1. INTRODUCTION

The separation of glass is an integral part in glass fabrication, where the desired shape and dimensions of the finial product are achieved. In this process of substrate transformation and modulation, several cleaving techniques are established. The mechanical scribe and break approach, as low-cost state of the art, has been used for decades. Here, a diamond wheel or point tool induces a kerf structure on the glass surface. Afterwards, the glass can be separated along the scribing path by applying the external force [1,2]. However, this method results in production defects on the glass edges [3,4], such as chipping [5], or mechanical breaking marks on the separation zone, such as Wallner lines [6,7]. Often these defects need to be removed because they significantly reduce the strength of a glass product [8,9]. Therefore, suitable post-processing techniques are required, such as grinding and polishing, which are very time-consuming and cost-intensive [10]. Other options for glass cutting are based on the application of various laser sources. One approach is the controlled fracture technique, based on a rapid heating and cooling. Therefore, a ${\mathrm{CO}}_2$ laser is focused onto the glass substrate surface and heats the material. A coolant directly cools it afterwards, producing strong tension stress. This thermally induced stress leads to the generation of cracks, guided along the laser trajectory and finally resulting in a substrate cleaving [2,11–13]. However, this process is limited to certain materials with high thermal conductivity [14]. Moreover, the high-quality cutting process is restricted to a maximum glass thicknesses of approximately 3 mm due to the finite penetration depth of the ${\mathrm{CO}}_2$ laser radiation [15,16].

Recent developments in the field of ultrafast lasers [17] in combination with novel beam-shaping methods [18–27] open up new approaches for glass processing. In particular, ultrashort laser pulses in the range of hundred femtoseconds to picoseconds offers the possibility for structuring inside-the-glass volume without damaging the surface. Due to the high intensities in the focal region and the resulting non-linear absorption processes, tailored modifications can be induced [28,29]. Thereby, this process is applicable to materials with various thermal expansion coefficients, substrate thicknesses, or lattice structures, and it can result in high-quality breaking edges.

In this paper, we use ultrashort laser pulses for inscribing in-volume modifications consisting of voids, cavities, and stress fields [23,30–32], which can be arranged as a breaking layer [23,24]. This approach has been demonstrated to work properly for 1-mm-thick samples. Here, we demonstrate how to scale the inscribing process of in-volume modifications to thicker glass samples up to 10 mm and more. Cumulative effects will be analyzed, and crucial parameters for homogeneous modification tracks will be identified.

2. EXPERIMENTAL SETUP AND METHODS

The laser source used in our experiment is a home-built ytterbium-doped ultrafast fiber laser system based on coherent combination of up to eight fiber laser amplifiers [17]. The laser operates at a central wavelength of 1030 nm, delivers up to 1 kW average power, and 300 fs pulse duration at a close-to-diffraction limited-beam quality with an $M^2$-value $<1.1$ on both axes. A burst energy of approximately 1.5 mJ can be extracted from each fiber in burst operation, allowing to create an output burst energy in excess of 10 mJ. The output beams from each amplifier are coherently combined using thin-film polarizers with adaptive phase control. This laser configuration allows to generate average-power and pulse-energy levels, inaccessible to single-emitter fiber lasers. For this particular experiment, the compressor was detuned to generate 19 ps pulses (full width at half-maximum) with pulse energies in the range from 1 to 10 mJ per burst, whereby the burst featured a decreasing energy slope. The pulse duration was experimentally identified to avoid filamentation [33] in air and inside the focusing optic. A typical extracted pulse train (measured with a photodiode) is depicted in Fig. 1 and features a decreasing energy envelope and an intra-burst pulse separation of 12.5 ns. In the field of ultrashort pulse laser material processing, bursts can be used to implement customized process strategies [34–37]. An intra-burst separation in the range of a few nanoseconds is particularly advantageous for in-volume processing compared to a pulse separation in the microsecond range [38]. One reason for this is a more favorable energy coupling of the following pulses into an excited material state, which results in an accumulative process [38–43].

 

Fig. 1. (a) Photodiode signal of the extracted pulse train with four (black) and eight (red) pulses per burst. Temporal separation between intra-burst pulses is 12.5 ns, and the burst features a decreasing energy slope. The red and black traces were shifted in time for better visualization. (b) Simulated ideal on-axis intensity distribution of a quasi-Bessel beam by a perfect axicon illumination with an opening angle of 20°, 400 μJ pulse energy, and 19 ps pulse duration.

