**Improvement of Etching Anisotropy in Fused Silica by Double-Pulse Fabrication**

#### **Valdemar Stankeviˇc \* , Jonas Karosas, Gediminas Raˇciukaitis and Paulius Geˇcys**

Center for Physical Sciences and Technology, Savanoriu Ave 231, LT-02300 Vilnius, Lithuania; jonaskarosas919@gmail.com (J.K.); g.raciukaitis@ftmc.lt (G.R.); p.gecys@ftmc.lt (P.G.)

**\*** Correspondence: valdemar.stankevic@ftmc.lt

Received: 20 April 2020; Accepted: 6 May 2020; Published: 8 May 2020

**Abstract:** Femtosecond laser-induced selective etching (FLISE) is a promising technology for fabrication of a wide range of optical, mechanical and microfluidic devices. Various etching conditions, together with significant process optimisations, have already been demonstrated. However, the FLISE technology still faces severe limitations for a wide range of applications due to limited processing speed and polarization-dependent etching. In this article, we report our novel results on the double-pulse processing approach on the improvement of chemical etching anisotropy and >30% faster processing speed in fused silica. The effects of pulse delay and pulse duration were investigated for further understanding of the relations between nanograting formation and etching. The internal sub-surface modifications were recorded with double cross-polarised pulses of a femtosecond laser, and a new nanograting morphology (grid-like) was demonstrated by precisely adjusting the processing parameters in a narrow processing window. It was suggested that this grid-like morphology impacts the etching anisotropy, which could be improved by varying the delay between two orthogonally polarized laser pulses.

**Keywords:** femtosecond; fused silica; double pulses; selective chemical etching

#### **1. Introduction**

Within the past two decades, femtosecond pulse processing of transparent materials has demonstrated outstanding results [1,2]. Many various processing technologies were developed for marking, dicing, cutting and internal modifications of transparent materials. Femtosecond laser-induced selective etching (FLISE) was demonstrated for the first time in 2001 by Marcinkeviˇcius et al. when a track recorded by a femtosecond laser inside fused silica was selectively etched along the modified region [3]. That was the start of subtractive manufacturing in transparent materials, mainly in bulk fused silica by forming microchannels and complex 3D structures. It was found, within two years, that the modifications are composed of self-organised nanogratings oriented perpendicular to the laser polarisation [4]. The detailed investigations showed that selective etching is related to the nanograting orientation [5] and molecular oxygen in nanopores within the nanogratings [6]. The latter discoveries prompted many FLISE optimisation works to be conducted [7–10], and FLISE has become a very promising technology for a wide range of complex applications, such as internal 3D structures for mechanical, optical and microfluidic devices [11–13]. The most used and investigated material for laser-induced chemical etching is fused silica.

New opportunities came from double-pulse fabrication approaches recently introduced for the processing of various materials. Double pulses with different wavelengths [14] or double-pulse laser technology for material ablation [15] gained the increased attention for the processing of transparent materials [16]. Pengjun et al. demonstrated that the temporal shaping of the femtosecond pulses could provide higher etching selectivity, and, at some defined pulse energies, the etching rate could be a few times higher compared to conventional single pulses [17] due to the better efficiency of photon absorption. The promising results were also achieved by combining double pulses with a linear and circular polarisation where the improvement in the etching rate for the formation of high aspect ratio channels was demonstrated [18]. In most publications it is assumed that the first pulse is responsible for the nanograting orientation [14,19]. However, the real situation could be more complicated. The nanograting formation in fused silica is usually related to the enhanced birefringence. Atoosa et al. demonstrated that orthogonally-polarised double-pulses could reduce the birefringence because the nanograting orientation is determined by the writing pulse with a higher intensity [20]. More recent work also confirmed that the second pulse could rewrite the nanograting orientation when the double pulses with non-equal energies are applied [21]. However, there are still limitations of the FLISE technology due to etching anisotropy (etching rate dependence on laser writing direction at a constant linear polarisation) and low processing (laser writing and etching) speed. The efforts to overcome this drawback was already made for single-pulse processing by varying the polarization direction during fabrication [22], or by changing laser pulse duration to the picosecond range [23]; however, the problem was not solved completely.

In this work, we investigate the double-pulse processing approach with crossed polarisations and variable inter-pulse delay. Initially, peculiarities of nanograting morphologies recorded with double pulses were investigated. Hereafter, we show the influence of the pulse duration and inter-pulse delay to the selective etching of the microchannels. In most applications, the laser writing trajectory has a curved shape. The nanogratings are statically orientated along the curved patch. Therefore, at different trajectory position, the nanograting orientation is shifted relative to the trajectory vector, and, as a consequence, the orientation-dependent etching takes place. To overcome this drawback, the double pulses with crossed polarisations were used, and the etching anisotropy and etching rate improvement were demonstrated. Particular experiments were arranged to fabricate the vertical bow-like structures—two-dimensional structures formed by raster scanning of a bow-like horizontal shape starting beneath the sample bottom surface and moving laterally through the laser focus up to the top sample surface. To our best knowledge, we demonstrate the new grid-like nanograting morphology for the first time and the possible explanation provided. That is an entirely new phenomenon that was not described earlier.

#### **2. Materials and Methods**

The setup for double-pulse experiments is schematically illustrated in Figure 1. The micromachining workstation with an integrated femtosecond laser (Pharos-6W, Light Conversion, Vilnius, Lithuania) was used. The ultrashort laser operated at a 1030 nm wavelength, and the 500 kHz pulse repetition rate was mainly used during the experiments. The laser was equipped with a tunable compressor which allowed the pulse duration to be tuned in the range of ~0.26 to ~10 ps with a positive or negative chirp. The laser power was controlled using an external attenuator which was calibrated to vary the mean laser power linearly. The laser beam was split into two pulses with a polarising beam splitter (PBS). The energy ratio between two pulses was controlled by rotating a λ/2 phase plate. During the tests, the energy ratio was set to 1:1. The Mach–Zehnder interferometer setup, composed from a variable arm (VA) and reference arm (RA), was used to combine two laser beams with a controllable temporal delay between pulses. The polarisation in the reference arm was set to *E<sup>y</sup>* and that in the variable arm to *Ex*. Afterwards, the beam diameter was reduced to ~2 mm using a two-lens telescope system and focused to the sample using the 100× microscope objective (M Plan NIR, Mitutoyo, Kanagawa, Japan) with a numerical aperture (NA) of 0.5. The microscope objective was translated with a linear translation stage (ANT130-L, Aerotech, Pittsburgh, PA, USA) along the Z direction. The 20× camera vision objective (Olympus Plan Achromat, Tokyo, Japan) was mounted on the same stage. A sample was mounted on a two-axis gimbal mount (GM200, Thorlabs, Mölndal, Sweden) and translated by a high-resolution (>500 nm) XY linear stage (ANT130-XY, Aerotech, Pittsburgh, PA, USA) with a scan speed up to 5 mm/s. The translation stages were controlled via the controller (A3200, Aerotech, Pittsburgh, PA, USA).

