1. Introduction

The widespread emergence of thin fragile panels, substrates and wafers in a wide range of industry and consumer products, e.g. optical display, microelectronic chips, cell phones, tablet computers, LED lighting, photovoltaics, and wearable health technology, has posed significant challenges and barriers in scribing and cutting with traditional mechanical scribing tools. As the result, a wide variety of laser processes have been studied and applied for contactless scribing and cutting, with the aim of reducing the forces applied on thin fragile workpieces, while improving precision, increasing speed, narrowing kerf width, eliminating grinding and polishing steps, and reducing collateral damage.

Front surface laser ablation is a common approach for strongly absorbing materials to cut directly through substrates or to scribe grooves to guide substrate cleaving [1–3]. Transparent materials further permit laser machining from the back surfaces [4] that may be applied for cutting or scribing transparent substrates. However, material removal rates are typically slow under ablative machining, and ablation plumes can cast debris or damage critical components near the cutting or scribing tracks. Laser heating induced mechanical or thermal stress has also been applied to cleave brittle materials with [5] or without [6] assisting water jet coolant, but such processes require large interaction zones that may be harmful to sensitive nearby components.

Internal laser scribing has rapidly evolved in research and application of cutting transparent materials where ultrashort pulsed lasers are favored for the nonlinear optical interactions that confine and shape the laser dissipation inside of the materials. In stealth dicing [7,8], the focal interaction volume is limited by small Rayleigh length to locally generate micro voids or cracks. Nevertheless, multiple scanning tracks have been successfully assembled in order to build up sufficient internal stress to guide a clean cleaving through a full wafer thickness. Alternatively, nonlinear beam propagation and beam shaping techniques can favorably extend the laser interaction volume over significantly longer range than the Rayleigh length to speed up and improve laser scribing and cleaving process [9–14].

Long laser filament tracks are readily formed in transparent media by balancing Kerr nonlinearity, plasma defocusing, and diffraction effects during ultrashort-pulsed laser propagation [15–17]. This unique spatial distribution of laser energy, confined over tens to hundreds of Rayleigh lengths, affords material processing benefits such as direct writing of optical waveguides and diffractive optical elements, conducting micro-welds, generating micro/nano-voids, and fabricating fiber Bragg gratings, as demonstrated in a wide variety of optically transparent materials [18–22]. Ahmed et al. applied laser filaments to open voids and guide the cleaving of display glass along arrays of such filament when formed near the rear glass surface [23]. Butkus et al. added a liquid surface layer to assist the filament formation promptly at the entry surface to guide the glass cleaving [24]. Burst train laser delivery [25,26] was adopted by Hosseini and Herman to enhance laser interaction with lower threshold, enabling formation of filament arrays with strong stressing forces to guide scribing in various transparent materials [9]. More recently, other groups have also applied burst train pulse delivery to improve glass cleaving [27].

Spatial shaping of the laser beam offers an alternative means for elongating the laser interaction zone without relying on Kerr nonlinear focusing. Drilling, scribing and cutting of various optically transparent or opaque materials have been demonstrated with Bessel and Airy beams [10–12,27,28], tailored absorption for elongated modifications [13], and multiple micro-fractures [14]. The focal volume can also be dramatically stretched by optical aberrations from a spherical focusing lens or the flat surface of a substrate when focusing deeply into a sample with large NA optics [29], for example, as demonstrated by Sercel and co-workers to improve laser cutting of glasses [30]. All these spatial shaping approaches rely on an angular dispersion of the laser beam, reducing the overall energy interaction efficiency in contrast with self-focusing Kerr effect. However, managing all of the laser energy within a single focal area is also challenging, possibly leading to over excitation, plasma defocusing, and other effects that distort and break apart the laser filament propagation. The spatial and temporal dynamics of such laser filament interaction and material response need to be well understood or studied.

In this paper, we report a systematic study of ultrafast laser scribing of transparent glass based on the formation of long filament tracks with moderately high pulse energy and burst train delivery. The filament interaction dynamics under burst train and single pulse laser exposure were characterized for the first time by time-resolved microscopic imaging of filament photoluminescence (PL). Various focusing conditions were examined to simultaneously form surface V-grooves with filament arrays in a single scanning track aiming to improve the quality and efficiency of glass scribing. The flexibility and contrasting benefits of laser filament scribing with and without burst trains or surface grooves are presented in the context of surface morphology and breaking force for soda-lime glass, which is widely used in laboratory equipment, windows, architectural glass, and containers.