Download Full Size | PDF

Preliminary investigations have shown that only weak material modifications can be achieved by the use of single and double pulses. This is in good agreement with other publications [24,35] in which bursts have also been used yielding stronger permanent modifications with significantly improved break results compared to single pulses. Therefore, the work carried out here only focuses on the use of bursts. In particular, we choose different burst regimes, consisting of either four or eight pulses.

To generate thin and long modifications inside the glass volume, as required for cleaving applications, the laser pulse has to be spatially shaped. Therefore, it passes an axicon with an opening angle of 20° (apex angle 140°), transforming the Gaussian beam into a Gaussian–Bessel beam with sufficient intensity to ionize the central region [19,21–23,44]. Neglecting aberrations, the ideal on-axis intensity distribution along the $z$-axis (propagation direction) can be expressed as [45]

 

Fig. 2. Schematic representation of the used setup, according to [47].

Download Full Size | PDF

The simulated ideal on-axis intensity distribution of the quasi-Bessel beam with perfect illumination by an axicon with 20° opening angle is displayed in Fig. 1(b), theoretically confined into a diameter of 1.22 μm. By using the described imaging system, a flipped intensity distribution is achieved compared to a single positive axicon [47]. The glass samples are polished fused silica blocks (High Purity Fused Silica standard grade, corning code 7980) with $15\mathrm{mm}\times 10\mathrm{mm}\times 20\mathrm{mm}$ as a typical glass with low thermal conductivity [8,48]. For each processing parameter set, the glass sample is replaced by a pristine fused silica block. The glass samples were translated underneath the laser beam with a three-axis positioning system (Aerotech ABL1000), with a maximum translation speed of 100 mm/s. To ensure single shot-modifications and avoid accumulating effects, modification-to-modification spacing of 10 μm was chosen by adjusting the laser repetition rate to 10 kHz. However, this could be easily scaled up to feed rates of 1 m/s or beyond by increasing the repetition rate to 100 kHz. While the laser system was able to provide pulses with the same parameters at such repetition rates, this was however beyond the capability of the positioning system used.

For post-process analysis, a bright-field microscope was used. This enabled an evaluation of the induced volume modifications before a separation process was initiated as well as a first analysis of the breaking edges after cleaving. A commercial strain analyzer (Ilis StrainMatic M4/60.13, wavelength 587 nm) was used to reveal the induced-stress fields expressed in differences of optical retardance. The strain analyzer spatial resolution $d_{\min }\approx 50\mathrm{\mu m}$ was too low to resolve single modifications, but it revealed large-scale material changes. As a result, possible positive and negative retardance values were measured in an integrated manner and resulted in an averaged measuring signal. The retardance error of our measurement was 1 nm.

3. RESULTS AND DISCUSSION

At first, we analyze the individual modifications inside the fused silica samples induced by single laser shots, using the axicon configuration described above and the two different burst regimes. As depicted in Fig. 3, varying the pulse energy up to approximately 3 mJ (energy of the first pulse within the burst), the other pulse amplitudes also rise with a fixed ratio according to Fig. 1(a), significantly increasing the modification length.

 

Fig. 3. Simulated and measured modification lengths obtained by applying 1030 nm, 19 ps pulses with different pulse energies. The graph was normalized to the energy of the first pulse within the burst.

Download Full Size | PDF

This dependence is in excellent agreement with a simulation based on Eq. (1). Therefore, we simply assume that modifications are obtained once a certain threshold intensity is overcome. By comparing the measurements with the theoretical prediction, a threshold of $I_0=1.8\times {10}^{17}\mathrm{W}/{\mathrm{m}}^2$ is obtained. This intensity threshold is overcome for a minimum pulse energy of 140 μJ (as can be seen by extrapolating the theoretical curve to zero modification length, see Fig. 3). However, in the experiments even for a pulse energy of 200 μJ, no visible modifications are obtained. This can be attributed to aberrations, alignment, and optics inaccuracies, especially due to the rounded tip of the axicon. Thus, a fraction of input beam is not properly converted into the Gaussian–Bessel beam and shows a modulation on the on-axis intensity distribution (Fig. 3), as described in [18,32,49]. This is in agreement with the observation that the measured modification diameter is $1.8\mathrm{\mu m}\pm 0.2\mathrm{\mu m}$ (constant over the entire modification track), which is slightly larger than the theoretically estimated value. It is interesting to note that the modification length is practically only dependent on the pulse energy of the first pulse within the burst. In addition, there are only marginal differences between the four and eight pulse bursts, yielding slightly longer modifications for the four-pulse burst.

