v., Hochschule Mittweida, University of Applied Sciences Mittweida, Germany Abstract Presently, there is a growing demand from the industry for microprocessing of materials. In particular, for applications in the field of microsystems technology it is necessary to produce structures with dimensions down to the micrometer scale especially in materials that could not be processed or processed well by conventional microelectronic technologies. We have been investigating the drilling of anodically bondable PYREX glass by means of laser microprocessing using the excimer laser mask projection technique (48 nm or 193 nm wavelength, 10 ns pulse duration, 8 mj pulse energy, 500 Hz repetition rate). We will show the dependence of the processing results on the laser parameters. The diameter of the holes ranges from 30 to 100 µm at the front side and from 1 to 50 µm at the rear side of the 500 µm thick wafer. We observed the formation of cracks in the laser processed region. Accordingly, we found distinct relationships between the process parameters and the quality of the walls of the drilled holes. Especially the change from 48 to 193 nm wavelength led to a distinct decrease of crack formation but the wall at the rear side of the wafer shows still a break off of some material. As a solution the drilling of the wafer from both sides of the wafer allows us to produce crack free holes. Introduction In the literature we can observe growing efforts to process various technical glasses (quartz glass, BK7, BGG31 and other glasses) using excimer laser wavelengths (193, 48 and 308 nm) 1,,3. Anodically bondable glass wafers are of special interest for applications in the field of microsystems technology since they allow the use of glass in combination with the silicon technology. That s why we have chosen 7740 borosilicate Pyrex glass for our investigations. A standard technology for producing holes in such wafers is ultrasonic drilling. The disadvantages of this method are the limited minimum diameter of the holes and the limited quality of the hole walls. In order to miniaturize components, holes with diameters of 100 µm or less are needed, which we wanted to produce using excimer laser microprocessing. The absorption coefficient and, therefore, the ablation mechanism of different glasses depends strongly on the composition of the glass. In contrast to quartz glass, Pyrex glass absorbs the excimer laser radiation with 48 nm wavelength, but we have find out in our investigations that the ablation of Pyrex glass and of different other glasses at 48 nm wavelength led to crack formation inside the irradiated area and, therefore, to a diminished quality of the hole walls. In consequence we used 193 nm wavelength for the actual investigations. Experimental details PYREX glass wafers were structured by means of ArF excimer laser microprocessing using the mask projection technique with the experimental setup given in Fig.1. We have used an ExciStar S-500 (TUILASER AG) and a LightBench System (ATL Lasertechnik GmbH). Table 1 shows the processing parameters used. The oxygen content of the air absorbs the laser radiation at 193 nm wavelength, resulting in a loss of laser energy of about 10 percent per meter length of the optical path and in the formation of ozone. To prevent the ozone formation it is necessary to flush the whole optical path with inert gas (N, Ar or He). That s why we have airsealed the optical assembly and connected one side to a gas supply and the other side to an exhaust exit. In order to flush the path between projection lens and sample surface we have used a nozzle with a mm bore (see Fig.1). In our experiments we have used Ar for flushing. Without flushing of the optical path the laser fluence at the sample surface is 0 percent lower. The use of the beam concentrator consisting of two lenses allows us to increase the laser fluence from 40 to more than 50 mj/cm at the surface of the mask and therefore to 5 J/cm at the sample surface if a reduction scale of 1:10 is used. A camera system enables the adjustment of the sample surface to the projection plane and the observation of the sample surface during the laser processing. The movement of the sample is done by an xyz-table (150 x 100 x 11 mm 3 ) with contouring control. The resolution of the system is 0.5 µm. The diameter of the processed area at the surface of the glass wafer was varied between 30 and 100 µm. In the text given diameters of conical holes are the diameters at the front side of the wafer. 1