2. Experiment

The experimental arrangement for laser filament scribing and real-time characterization is illustrated in Fig. 1(a). A diode-pumped Nd:YVO₄ laser (Coherent, Hyper Rapid 50) with 1064 nm wavelength and ∼10 ps pulse duration delivered single pulses (Fig. 1(b)) or burst trains of pulses (e.g. 4-pulse burst in Fig. 1(c)) with variable repetition rates up to 1 MHz (i.e. t₂ ≤ 1 µs). Burst trains with up to 10 pulses per burst could be generated at 50 MHz frequency, with burst-to-burst temporal separation of t₁ = 20 ns. The expanded laser beam (6 mm 1/e² diameter) was focused with a spherical lens (25 mm focusing length) into soda-lime (Corning 0215, 75 mm × 25 mm × 1 mm) or fused silica glass (Corning 7980, 50 mm × 25 mm × 1 mm) substrates to 13 µm spot size (1/e² intensity) and 290 µm depth of focus, when accounting for lens and glass surface aberration effects. The side facets of fused silica substrates were optically polished to enable time-resolved microscopic imaging of the filament formation dynamics as illustrated in Fig. 1(a). Images were recorded with a time-gated intensified CCD (ICCD) detector (Andor, iStar DH734-18U-03) through a 50× objective lens with time synchronized to laser pulses with an adjustable electronic gate delay (t_delay). Glass samples were translated with linear XY motion stages (Aerotech ABL1000) to form filament arrays and surface V-channels with focusing position controlled by a Z motion stage (Aerotech ALS130). Laser modified samples and cleaved substrates were assessed by optical microscopy and atomic force microscopy (AFM). Laser scribed and cleaved samples were subjected to four-point flexural testing (Shimatsu AG-I) to record the breaking force and flexural strength, respectively.

 

Fig. 1. Schematic arrangement (a) for ultrafast laser filament formation and in-situ photoluminescence characterization, driven by single-pulse (b) and burst train (4 pulses/burst as an example) (c) exposure, under various combinations of V-channel and/or filament array scribing structures (VB-nF, VT-nF, VB-F, VT-F, and I-F, see text for definitions). ICCD: intensified CCD detector; t_gate: ICCD gate width; t_delay: ICCD gate delay time; t₁: time period (20 ns) between laser pulses in a laser burst; t₂: time period between single laser pulses or burst trains.

Download Full Size | PDF

Five different combinations of laser filament and V-groove scribing were considered as depicted inside of the glass substrate in Fig. 1(a). Designations VB-nF and VT-nF represent ablative formation of V-grooves at the bottom (B) or top (T) surface, respectively, without filament modification tracks (-nF). VB-F and VT-F denote the addition of a filament array that connects with the bottom or top V-grooves, respectively, with or without ((nB)) the burst train mode. In contrast, I-F represents formation of an internal filament array with burst train exposure but without laser ablation or any other form of modification at the bottom and top surfaces.

Photoluminescence images of the laser filament formation were recorded as shown in Fig. 2 for single pulse exposure of fused silica at varying focal position (Figs. 2(a)–2(g)) and pulse energy (Figs. 2(h)–2(l)). In the sequence of PL images from Figs. 2(a) to 2(g), a single laser pulse of 74 µJ energy was focused to a varying focal plane positions (green dash lines) starting at 350 µm below the glass bottom surface in Fig. 2(a) and rising to 200 µm below the top surface in Fig. 2(g). The PL was recorded with a narrow 5 ns gate width and synchronized with the peak PL emission time to capture ∼20% of the total PL signal. All the PL tracks (up to ∼800 µm long) are seen to extend longer (up to ∼2.5×) than the effective Rayleigh length (≤ 290 µm), stretching both above and below the focal plane, when focused inside of the glass. This stretching and upward movement of the filament PL toward the laser attests to the strong underlying role Kerr-lensing plays to form a highly elongated and narrowly focused beam. Moreover, higher resolution images of the PL tracks reveal a narrowing of the laser interaction to ∼2 µm width (3 dB) that falls far below the 13 µm beam waist expected by ray-tracing simulation (Zemax). This 74 µJ pulse exposure also exceeds the theoretical threshold energy of ∼40 µJ expected for nonlinear self-focusing in fused silica with 10 ps laser pulses [31].