−

*‖*

**Figure 1.** The double-pulse experimental setup. P—Brewster angle polarizer, PBS—polarising beam splitter, BS—beam splitter, D—dump, OB—microscope objective.

*λ θ* The delay range from –20 ps to 20 ps was possible to achieve, between the two pulses, a temporal delay resolution of ~7 fs. The delay was estimated relative to the reference arm; consequently, the negative delay was achieved when the variable arm was shorter with respect to the reference arm (the reference-arm pulse was the second). During the tests, the delay was varied from −10 ps to +10 ps. In further text, to avoid confusion, the following indication will be used: *E*k: polarization parallel to the scanning direction, and *E*⊥: polarization perpendicular to the scanning direction. The pulse order is indicated as the first or second one. To obtain the 0 fs delay, the double-pulse setup was calibrated temporally (Figure 2). Initially, the temporal delay was calibrated by changing the length of the delay line with a slight angular misalignment between the beams and registering the intensity profile with a CCD camera (Spiricon SP620U, Ophir, Jerusalem, Israel). When the delay in the variable arm approached the ~0 fs delay (i.e., two laser pulses overlap entirely in time), the clear interference intensity pattern was registered. When the laser pulses overlapped only partly (~160 fs), the interference was still observed. However, the contrast was very poor. For relatively long delays (>300 fs), the laser beams did not interfere. The interference condition also was not satisfied for cross-polarised beams, and the non-interfering double-pulse beam profile was observed even at the 0 fs pulse delay.

The spatial beam position adjustment was performed by recording the interference pattern at the 0 fs delay and measuring the interference period. According to the relation of the interference pattern and the beam angle, *d* = λ/2*sin*θ, it was possible to calculate the angle between two beams. The interference period increased when the angle decreased. Therefore, two beams appeared almost parallel after a few iterations of mirror angle adjustment.

**Figure 2.** Spatial and temporal beam alignment: (**a**) two-beam angle adjustment principle; (**b**) intensity profiles of two beams with different temporal delay and polarisation orientation; (**c**) interference patterns of two beams with various misalignment angles at the 0 fs delay.

Commercially available fused silica (JGS1, 20 mm × 3 mm × 2 mm, Eksma Optics, Vilnius) samples were used for the experiments. The modifications were recorded by focused laser beam ~20 µm below the sample surface at the regimes when nanogratings were formed (Type II modification) [24]. A few parallel lines shifted by a z-step (1 µm) were recorded to ensure that the right vertical cross-section of the nanogratings would be investigated after the cutting and polishing procedure (Figure 3a). The exposed nanogratings were etched for 1 min in 5% hydrofluoric acid (HF) acid and coated with a ~10 nm gold layer. The prepared samples were investigated by scanning electron microscope (SEM) (JEOL, JSM6490LV, Tokyo, Japan).

**Figure 3.** Schematic illustration of experiments: (**a**) recording and polishing of the lines written with double pulses for nanogratings observation by scanning electron microscope (SEM); (**b**) the line writing procedures for the investigation of the double-pulse influence on the chemical etching rate; (**c**) recording of the vertical bow-like structures for the investigation of directional etching dependence.

To investigate the double-pulse delay effects to the chemical etching rate of the fused silica samples, a group of lines was written by varying the pulse energy and scan speed at a constant repetition rate. The separate lines were recorded at different focusing depth (50–500 µm below the sample surface) with the vertical raster-scanned surface ending at the sample top surface in the middle of the recorded track to get the acid access to the modified area (Figure 3b). Three lines in each group were recorded under the same conditions and the measured data were averaged. The directional etching dependence on the double-pulse delay was investigated recording the bow-like vertical structures (Figure 3c) by a raster scanning the single bow-like (curved) trajectory laterally moving through the laser focus with a 7 µm vertical z step, starting beneath the bottom sample surface and ending on the sample top surface.

After the irradiation with the focused ultrashort laser pulses, the fused silica samples were immersed into a wet-etching bath of HF with a concentration of 10% and 30 ◦C temperature for 30 min. After the chemical etching, the samples with the vertical bow-like structures were polished in the XZ plane with a 0.3 µm grade colloid silica to recover a smooth, transparent surface for the high-contrast microscope measurements.

#### **3. Results and Discussion**

#### *3.1. Etching Rate of Modifications Recorded with Double Femtosecond Pulses*

Multiple lines were inscribed in the bulk of fused silica, as shown in Figure 3a to investigate the effects of the inter-pulse delay, pulse energy, and focusing depth to the etching rate of the double-pulse modified material. The group of lines were recorded with different pulse density ranging from 100 to 2000 ppµ (pulses/µm), keeping the laser pulse energy constant. Various laser pulse energies were applied from 100 nJ to 600 nJ for each group of lines. It should be noted, that the indicated pulse energy was measured when two pulses were combined into one optical path, so the pulse energy for separate beams was twice lower. That was done to allow a direct comparison of the results with the single-pulse processing. In the single-pulse processing experiments, one of the beams was blocked.

In our previous works, the maximum etching rate of the fused silica was ~1300 µm/h with the leading etching selectivity of ~120:1, when processing with the 1030 nm wavelength was applied [25,26]. The results were achieved with ~400–600 nJ pulse energy and 1000 ppµ density. The results of the double-pulse processing in fused silica are demonstrated in Figure 4. The visual observation showed that even for double-pulse processing, the pulse duration had a strong influence on the etching rate. At 0 fs pulse delay and various pulse durations, we observed the etching rate variation when the pulse density was changed. A lower etching rate variation was observed for longer pulse durations independently on the chirp direction.

**Figure 4.** The etching rate dependence on the delay between two pulses for modifications recorded in fused silica with the 200–400 nJ pulse energy and orthogonal polarisations: (**a**) 600 fs pulse duration with a negative chirp; (**b**) 290 fs pulse duration; (**c**) 600 fs pulse duration with a positive chirp; (**d**) The microscope picture of etched channels for the 0 fs pulse delay at 290 fs pulse duration. Red curve: the etching rate dependence on the pulse energy for a single pulse. Samples were etched 30 min in 10% diluted hydrofluoric acid (HF) acid. The inset shows the SEM pictures from the side (**b**,**c**) and top (**d**).

