Optical fiber lasers using a MOPA architecture offer key advantages, such as high beam quality, high repetition rate, and short pulse duration, for the development of high-throughput laser processes. Cutting strengthened glasses, among other materials, is one of these. Furthermore, increasing the efficiency of picosecond laser sources can be performed by bunching picosecond pulses into bursts. In this paper, we show that rapid bursts of picosecond pulses generated by an INO’s MOPA fiber laser can be a promising regime for creating defects of the proper size in transparent materials. Combined with an appropriate cleaving method, the process will allow cutting strengthened glass with edges that show a good visual appearance. The quality of the edges is controlled by the energy of the pulse burst at each targeted depth and by the number of lines in which the defects are created.
Keywords: strengthened glass, laser cutting, pulse shaping, fiber laser, picosecond, burst
1. Introduction
Strengthened glass has become very popular for the manufacture of protective windows for mobile electronic devices, thanks to its high resistance to damage and flexibility. However, the increased strength of the glass makes it difficult to cut, due to structural stresses within the material resulting from the post-production chemical treatment of the surfaces. The stealth dicing technique, initially developed for semiconductor wafer dicing, is widely used to cut glass sheets due to its high throughput and high-quality edges.
In our setup, ultra-short-pulse lasers are used to modify the structure of the transparent material via nonlinear absorption, achieved by tightly focusing the laser beam to high peak power. Internal minor defects can then be generated to weaken the material and cause it to separate along a line. The delimited parts are then separated by applying a mechanical force or a thermal stress along the line of defects. This technique does not eject material outside the sample, enabling the development of a laser process free of debris and producing no kerf. Furthermore, only a portion of the total energy required to break the chemical bonds across the entire thickness of the material comes from the optical laser beam, enabling high-throughput processes with relatively low-power lasers. Laser processes that use pulse energies just above the damage threshold result in the formation of micrometric defects, enabling the production of well-defined edges.
For laser processes like stealth dicing, throughput is mainly determined by the repetition rate. Fiber lasers are inherently designed to emit at high repetition rates and can produce very short pulses with appreciable energy; they are therefore ideal laser sources for such processes.
The fiber lasers developed at INO feature high beam quality, polarized output, nanosecond pulse shaping, adjustable repetition rate (single shot to 500 kHz), and enslaved pumping to keep the output pulse energy constant as the repetition rate is varied.
They also show a unique regime of pulses called picosecond-burst mode, which consists of closely spaced short pulses (separated by 625 ps) with durations of tens of picoseconds, emitted under an arbitrarily shaped, programmable envelope of nanosecond duration. For processing transparent materials, the ps-burst mode has the advantage of producing high peak intensities within short pulses that alter the material via nonlinear absorption, while simultaneously providing substantial energy by cumulating laser shots in the same targeted area during the short burst time.
INO’s MOPAW laser (Master Oscillator Programmable Arbitrary Waveform) shows extraordinary potential for the exploration of laser-matter interactions and for the development of laser processes on a fine scale [1]. The combination of fiber amplifiers with advanced digital electronics enables temporal and amplitude shaping of nanosecond pulses, bursts of picosecond pulses, and the production of single picosecond pulses. Figure 1 shows a few examples of the numerous shapes that can be generated by directly programming the INO MOPAW laser. Whatever the pulse regime selected, the MOPAW laser can be triggered to emit pulses on demand, which is of great importance for applications involving single-pulse laser processes, where synchronization between the laser source and the sample position under the laser beam at a given time is required. Also, the duration of nanosecond pulses or picosecond bursts can be programmed from 1.25 ns to 500 ns, enabling fine-tuning of a laser process depending on the target material [2, 3].
INO’s laser platform adopts a MOPA (Master Oscillator–Power Amplifier) configuration [4], as shown in Figure 2, that efficiently leverages the versatility of an advanced, electronically driven, fiber-based picosecond optical pulse generator. The master oscillator consists of a short-pulse generation module based on spectral filtering of a phase-modulated pulsed laser diode. The power amplifier module consists of two cascaded fiber amplifiers using depressed-clad, large-mode-area (LMA) fibers [5]. Ytterbium-doped fiber amplifiers offer the advantages of optical efficiency, high average power, and high beam quality, independent of operating conditions. Polarization-maintaining (PM) fibers throughout the laser chain produce a stable, linearly polarized laser beam with an extinction ratio exceeding 20 dB, making the MOPAW laser compatible with harmonic generation for material processing at shorter wavelengths.
