1. Introduction

The invention and continuous development of ultrashort pulsed lasers has led to miscellaneous applications in a very short period of time. The combination of short pulse durations in the range of femto- to picoseconds and high intensities facilitate the processing of almost every material, not limited by its aggregate state or band gap [1–3], and even opens the field to novel research technologies such as time-resolved measurements of brain cells or 3D printing of nanostructures to the meso scale [4,5]. Especially the demand to process transparent materials has increased vastly due to the fast-growing markets of semiconductors and consumer electronics. Beside ablation and structuring processes, the cutting of displays and safety glasses is enabled [6–8]. Traditional separation techniques such as water-jet-cutting deliver reliable and constant results, but laser-based processing offers the possibility to reduce extensive and expensive post-processing [9].

The initiating process is the nonlinear absorption of laser radiation to bridge the material’s band gap. The interaction zone is spatially confined to few micrometers due to the high intensities, typically achieved by focusing of the laser radiation and short interaction times. The interaction steps consist of absorption and ionization mechanisms, that result in a free electron plasma. The type of permanent modification of the material is authoritatively determined by the choice of laser parameters, which in turn determine the relaxation channels [1,10,11]. Therefore, sensitive adjustment of the laser parameters is necessary to create the desired modification of the material. Several successive modifications can be inscribed in the material. There they form a contour along which the sample can be separated. While straight lines present the simplest form of a cutting edge, complex contours like circles or inclined cutting edges for tilted glass samples are also possible [8,12,13]. Recent studies show the targeted weighting of beam profiles by spatial frequency filtering techniques or by tilting individual optical components to create a preferential crack orientation [14,15]. Although preliminary modifications tests on glasses with a thickness of several hundred micrometers indicate the high potential of a tailored crack orientation [14–16], additional concepts for the treatment of thicker glasses are needed.

Separation occurs typically in a second process step and is of equal importance to sustain chipping-free cutting edges and high success-rates. Possible separation methods utilize the precise application of mechanical force, localized thermal heat gradients or chemical interaction from wet-etching-processes [17–19]. In this paper we purposefully investigate the separability of laser-modified material with different glasses of different thicknesses. The study implies the mechanical separation of glasses beforehand processed by single pass laser modification with Bessel-like beam profiles. Simulations show the impact of changing the modifying laser beam’s geometrical shape. We use pump-probe microscopy to further elucidate the modification process enabling the subsequent separation process, to understand the interaction of newly generated beam profiles with the material and crack generation at its heart. The microscope is adapted to transient absorption and polarization measurements under sample movement as well as tomographic reconstruction of the advanced beam profiles. The successful facilitated separability concludes the paper.

2. Experimental setup

2.1 Laser systems

Two laser systems have been used to perform the cleaving modifications. A 20 W lab-version based on the TRUMPF TruMicro Series 2000 offers the flexibility to use repetition rates up to 1 MHz and pulse trains of up to 8 pulses in a burst configuration with 20 ns between consecutive pulses. The pulse duration can be adjusted electronically between 300 fs and 20 ps. The second system comprises a TRUMPF TruMicro Series 5000 disk laser modified to generate pulse energies up to 2 mJ at a repetition rate of 50 kHz. This system is set to a pulse duration of 3 ps and is primarily used to process glasses with thicknesses of 2 mm and above. The wavelength of both processing lasers is $\def\upmu{\unicode[Times]{x00B5}}\def\uppi{\unicode[Times]{x03C0}}\lambda \,=\,1030 \,\textrm {nm}$.

In situ diagnostics by means of pump-probe microscopy is realized by combining either one of the processing lasers with a third laser system. The latter is based on a frequency doubled TRUMPF TruMicro Series 2000 system operated at $\lambda \,=\, 515 \,\textrm {nm}$ with a pulse duration of 300 fs and used for observation (probe). Key feature of this arrangement is the internal synchronization of the lasers due to a common seed module with independently selectable pulses. This provides the possibility to temporally shift the pulses in multiples of 20 ns whilst maintaining synchronization. An additional mechanical delay line is used to thoroughly cover the remaining time span between two seed pulses.