− The more surprising result was observed when the pulse delay was varied from −10 ps to +10 ps. As demonstrated in Figure 4a–c, the etching rate significantly reduced when the delay between two pulses approached 0 fs. For the positive and negative delays, the etching rate grew almost symmetrically when the pulse duration was 600 fs. For the shortest pulse duration (~290 fs with compensated chirp)

and negative delays, the etching rate increase was higher compared to the positive delays. In the case of negative delay, the first beam coming from the variable arm was with the parallel polarisation (*E*<sup>k</sup> **first**), and only then the second beam from the reference arm with the perpendicular polarization (*E*<sup>⊥</sup> **second**) was arriving. The etching rate difference could be related to the different modification thresholds for the nanograting formation, depending on the beam polarisation [6]. Therefore, the significant influence of the second pulse (*E*<sup>⊥</sup> **second**) to the nanogratings orientation in the negative delay range was observed. Saturation of the etching rate was observed when the pulse delay approached ~10 ps independently on the delay direction. More objective etching rate comparisons could be made by introducing the etching rate contrast parameter ∆*R*, which is described as the ratio of the difference between the maximum and minimum etching rate and maximum etching rate as follows <sup>∆</sup>*<sup>R</sup>* <sup>=</sup> (*Rmax* <sup>−</sup> *<sup>R</sup>min*)/*Rmax*. According to this description, the etching contrast for the 600 fs pulse duration was consequently <sup>∆</sup>*R*−600fs <sup>=</sup> 0.25 and ∆*R*+600fs = 0.44. The maximum etching rate contrast was achieved for the shortest pulse duration ∆*R*290fs = 0.83. The etching rate contrast showed how much the pulse delay influenced the etching rate: the lower etching rate contrast value, the weaker the pulse delay influence on the etching rate was observed. For longer pulse durations, the broader pulse delay range with a lower etching rate was found compared to the shorter pulse duration. Longer pulses had a wider temporal overlap range that confirms the observations of the highest etching rate drop only in the area where the temporal pulses overlap. It could be noted that the etching rate contrast drop was usually observed for higher pulse energies. By comparing the etching results of the single-pulse and double-pulse experiments, we can distinguish a significant difference in the etching rate. While for the single-pulse processing at the same pulse energy, the 1300–1400 µm/h etching rate was the maximum value that could be achieved, the double-pulse processing enhanced the etching rate up to 2000–2100 µm/h. That yielded >30% of the etching rate increase. This achievement is even more attractive as this rate enhancement can be achieved with a lower pulse density of 100–500 ppµ (pulse per micrometre), which allowed us to speed up the processing more than two times. Murata et al. demonstrated that for the double-pulse processing, the diameter of emerging nanopores is twice time larger comparing to the single-pulse processing [27]. We believe that such behaviour had substantial input to the etching rate increase in our case due to the higher area of nanopores. *‖* Δ Δ *−* Δ Δ Δ

To understand the etching rate enhancement, the in-depth investigations of the Type II modifications induced using double pulses at various delay times were performed. Figure 5 shows the nanograting morphology dependence on the delay between two pulses.

**Figure 5.** SEM pictures of the nanograting morphology evolution depending on the delay between two pulses. The inset shows the enlarged area of the grid-like nanograting structure. The nanostructures recorded with a total 400 nJ pulse energy (200 nJ + 200 nJ).

For the 0 fs delay between two pulses, the nanogratings consisted of parallel lines rotated by ~45◦ relative to the scanning direction. When temporal pulse overlap was in the range of laser pulse duration (±250 fs, for ~290 fs pulse duration), the 45◦ nanograting orientation was still observed. The nanogratings were slightly shifted from the straight line and appeared bow-shaped indicating the nonperfect spatial beam overlap. When the positive delay exceeded the pulse duration, the evident influence of the first pulse (*E*<sup>⊥</sup> **first**) on the nanograting formation was observed, that confirmed the results described by Rohloff et al. [28]. They found that, at low fluences, the orientation of the laser-induced periodic surface structures (LIPSS) structures on the fused silica surface changed their direction by 90◦ when delay direction changed, and it was determined by the first pulse polarization.

In our case, when the delay between laser pulses approached to −10 ps (*E*<sup>k</sup> **first**), the grid-like structures were observed, which was an unexpected result considering the previous investigations [19,29]. This result was replicated a few times and was repeatable. To our best knowledge, the grid-like structures in the bulk fused silica were demonstrated for the first time, and there are no valuable phenomena to explain. We can speculate that the mentioned structures appear as a consequence of the different modification thresholds for two polarisations. According to previous investigations, the Type II modification threshold for perpendicular polarisation is ~2 times lower compared to the parallel polarisation [6]. The electron plasma generated by the first pulse absorbs more efficiently the second pulse due to the reduced ionisation threshold [30]. As a consequence, the already created point defects, such as colour centres (E') and nonbridging oxygen hole centres (NBOHCs) [31], interact with the electrical field of the second pulse. Such interaction could enhance the electron plasma more effectively due to the lower modification threshold for perpendicular polarisation and a memory effect [30]. However, as it is known, at least a few tens of pulses are required to rewrite the nanogratings [32]. The polarisation changed after each pulse in the double-pulse regime, and that was not sufficient to overwrite the nanograting orientation. When the delay approached the −10 ps, the orientation of the inhomogeneities was created and defined by the electrical field orientation of the first pulse (*E*<sup>k</sup> **first**) due to formed STE (self-trapped excitons) and not relaxed states still at a high temperature which involves the viscous flow of the silica. Hence, the induced periodic electron plasma pattern by the second pulse (*E*<sup>⊥</sup> **second**) with the perpendicular polarisation records the periodical nanopattern on top of the already oriented nanogratings partly covering them due to the lower modification threshold.

From another point of view, according to the numerical solution of Maxwell's equation in the vicinity of subsurface planar cracks oriented normal to the surface of fused silica [33] (created nanogratings in our case), the higher field enhancement is predicted (approx. by a factor of 2) inside narrow cracks (nanoplates) which are oriented perpendicular to the laser polarisation. In the case of negative delay (*E*<sup>k</sup> **first**), the more significant field enhancement is predicted for the first pulse (*E*<sup>k</sup> **first**) with the parallel polarisation (it first creates nanogratings perpendicular to the polarisation). However, due to a higher modification threshold for the same polarisation, the second pulse (*E*<sup>⊥</sup> **second**) should generate a higher amount of free electrons and create the grid-like nanostructures. In the case of positive delays (*E*<sup>⊥</sup> **first**), the influence of the first pulse strongly prevails due to higher light enhancement factor for the first pulse polarization (it first creates nanogratings perpendicular to the polarization) and lower modification threshold. Therefore, only a small part of the second pulse (*E*<sup>k</sup> **second**) induced nanogratings is noticeable. However, for longer delays, the second pulse (*E*<sup>k</sup> **second**) could have a stronger influence. As follows from Figure 4, the reduced etching rate for positive delays was observed.

