1. Introduction

Processing of dielectrics is a growing technological need. Hard and brittle materials such as sapphire are important for next generation touchscreens or to be used as substrates for LED growth [1]. While the ablative processes with lasers [2–4] are slow and generate dust particles, infrared ultrafast lasers can be used to process dielectrics within the bulk and allow for material separation [5]. In this context, structured light has already show high benefits to control how the energy is deposited within the material bulk. Diffraction-free Bessel beams allow for depositing high energy density almost all along the material thickness in a propagation-invariant way [6]. The remarkable property of Bessel beams is that they resist to nonlinear distortions, in contrast with Gaussian beams that undergo complex spatio-temporal dynamics and reshaping [7, 8]. Single shot illumination at moderate pulse energy (typ. μJ scale) already allows for drilling high apect ratio nanochannels in glass. In sapphire crystal, ultrafast Bessel beams generate high aspect ratio nano-voids that are completely enclosed within the material’s bulk [9]. In amorphous glasses, drilling a series of nanochannels very close one to each other defines a weakened plane in the material which can make it cleave at very high processing speed [10–13].

Here, we develop a novel concept of laser separation of crystalline materials based on the control of fracture planes induced by structured light, ie nondiffracting zeroth order Bessel beams. With energies on the order of tens of μJ, Bessel beam can create fractures in sapphire, consistently all along the Bessel zone. Specifically, we show that femtosecond pulses can generate cracks in a single direction, in contrast with longer pulses yielding cracks along several axes. Orientation-controlled crack formation in glass with Bessel beams has been recently reported but its formation was linked to the ellipticity of the axicon [14]. Here, in sapphire crystal, the orientation of the crack can be controlled with the polarization of the beam with respect to the crystalline axes and the scan direction. We demonstrate that the cracks created by subsequent pulses can join to form a plane and eventually cleave the material.

Crystalline materials can undergo cleavage cracking after single shot irradiation and for sufficiently high density of deposited energy, along the crystallographic axes [15,16]. It originates from a tensile stress created within the material [17]. This approach can become an enabling technology because the extent of the cracks can be on the order of tens of microns for a single illumination pulse. If the cracks can be joined in a line, the processing speed could be very high (tens of cm per second), even for lasers with low repetition rates (1–10 kHz). This overcomes the speed of standard mechanical scribe-and-cleave processes with more flexibility on the geometry.

2. Experimental setup

Our experimental setup is described in detail in reference [18]. The laser source is a chirped pulse amplified laser (SpectraPhysics Spitfire Pro V) emitting ~140 fs pulses at 800 nm central wavelength with 5 kHz repetition rate. Beam shaping was realized with a spatial light modulator (Hamamatsu PAL-SLM) associated to a telescope arrangement allowing both Fourier filtering and precisely controlling the position of the Bessel beam inside the sapphire samples. The Bessel beam is ~ 32 μm long in air, with a cone angle of 26 degrees, which produces a central lobe of ~ 0.7 μm in diameter. The polarization of the central lobe of the Bessel beam is linear. The polarization orientation was controlled after Fourier filtering by a half-waveplate. The pulse duration was controlled between ~140 fs and 3 ps by chirping the pulse with the laser compressor. Pulse duration was measured at sample site. Single shot operation was controlled by an independent pulse picker inserted after the laser source. The samples are monocrystalline sapphire (Shinkosha Co), C-cut, double-side polished with a thickness of 150 μm.

Figure 1 shows the beam formation in the sample and the relative positioning of the crystalline axes with the sample translation. The sample was translated between each laser shot. The inter-damage distance was varied between 15 and 50 μm for a series of damages processed under identical conditions. Between 2 series, the spacing is 250 μm so that the series are fully independent one to each other.

 

Fig. 1 Concept. The Bessel beam creates a crack extending all along the length of the Bessel zone. In the ultrashort pulse regime, this generates a fracture along a plane with a single pulse. We show how these can be aligned to form a cleavage plane along the full sample or wafer. The crystallographic axes A₁,A₂,A₃ are shown as solid black lines, as defined in reference [1].

Download Full Size | PDF

3. Results

For 140 fs pulse duration, the threshold for damage formation is 1.2 μJ and the threshold for the onset of crack formation is typically 2.4 μJ (corresponding to launched intensities ~ 10¹⁴ W/cm²). With 3 ps pulses, the threshold for damage is ~0.6 μJ and it is ~2.3 μJ for crack formation. We have investigated the regime of crack formation for energies up to 60 μJ. Crack formation was observed whatever the beam position: inside the sample or crossing the top or exit surfaces. All further Figs show the results for a Bessel beam fully within the material bulk.

In a first experiment, the translation is along x axis, the polarization is set orthogonal to the translation direction, ie along y direction and the crystallographic axes are oriented as shown in Fig. 1. We will see later how these were determined. Figure 2 shows the evolution of the cracks morphology with the pulse duration for a pulse energy of 20 μJ and a shot to shot spacing of 30 μm. Two important conclusions can be drawn. First, it is apparent that the morphology strongly evolves with pulse duration, which was never reported before. We observe that short pulse durations create cracks in a single direction while long pulse durations create more cracks, up to 3 different directions. The ultrashort pulse regime is valid up to ~650 fs.

 

Fig. 2 Evolution of the crack morphology with pulse duration. The pulse energy is 20 μJ, the shot to shot separation is 30 μm. The laser polarization and crystallographic axes are oriented as in Fig. 1.