2.2 Methodology

The laser-matter-interaction zone is recorded transversely to the processing laser radiation in a transmission light microscope setup. The region of interest is imaged by an objective and tube lens directly onto a camera chip. Further, the microscope can be extended to perform polarization sensitive measurements, which allows the transient detection of birefringence, e.g. pressure waves or thermal stress. More information about the pump-probe setup can be found in [20].

Depending of the prevailing emphasis of the respective measurement, the system offers the possibility to switch between a complementary metal-oxide-semiconductor (CMOS) camera and a high-speed camera (HSC) with recording rates in the range of 100 kHz. The single-shot measurements with the CMOS camera provides the finer resolution and pixel density compared to the HSC and is thus used to analyze the beam profile and its absorption mechanisms. Accordingly, the HSC is chosen, if the emphasis is on the interaction and cumulative action of the individual pulses and the immediate material reaction, such as the laser processing under sample movement. In order to enable the reader to assign the displayed graphs as easily and intuitively as possible, each measurement configuration or simulation method is assigned its own color scheme representation. The schemes are created based on the cubehelix approach presented by D. Green [21].

2.3 Beam shaping elements

The concept for shaping non-diffracting beams with a variety of transverse intensity profiles (cf. Sec. 5) are realized by different beam shaping elements. The high-energy and high-power experiments for cleaving thick glasses, cf. Sec. 3, are carried out by static diffractive optical elements [22]. The efficacy proof of our concept by pump-probe microscopy (Sec. 5), however, required greater flexibility, e.g. in the variation of beam orientation. The liquid crystal-on-silicon-based spatial light modulator (SLM Hamamatsu X1313803-03) used here was illuminated with significantly lower optical power (highest average powers $< 50 \,\textrm {W}$, peak fluences $< 0.01\,\textrm{J}/\textrm{cm}^2$). However, the ability to flexibly change the holographic transmission function comes at the expense of reduced efficiency. The power loss in unwanted diffraction orders depends strongly on the phase distribution and reaches $\approx 20 \%$ in our cases. The shaped non-diffracting beams are imaged with demagnification into the transparent workpiece [22]. The advantage of using a telescope lies in the variability of the exact length of the Bessel-like beam, which we adapt to the thickness of the material to be cut [22].

3. Cleaving of glass

The excellent features of Bessel-like beams for glass cutting applications have been widely discussed in literature. [8,13,17,23]. The wave-optical simulation of such a beam-profile is depicted in Fig. 1 (a) and shows the most distinctive feature, i.e., the ratio between transversal and longitudinal extension in the micrometer- and millimeter-regime, respectively. Therefore, it is possible to precisely modify the whole thickness of glass along a contour by stringing together multiple pulses. Recent literature reports on modification and cleaving for glass thicknesses in the range of $(8 \dots 10) \,\textrm {mm}$ [17,23,24]. We performed a feasibility study of aluminosilicate, borosilicate and soda-lime glass with different thicknesses. Here, we were able to modify and break glasses of up to 12 mm. The result of a single pass modification of 12 mm soda-lime glass and subsequent separation with a typical cutting pliers is shown in Fig. 1 (b). The mean roughness $S_a$ of less then ${\upmu}$m confirms consistent quality throughout the sample thickness and suitability of Bessel-like beams for cleaving applications. A detailed description of the optical setup used to perform these experiment can be found in [22]. The processing occurred at a pulse energy of $1.5 \,\textrm {mJ}$ evenly spread over a 4-pulse burst train ($17 \,\textrm{ns}$ between consecutive pulses) at a pulse duration of $3 \,\textrm{ps}$. The spatial pulse distance was set to 3 ${\upmu}$m at a processing speed of $130 \,\textrm{mm}/\textrm{s}$.

 

Fig. 1. (a) Simulated intensity distribution of a symmetrical Bessel-like beam used for glass modification with information about the possible size ranges. The result of its application to modify a 12 mm thick soda-lime glass in a single pass configuration and subsequent separation (b) shows an homogeneous cutting edge. (c) Measurements of bending stress (red) and force necessary for separation (blue) for different glass thicknesses.