The mechanism of the tilted nanogratings for the 0 fs delay can be simply explained in the following way: for spatially and temporarily adjusted beam the nanogratings are ~45◦ tilted relative to the X or Y polarisations, this is caused by the superposition of two pulses with orthogonal polarisations with no phase delay between two pulses, and the resultant vector is rotated by 45◦ [34,35]. The smallest step size of the delay line micrometre stage was 1 µm that changed the length of the delay line for double-length and corresponded to the ~7 fs delay. The resolution is not sufficient to precisely control the phase delay between two pulses, however for the exceptional case of 0 fs delay, the phase between two pulses could be set accurately due to the registration of the interference pattern of the slightly misaligned pulses. In this case, even the phase delay is set not accurately, and the elliptical polarization dominates, the morphology of the nanogratings still was normal to its major axis [36].

For the small beam misalignment in the +Y or −Y directions, the nanogratings were of the bow shape (Figure 6) due to partial interaction of the beams and domination of the opposite polarisations in the modification of top and bottom areas. For the intermediate case, the nanogratings were the result of both beams. For longer delays, the pulses did not interact coherently. Peculiarities in the nanograting morphology were defined by the interaction of the second pulse with the material excited by the first pulse. −

**Figure 6.** The pictorial explanation of the nanograting shape only for temporally adjusted pulses (0 fs delay) when a slight spatial beam misalignment is induced in the Y direction.

#### *3.2. Directional Etching of Vertical Bow-Like Structures*

. Th In previous sections, we discussed the etching of the single microchannels recorded with the polarisation normal to the scan direction or utilising double-pulses with orthogonal polarisations. To investigate the double-pulse effect to the writing direction, bow-like vertical structures were inscribed in fused silica (Figure 3c). This way, the beam polarisation angle was constantly shifting relative to the writing direction of the bow-like curvature. Therefore, the different etching rate was obtained for the single-beam case depending on the location on the curve [22]. We were expecting that double pulses with orthogonal polarisations could suppress this effect due to grid-like structure formation and enhanced nanopores formation [27]. It should be considered, that in this recording configuration, additional influence of the line stacking along the beam propagation direction had an impact on the etching rate.

− The vertical bow-like structures were recorded with the different double-pulse delay from −10 ps to +10 ps in fused silica. The separation between single vertical lines was set to 7 µm and was constant during all experiments. For one set of the double-pulse delay, an array of vertical surfaces with two different pulse densities of 500 and 1000 ppµ and four different pulse energies 200, 400, 600 and 800 nJ were recorded. The structures were etched 30 min in 10% HF acid. The etching behaviours of the vertical structures are demonstrated in Figure 7. Due to disturbance in the laser beam profile by the spherical aberration, a weak sample etching from the bottom surface was observed for the lower pulse energies (200–400 nJ). For higher pulse energies, the etching rate from both bottom and top surfaces was comparable at a delay in the range from −10 ps to 0 ps. However, the bottom structure etching rate

*‖*

−

was lower at the delay from 0 ps to 10 ps. The etching dependence on the direction was minimised for negative delays (*E*<sup>k</sup> **first**) (for etching from the top and bottom surfaces). In the case of positive delays (*E*<sup>⊥</sup> **first**), the directional etching effect was significant, demonstrating the lower etching in the middle part of the vertical structure (Figure 8b, left inset).

**Figure 7.** The microscope measurement of the etching dependence on the double-pulse delay for the bow-like vertical structures from the sample side (ZX plane): (**a**) the structures recorded with 200 nJ and 400 nJ pulse energy and (**b**) the structures inscribed with 600 nJ and 800 nJ pulse energy. Two channels with 500 and 1000 ppµ density were recorded at the constant pulse energy. Etching performed for 30 min. in 10% HF. The inset shows the bow-like trajectory view on the XY plane with a predicted nanogratings orientations.

− *‖ ‖* For a more objective analysis, we introduce the etching isotropy factor *K*, which demonstrates the etching difference of the bow-like vertical structures in the middle part and corner parts (Figure 8a) *K* = 2*L*M/(*L*<sup>R</sup> + *L*L), where *L*M, *L*<sup>R</sup> and *L*<sup>L</sup> are consequently the etched depth at the middle, right and left structure part. When *K* is approaching 1, uniform etching is obtained. For comparison, we recorded the bow-like vertical structures with a single pulse (Figure 8a). Usually, for this regime, the etching isotropy factor was below 0.7 and going down when the pulse energy or pulse density was decreased. For the 600 nJ pulse energy, the etching isotropy factors for single and double-pulse recording were 0.62 and 0.84, respectively. The 26% gain in the etching isotropy was obtained. The etching isotropy factor increased by raising the pulse energy for the positive pulse delays (*E*<sup>⊥</sup> **first**) (Figure 8b). However, *K* > 1 value was observed in the delay range from −0.3 ps to 0.3 ps, i.e., within the temporal pulse overlap range. The etching rate measurement also demonstrated that the maximal etching rate of ~ 1500 µm/h was achieved usually for negative delays (*E*<sup>k</sup> **first**). The difference in etching rate was marginal and close to the maximal value for the positive delays (*E*<sup>⊥</sup> **first**) in some cases. In comparison for the single-pulse regime, the maximal achieved etching rate was ~1000 µm/h, which is ~33% slower and agrees well with the single channels etching results. Therefore, the nanograting orientations in the double-pulse regime played a significant role and enabled easier etchant penetration to the modified area. The etching behaviour for different delays could be confirmed by the nanograting investigation in Figure 5, where the grid-like nanogratings were demonstrated. On the left or right side of the bow-like structure, the speed vector made up the different angle with the polarisation. When on the middle part the first pulse polarisation is parallel (*E*<sup>k</sup> **first**), on the left or right side it tends to be perpendicular, that made the nanogratings to be along the scan direction. That behaviour suppressed the directional

etching dependence. Contrary, in the positive delays (*E*<sup>⊥</sup> **first**), the first pulse influence should be leading in the middle part of the surface, where the nanogratings along the scan vector should be formed, consequently at right and left side the grid-like nanogratings were pronounced. According to this description, the faster etching rate was predicted in the middle structure part, as we can observe for the pulse delays from 0 ps to +0.47 ps. However, for the longer positive delays, the opposite behaviour was found. We could speculate, that the second pulse (*E*<sup>k</sup> **second**) also had a substantial impact on the nanograting formation, where the middle structure part was less etched. Especially the second pulse influence to the nanogratings direction was observed for the longer delays (>10 ps) and higher second pulse energy [21].

*‖*

**Figure 8.** Etching of bow-like structures fabricated with the single-pulse regime at 600 nJ and 800 nJ pulse energy (**a**) and dependence of the etching isotropy factor *K* on the pulse delay for etching from the top sample surface of the vertical surface. (**b**). The insets show the microscope pictures with etched vertical surfaces.