 

Fig. 2. Photoluminescence images (a)-(g) of laser filament tracks recorded in soda-lime glass with 5 ns ICCD gate width at varying laser focal positions illustrated by the green dashed lines except for (a) where the focal plane is 335 µm below the bottom surface. Photoluminescence images (h)-(l) of laser filaments recorded with 5 ns ICCD gate width, at fixed focal position (green dashed lines) with varying laser pulse energy: 34 µJ (h), 50 µJ (i), 74 µJ (j), 104 µJ (k), and 153 µJ (l). TS: glass top surface; BS: glass bottom surface.

Download Full Size | PDF

It is apparent in Figs. 2(a)–2(g) that the laser focusing position offers a well-controlled placement of such highly elongated filaments internally into the 1 mm thick silica glass, to be isolated from surfaces and fully buried (Fig. 2(e)), or to be weakly or strongly connected with the bottom (Figs. 2(d) and 2(a), respectively) or top (Figs. 2(f) and 2(g), respectively) surfaces. These examples of strong filament-to-surface connection are associated with observation of surface ablation, permitting the simultaneous formation of V- or U-shaped surface channels that were connected with internally modified filament array tracks when scanning the sample laterally.

Optical microscopy images of single filament tracks showed that their axial position, length, width, and overall contrast to be strongly correlated with observed position, length, width, brightness and persistence of the PL track images such as presented in Fig. 2. The axial breaking of laser focus, for example, into two widely separated filament tracks as seen in Fig. 2(a), is attributed to strong Kerr-lensing and plasma defocusing effects [17]. In addition, the filament tracks can break apart into an array of modification zones as shown in Fig. 2(g) when strong plasma generation is anticipated at the top surface to disturb the beam propagation to below. This type of in-situ and time-resolved PL microscopy thus provides a powerful tool for real-time characterization of the anticipated modification track, enabling prompt and reliable assessment of the permanent filament modification by length, width, uniformity, and contrast. Moreover, the proximity of the filament PL to the top or bottom surfaces can be used to predict or avoid the onset of surface ablation. For example, evidence of such ablation is noted by the ablation plume emission observed at 0 ns delay time from the space below the bottom and above the top surface as noted in Figs. 2(a) and 2(g), respectively.

By fixing the laser focal plane at 550 µm below the top surface, Figs. 2(h)–2(l) show that the laser filament track length increases with pulse energy from ∼400 µm (h) at 34 µJ/pulse to ∼900 µm (k) at 104 µJ/pulse, or can be made to cross the full 1 mm glass thickness (l) at 153 µJ pulse energy. Therefore, the collective results of Fig. 2 are compelling in controlling filament structuring – position, length, contrast, and near-surface connection – for potential application in glass scribing (i.e. Figs. 2(b) to 2(f)) or near-surface micro-machining (Figs. 2(a) and 2(g)).

The delivery of burst train pulses was found to amplify the laser filament formation dynamics as noted by comparing the temporal sequence of PL images presented in Figs. 3(a) and 3(b). In Fig. 3(a), the PL filament was recorded with a 5 ns gate width for single laser pulse exposure of 50 µJ and 550 µm focal depth (green dashed lines). The laser exposure was shifted to a new glass zone for each pulse (0.5 Hz repetition rate). The PL peaks promptly with the arrival time of the laser (0 ns gate delay) and decays rapidly to the noise floor in ∼40 ns. A plot of the total PL intensity as a function of gate delay time (Fig. 3(c)) indicates an exponential like decay of the PL following after the laser exposure (red dashed line) with a fitted decay time of 5 ns that is at the time resolution limit of the 5 ns ICCD gate width.