Download Full Size | PDF

The forces necessary to cleave borosilicate glass (SCHOTT borofloat 33) and the corresponding bending stress are presented in Fig. 1 (c). The beam length and other processing parameters were set similar to parameters described above to cleave the $12 \,\textrm{mm}$ thick glasses in order to ensure the comparability between the thicknesses. We used a four point bending tester by Instron (5944) to perform the measurements. The maximum applicable force to separate the samples was limited to 1 kN which set the maximum separable thickness to $6.5 \,\textrm{mm}$. The glass sample geometry is $60 \,\textrm {mm} \times 25 \,\textrm{mm} \times d$, whereby d denotes the glass thickness [cf. Fig. 1 (c)]. The modifications were inscribed along the $20 \,\textrm{mm}$ axis. The $y$-scale shows the logarithmically displayed force necessary for separation, hereafter referred to as separability, and the corresponding bending stress until separation. We point out, that the separability ($N$) is dependant on the sample geometry and the force-applying lever. Each parameter was repeated three times as a trade-off between diminishing variations and evaluation times. The measurements show that bending stress remains approximately the same, but force scales linearly with the increasing thickness. We expect the enhanced bending stress at thickness of $d = 1 \,\textrm{mm}$ due to the increased flexibility of thinner glasses. Recent calculations by Feuer et al. show that the modification length can even be further increased [23]. In combination with state-of-the-art USP laser systems reaching multiple milli-joules per burst and average output powers up to 1.4 kW [25], glass modification of several centimeter thickness might be possible soon. This further underlines the necessity to find a way to reduce the force necessary for separation with otherwise the same quality conditions.

4. Stress simulations

The shape of the beam profile takes an important role in search for an universal approach to reduce separability. Thus, we simulated the response of glass, which was previously modified by different beam profiles, being subject to mechanical stress. The profiles comprise a typical symmetrical shape as well as an elliptical cross-sectional area and different orientations thereof. The material of the simulations is silica glass, since it is the basic component for all the glasses processed within this paper. Although the simulations show noticeable differences dependent on the chosen beam profile, we point out that fine tuning of laser parameters and processing strategies, e.g., pulse duration and pulse energy or repetition rate and velocity of the sample, will nevertheless be necessary to optimize the process in accordance to the glass composition and thickness. The cleaving application can be divided into the process steps modification and subsequent separation. Single pulse modifications by symmetric beam profiles as shown in Fig. 1 (a) will generate statistically distributed cracks in an isotropic medium such as glass if the deposited energy is higher than the material threshold to withstand damage. By inducing an asymmetry to the beam profile, the interaction with the material will experience an asymmetry-dependent preferential direction. We will discuss this immediate material-response in the following section.

We use the finite element method (COMSOL) to analyze stress and strain during the mechanical separation as the second process step. The results for different modification orientations in a thin fused silica plate of $(50 \times 50 \times 0.55) \,\textrm{mm}^{3}$ are shown in Fig. 2. Thereby, we assume that the absorption and subsequent free electron plasma formation and relaxation of a single laser pulse will modify the material homogeneously according to the incoming beam profile. The base area of the circular (radius of $\sqrt {2} \,\textrm{mm}$) and elliptical (semi-axes of $1 \,\textrm{mm} \times 2 \,\textrm{mm}$) profiles are chosen to be equal in order to ensure the comparability between the simulations. The profile geometry was stretched by a factor of $\approx 1000$ compared to the real dimensions of the expected beam profile’s cross section. Thus, the qualitative stress distribution will be the same. The interaction of ultrashort laser pulses with matter can alter its features [26]. The behavior under stress impact is key element in our simulation. In particular, the elasticity of the glass in relation to the modified glass determines the internal stress build-up. This feature is described by Young’s modulus. Here, we assume that Young’s modulus of fused silica will decrease after ultrashort laser interaction based on the studies of Athanasiou and Bellouard [27].