The goal of this study was to evaluate the performance of the ps-burst regime for cutting strengthened glass sheets, like Corning Gorilla® glass. A series of tests was carried out to determine the operating conditions (cutting speed, defect depth, cleaving method), taking into account the characteristics of the pulse regime (laser power, pulse duration) that affect cut quality. A prototype of the MOPAW laser capable of operating in the ps-burst regime has been used for the evaluation.
2. Experimental setup
The dedicated motion system, tailored to the laser prototype's characteristics, is illustrated in Figure 3. It consists of a fast, motorized translation stage (up to 500 mm/s) mounted above manual stages for precise adjustment in all angular directions. An accurate Z-axis motion system allows positioning the beam focal point within the material with ±1 µm accuracy. This is performed with the help of an in-line camera located behind a dichroic mirror. The incident beam size and lens focal length are chosen to maximize the pulse peak intensity and achieve efficient multi-photon absorption. The focal length must be very short but is limited by the remaining working distance and practical considerations arising from industrial laser processes. An 8 mm focal-length aspheric lens, in conjunction with a beam expander and a near-Gaussian laser beam (M2 < 1.3), is used to create a focal spot of approximately 3.5 µm inside the material. The sample is placed on a holder with high flatness to ensure that the focal spot depth within the material remains nearly constant during sample motion.
Our laser prototype provides burst durations of 5-80 ns at a repetition rate of 100 kHz. Every pulse in the burst has a duration of 60 ps and is separated by 555 ps. The output beam is linearly polarized. In these conditions, the sample, moving at maximum speed, will be stroked by a burst of short pulses every 5 µm along a straight line.
3. Results
The tests have been conducted on 560 µm-thick Gorilla® glass samples. The influence of various parameters on defect formation in the material and on the quality of cleaved edges has been studied. Namely, the polarization state, the number of defect lines (or passes), the laser pulse energy per pass, and the resulting depths of the created defects in the material have been studied.
At first, we estimated the Gorilla® glass damage threshold as a function of defect depth. With a repetition rate of 100 kHz and 5 ns ps-burst pulses, the damage threshold is approximately 1 W near the second surface of the glass sheet and approximately 0.5 W near the first surface. As expected, the threshold increases with depth due to spherical aberrations that spread the energy along the propagation axis as the focal point moves away from the first surface of the glass. With incident powers slightly above threshold, only minor defects are created, forming broken lines. The irregularities in the advent of defects are thought to be mainly caused by the inhomogeneities found inside and on the surface of the material. By striking the material with energies sufficiently above threshold (by ~60%), not only is a continuous line of defects is created, but microcracks connect them.
The best pulse shape we have found so far for obtaining the narrowest line of microcracks is shown in Figure 5. It was the shortest burst of picosecond pulses that could be produced by the MOPAW laser used at the time we conducted the tests. The burst has a 5-ns FWHM, Gaussian-like shape, containing about 7 to 8 ps pulses with sufficient intensity to be nonlinearly absorbed by the material. The use of longer bursts increases, all other parameters being equal, the length of lateral microcracks and reduces the force needed to cleave the sample, but at the expense of the edge quality. Since the peak intensities must remain relatively high, longer bursts imply a higher amount of energy per burst to obtain continuous lines of defects. Increasing energy ultimately led to larger microcracks and, eventually, to an uncontrolled break-up of the sample into several parts.
Tests have also been conducted when changing the polarization state of the laser beam from circular to linear, perpendicular, and parallel to the cutting direction. No significant influence of the polarization state on defect creation has been observed, except for a slight advantage for parallel linear polarization, which seems to create microcracks more oriented along the cutting direction. More tests are needed to confirm this effect.