 

Fig. 3. Photoluminescence images of laser filament tracks recorded in soda-lime glass with 5 ns ICCD gate width for single pulse (a) and burst train (b) exposure. The gate delay time is as labeled and focal positions are illustrated by green dashed lines. The total filament areal photoluminescence intensity is plotted as a function of gate delay for single pulse (c) and burst train (d) exposure, with the laser coincidence time depicted by the vertical red dashed lines. TS: glass top surface; BS: glass bottom surface.

Download Full Size | PDF

In strong contrast, the PL images are substantially brighter and longer lasting in Fig. 3(d) where laser exposure of 4-pulse burst trains was applied at similar single pulse energy (51 µJ/pulse or ∼200 µJ/burst). The total filament PL intensity from such images were plotted in Fig. 3(d), and showed a strong 3-fold buildup of peak PL from the first pulse to the third and fourth laser pulses (laser arrival marked by red dashed lines). In contrast, a repeating exposure of the single pulse exposure in the same interaction volume yielded a nearly identical PL brightness and decay time (Fig. 3(c)) when applied at low, 0.5 Hz repetition rate. The PL of the first pulse exposure for the burst train in Fig. 3(d) is seen to follow the PL profile generated by the single pulse exposure in Fig. 3(c), rising and then decaying until the arrival of the second pulse at 20 ns. Thereafter, the PL decay is brighter and longer lasting with the arrival of each additional pulse, and without decaying to zero in the 20 ns interval between pulses. After the full 4-pulse burst exposure, the PL followed a double exponential decay consisting of a fast τ_f = ∼5 ns decay time followed by a much slower component of τ_s = 70 ns. This latter decay time arises from heat accumulation effects when the heat dissipated by a previous pulse does not have sufficient time to diffuse away in the 20 ns pulse-to-pulse interval [32,33]. In this time interval, the thermal diffusion scale length of ∼0.2 µm is small enough to build up a higher peak temperature and larger heating volume by the end of the fourth pulse. As such, the time-integrated PL intensity observed with the 4-pulse burst exposure is approximately 20 times larger than with a single pulse exposure, manifesting in an overall 5× brighter PL emission. Hence, the burst mode exposure drove a stronger laser interaction, created higher peak temperature and sustained longer interaction time, generating filament modification tracks with higher morphological contrast and longer overall length than in the single pulse exposure case.

The relative merits of burst and non-burst filament formation were tested in laser scribing of soda-lime glass plates over various (Fig. 1(a), VB-nF, VT-nF, VB-F, VT-F, I-F) focusing positions, pulse energies, and scanning speeds, optimized to reduce the mechanical separation force and to minimize the cleavage facet damage. The underlying objective was to retain the original high flexural strength of the glass in the separated glass samples. Sample scanning speed was tested up to ∼20 mm/s scanning speed, limited by the motion stage acceleration rate to generate uniform filament arrays along a 75-mm sample length.

Figure 4 presents the optical and AFM images of soda-lime glass scribed and cleaved by laser filament arrays under the focusing geometry of VB-F, comparing the merits of single pulse Figs. 4(a) – 4(f) and burst train Figs. 4(g) – 4(l) exposure. A long internal filament was formed to connect with the bottom surface, and tuned to use remaining laser light to ablate a trench at the bottom surface (Figs. 4(a) and 4(g)). The filament array was formed at 5 kHz laser repetition rate. The optimized glass scanning speed was found to be 3.5 and 5 mm/s for ideal focal positions of 365 and 375 µm below the sample bottom surface, for single pulse and burst train case, respectively.

 

Fig. 4. Optical microscope images in end ((a), (g)), top ((b), (h)), and bottom ((c), (i)) surface views of filament arrays formed in soda-lime glass and connected with a simultaneously formed bottom surface V-channel (VB-F), comparing morphology of single pulse ((a)-(c)) and burst train ((g)-(i)) exposures prior to cleaving, and optical microscope ((e), (j)), camera ((d), (k)), and AFM images ((f), (l)) of the resulted cleavage facets, comparing facet quality under single pulse ((d), (e), (f)) and burst train ((j), (k), (l)) exposures. TS: glass top surface; BS: glass bottom surface.