 

Fig. 2. FEM simulation of von Mises stress distribution at the surface for symmetrical profile and elliptical orientations $(0^{\circ }, 30^{\circ }, 60^{\circ }, 90^{\circ })$. The arrows indicate the direction of applied force.

Download Full Size | PDF

Depicted are the von Mises stress gradients on the surface for planar force application of $F = 1 \,\textrm{Pa}$ in opposite y-direction along the two edges of the sample. Therefore, only tensile forces are applied. In addition to the simulations presented in Fig. 2, we performed similar calculations for boundary limitation of perfect interaction between the laser pulse and the material with a vanishing small Young’s modulus, i.e. a complete void inside the material. These calculations show basically identical results as already presented and are therefore not shown separately. The simulations reveal a distinct dependency of the geometrical shape and its orientation to the predominant direction and intensity profile. For perpendicular orientation of the long axis of an ellipse to the applied force, the maximum stress can be enhanced compared to the distribution induced by a cylindrical profile. Thus, the probability to generate a crack along this axis increases. Advanced processing strategies for separation along arbitrary contours will depend on this fact.

5. Asymmetric elongated beam profiles causing preferential crack orientation

The implementation of asymmetric elongated beam profiles has been discussed in the context of glass processing for diffractive and non-diffractive beam profiles [14,15,28]. The given examples use different techniques to generate the asymmetrical beam profile. One approach is spatial frequency filtering which results in the central mode as well as corresponding side-lobes, leading to an elliptical shape. Unfortunately, this is at the expense of efficiency because parts of the initial laser power are blocked [14]. Other approaches use controlled introduction of aberrations [15,28]. These processes would be challenging to handle in an industrial environment in terms of process stability. In the following, we present a novel concept for the efficient and robust generation of Bessel-like beams whose radial symmetry of the transverse intensity profile is intentionally broken.

A simple method for the flexible generation of Bessel-like beams is given by digital-holography [22,29] where the SLM acts as axicon hologram. A radial-symmetric transmission according to $T^{\textrm{axi}}\left (r\right )=\exp {\left [\imath \Phi ^{\textrm{axi}}\left (r\right )\right ]}=\exp {\left (\imath k_r r\right )}$ is displayed. In thin element approximation the radial component of the wavevector $k_r$ allows to directly assign a holographic axicon to its refractive counterpart via $k_r = {2}$${\uppi}$$\left (n-1\right )\gamma /{\lambda}$, with axicon angle $\gamma$ and refractive index $n$ [22,29]. The axicon-like elements used within this work exhibit opening angles of $\gamma = \left (1\ldots 3\right )^{\circ }$ which corresponds to $k_r = (50 \dots 150) \,\textrm{mm}^{-1}$. Illumination of these holograms by fundamental Gaussian beams generates Bessel-like beams of zero order. It is well known that minor modifications of $T^{\textrm{axi}}$ such as the multiplexing of phase vortices or $\uppi$-phase jumps allow the generation of Bessel-like beams of higher-order or superposition thereof [22,30]. Here, the required phase modulation $\Phi$ is no longer radial symmetric but exhibits additional azimuthal dependencies such as, e.g., constant azimuthal slopes $\Phi \left (r, \phi \right ) = k_rr+\ell \phi , \ell \in \mathbb {Z}$ [22,30].

The beam shaping approach pursued within this work to generate asymmetric Bessel-like beams represents a generalization of this concept. Now, the axicon-like phase transmission $T^{\textrm{gen}} = \exp {\left (\imath \Phi ^{\textrm{gen}}\right )}$ is characterized by a phase modulation with arbitrary azimuthal dependencies $\Theta \left (\phi \right )$