Results and discussion In a first step we have determined the ablation depth per laser pulse in dependence on the laser fluence and on the number of pulses at different repetition rates. The total ablation depth was determined with a Dektak 3030 profilometer. The results are shown in Fig. for 50 Hz repetition rate and in Fig.3 for 500 Hz repetition rate. After the first laser pulse no ablation depth could be measured, since it led only to a laser induced change of the sample surface, e.g. to the ablation of impurities or water as well as increasing roughness and absorption. This behaviour is known from former investigations 4 of the ablation of different materials at 48 nm wavelength. After ten laser pulses relatively high ablation depths between 100 and 160 nm per pulse were observed. The following ten laser pulses show only half the ablation depth per pulse. For some laser fluences was measured a slight increase of the ablation depth at 100 laser pulses. This is the maximum of laser pulses applied in the actual investigations because of the limited measurement range of the profilometer. In the next step of our investigations we have drilled the holes through the 500 µm thick glass wafer. The moment the hole is drilled through was determined by the UV-laser-induced fluorescence observed on a white sheet of paper placed beneath the wafer. From this experiment we calculated the average ablation depth per laser pulse for different laser fluences, repetition rates and diameters which we show in Fig.4. We have found that for drilling a 50-µm diameter hole through the 500 µm thick wafer there exists a laser fluence limit. Below that limit of about 4 J/cm the drilling process stopped before drilling through. This behaviour is due to the conical shape of the holes. The slope of the walls and consequently the maximum thickness that can be drilled through depends on the laser fluence. The average ablation depth for drilling the whole wafer was distinctly higher than the ablation depths measured during the first 100 laser pulses at all laser fluences above the limit. In our former investigations 4 an increase of the ablation depth per laser pulse with growing total ablation depth was not observed. This may have various reasons: At first we have used 193 nm wavelength and 10 ns pulse duration compared to 48 nm and 30 ns, producing a threefold peak power at the same laser fluence. For the drilling of small holes we discuss a special ablation mechanism. The steep walls (<.3 ) reflect a large part of the radiation down to the ground of the hole though the laser fluence can increase with growing total ablation depth with an decreasing area resulting in larger ablation depths per laser pulse. The influence of the laser fluence on the ablation depth per pulse above the limit is relatively small: The threefold increase of laser fluence results only in the 1.3 fold increase in ablation depth. We assume that the excess of energy is mainly absorbed by the ejected material resulting in a stronger laser plasma. In our investigations with 30 µm diameter we observed a similar behaviour. But the limit for drilling through increases to 7. and 10. J/cm (50...00 Hz and 500 Hz, respectively). The higher limits can be explained with the increasing heat abduction from the smaller laser irradiated area. The ablation depth per laser pulse above the limit is nearly the same as we have determined with 50 µm diameter. During the drilling experiments with both diameters we can observe differences in the ablation process at different laser fluences: At a laser fluence just above the limit the observed fluorescence light becomes visible only after drilling through the wafer. If we used higher laser fluences, we have observed some violet fluorescence already at nearly half the drilling time. However, only when the hole is drilled through the fluorescence area becomes clearly defined and the observed fluorescence became more intensive allowing us to determine exactly the drilling time. This behaviour indicates that some laser light must be shining through microcracks in the glass already after half the drilling time. The crack formation should be due to the high pressure at the ground of the hole caused by the expansion of the strong laser plasma. Another reason for the cracks might be the thermally induced mechanical stress in the glass, but since the behaviour is nearly independent on the repetition rate this mechanism is not probable. The influence of the repetition rate on the ablation depth per pulse is very small. Only at the largest repetition rate of 500 Hz the ablation depth per laser pulse is a bit smaller. In Fig.5 we present Scanning Electron Microscopy (SEM) micrographs of holes drilled with different repetition rates at different laser fluences and 50 µm diameter. The wafer was cleaned with a standard process (½ min with 1: HF / NH 4 F) in order to remove the material redeposited during the drilling process from the surface. The holes are also slightly overetched resulting in a small increase of the diameter. The results agree with the above explained relationships: The repetition rate has only a minor influence on the quality of the holes at the rear side of the wafer, whereas increasing laser fluence leads to an increasing crack formation and break off of material at the bottom of the wafer. Best results were obtained with low laser fluences were no crack formation and only little break off of material but also a decreasing hole diameter at the rear side of the wafer was observed. The reason for this behaviour should be the force that is caused by the high pressure of the expanding laser plasma. That force increases with growing diameter of the bottom of the hole. That means we can minimize or prevent the damage inside the glass when the hole diameter at the rear side of the wafer is small.