Download Full Size | PDF

For single pulse exposure, a V-groove trench and filament combination was formed with 123 µJ pulse energy, yielding the end view as shown in the microscope image of Fig. 4(a). The filament modification shows moderate contrast inside of a 2 µm wide modification zone that is slightly broken in a length extending over 70% of the sample thickness, defining a stressed zone with high aspect ratio (350:1). Figures 4(b) and 4(c), respectively, show an unmodified top surface above the filament array and a ∼50 µm wide V-groove opening on the bottom surface. The glass was cleaved to follow along the scribed plane by a mechanical force, yielding a visually smooth cleaving facet as seen in both low (Fig. 4(d)) and high magnification (Fig. 4(e)) images. An AFM assessment (Fig. 4(f)) confirms low surface damage with only 3 nm (RMS) roughness.

A similar focusing geometry (VB-F) was optimized for a 4-pulse burst train of nearly equal energy pulses tuned to a similar total energy of 118 µJ. A small ablation trench was formed at the bottom surface and connected to a higher contrast filament modification track as illustrated by the end view image of the scribed glass (Fig. 4(g)). The filament length and width were similar to the case of single pulse exposure (Fig. 4(a)). The side (Fig. 4(g)) and bottom surface images (Fig. 4(i)) reveal a much shallower (∼3 µm) and narrower (∼7 µm) trench than formed in the single pulse case, possibly owing to the nonlinear fall-off of ablation rates with ∼4× lower energy pulses. Nevertheless, this combination of narrow groove with a connected filament modification track led to reproducible glass cleaving, generating a moderately smooth cleavage facet (Fig. 4(j)). A smooth cleaved facet appears to be disrupted by micro and/or nano-fractured tracks that aligned spatially with the high contrast filament array zones (Fig. 4(g)). AFM images (Fig. 4(f)) revealed a sub-micron surface roughness of 0.5 µm (RMS) in these strongly sheared zones. The burst exposure enables heat accumulation and incubation effects [32,33], resulting in stronger stresses as well as formation of micro or nano-voids [19,20] that was not apparent in the single-pulse exposure case. These morphological differences cannot be attributed to the laser energy as total exposure per filament track was similar within 5%. The more aggressive morphological changes must arise from the stronger dissipation and accumulation effects when laser energy is divided over several pulses in high-repetition-rate (50 MHz) burst trains.

The burst train exposure was also able to simultaneously form and connect filament arrays with an ablation trench at the top surface (Fig. 1(a), VT-F). Cleaving with minimal mechanical separation force was obtained by focusing to a higher position, ∼600 µm below the top surface, and using a slightly stronger total energy (148 µJ/burst) divided almost equally over a 2-pulse burst. The filament array was formed with 4 kHz repetition rate and 10 mm/s glass scanning speed, yielding a highly contrasting modification zone of up to 28 µm width that stretched to cover the top 60% cross-sectional area of the glass plate (Fig. 5(a)). The filament was connected with a weakly formed trench of ∼2 µm depth and ∼2.5 µm width (Figs. 5(a) and 5(b)). After mechanical cleaving, the filament modification zone was seen to generate a dense network of micro and/or nano-fractures (Fig. 5(d)) with sub-micron surface roughness. A close inspection of the filament array (Fig. 5(a)) shows a high contrast modification zone of ∼2 µm width, defining a filament similar to the VB-F case (Fig. 4(g)), but surrounded by a darkened modification zone up to 28 µm wide (Fig. 5(a)). The nearly flat surface (Fig. 5(d)) suggests that a narrowly defined filament-modification zone had formed to guide the cleaving, while surrounded by a weak modification volume of low stress defects such as color centers, in contrast with the VB-F cases described above (Figs. 4(g)–4(l)).

 

Fig. 5. Optical microscope images in end ((a), (f)), top ((b), (g)), and bottom ((c), (h)) surface views of filament arrays formed in soda-lime glass under burst train exposure prior to cleaving, and optical microscope ((d), (i)) and camera ((e), (l)) images of the resulted cleavage facets, as well as optical microscope images of the top ((j)) and bottom ((k)) surfaces showing the sharp and smooth corners of the cleavage facet. Filament arrays were formed simultaneously with (VT-F; (a)-(e)) or without (I-F; (f)-(l)) ablation of a top surface V-channel. TS: glass top surface; BS: glass bottom surface.