The laser process, which involves weakening the glass along a line, is only part of the overall cutting process. Complete separation of the parts is obtained by cleaving along the weakened line. As such, we have compared different cleaving methods, both mechanical and thermal. Figure 6 illustrates the mechanical cleaving process used to separate the parts. The sample is held on a flat surface with the lines of defects slightly overhanging. A perpendicular force is applied on one end of the sample, uniformly distributed along the line of defects.
For thermal cleaving, the sample is held in a similar manner, but mechanical force is replaced by localized heating from a heat gun, a steam jet, or a heating plate. In all cases, a significant temperature gradient must be created across the lines of defects, so that differential thermal expansion forces the parts to cleave. Placing a film of water on the side not directly heated helps maintain a high temperature gradient (~100 °C) across the weakened line and facilitates cleaving. With the heat gun and the steam jet, it took approximately 30 s to initiate cleavage, whereas only 5 s were required with a heating plate.
3.1 Straight cutting
We have observed that 560 µm-thick samples of strengthened glass were somewhat challenging to cleave, with only a few lines (up to 3) of defects distributed throughout the volume. Thermal heating processes have yielded no interesting results, and the high mechanical pressure required to split the parts occasionally caused the cleaved line to drift. Also, the sharpness of the cleaved edges did not meet the target criteria. To further weaken the glass sheet, the incident power could be increased, but this would degrade the edge quality.
For very high edge quality, it is preferable to increase the number of lines of defects distributed in the volume and to limit their individual affected zones by maintaining the incident power just above threshold, as evaluated earlier. Using 4 or more lines of defects significantly facilitates cleaving for short lengths of glass (< ~ 75 mm). However, the required force for mechanical cleaving increases significantly for lengths greater than 100 mm. For such lengths, the number of defect lines needs to be increased to preserve cleaving ease. The creation of six or seven lines of defects in the volume enables a 560 µm-thick glass sheet to be cleaved more easily, producing better-quality edges. Particular attention should be paid to the first and last lines located near their respective surfaces. They mainly determine the ease of cleaving and the sharpness of the corner edges.
As schematically illustrated in Figure 7, for a focal spot positioned very close to the surface (z < zth) and for energy density above threshold, there is not enough material above the created defect to contain the increased internal stress. Therefore, surface chipping occurs. At higher energy densities, surface ablation begins. Scribing processes become possible, and if the energy density reaches a specific value, the sample will break into several pieces. For a focal spot positioned far enough from the surfaces (z > zth), minor defects and microcracks can be created without breaking the sample. However, as the focal spot approaches zth, the range of energy densities that allows their creation becomes narrower. The energy density at these depths should then be chosen carefully.
The first and last lines of defects shall be positioned close enough to their respective surfaces to facilitate cleaving, but sufficiently far apart to avoid chipping at the surfaces. This requirement implies that the flatness of the sample and its parallelism with respect to the motion axis must be controlled to prevent these extreme lines from reaching the surfaces as the sample is moved under the focused laser beam. The dimensions of our samples (~200 mm long) require only tilt adjustments before the laser process.
With the laser parameters and setup adjustments optimized to produce sharp, straight edges, as described earlier, we cut rectangles measuring a smartphone's dimensions from a 560 µm-thick Gorilla® glass sheet. Figure 8 shows views of one edge after being cleaved along a weakened straight line that has been laser processed with seven lines of defects. The force applied to cleave along the lines was approximately 1.4 kg / 100 mm.
The edge quality may be defined as the deviation of the cleaved line viewed from the top from the targeted line. In the example shown in Figure 8(b), we obtained a maximum deviation of 25 µm (50 µm peak-to-peak), with an average deviation of about 10 µm. By comparison, for two lines of defects in the volume, the maximum deviation is 50 µm (100 µm peak-to-peak). The difference can easily be seen. On the other hand, increasing the number of lines beyond 7 does not provide any real advantages. Instead, it increases process time and, if the lines are uniformly distributed, microcracks become too close to the surface, leading to chipping.