For the determination of $\Theta \left (\phi \right )$ an optimization algorithm is applied which starting from $\Phi ^{\textrm{gen}}\left (r,\phi \right )$ calculates the optical field at a well defined propagation distance. Here, an adapted merit function determines the deviations of the actual beam profile and a target intensity distribution. In our case, e.g., an elliptical on-axis intensity distribution, see Fig. 3 (d). To retrieve $\Phi ^{\textrm{gen}}\left (r,\phi \right )$ that minimizes the defined merit function it may be useful to implement a number of $j_{\textrm{max}}$ phase disturbances by defining angular segments of position $\alpha _j$, width $\Delta \beta _j$ and constant phase offset $\Theta _j$. As depicted by the selected example of Fig. 3 (c), $\Phi ^{\textrm{gen}}\left (r,\phi \right )$ is completely determined by the set of parameters $\left \{ \alpha _j, \Delta \beta _j, \Theta _j \right \}$ [32].

 

Fig. 3. The sketch (a) of the experimental setup used for modification and in situ diagnostics. Simulated transverse intensity distribution for a symmetrical Bessel-like beam (b). Schematic phase mask (c) to create nondiffracting beam profiles with arbitrary transverse intensity distribution. Examples of simulated intensity distributions for elliptical Bessel-like beam profiles (d) and (e) with a ratio of 1.5 ${\upmu}$m : 1 ${\upmu}$m for major to minor half-axis. The three-beam configuration (f) uses an asymmetrical envelope to achieve preferential directions. Left side shows the corresponding phase mask, the middle column the distribution in x-y-direction and right-sided to it the corresponding simulation along the z-axis.

Download Full Size | PDF

Examples of elliptical Bessel-like beams (elliptical central spots or an elliptical envelope of main and side lobes, cf. intensity threshold discussion of Sec. 5) and corresponding axicon-like phase transmissions $\Phi ^{\textrm{gen}}\left (r,\phi \right )$ are depicted in Fig. 3 (d)–3 (f). Here, linear field simulations in air show the transverse distributions $I\left (x, y\right )$ as well as corresponding longitudinal intensity cross sections $I\left (x, z\right )$ proving the non-diffracting propagation behavior along several hundred micrometers. Additionally, in order to demonstrate our concept’s high efficiency, we state peak intensity values relative to the fundamental Bessel-like beam [cf. Fig. 3 (b)] normalized to unit intensity $I_{\textrm{peak}}=1$. Please note, that depicted asymmetric beam profiles represent only a small selection from a plethora of Bessel-like solutions exhibiting almost arbitrary transverse intensity profiles, see Chen et al. [32] for additional information.

We employed pump-probe microscopy to gain access to the transient and dynamic interaction of the asymmetric beam profiles inside the material. The measurements in Fig. 4 (a) show the optical depth $\tau =\ln {\left ({I_S}/{I_0}\right )}$, i.e., extinction according to Lambert-Beer’s law [33], for observation transversely to the z- and either x- or y-axis (top or bottom) shown in Fig. 3 (f). All in situ-images throughout this paper are background corrected. The absorption behavior of ultrashort laser pulses inside glass is driven by nonlinear and linear absorption processes [2], whereas the intensity distribution simulations presented in Fig. 3 are solely linear. We expect the absorption to occur at positions where the simulations predict high intensities. By comparison of Fig. 4 (a) with Fig. 3 (f) the excellent agreement between simulations and experiments becomes clear and justifies the linear simulation approach. We used a double pulse (pulses $20 \,\textrm{ns}$ apart) with total energy of 60 ${\upmu}$J, $3 \,\textrm{ps}$ pulse duration and a temporal observation delay of $20 \,\textrm{ps}$ after the last of both pulses for the measurements presented in Fig. 4. The 3 partial beams can be detected clearly separated and show an homogeneous intensity distribution for the optical depth $\tau (x, z)$. The accumulated total intensity along the z-axis presented in Fig. 3 (f) shows a slightly reduced intensity for the 2 outer beams by a factor of $\approx 0.87$. This is also apparent by the reduced beam length of the outer beams in the measurements visible in Fig. 4 (a) and Visualization 1. By extending the pump-probe setup with a tomographic measurement technique [34] the asymmetric nature of the created profile can be thoroughly captured, cf. Figs. 4 (b)+4(c). The sinogram in Fig. 4 (b) shows optical depth representation $\tau _\theta \left (\theta , x_\theta \right )$ along the cut-line sketched in Fig. 4 (a) for 50 different rotation angles $\theta = 0^{\circ }..180^{\circ }$. Figure  4 (c) and Visualization 1 show the measured 3D distribution of the extinction coefficient $\kappa (\boldsymbol {r})$ reconstructed by inverse radon transformation.