#### **4. Conclusions**

**‖** The etching peculiarities of modifications in the fused silica recorded with the variable delay double-pulses were observed and discussed. The double-pulse fabrication enhanced the etching rate by >30% compared to the single-pulse fabrication. The detailed analysis of the nanostructures in the fused silica revealed the appearance of the grid-like nanostructures when the first pulse polarization was along scan direction (*E*<sup>k</sup> **first**, negative delay). The nanogratings morphology explains the etching rate dependence on the pulse delay. The etching rate raised for negative and positive delays and was

suppressed at the 0 fs delay, where the 45◦ oriented nanostructures are formed. The application of double pulses significantly improved the etching isotropy of the vertical bow-like structures. It was measured qualitatively by introducing the etching isotropy factor that showed the value near 1 for negative delays. We demonstrated that the double-pulse processing technique is a simple fabrication method that improves the etching isotropy without a need for phase plate rotation during processing. The etching rate improvement is even more attractive as this rate enhancement can be achieved with lower pulse density of 100–500 ppµ, which allows to speed up the processing more than two times.

**Author Contributions:** Conceptualization: V.S.; methodology: V.S. and J.K.; investigation: V.S. and J.K.; validation: V.S., P.G. and G.R.; formal analysis: V.S.; resources: V.S.; data curation: V.S.; writing—original draft preparation: V.S.; writing—review and editing: P.G. and G.R.; visualization: V.S and P.G.; supervision: P.G. and G.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **One-Step Femtosecond Laser Stealth Dicing of Quartz**

#### **Caterina Gaudiuso 1,\*, Annalisa Volpe <sup>2</sup> and Antonio Ancona <sup>1</sup>**


Received: 24 January 2020; Accepted: 21 March 2020; Published: 22 March 2020

**Abstract:** We report on a one-step method for cutting 250-µm-thick quartz plates using highly focused ultrashort laser pulses with a duration of 200 fs and a wavelength of 1030 nm. We show that the repetition rate, the scan speed, the pulse overlap and the pulse energy directly influence the cutting process and quality. Therefore, a suitable choice of these parameters was necessary to get single-pass stealth dicing with neat and flat cut edges. The mechanism behind the stealth dicing process was ascribed to tensile stresses generated by the relaxation of the compressive stresses originated in the laser beam focal volume during irradiation in the bulk material. Such stresses produced micro-fractures whose controlled propagation along the laser beam path led to cutting of the samples.

**Keywords:** ultrashort laser pulses; heat accumulation; transparent materials; quartz; stealth dicing

#### **1. Introduction**

Transparent materials are used in an increasing number of applications, ranging from microelectronics [1], to microfluidics [2,3] and optoelectronics [4]. In particular, glass, quartz and sapphire, due to their broad spectral band of light transmission, hardness and scratch-resistance, are good candidates for display components of portable mobile electronics, as light emitting diode (LED) substrates, protection mirrors of mobile phone cameras, smartwatches, etc. Despite their growing use in consumer electronics products, cutting these materials is still a challenging task, due to their brittleness. Obtaining a high-quality cut edge completely free from micro-cracks or chips is not easy, particularly when employing thinner substrates. The state of the art cutting technologies most widely used include traditional mechanical cutting, chemical etching, electrochemical machining and laser-based methods.

Traditional machining methods are mainly based on diamond cutting [4]. Here, the material is first marked and scribed with a diamond tool and then an external force is applied to break the substrate along the scribing path [5]. Unfortunately, with this method, at high cutting speeds the diamond blades can generate chipping and cracks, which compromise the quality of the cutting edges, reducing the resistance of the materials by up to 60% in the case of glass [6]. Moreover, tool wear affects the repeatability and the efficiency of the dicing process [7], and additional grinding and polishing steps are needed to achieve the required smooth finish.

Another method for cutting transparent materials is based on the selective chemical etching of laser modifications inside the volume [8]. The main advantage of this technique, besides the lack of debris, is the possibility of designing complex shapes. However, several modifications are required along the entire thickness of the substrate, thus limiting the processing speed. Furthermore, hydrofluoric acid (HF) is the only etchant which attacks amorphous SiO2, quartz or glasses at significantly high etching rates [9]. Unfortunately, due to its high toxicity and corrosiveness, when hydrogen fluoride (HF) is used for large scale productions, many countries require very strict safety regulations, see e.g., [10,11].

Electrochemical methods for machining quartz were reported by several authors [12–15]. Jain et al. [12] exploited electrochemical spark machining (ECSM) showing that, depending on

the polarity and the applied voltage, a cut kerf ranging approximately from 0.5 to 0.9 mm and a surface roughness from 3 to 14 µm were obtained on a 2 mm-thick quartz sample. Wang et al. [15] demonstrated the shape cutting of quartz glass by wire electrochemical discharge machining (WECDM), but with this technique, significant bulges were generated on the edge of the cut circle.

In laser processing methods, the energy of a focused laser beam is exploited to modify the substrate from its surface or in the bulk, thus leading to its separation. Laser-based techniques are making huge advances in the field of cutting transparent materials, as they have numerous advantages over mechanical methods. Being non-contact, laser processes avoid any effect due to tool wear and mechanical stress, achieving high quality and precision of the cut at reasonable costs. Furthermore, the use of the laser prevents any contamination of the materials being processed. However, the process parameters, and in particular the wavelength of the laser source and its pulse width, must be carefully selected and adjusted to obtain the desired results and cutting-edge quality [16].

One of the first laser-based methods was introduced by Garibotti in 1963 [17] and consists in laser scribing and dicing. It is a two-step process: the laser is focused on the surface of the workpiece to generate a groove, which is subsequently fractured by applying a tensile stress. The mechanical stress that initiates the cutting process can be generated either by an external mechanical force or by a rapid heating-cooling cycle [10]. Serdyukov et al. [18] reported on numerical simulation about the controlled thermal cleavage of crystalline quartz, produced by laser heating and exposure to coolant. The cut was ascribed to the thermoelastic stress produced in the quartz crystals, due to the anisotropy of the thermal conductivity and thermal expansion. Such interpretation was confirmed through experiments performed using a CO<sup>2</sup> laser, which led to the successful cut of quartz crystal, though with the presence of evident chips and micro-cracks.