Download Full Size | PDF

The second important conclusion is that crack formation can be controllable so as to follow only a single line. We note that the shot to shot spacing does influence the morphology of the crack. But in the long pulse regime, a reduction of the shot to shot distance never allowed to reduce the number of cracks below 2.

For a pulse energy of 2.4 μJ, the cracks were typically few μm long and ~25 μm at an energy of 20 μJ. We note that this is of the same order of magnitude as the cracks reported in LiF and MgO produced with highly focused Gaussian beams [15, 17].

The picosecond regime generates cracks mostly along 3 axes at 120 degrees one to each other, which is compatible with the orientations of the crystalline directions of the hexagonal lattice of oxygen ions in sapphire [1]. In contrast, the femtosecond regime leaves only one direction of cracking, which is one of the previous three. This comparison was repeated for pulse energies up to 60 μJ with the same qualitative results. Increasing the pulse energy only increases the crack length, but does not change the number of cracks.

The difference between the two regimes can originate from the difference in the plasma morphology. Indeed, at lower pulse energy, we reported in reference [9] that the cross section of nanochannels created in sapphire by femtosecond pulses is elliptical with the major axis following the polarization while under picosecond illumination, circularly symmetric damages were created.

In the following, we investigate the influence of the polarization direction in the 140 fs regime for the application to material separation. Figures 3(a),3(b) and 3(c) show the evolution of the crack morphology with two different relative directions between the crystalline axes and the sample translation direction for three different polarization directions. We first note that whatever the linear polarization orientation, the crack always follows the crystallographic axes. We observe that either one or two cracks form, but it is only in the case where the sample translation follows a crystallographic axis with polarization orthogonal to the translation that a single crack is always unambiguously observed. In the second part of the same Fig. 3(d), 3(e) and 3(f), the material has been rotated by 90 degrees. While the respective angle between the laser polarization and the crystallographic axis is the same in 3(d) and 3(e) as respectively in 3(b) and 3(a), the crack direction is different. We note that similar results were obtained when writing along y direction without rotating the sample. The first impact does not show any difference with the other ones. We conclude that the cracking orientation is also guided by the cracks formed previously in the material at surrounding sites, as was observed in reference [17]. Additional experiments are needed to fully uncover the effect of the scanning direction and speed, we anticipate that the kinetics of crack formation is also an important parameter. As a summary of our present results, the cracking direction is determined by 3 parameters in the femtosecond regime: polarization direction, crystallographic axes directions, scanning direction.

 

Fig. 3 Influence of the polarization and crystalline axes directions. Optical microscopy view of the cracks generated for 140 fs, 20μJ pulses with polarization orientation of 90 deg in (a)and (d), 0 deg in (b)and (e) and 120 deg in (c) and (f) with respect to x axis, shown in Fig. 1. In (a),(b),(c) the main axes are oriented as in Fig. 1, while in (d),(e),(f), the sample was rotated by 90 degrees.

Download Full Size | PDF

We further investigate the possibility of joining the cracks by increasing the pulse energy and reducing the shot to shot distance. The results are shown in Fig. 4 where we varied these two parameters. The geometrical configuration is the same as the one showed in Fig. 4(a). It is apparent that a line of joining cracks can be processed in several configurations. However, we note that reducing the distance below the one where the cracks would naturally join, creates additional shorter cracks in other directions (e.g. see cases with 15 μm spacing). An optimal case for single crack joining in a line was found as pulse energy of 20 μJ with 25 μm spacing. We processed a line in these conditions over the whole width of 20 mm samples. The samples cleaved under extremely weak mechanical bending or vibration. Under a 3 points bending test, the deflection at fracture was less than 50 μrad.

 

Fig. 4 Dependence of the crack morphology with spacing and pulse energy. The pulse duration is 140 fs. Polarization and crystallographic axes are oriented as in Fig. 1.

Download Full Size | PDF

Figure 5 shows SEM images of a sample cleaved with these parameters, from top 5(a) and side 5(b) views. It is very apparent that the cleaved edge has no chipping. These images confirm that cracking is effective along the full ~ 50 μm length of the Bessel beam inside sapphire. The largest surface roughness is observed around the laser-induced damages. The maximal roughness is estimated from SEM images to a few μm.

 

Fig. 5 Scanning Electron Microscopy (SEM) imaging of a cleaved sample (a) Top view (b) Side view. The pulse separation is 25 μm. The energy per pulse is 20 μJ. Polarization and crystallographic axes are oriented as in Fig. 1. (c) Triangular cut out of a sapphire wafer observed by optical microscopy. (d),(e),(f) show different views of triangular sample under SEM imaging.

Download Full Size | PDF

This technique is therefore applicable to material cleaving, which we implemented to create a triangular cut of the material along 3 of the crystallographic axes as can be observed from Figs. 5(c) to 5(f). The polarization and the processing direction were rotated by 120 degrees to produce each segment of the triangle. Figure 5(f) features the extreme sharpness of the angle created : the radius of curvature of the angle is below 0.5 μm.

4. Conclusion

We have reported that femtosecond pulses can be used to control single crack orientation in sapphire, unlike picosecond pulses. The crack morphology depends on the relative orientation of three parameters: polarization, crystalline axes and scanning direction. When the scanning is operated along one of the directions of the crystalline axes with a polarization orthogonal to it, a single crack is formed along the scanning direction. Bessel beams allow for creating homogeneous damages all along the depth of the material. The control on the crack formation enabled processing a fracture planed, used for sapphire separation. This opens the route for high speed processing and separation of sapphire substrates. We expect that longer Bessel zones can be used to process thicker samples. We believe this approach can be extended to the case of other transparent crystals.