 

Fig. 4. (a) Optical depth $\tau (z)$ measurements for an asymmetric beam profile consisting of 3 partial beams as shown in Fig. 3 (f) next to each other (up) and in a row (bottom). We used the tomographic setup to derive the sinogram (b) of $50$ projection-angles of the optical depth $\tau _\theta$ at the position indicated by the dotted line in (a). The reconstruction over the whole beam profile length (c) shows the 3D-extinction-measurement for single laser pulses at 20 ps after first laser-matter-interaction.

Download Full Size | PDF

To prove our considerations concerning the glass cutting, we imaged the elliptical beam profile [cf. Fig. 3(e)] $20\times$ demagnified by a $4f$-setup into glass. The transmission light microscope images in Fig. 5 (a) show permanent modifications for single shot laser pulses. The profile has been rotated counterclockwise in steps of $30^\circ$ over a total range from $-90^\circ$ (left) to $90^\circ$ (right). The nomenclature is given as in Fig. 2. Thus $90^\circ$ implies orientation of the ellipse’s major axis and crack generation parallel to the movement direction (x-axis).

 

Fig. 5. Permanent material modifications at the glass surface for single shot (a) and multi shot experiments (b) at different orientations of the asymmetric beam. Pump-probe images (c) show optical depth (left) and retardance (right) measurements for preferential orientation parallel ($90^{\circ }$) and perpendicular ($0^{\circ }$) to the axis of sample movement. The dashed line guides the eye along the area of the preceding crack. The scale bar in each picture represents 50 ${\upmu}$m.

Download Full Size | PDF

Each shot consisted of 8 pulses in a train configuration (burst) with total energy of approximately 180 ${\upmu}$J and pulse duration of 3 ps. The temporal delay between two pulses in a burst is $20 \,\textrm{ns}$. The permanent modifications visible in Fig. 5(a) show two distinct areas: a central spot and a narrow line that runs through this spot. We assume the latter to be cracks with a length of approximately $l \approx$ 10 ${\upmu}$m. A typical cleaving configuration is achieved by merging the single shots close to each other along a line. Figure 5 (b) shows different spatial pulse distances (14 ${\upmu}$m, 16 ${\upmu}$m and 18 ${\upmu}$m) between consecutive burst trains for either cracks orientated in-line with the sample movement direction ($90^{\circ }$) or perpendicular ($0^{\circ }$). The images with $90^\circ$-orientation show a continuous modification in the observation plane even for a spatial pulse distance of 18 ${\upmu}$m, which is larger than the permanent modification length $l$ for single shots would suggest.

Again, with the help of pump-probe microscopy this observation can be elucidated. The measurements in Fig. 5 (c) show the optical depth and polarization dependent measurements of the retardance to analyze the stress distributions [20]. The depicted data belongs to a measurement series of 240 consecutive burst trains, each recorded by the HSC. Each burst consists of 4 individual pulses (temporal delay of $20 \,\textrm{ns}$ between two pulses) with a total train energy of 120 ${\upmu}$J and $1 \,\textrm{ps}$ pulse duration. The time of observation is set to $5 \,\textrm{ns}$ after the last pulse in the burst train. Figure 5 (c) shows exemplarily the measurement after pulse train #80. The whole series is presented in Visualization 2. To further enhance the visibility by multiple pulses, the movement of the sample was set to $75 \,\textrm{mm}/\textrm{s}$ at a fix repetition rate of $50 \,\textrm{kHz}$. The optical depth measurements show higher values of $\tau (x, z) = \left (1 \dots 2\right )$ where the laser radiation was absorbed and decay processes, e.g., self-trapped excitons [35], are still ongoing [2]. While the $0^{\circ }$-orientation of the beam profile shows a smooth front of the absorption zone, the $90^{\circ }$-orientation has small irregularities, which may belong to still open cracks generated by the previous pulses [36]. Another difference is the preceding area indicated by the dotted line, slightly above the background level. This feature is even more pronounced if we compare the retardance measurements. Again, the $0^{\circ }$-orientation shows a smooth transition zone, where absorption is located. In addition, we see a build-up stress field with increased retardance at the position of the previous pulses (right side). This area is absent in the comparative measurement with rotated beam profile. Intriguingly, there is also an area travelling ahead of the absorption zone, which we interpret as a transient crack. This crack is aligned parallel to the long axis of the beam profile and prevents stress build-up. The formation of the crack and its propagation over 200 laser pulses is shown in Visualization 2.