In some cases, the laser irradiation itself provides enough thermal stress to cause the cutting of the sample. Such cutting process is referred to as thermal cleavage [19]. The focused laser spot acts, indeed, as a localized heat source generating a thermo-mechanical compressive stress, whose relaxation causes the material to separate along the laser scanning path [20]. The fracture mechanism is similar to a crack extension, which can deviate from the desired path, especially for fast cutting speeds and long cutting lengths [11]. After being optimized in order to control the path of the stress-induced fracture, such a technique has been successfully applied to cut various materials, including silicon [21], alumina ceramics [22] and glass [23]. Xu et al. [19] demonstrated the laser thermal cleavage of sapphire substrate wafers using a CO<sup>2</sup> laser source. Here, a groove was engraved by laser ablation on the substrate surface. Laser irradiation caused localized and rapid heating and cooling, which led to the generation of local micro-cracks that propagated along preferential directions, with consequent cleaving and cutting of the substrate. Although attractive, this experimental procedure is very complex, since it requires cooling of the substrate and accurate alignment of the laser engrave to achieve the desired cutting path. Moreover, a protective layer must be applied to avoid surface contamination by redeposited laser ablation debris. An indirect, two steps thermal cleavage process has been developed by Choi et al. [16] for cutting glass, using a near-infrared (NIR) nanosecond laser. In this case, a laser-induced plasma plume was produced by irradiating a sacrificial absorbing layer, positioned a few hundred microns far from the glass target. Such a plume allowed the improvement of the absorption of laser energy by the target substrate, thus achieving localized and rapid heating and cooling. By delivering the appropriate amount of laser energy generating the plasma plume, controlled local microcracks were induced, which resulted in the cutting of the glass target.

A different approach was used by Russ et al. for cutting thin and ultra-thin glass plates (i.e., a thickness of a few hundred micrometers). Using ultrashort laser pulses, they ablated, layer-by-layer, the entire thickness of the substrate along the desired contour path [24]. This approach does not require applying any tensile stress and is suitable for any geometry, however it does not produce perpendicular cutting edges, has limited processing speed and produces a large amount of ablation debris [25]. Vanagas et al. used an analogous approach for the cutting of quartz and borosilicate cover glass samples, using 150 fs laser pulses at the wavelength of 800 m and repetition rate of 1 kHz. They demonstrated the feasibility of the laser cutting of quartz, although the processing speed was 200 µm/s and the laser-induced damage on the rear surface, caused by the multiple overlapped scanned paths, was quite extended [26].

Stealth dicing, instead, consists of focusing the laser beam inside the bulk material, transparent to the laser wavelength, and moving it along the desired path that acts as the initial division line, when an external tensile stress is subsequently applied. The process is ablation-free, does not generate any debris, and is extremely fast. The successful stealth dicing of thin sapphire wafers (350 µm) has been demonstrated using a fs-laser in the NIR wavelengths focused through a microscope objective [1]. However, single-focus stealth dicing laser processing does not always ensure a precise control of the taper, thus compromising the quality of the cutting edge. The laser cutting of thin borosilicate glass slides using a commercial fs-laser has been tested [27], finding that the fine cut can be successfully obtained by carefully adjusting the scan speed, in order to trigger the formation of micro-cracks at the exit of the laser beam from the sample. The generation of such microcracks was ascribed to the damage induced by filamentation and the consequent mechanical stress built-up in the material. Moreover, many studies have explored multifocal laser processing, mainly using a combination of diffractive optical elements (i.e., Fresnel lens) and Bessel beams [22,23]. Tsai et al. [28] demonstrated the cutting of a thin glass with a thickness of 100 µm through modification in the bulk volume, using a femtosecond laser Bessel beam and applying a breaking stress. A smooth cut edge with chipping <1 µm was obtained. Dudutis et al. [29] proved the possibility to use a picosecond laser for glass dicing using an axicon-generated asymmetrical Bessel beam. A tilt movement was applied to the axicon holder in order to add an astigmatic aberration, producing a Bessel beam with controlled asymmetry. This solution allowed the achievement of 2.8 times faster dicing speed and the lower propagation of cracks into the bulk material, compared to dicing with a symmetrical Bessel beam. The stealth dicing of sapphire using ultrashort pulsed Bessel beams has also been demonstrated by Lopez et al. [30], but achieving a sidewall roughness of around 2 µm.

In this work, we investigate the cutting of quartz by ultrashort laser pulses. Quartz is a material of relevant interest in many fields like e.g., optoelectronics, fibre-technology and photoacoustics [31,32], thanks to its high optical transmission in a broad range from ultraviolet (UV) to mid- infrared (MIR), unique thermal and electrical properties, besides an excellent chemical resistance. The high precision cutting of quartz is crucial for many applications that require the fabrication of miniaturized quartz-based devices. The traditional method for cutting quartz is based on lithography followed by wet etching, using highly toxic agents like ammonium bifluoride [17] or HF acid [8]. The laser-based cutting method proposed in this work is a clean, single pass stealth dicing process, which does not involve any chemical agent or external tensile stress. The dependence of the main laser process parameters as the repetition rate, the pulse overlap and the pulse energy on the cutting efficiency were experimentally investigated. The quality of the cut edge has been thoroughly analyzed by optical microscopy. Optical profilometry has been exploited to evaluate the roughness of the cut edge and the flatness of the final cuts.

#### **2. Materials and Methods**

The set-up used for these experiments is shown in Figure 1. The laser source was the Pharos SP 1.5 from Light Conversion, providing 200 fs pulses with variable repetition rate from a single pulse to 1 MHz. The almost diffraction limited laser beam (M2<1.3) was characterized by a central wavelength of 1030 nm and had a maximum average power of 6 W and maximum pulse energy *E<sup>p</sup>* of 1.5 mJ.

The linearly polarized exit beam passed through a half-wave plate and a polarizer, which allowed one to tune the average power by rotating the half waveplate. Next, the laser beam was sent to a microscope objective with a focal length of 8 mm (the estimated focused spot diameter *d* in air was 1.3 µm) mounted on a PC controlled motorized axis (Aerotech, ANT130 LZS, Pittsburgh, PA, USA). This enabled the beam focus to be finely positioned in the bulk of the transparent samples that were moved on a XY plane, perpendicular to the beam axis by two Aerotech ABL1500 motorized stages with

sub-micrometer resolution. As samples, 250 µm thick Z-cut quartz plates from Nano Quartz Wafer were used. In the present experimental conditions, self-focusing is likely experienced at a distance from the sample surface estimable as [33]:

$$z\_{\rm sf} = \frac{2n\_0(d/2)^2}{\lambda} \frac{1}{\sqrt{P/P\_{\rm cr}}} \tag{1}$$

where *n*<sup>0</sup> is the linear refractive index and λ the wavelength, P is the applied peak power and *P*cr=11 MW is the critical power for having self-focusing, determined according to Eq. 7.1.1 in [33]. Considering that the applied peak power ranged from 75 MW to 175 MW, the calculated self-focusing distance ranged between 0.31 µm and 0.47 µm. Correspondingly, the spot radius at such distance was determined according to [34]:

$$w\_{\rm sf}^2(z\_{\rm sf}) = (d/2)^2 \left[ \left( 1 + \frac{z\_{\rm sf}}{R\_0} \right)^2 + \frac{4\gamma}{k^2 (d/2)^4} z\_{\rm sf}^2 \right] \tag{2}$$

where *R*<sup>0</sup> is the curvature radius, *k* is the wave number in linear media and γ = 1 + α −2 <sup>1</sup> <sup>−</sup> *P P*cr , with α the degree of global coherence. In the present case, considering α→∞, the estimated beam spot radius due to self-focusing varied from 0.42 µm to 0.52 µm at the self-focusing distances, according to the applied peak power. However, it is important to highlight that the aberration occurring from focusing the laser beam deep inside the quartz sample could cause deviations from a diffraction-limited beam spot, by extending the focal region along the optical axis. 