Download Full Size | PDF

The key objective in this study was to examine filament-only guided scribing (Fig. 1(a), I-F) and avoid trench ablation on either the bottom or top surface. Filament guided scribing was found effective with 2-pulse burst trains having 102 µJ/burst energy, focused 825 µm below the top surface, and with 7.5 mm/s scanning speed at 4 kHz repetition rate. Figure 5(f) shows the end view image of the resulted modification track, consisting of a darkened modification zone up to ∼ 22 µm wide that overshadows a narrow (∼2 µm wide) filament zone with features similar to the VT-F case. The images of the top and bottom surface (Figs. 5(g) and 5(h), respectively) confirm the absence of surface ablation even with the top end of the filament track positioned only ∼10 µm from the top surface. The cleaved facet (Fig. 5(i)) shows micro and/or nano-fracture features to align with the filament track (Fig. 5(f)) as found in the prior burst-exposure cases (Figs. 4(j) and 5(d)), but now extending over ∼90% of the cross-sectional surface. Without surface ablation trenches, the cleaved surface terminates into smooth and sharp corners on both the top (Fig. 5(j)) and bottom surface (Fig. 5(k)) with the sharper corner on the bottom where the filament track is farther from the surface.

The laser filament scribing was assessed quantitatively with four-point flexural test to first measure the breaking force and then record the flexural strength of the cleaved glasses. During testing, the V-groove side or the top surface in the I-F case was placed under tension. The breaking force (red circles) and flexural strength (black circles) are presented in Fig. 6 for various cleaving cases reported above (VB-F(nB), VB-F, VT-F and I-F) and compared with non-scribed glass plates having standard ground side facets (Ref). The filament-only case (I-F) offered a marginal decrease of ∼16% in the breaking force over the reference sample, while sharper reductions of 49% for VB-F(nB) case, 69% for VB-F case, and 39% for VT-F case were found. The presence of V-grooves in the three latter cases is attributed to this ∼3-fold drop in breaking force. The stronger relative decrease in breaking force of 69% for VB-F in contrast with 39% for VB-F(nB) attests to the stronger influence of burst-train modification over single-pulse filament formation, since both samples similarly had bottom surface V-grooves. Higher stresses and/or more defects are thus anticipated in the filament arrays formed with laser burst trains, which indicates that the internal filament structure was also key to facilitating a strong guiding effect that could reproducibly direct the glass fracturing in a straight and orthogonal plane as defined by the laser filament geometry.

 

Fig. 6. Comparison of breaking force (red circles) of non-scribed glass (Ref) and glass samples scribed by the following laser structures: filament array with bottom V-groove under single pulse (VB-F(nB)) and burst train (VB-F) exposure, and burst-train exposure of filament arrays with (VT-F) and without (I-F) a top surface V-groove. Flexural strengths of the pristine and cleaved glass samples are shown by black circles. Error bars are standard deviation over ∼5 samples.

Download Full Size | PDF

Figure 6 shows that the flexural strength of the separated samples to have degraded for all samples having V-grooves. Statistically significant decreases of 12%, 13%, and 7% were seen in the flexural strength for the V-groove cases (VB-F(nB), VB-F, and VT-F, respectively) in comparison with the reference sample (Ref). In stark contrast, the filament-only sample (I-F; 112.1 MPa) yielded only less than a 1% decrease in flexural strength that is statistically comparable with the strength of the reference sample (Ref; 112.9 MPa). The filament-guided cleaving without assistance from surface trenches (I-F) thus offered the best flexural strength together with the highest cleaving edge quality that is comparable with the ground surfaces typically found in commercial optics. Hence, one may bypass post-processing steps such as facet grinding when cleaving glass plates with laser filaments.