To demonstrate the feasibility of our presented approach, glasses with different thicknesses were modified by asymmetrical beam profiles and separated. The arithmetic mean and standard deviation for measurements of 550 ${\upmu}$m, 1.1 mm Gorilla glass (Corning) and 3.8 mm borofloat glass (SCHOTT) presented in Fig. 6 are conducted with the same setup as presented in Fig. 1 (c). The modifications were inscribed at a constant repetition rate of 50 kHz for the different spatial pulse distances presented in Fig. 6. The pulse duration is $3 \,\textrm{ps}$ for each experiment. The pulse energy is equally split to a 4 pulse burst configuration and is adapted to the calculated peak intensities presented in Figs. 3 (b), 3(e) and 3(f). In addition, the energy for the profiles was set to guarantee modification along the whole glass thickness. Based on a burst energy of 40 ${\upmu}$J for the symmetrical profile at 0.55 ${\upmu}$m glass thickness, the elliptical profile (axis-ratio 1.5:1) and 3-spot profile use a burst energy of 90 ${\upmu}$J and 100 ${\upmu}$J, respectively. The energy was increased to approximately 500 ${\upmu}$J and $1.5 \,\textrm{mJ}$ for glass thicknesses of $1.1 \,\textrm{mm}$ and $3.8 \,\textrm{mm}$. In these cases a DOE generates the nondiffracting beam profiles instead of the SLM, due to the SLMs limited damage threshold as discussed in Sec.2.3. Each experiment of the thinner glass with 0.55 ${\upmu}$m was repeated 3 times, for 1.1 mm Gorilla glass samples 5 times and the 3.8 mm SCHOTT borofloat glass samples 2 times, respectively. The geometry of each glass sample was set to $100 \,\textrm{mm} \times 10 \,\textrm{mm} \times d$ for all the thicknesses d.

 

Fig. 6. Separability measurements of different modification distances for adapted pulse energies to the different thickness of 0.55 mm, 1.1 mm Corning Gorilla glass and 3.8 mm SCHOTT Borofloat glass. The complementary reflected light microscope images (i)-(iv) show the quality of the cleaving edges. The transverse intensity distribution of the symmetrical, elliptical and 3 spot profile is comparable to those shown in Figs. 3(b),3(e) and 3(f). The inset (*) compares the modifications induced by the central spot and side lobes with the expected instensity distribution from simulation.

Download Full Size | PDF

The inset shown in Fig. 6 (*) compares the permanent modification at the surface and the simulated intensity profile presented in Fig. 3 (d) with a binary colormap set to a threshold of $6 {\%}$ maximum intensity. We aligned the longitudinal position of the Bessel-like beam, that its highest intensity is located at the surface level. The comparison between experimental data and simulated intensity distribution match the elliptical central spot as well as various higher order side lobes. Here, the side lobes protrude more dramatically because the pulse energy was increased by a factor of 2 compared to the necessary energy to modify the sample and because of the shifted focal position with reduced damage threshold at the surface.