The appearance of the front side of the holes shows only little dependence on the laser fluence but a slight increase of the thrown up region around the holes can be observed with increasing laser fluences. Again, best values were obtained at the low fluences: The thrown up region is then less than 0 µm wide and less than 1 µm high. In Fig.6 we show optical micrographs of 30-µm holes drilled with laser fluences at 7 and 11 J/cm,i.e., near the limit for drilling through and far above it. The wafer was not cleaned after drilling, so that the redeposit of ablated particles is clearly visible at the front side of the wafer. In accordance with the results obtained for 50-µm diameter holes we can observe a better quality, that means less break off of material at the rear side of the wafer, if a laser fluence just above the limit for drilling through is used. Unfortunately, the diameter at the rear side of the wafer is much smaller then (only 1.5 to µm). Our excimer laser drilled holes are more or less conical and have a relatively small diameter at the rear side of the wafer provided the optimized parameters are used. The slope of the walls varies between 3.5 and 1.6 (100 and 30 µm diameter, respectively). Up to now it seems not to be possible to avoid the break off of material in the moment of drilling through completely merely by optimizing the laser parameters. In our experiments we have also tested a number of other possibilities to avoid such break off: The wafer was coated with a photo resist or with a 1.5 µm thick Physical Enhanced CVD SiO film in order to reduce the force on the wafer during drilling through. However, those films did not distinctly improve the quality of the holes. In another experiment we put a second Pyrex glass wafer beneath the drilled wafer, but also without any positive result. Only after putting an alcohol film between the pressed wafers did we observe less break off of material. This might be caused by the generation of a counteracting force at the rear side of the wafer and/or by the cooling effect of the alcohol. Another possibility to improve the quality of the holes is the drilling from both sides of the wafer. The optical micrograph in Fig.7, shows a cross-section of such holes. It demonstrates that this method allows us to produce holes without break off of material on the rear side. Additionally, it is possible to enlarge the minimum diameter inside the hole. Summary and conclusion The excimer laser microprocessing system presented in this paper is suitable to produce holes with diameters from 30 to 100 µm in 500 µm thick Pyrex glass. Smaller hole diameters require larger laser fluences due to the increasing heat abduction. The quality of the holes at the front side of the wafer is very good without cracks in the glass. The formation of cracks and a break off of material at the rear side of the wafer depends on the processing parameters. At the optimized parameters (laser fluences just above the limit for drilling through) no cracks were observed but a slight break off of material may still occur. The ablation depth per laser pulse is nearly independent of the repetition rate. Only at 500 Hz it decreases slightly, but the advantage of the high repetition rate still exists since the mean ablation depth is higher than at lower repetition rates. The holes are conical with a slope of the walls between 3.5 and 1.6. Holes with the best quality (without break off of material) can be made by drilling from both sides of the wafer. Moreover, this method offers the possibility to enlarge the minimum diameter inside the conical hole, too. Acknowledgements The authors would like to thank the Dr. Teschauer & Petsch AG, Chemnitz for the financial support. The research project