A closer look at the modifications reveals an energy influence on the structure of the track. Applying a four-pulse burst with the first pulse energy of 365 μJ leads to an almost continuous and homogeneous modification track (length $\approx 7.4\mathrm{mm}$) along the $z$ direction, as depicted in Fig. 4(a). These tracks consist of precisely aligned micro-voids and cavities. Only at the beginning and end of the modification track, some imperfections (no or minimal void generation with $\approx 200\mathrm{\mu m}$ in length) can be observed [see Fig. 4(b)]. This behavior can be attributed to aberration induced deviations of the on-axis intensity distribution [18,49], yielding an intensity modulation and regions below the modification threshold. Applying bursts with first pulse energies around 930 μJ removes theses imperfections at the beginning and end of the modification tracks due to the increased on-axis intensity distribution. Furthermore, the overall length is increased to approximately 9.2 mm. However, the formerly smooth and continuous void formation now exhibit interruptions within the modification track. Noticeable are disruptions and bulges within single tracks. As a result of the void generation, additional material modifications like stress fields [31,41], heat-affected zones [29,39,50,51], and material rarefaction [28,52] are induced around the void. In our particular case, the material changes around the voids exhibit a range from 2 to 5 μm, caused by the injection of energy significantly above the modification threshold. As a result, these alterations influence the laser propagation on the next intended pristine modification position due to the induced stress fields, refractive index changes, and shielding effects on the larger bulges and disruptions. Maximum modification lengths of approximately 10.8 mm are obtained with the four-burst regime, where the first pulse carries an energy of 2.67 mJ. A diameter increase of the entire modification zone up to $\ge 5\mathrm{\mu m}$ is observed.

 

Fig. 4. (a) Microscopic images of homogeneous modification tracks before cleaving obtained with a four-pulse burst where the first pulse carries 365 μJ pulse energy. (b) Imperfections at the beginning of the modification track due to alignment and optics inaccuracies.

Download Full Size | PDF

In the next step, we inscribe modifications with eight pulses per burst. Here, the pulse energy for the first four pulses is comparable with the four-burst regime. Again, no modifications were obtained by applying pulses with a first pulse energy (215 μJ) slightly above the theoretical threshold. This is interesting to note, as the single pulse modification thresholds can be lowered by using several pulses within a burst train and more energy input. The change of the modification threshold value can be attributed to cumulating effects [39,40] or the generation of slow-relaxing defect states [43,53–55]. However, it becomes apparent that a further energy input of four additional pulses does not change the threshold for remaining modification.

At higher energies, permanent modifications are induced, which grow in length with energy, again in good agreement with the estimated trend (Fig. 3). Here, the extra energy supply of four additional pulses does not increases the overall modification length. However, cumulative effects on the permanent modifications are intensified at the eight-burst regime. As depicted in Fig. 5(b), disruptions are observable within every modification track at pulse energies of 980 μJ. As a consequence, the discontinuity of the inscribed structures becomes worse, and the diameter of the modification zone is $\ge 5\mathrm{\mu m}$.

 

Fig. 5. Microscopic images of modification tracks before cleaving obtained with a (a) four-burst train, where the first pulse carries 930 μJ pulse energy and (b) eight-burst train, where the first pulse carries 937 μJ first pulse energy. Because an extensive energy contribution above the modification threshold disruptions distorts the modification track and influences follow-up pulses, these disruptions are increased in diameter and number in the eight-burst regime.

Download Full Size | PDF

Finally, these disruptions accumulate to macroscopic defects with further energy rise (Fig. 6). Thereby, global changes are induced that affect several modifications tracks, yielding extensive discontinuities, unmodified areas, as well as several modification tracks that have melted together. It becomes obvious that these macroscopic disruptions hinder the undisturbed laser-glass interaction and results in an inhomogeneous intensity distribution. As mentioned before, the void generation leads to additional material modifications such as stress fields. To quantify the induced stress, we measured the optical retardance over the entire sample (Fig. 7). The detected optical path difference $\varphi$ can be converted into normed stress $\mathrm{\Sigma }=\varphi /(d_m\times \beta )$, where $d_m$ is the size of the modified region, and $\beta$ is the photoelastic coefficient of the material ($\beta =3.5{\mathrm{TPa}}^{-1}$ for fused silica [48,56]).

 

Fig. 6. Microscopic images of modification tracks before cleaving obtained with an eight-burst train, where the first pulse carries 2.85 mJ pulse energy. Due to an extensive energy contribution above the modification threshold, global disruptions interrupt the modification track and influence follow-up pulses.