3. Discussions

A comparison of various scribing methods, with and without filament arrays, is summarized in Table 1. All methods involving burst-train formation of filaments – with or without surface V-grooves – have shown benefits of reproducible cleaving guidance while maintaining a moderate flexural strength of 88% (VB-F(nB)), 87% (VB-F), 93% (VT-F), and 99% (I-F) of the reference sample. Cleaving along laser-formed surface V-channels without internal filament arrays was found to offer a non-reproducible and weak guidance, frequently yielding multiple fracture zones and non-flat facets. On the other hand, reproducible cleaving with the best cleavage facet quality (3 nm RMS roughness) was observed when bottom surface V-grooves were connected with single-pulse filament arrays. However, a reproducible and strongly guided cleaving was not observed when single-pulse filament tracks were formed without connection to a bottom or top surface V-groove. This could be mitigated with burst-train formation of filament track arrays, where a stronger modification morphology guided fracturing with high reproducibility, enabling moderately smooth facets (∼500 nm RMS roughness) to form with the highest corner quality and best flexural strength (99% of the reference sample) without the need of surface V-grooves. For scribing of 1 mm thick glasses under the four different focusing cases (VB-F(nB), VB-F, VT-F, and I-F), the application requires a relatively small clear beam aperture of 328, 330, 99, and 136 µm diameter, respectively, on the glass top surface.

Table 1. Comparison of ultrashort-pulsed laser glass scribing methods

View Table

The heat accumulation effects expected from laser burst trains [33] appear to have generated stronger internal stresses together with nano-voids or micro-fractures that strongly guided the cleaving (Figs. 4(g)–4(l), Fig. 5) in contrast with the weaker morphology observed with filament arrays formed by single pulses (Figs. 4(a)–4(f)). The shearing dynamics under burst train exposure resulted in a higher ∼500 nm (RMS) surface roughness (Figs. 4(j), 4(l), 5(d) and 5(i)). However, these surface features do not appear to diminish the flexural strength (99% of reference) in the case of filament-only scribing (Fig. 1(a), I-F). Such burst-filament scribing offered the highest quality facet corners while also eliminating laser ablation debris and minimizing collateral damage that would be attractive to scribing of substrates with thin-film coatings.

The time-resolved PL observation (Figs. 3(b) and 3(d)) pointed to much longer duration (14×) and stronger (5× PL intensity) emissions under burst-train interaction, manifesting in stronger morphological changes and improved guidance for filament cleaving. The filament structure was shaped mainly in the domain of Kerr lensing while weak lens focusing (NA = 0.1) and relatively long pulse duration (10 ps) minimized plasma defocusing effects to enable strong laser energy coupling into thin and long filament tracks. In this Kerr effect domain, a higher portion of laser energy will follow along the filament track in contrast with methods that angularly disperse or break apart the beam such as with axicons [28], spherical aberration [30], and liquid layers [24]. The further combination of burst-train pulses with more advanced beam delivery methods is promising in tuning an ideal mixture of nonlinear and geometric optical effects that can improve on filament scribing speed, kerf width, edge and surface quality, and flexural strength over a wider range of substrate thicknesses and transparent materials.

For the present study, laser repetition rates were down counted from 1 MHz to ∼5 kHz to accommodate low acceleration motion stages. The filament scribing speed could therefore be scaled up to over ∼2 m/s scanning speed with higher speed stages or laser scanners to meet today’s industry standards [34].

In the present study, 2-pulse and 4-pulse burst trains provided a marked increase in filament morphology and scribing efficacy when using similar total energy as to the single pulse case. Further improvements in scribing speed and scribing quality can be anticipated by exploring higher number of pulses and different shaped (i.e. non-uniform) energy envelopes. The GHz pulse repetition rate regime is also of special interest where new physical interactions such as laser ablation cooling may be harnessed as reported in [35].

4. Summary

Time-resolved in-situ microscopy has unveiled the dynamics of laser filament responses inside of fused silica glasses, demonstrating the stronger or more effective interaction and morphological changes made possible with burst-train pulse over single pulse exposures. Several exemplary combinations of laser-formed trenches and filament arrays were presented that offered reproducible cleaving guidance. Cleaved facets with sub-micron roughness in the range of 3 to 500 nm (RMS) were presented without reducing flexural strength by more than 15%. Filament-only cleaving, enabled by burst-train interaction, offered the best edge quality, providing the highest flexural strength matching within 1% of the pristine reference sample. Such all-internal filament formation is promising for scribing applications in a wide range of materials including high-strength glass panels.