‐ **Figure 1.** Experimental set-up used for the cutting experiments.

‐ The translation speed *v* and the repetition rate *f* defined the average number of pulses pps (pulses per spot) impinging on the same focal area, calculated as pps = *w*sf*\*f*/*v*.

 μ The repetition rate, the scan speed and the pulse energy were varied and their influence on the cutting process was evaluated, in order to better understand the underlying physical mechanisms.

#### ‐ μ ‐ **3. Results and Discussion**

‐

#### *3.1. Influence of the Repetition Rate and pps*

ୱ ଶ

ሺୱሻ ൌ ሺ/2ሻ

ୱ ൌ ଶబሺௗ/ଶሻ మ ఒ ଵ ඥ/ౙ౨ ሻ λ ‐ It was found that the repetition rate plays a fundamental role to get the laser-induced dicing of the quartz plates in a single step. Indeed, the repetition rate defines the time separation between two successive laser pulses and together with the travel speed and the pulse energy, determines the amount of laser energy and heat released in a given material volume per unit time.

‐

ൌ ሺ1ିଶሻ ቀ1 െ

‐

ౙ౨ ቁ

‐ μ μ Using a repetition rate of 25 kHz, the single-step cutting of quartz was never achieved regardless the travel speed and, thus, the number of pps. At higher repetition rates, stealth dicing of the plates

 ൰ ଶ 

‐ μ μ ‐

4 <sup>ଶ</sup>ሺ/2ሻ ସ ୱ ଶ 

ଶ

*α α*

occurred with quite different cutting edge qualities, depending on the pulse energy and the number of pps.

‐

In Figure 2, the optical microscope images of the cut edges obtained on quartz at 50 kHz and 100 kHz, with a pulse energy of 20 µJ and pps of 48 and 96, respectively, are shown. All the cuts exhibit a damaged area beneath the surface, in the region where the laser focus was placed. The areas above and below appear to be rather clean. μ

 μ **Figure 2.** Optical microscope images of the cut edges obtained with a fixed pulse energy *Ep*=20 µJ and different repetition rates of 50 kHz and 100 kHz and pulse overlap pps= 48 and 96 (**a**–**d**).

‐ ‐ ‐ ΔT Several physical mechanisms, generally taking place when irradiating a transparent material with intense ultrashort laser pulses, might have concurred to get the single-step stealth dicing of quartz. Besides self-focusing, also aberration from focusing deep inside the bulk quartz and filamentation typically occur at the peak power levels used in our experimental conditions [35]. However, the laser damaged area is quite confined inside the bulk material. Therefore, beam filamentation unlikely occurred along the whole quartz thickness. It is much more plausible that the physical mechanism originating the single pass stealth dicing is an accumulation of laser-induced stresses. In fact, in agreement with [36], it can be estimated that the peak temperature reached by the quartz lattice after each laser pulse is around 10<sup>3</sup> K. At such a high temperature, a transient tensile stress is generated, whose magnitude can be estimated by the following formula [16]:

ൌ

ν

$$
\sigma = \frac{Ea\Delta\Gamma}{1-\nu} \tag{3}
$$

‐ ‐ μ where *E* is the Young modulus, α is the coefficient of thermal expansion, ∆T is the temperature increase due to irradiation, and ν is the Poisson's number. The magnitude of such stress, which has been calculated taking into account the specific experimental conditions, is of the order of some hundreds of MPa. Literature reports a formation of cracks when the stress exceeds 1 GPa [37]. Therefore, the generation of a crack is not expected after each single pulse. However, since the single-pulse laser-generated mechanical stress lasts longer than 10 µs [38], if a second pulse arrives before such stress is released, then the two stresses accumulate. Pulse after pulse, this stress accumulation mechanism leads to the generation of cracks. The joining and propagation of cracks following the laser path and throughout the entire thickness of the quartz plate originates the single step stealth dicing process.

This explanation justifies the different results obtained at the three investigated values of repetition rate. In fact, at 25 kHz, the time delay between subsequent pulses is much longer than the relaxation time of each laser induced stress. Therefore, the stresses of following pulses do not overlap, and dicing of the sample does not take place. As far as the repetition rate increases, the pulse-to-pulse time interval shortens and the laser-induced stresses start to accumulate, finally resulting in the formation of cracks and dicing. At a repetition rate of 50 kHz, the time delay between consecutive pulses is half of the previous case and the self-induced stealth dicing of the samples was successfully obtained, with overall acceptable quality of the cut edges, especially for pps = 48. 

In this case, the area modified by the laser interaction is well confined in the bulk volume and can be easily recognized, as shown in Figure 2a. Here, the average areal roughness Sa was found to be around 1 µm. No significant collateral damage is noticed above or below the laser modification trace and a much lower surface roughness of around 0.05 µm was found. By increasing the number of pps to 96, the quality of the cut edge is still acceptable, but some scratches begin to be noticed above and below the laser modified volume, see Figure 2b. ‐ ‐ ‐ ‐

At an even higher repetition rate of 100 kHz, the laser induced crack propagation mechanism is no longer under control, owing to the excessive thermal load released into the focal volume [39]. This causes significant collateral damage around the laser absorption area, with unacceptable quality of the cut edges, as shown in Figure 2c,d. In particular, in the top part of the cut edge, long cracks propagating all the way towards the surface can be noticed at the lower number of pps = 48. For higher pps = 96, a large number of very dense erosion lines, accompanied by microcracks bridging them, appear in the upper part of the cut edge. For both investigated pps values, at 100 kHz of repetition rate, the lower part of the cut edge is completely destroyed, with big chips of quartz that have detached from the bottom. μ μ