Download Full Size | PDF

 

Fig. 7. Retardance measurement of the modification tracks before cleaving obtained with a 20° axicon and four- and eight-burst trains, respectively. The spatial resolution $d_{\min }\approx 50\mathrm{\mu m}$ is too low to resolve single modifications, which results in an averaged stress measurement of the global changes.

Download Full Size | PDF

The induced stress field with a four-burst pulse and 365 μJ first pulse energy is smooth with an averaged positive optical retardance of 4 to 5 nm. Only at the bottom of the modification track, a slightly increased retardance of 6 to 7 nm is obtained. This indicates higher intensities at the end of the interaction region, as simulated in Fig. 1(b). Looking at the entire interaction zone, a cumulated stress field of approximately 20 MPa results, almost homogeneous over the entire length. However, the low spatial resolution of the strain analyzer causes an averaged measurement over several modification tracks. Thus, it can be assumed that the measured values differ significantly from high-resolution images, and the interpretation of these data requires some caution. For example, it is known that the laser-damaged region in the bulk of fused silica can locally exhibit positive and negative refractive index changes (corresponding to largely varying strain fields) [29,39,57–59].

Modifications inscribed with the maximum first pulse energy of 2.67 mJ and a four-burst pulse reveal a doubled retardance values over the entire modification. It can be seen that slightly increased retardance values are measured especially in regions with microscopic disruptions [comparable to Fig. 5(a)]. At the end of the modification, an increased retardance of approximately 16 nm is measured, which correlates with the intensity distribution at this position.

Comparing the stress fields that are induced with an eight-pulse burst, clear differences can be seen [at Figs. 7(c) and 7(d)]. Instead of a homogeneous stress field alternating stripes of weak ($0\dots 4\mathrm{nm}$) or strong ($8\dots 16\mathrm{nm}$) retardance can be observed along the laser propagation direction, almost like a sinusoidal wave.

These stripes correlate with the transversal regions with increased energy input, where several tracks showed extensive disruptions, molten as well as non-modified regions. As described above, an energy supply significantly above the modification threshold leads to a local laser beam influence, which means that some areas cannot be adequately structured. Other effects like diffraction on the sample edge can be excluded. A further increase in the first pulse energy to 2.85 mJ results in an enhanced uneven stress distribution.

These extensive disruptions influence the cleaving process and induce uneven breaking edges with large laser-affected areas. First cleaving tests show a good agreement of the resulting breaking edges (see Fig. 8) with the microscope images of the modification traces.

 

Fig. 8. Microscopic images of the breaking edge after laser inscription obtained with (a) a four-pulse burst, where the first pulse carries 365 μJ pulse energy and (b) with an eight-burst train, where the first pulse carries 2.85 mJ pulse energy.

Download Full Size | PDF

As shown in Fig. 8(a), an in-volume modification with moderate laser parameters (four-pulse burst where the first pulse carries 365 μJ pulse energy) leads to a uniform fracture pattern, whereby the individual modification traces are still visible. However, an energy input significantly above the modification threshold (eight-burst train, where the first pulse carries 2.85 mJ pulse energy) leads to very rough fracture edges with re-melted areas [see Fig. 8(b)]. To evaluate further fracture and glass-relevant properties (roughness, breaking load, chipping, etc.), fracture tests with high statistical significance have to be carried out in future.

4. CONCLUSION

The controlled and uniform energy deposition is a key factor for high-aspect and high-quality in-volume modifications of glass. In this paper, an ultrafast fiber laser system with pulse energies up to 3 mJ and two burst regimes consisting of four or eight pulses, was used. The laser beam was transformed to a Gaussian–Bessel beam via an axicon with a 20° opening angle. Smooth and even modification traces consisting of voids and cavities can be inscribed in a length of approximately 7.4 mm, with four pulses and the first pulse energy of 365 μJ. Experiments with pulse energies close to 3 mJ result in modification lengths of approximately 10.8 mm. In this process, disruptions were generated when the injected energy significantly exceed the modification threshold. Thus, homogeneous modification tracks, obtained with lower pulse energies, were interrupted and show enlarged modification diameters and bulges. The injection of eight pulses boost the cumulative disruptions and the transversal modification zone exceeding a range of 5 μm. In turn, these irregular modifications interfere with the injected laser pulse and lead to large scale disruptions and molten tracks along the translation direction. A subsequent investigation of the optical retardance confirms the microscopically obtained findings.