that was the basis for this publication is supported with EFFRE-means of the EU and with means of the Freistaat Sachsen (Projekt-Nr.: 4355/679). We thank Ina Schubert from the Technical University Chemnitz for the SEM and optical microscopy investigations and Andreas Eysert from the University Mittweida for carrying out the cross-section preparations. References 1 B. Wolff, P. Simon. Nanosecond and Femtosecond Excimer Laser Ablation of Fused Silica. Applied Physics A 54, (199) pp. 363-368 R. Nowak, S. Metev, G. Sepold. Excimer laser processing of BK7 and BGG31 glasses. Glastechnische Berichte, 66 (1993) No. 4, pp. 7-33 3
3 C. Buerhop; N. Lutz; R. Weißmann; G. Tomandi. Surface treatment of glass and ceramics using XeCl excimer laser radiation. Glastechnische Berichte, 66 (1993) No. 3, pp. 61-67 4 Horst Exner, Bernd Keiper, Peter Meja. Microstructuring of materials by pulsed laser focusing and projection technique. Presented at Photonics West `99, 3.-9. January 1999, San Jose, California, published in Laser Applications in Microelectronic and Optoelectronic Manufacturing IV ed. by Jan J. Dubowski, H. Helvajian, E. W. Kreutz, K. Sugioka, SPIE Proceedings Series, Volume 3618, (1999) 340-347. Figures Beam concentrator Mask Camera LightBench (Airsealed) Lighting Mirror (visible) Laser ExciStar S-500 Optical rail Dielectric mirror (193 nm) Flush gas (Argon) Exhaust exit 3 lens projection system Nozzle Sample xyz-table Fig.1: Experimental set-up for excimer laser microprocessing Ablation depth per laser pulse [nm] 160 140 10 100 80 60 40 0 0 7. J/cm 8.8 J/cm 10. J/cm 11.6 J/cm 1.8 J/cm 0 15 30 45 60 75 90 105 Number of pulses Fig.: Dependence of the ablation depth per laser pulse on the number of laser pulses for different laser fluences at 50 Hz repetition rate and 100 µm diameter. 4
Ablation depth per laser pulse [nm] 160 140 10 100 80 60 40 0 0 0 15 30 45 60 75 90 105 Number of pulses 7. J/cm 8.8 J/cm 10. J/cm 11.6 J/cm 1.8 J/cm Fig.3: Dependence of the ablation depth per laser pulse on the number of laser pulses for different laser fluences at 500 Hz repetition rate and 100 µm diameter. Average ablation depth per laser pulse [nm] 300 50 00 150 100 50 0 0 4 6 8 10 1 14 Laser fluence [J/cm ] f=100 Hz; d=50 µm f=500 Hz; d=50 µm f=100 Hz; d=30 µm f=500 Hz; d=30 µm Fig.4: Dependence of the average ablation depth per laser pulse on the laser fluence for different repetition rates and diameters of the holes refer to a total ablation depth of 500 µm in Pyrex glass. Also included the limits for drilling through (Points on the horizontal axis, where 50000 pulses were used). 5
Laser fluence: 4.4 J/cm 5. J/cm 7. J/cm 11.6 J/cm Repetition rate: 50 Hz back side 0 µm front side 0µm 00 Hz back side 0 µm front side 0µm 500 Hz back side 0 µm front side 0µm Fig.5: SEM images of 50 µm holes in 500 µm thick Pyrex glass drilled with different laser fluences and repetition rates. 6
10 µm a) b) c) Fig.6: Holes with 30 µm diameter in 500 µm thick Pyrex glass. a) shows the typical look of the front side, b) and c) show the rear side of the holes drilled with different fluences. The hole a), b) was drilled with 7 J/cm laser fluence and c) with 11 J/cm laser fluence. The drilled through region is marked in b) with a white circle to make it visible in the printed image. 100 µm Fig.7: Cross-section preparation of holes in 500 µm thick Pyrex glass. Both holes were drilled with 100 µm diameter from the front side of the wafer. The left one was drilled with 50 µm and the right one with 100 µm diameter from the rear side of the wafer. Table 1: Processing parameters Excimer laser Wavelength Pulse duration Repetition rate Pulse energy Laser fluence Reduction scale [nm] [ns] [Hz] [mj] [J/cm ] 193 10 50 to 500 Up to 8 4 to 5 1:10 7