The role of the overlap between pulses has been further investigated by performing additional experiments at the repetition rate of 50 kHz, keeping the same pulse energy of 20 µJ and reducing the number of pps to 24 and 10, respectively. Even in these two cases, the single step laser stealth dicing of the quartz samples was achieved. The corresponding optical microscope images of the cut edges are shown in Figure 3. As for pps = 48, the average areal roughness Sa of the laser modified zone was equal to 1 µm for both samples, while the surrounding area showed a surface roughness around 0.05 µm. However, compared to the case of pps = 48 shown in Figure 2a, where the cut edge was almost perfectly flat without any evident imperfection, a reduction of the number of pps has not led to further improvement of the cut quality. Unexpectedly, some small scratches or micro-cracks appear close to the laser modified area, suggesting that besides the repetition rate, the number of pps must also be selected within an ideal range to obtain a clean single step cut. On the other hand, a slight decrease of the laser damaged area depth, from approximately 50 µm to 35 µm, is noticed when decreasing the pps from 48 to 10. This is ascribable to the well-known incubation effect [40]. As the number of impinging pulses increases, the damage threshold decreases through the creation of point defects. Those defects enhance absorption of subsequent pulses, thus improving the coupling of laser energy into the lattice [40]. As a consequence, the dimensions of the laser-modified area increase with the number of pps. In addition, an increase of the pps leads to an increment of the accumulated fluence (defined as the pps multiplied by the fluence of the single pulse), which may change the morphology of the laser damage trace [41]. μ μ μ ‐ μ μ ‐ ‐

 μ **Figure 3.** Optical microscope images of the cut edges obtained with pps = 24 and 10, at 50 kHz repetition rate and a pulse energy of 20 µJ. The reduction of the pps from 24 to 10 causes the cut edge not to be perpendicular to the target surface. In (**a**), the image is almost entirely on focus. In (**b**), the part of the cutting edge above the laser damage trace is out of focus, thus indicating a different height with respect to the part below.

#### *3.2. Influence of the Pulse Energy*

The influence of the pulse energy on the overall cutting mechanism and the cut edge quality has been investigated by carrying out experiments at 50 kHz of repetition rate, 1 mm/s of translation speed and varying the pulse energy from 15 to 40 µJ, with increments of 5 µJ. Indeed, below 15 µJ, the pulse energy was too low to get the single step cutting of the samples, thus indicating that it is a threshold process. μ μ μ

The corresponding cut edges are shown in Figure 4, where the top and bottom views are also presented. For Ep = 15 µJ (Figure 4a) a single step stealth dicing cut with acceptable quality of the cut edge was obtained, except for some small imperfections on the top of the laser modified area, which had a depth of 45 µm. The best cut quality was obtained at 20 µJ, where a clean cleavage is observed, with a 50-µm-deep laser modified zone buried inside the sample thickness. The cut is straight with well-defined edges following the laser path, and very few surface defects, as can be noticed from the top and bottom view of the sample, unlike the case reported by Vanagas et al. [26], where a damage area on the sample rear surface of approximately 500 µm wide was observed. A comparison with the top and bottom view of the same sample before irradiation (Figure 5) does not show any significant difference. Therefore, it can be excluded that when focusing the laser beam inside the bulk material surface, defects are generated on the top and the bottom of the sample. μ μ μ ‐μ ‐ ‐ μ

**Figure 4.** (**a**–**d**) Side view of the cut edges obtained at 50 kHz, at the scan speed of 1 mm/s and four different pulse energies. The top (**e**,**g**,**i**,**k**) and the bottom (**f**,**h**,**j**,**l**) views are also shown.

**Figure 5.** Top (**a**) and bottom (**b**) surface of the target before laser irradiation.

 μ μ μ As the pulse energy was increased to above 25 µJ, a clear ablation occurred at the bottom of the sample. At an even higher value of pulse energy of 35 µJ, the damaged region merged with the laser trace, reaching an extension of almost 160 µm and resulting in cuts of poor quality. μ μ μ

 μ μ In Figure 6, the 3D and line profiles of the cut edges obtained for 50 kHz of repetition rate, 1 mm/s of scan speed and pulse energies of 35 µJ and 20 µJ, respectively, are shown. μ μ

‐ μ μ ‐ μ μ **Figure 6.** Three-dimensional profiles and line profiles of the cuts obtained at 50 kHz, scan speed of 1 mm/s and pulse energy of (**a**,**c**) 35 µJ and (**b**,**d**) 20 µJ, respectively.

‐ ‐ At the higher pulse, energy evident laser-induced damage is present, as also clearly visible in Figure 4d,l. The cut edge obtained with the lower pulse energy has a significantly smoother surface, except for slight erosion lines on the right side and a small depression, only a few micrometers high, positioned at the laser trace in the middle of the cut edge.

#### **4. Conclusions**

‐ ‐ We performed a systematic study on the stealth dicing process of quartz plates using ultrashort laser pulses. The influence of the main laser parameters as the repetition rate, the pulse energy, the translation speed and, as a consequence, the pulses overlap, on the cut efficiency and quality has been thoroughly investigated.

 μ ‐ μ ‐ We have found that the single pass self-induced laser cutting of quartz is possible within the range of plate thickness explored in this work (250 µm). The physical mechanisms leading to the cleavage have been ascribed to the accumulation, pulse after pulse, of tensile stresses, generated by the rapid increase of temperature experienced in the focal volume after the nonlinear absorption of laser energy. Such stresses cause micro-fractures that produce the cut, while propagating along the laser path and throughout the entire thickness of the plate.

The repetition rate has been found to be a key parameter to generate a controlled and clean cut, since this parameter defines the time delay between successive pulses and thus the amount of stress accumulated in the focal volume. In fact, a repetition rate of 25 kHz was found to be too low to initiate the crack, while at 100 kHz, the bottom part of the plates was completely disrupted, due to the

excessive laser induced stress. The optimal value of repetition rate, which has allowed obtaining a neat cut without any significant damage above or below the laser modified zone, was 50 kHz.

Increasing the overlap between pulses or the pulse energy and keeping the same repetition rate resulted in reducing the quality of the final cut, with the appearance of erosion lines and chipping around the laser modified zone. This indicates that a higher energy load causes, once again, excessive laser induced stresses and/or the initiation of ablation from the top or the bottom surface, thus making the stealth dicing process unstable.

**Author Contributions:** Conceptualization, C.G., A.V. and A.A.; methodology, C.G. and A.V.; formal analysis, C.G. and A.V.; investigation, C.G. and A.V.; data curation, C.G.; writing—original draft preparation, C.G and A.V.; visualization, C.G., writing—review and editing, C.G., A.V. and A.A.; supervision, A.A.; funding acquisition, A.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Italian Ministry of Education, University and Research (MIUR) within the Project AIM184902B-1- ATT1.

**Acknowledgments:** The authors gratefully acknowledge the PolySense Lab (http://polysense.poliba.it/) for the provision of the quartz samples and Pietro Paolo Calabrese for his technical support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