3.2 Curve cutting
Cutting of curved lines with our laser has also been explored. A compact rotating platform has been added to the experimental setup between the tilt platform and the linear translation stage, as described in Figure 3. The laser process developed for cutting straight lines as shown in Figure 8 has been applied to curved lines. Figure 9 shows an example of a right-angle corner of a strengthened glass sheet, as used previously, that has been rounded using the same laser process as for straight cuts. For rounded corners, the parts have been separated by a thermal cleaving process. In the rounded-corner case, mechanical cleaving yielded poor results because the applied force tends to initiate cleaves along straight lines. The sharpness of the edges and the roughness on the side of the corner are comparable to those of the straight lines. As it can be seen in Figure 9, the cleaving process has left uncut tiny parts of glass at the junction between the curved and the straight sides. This is due to the grazing angle between the lines of defects and the sample edges. Regardless of the cleaving method used, the cleave of round corners has a tendency to deviate away to reach faster the exterior of the material. In an optimized process, every line of defects would be laser-processed along the entire contour before cleaving to avoid protrusions.
4. Conclusion
We have shown that a burst of picosecond pulses from a high-repetition-rate fiber laser can be part of an efficient cutting process for strengthened glass sheets. Repeatable performance has been achieved in cleave straightness, ease of cleaving, and the visual appearance of cleaved edges. This study has identified key parameters to control to improve both the laser and cleaving processes. We have also shown that the laser process developed can be used to cut curves with small radii of curvature. Some methods for cleaving curved lines were tested and show interesting potential worth investigating further.
The experimental results show that strengthened glass sheets cut using an in-volume laser process followed by cleaving can produce edges of good quality if the number of defect lines is sufficient and the pulse energies remain near the damage threshold, especially for defects located close to both sample surfaces. A compromise can then be made between the time of the laser process and the performance of the cleaving process for a given application. The mechanical and thermal cleaving processes we have tested produce similar edge quality but differ in process time and in their ability to cleave shapes other than straight lines. A heat source capable of targeting a more localized area, such as a CO2 laser, a narrow flame, or a laminar flow of hot air or steam, would probably trigger faster, more efficient separation of the parts.
The overlap of the microcracks observed in Figure 4 suggests that they could be more widely spaced, potentially resulting in higher edge quality and a faster process if they do not deviate much from the plane defined by the defects within the glass sheet. This assumption will be validated with the latest version of INO’s MOPAW laser, which can operate at a lower repetition rate. On the other hand, its ability to run at repetition rates up to 500 kHz will enable the development of laser processes in which the laser beam is scanned over the sample at speeds achievable with state-of-the-art laser scanners, which motorized translation stages cannot match. Also, INO is now developing a PM, LMA, active fiber with an effective mode area of 1000 µm2, twice as large as the one used so far. This will enable the production of high peak intensities (or energy densities) at the focus plane when using lenses with longer focal lengths. The use of F-theta lenses will be possible, and more complex patterns will be processed more rapidly with laser scanners.
Acknowledgments
This research was supported by grants from the Centre de Développement d’entreprises technologiques (CDET), Gatineau, QC, Canada.
References
[1] L. Desbiens, P. Deladurantaye, A. Cournoyer, D. Gay, C. Paré, S. Eiselen, V. Roy, Y. Taillon, Flexible and Programmable Pulse Shaping MOPA Fiber Laser Platform, Performances and Applications, The Review of Laser Engineering, vol. 41, pp.691-696 (2013).
[2] L. Desbiens, M. Drolet, V. Roy, M.M. Sisto, Y. Taillon, Arbitrarily-shaped bursts of picosecond pulses from a fiber laser source for high-throughput applications, Proceedings of SPIE, vol. 7914, 791420-1 (2011)
[3] P. Deladurantaye, A. Cournoyer, M. Drolet, L. Desbiens, D. Lemieux, M. Briand, Y. Taillon, Proc. of SPIE, vol. 7914, 791404-1 (2011).
[4] L. Desbiens, M. Drolet, V. Roy, M. M. Sisto, Y. Taillon, Proc. of SPIE, vol. 7914, 791420-1 (2011)
[5] V. Roy, C. Paré, H. Zheng, P. Laperle, L. Desbiens, Y. Taillon, CLEO/Europe – IQEC 2013.
D. Gay1, S. Lavoie2, Y. Taillon1
Main author email address: david.gay@ino.ca
INO, 2740 Einstein Street, Quebec City, QC, Canada, G1P 4S4
Allied Scientific Pro (ASP), 815 La Carrière Blvd suite 202