However, our presented approach to arbitrarily distribute the intensity in transverse direction also allows to make use of the side lobes. One possibility is to create an asymmetrical envelope instead of one central elliptical spot, as shown in Fig. 3 (f) and Fig. 6 (3-spot-profile). Remarkably, this may lead to a more efficient energy conversion in terms of summarized peak intensities. Another option is to use the side lobes to create an intensity-dependant preferential orientation by either addressing the elliptical central spot or the envelope by combined interaction with several side lobes. Our presented beam profiles are designed to keep the ratio of central spot to side lobes at a minimum. Thus, a strong influence of the side lobes, exemplarily shown in Fig. 6(*), on generating cracks with undesired orientation is not expected for the pulse energies applied during the cleaving applications. For the following separation measurements we have restricted ourselves to 3 different profile types: a rotationally symmetric profile acting as a reference and two asymmetric profiles, where the asymmetry is achieved either by an elliptical central spot or by the envelope of a 3 spot configuration, comparable to the profiles visible in Figs. 3(b), 3(e) and 3(f), respectively.

The separation results for the thinnest of the glass samples tested show the best separability for the elliptical profile. In addition, the spatial pulse distance between two modifications could be increased in the range of up to 15 ${\upmu}$m (20 ${\upmu}$ m) for the elliptical (3-spot) profile with only negligible deviations in separability. An increased pulse distance can be directly converted to an increased processing speed. Notably, these high spatial pulse distances are not possible to achieve with rotationally symmetric profiles, to the best of our knowledge. In the previous Section 2.3 we addressed the advantages and disadvantages of the SLM. Due to the damage problem at high-power applications, we used a DOE as beam shaping element to cleave the glass thickness of $1.1 \,\textrm{mm}$ and $3.8 \,\textrm{mm}$. This limited us to the comparison of the symmetrical and elliptical beam profile, cf. Fig. 6. Here the advantages of the tailored crack alignment of the elliptical beam profile become obvious. The force necessary to separate the modified glass sample is reduced by a factor of $\approx 2$. Additionally, the spatial pulse distance can be increased as well. Even distances up to 30 ${\upmu}$m were possible for glass thickness of $1.1 \,\textrm{mm}$ and still easier to separate compared to the reference.

Intriguingly, the quality by means of mean surface roughness $S_a$ shows similar values in the range of $S_{a} \approx$ 1 ${\upmu}$m throughout the range of tested spatial pulse distances. However, we measured an increasing peak-to-valley height for increasing spatial pulse distance. Furthermore, the insets in Fig. 6 (iii)+(iv) show that the cleaving edge has several unmodified areas. We would like to point out, that those areas feature a smooth cleaving edge after separation and will significantly contribute to the good surface roughness. Nevertheless, the investigated profiles promise either easier separation of the glass with similar quality to existing applications, or a high velocity laser-inscribe and break method, comparable to its mechanical antagonist. Of course, combinations in between the mentioned parameter regimes and further optimization of this new class of Bessel-like beams and process strategies are ongoing and may further extend the prevailing benefits.

6. Conclusion

To conclude, we analyzed the separation process of glass substrates with different thicknesses in this work. Incipient elementary studies demonstrated the possibility to cut glasses up to $12 \,\textrm{mm}$ thickness with elongated beam profiles, such as Bessel-like beams. However, we detected a linear increasing force required to successfully separate the modified samples. In order to overcome this detrimental feature, we optimized the targeted orientation of stress distribution inside the material by usage of asymmetric beam profiles using finite element method simulations. A new approach for maintaining the propagation properties of Bessel-like beams under arbitrary formation of the transverse intensity distribution is presented and used to generate new profiles with the desired asymmetric features implemented by digital-holography. The profiles can be used to intentionally orientate permanent cracks inside the glass. In situ diagnostics further revealed additional transient crack formation during the modification process. Final separation studies, validated with different glass compositions and thicknesses, showed an reduced force necessary to separate the sample up to a factor of 2 and the potential to increase processing times by an order of magnitude. These results underline the high potential introduced by this new class of Bessel-like beams and further optimization thereof in combination with new processing strategies could have a major impact to tailored cleaving applications.