Crack-free high-aspect ratio holes in glasses by top–down percussion drilling with infrared femtosecond laser GHz-bursts

We report novel results on top-down percussion drilling in different glasses with femtosecond laser GHz-bursts. Thanks to this particular regime of light–matter interaction, combining non-linear absorption and thermal cumulative effects, we obtained crack-free holes of aspect ratios exceeding 30 in sodalime and 70 in fused silica. The results are discussed in terms of inner wall morphology, aspect ratio and drilling speed.


Introduction
Laser micromachining with femtosecond lasers in GHz-burst mode has become a hot topic in the very recent years. With ablation rates on metals and semiconductors exceeding those obtained with single pulses [1][2][3][4][5][6][7] this method has generated significant interest. However, some of the studied laser configurations, notably those with only few pulses within the GHzburst and at pulse fluences above the threshold for singlepulse ablation, showed a disadvantageous behavior in ablation experiments [8,9]. So far, most studies have been realized on metals and silicon and only few on dielectrics which are focused on milling ablation removal rates [10][11][12]. These three papers point out the fact that the ablation rate increases with * Author to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. the rise of the number of pulses per burst but at the expense of surface quality.
Glass modifications have been reported by different approaches including drilling with a Bessel beam [13] and using a CO 2 laser [14]. Very-high aspect ratio holes can be achieved using UV technologies [15][16][17]. However, a huge potential has been predicted for GHz-burst micromachining in dielectrics in a recent review [18], and very recently, first results have been reported on percussion drilling in dielectrics [19].
In this contribution, we report novel results on top-down percussion drilling with a GHz-burst femtosecond laser in sodalime and fused silica glass. The heat diffusion time in fused silica is reported to be in the range of 1-10 ns [20]. Therefore, the GHz-burst regime offers an interesting approach as the time spacing between two consecutive pulses inside the burst corresponds, in our case, to 1 ns. So, this technique, thanks to non-linear absorption and thermal cumulative effects, creates taper-free holes in glass with a smooth inner surface in contrast to single pulse laser-based processes. The term single pulse process is referring to the standard regime where multiple single pulses without bursts are applied. We studied the drilling time for both materials and visualize the quality and aspect ratio of the holes by microscope images. We discuss the influence of the number of bursts and of the laser burst fluence, which are important parameters in the drilling process. Moreover, we present the maximum depths and aspect ratios of completely crack-free holes that we were able to drill in the two glasses.

Experimental setup
For the experiments, we used a commercial Yb-doped femtosecond laser (Tangor 100, Amplitude) emitting 500 fs pulses with a maximum average power of 100 W at 1030 nm. It can be operated in single-pulse mode at repetition rates from 1 Hz to 2 MHz or in GHz-burst mode at inter-burst repetition rates from 1 Hz to 200 kHz. In this study, we worked in GHz-burst mode with 50 pulses per burst at an intra-burst repetition rate of 1 GHz, thus creating 50 ns bursts. The burst parameters are chosen in a way that the intensity of the individual pulses in the burst allows for non-linear absorption and that the burst duration is sufficiently long to attain the beneficial accumulative thermal regime that offers enhanced ablation rates [10]. The inter-burst repetition rate was set to 1 kHz in order to avoid the appearance of any heat affected zone around the holes [19].
A microscope objective (Mitutoyo NIR APO 5×) with an effective numerical aperture of 0.10 was used to focalize the laser beam resulting in a spot diameter of 9.3 µm at 1/e 2 , which we measured using a CCD camera (DAT-WinCamD LCM4) along with a calibrated home-made magnification system. A Basler camera from top view allows for visualizing through the microscope objective and verifying the position of the laser focus at the front surface of the different glass samples (figure 1). During the drilling, we used a side-view system composed of a green diode emitting at 523 nm and a Basler camera (Basler acA1920-25mu, 1/3.7 ′′ sensor, resolution of 1920 × 1080p, pixel size 2.2 µm × 2.2 µm, 25i/s, rolling shutter) coupled with a long-distance microscope (InfiniMax KX with MX-5 Objective) with a 520 nm bandpass filter in order to visualize directly horizontally through the samples and not being blinded by the processing laser wavelength. The TTL signal from the external acousto-optic modulator is filtered thanks to a dedicated electronic card (Tombak, Aerodiode) and delayed by the camera software (Pylon, Basler). The latter allows us to set the exposure time and the acquisition rate to visualize the dynamics of the drilling. The focusing head is mounted on a Z-motorized stage (VP25X, MKS Instruments) whereas the sample is fixed on a motorized XY-monolithic stage (One-XY60, MKS Instruments). The XYZ-stages and the laser gate are controlled by the DMCpro software (Direct Machining Control). The workstation has a granite base and gantry ensuring a high stability and an excellent repeatability of the experiments. An optical measuring microscope (MF-B1010D, Mitutoyo) is used for post-mortem imaging and high accuracy measurements of the hole depths and diameters with a precision of ±2.2 µm + 0.02L, with L in mm.

Hole depth
We studied the evolution of the hole depth as a function of top-down drilling time from 20 ms to 100 ms at an inter-burst repetition rate of 1 kHz, where the drilling time was increased by 10 ms for consecutive holes. It corresponds to an increasing number of bursts from 20 to 100, each containing 50 pulses at an intra-burst repetition rate of 1 GHz, which are received by the sample. The applied burst-fluence was 52 J cm −2 in the case of sodalime and 136 J cm −2 for fused silica. The resulting holes are depicted in figure 2. This figure illustrates that the general shape of the holes is essentially the same for both materials, the holes are cylindrical and there is a linear evolution of the depth ranging from 73 µm to 298 µm in sodalime and from 110 µm to 292 µm in fused silica.
Drilling fused silica required a higher burst fluence compared to sodalime because fused silica has a higher bandgap value (3.9 eV for sodalime [21], 9.0 eV for fused silica [22]), and consequently a higher ablation threshold (3.6 J cm −2 for fused silica, 2.9 J cm −2 for sodalime [23]). Furthermore, fused silica has a lower thermal expansion coefficient which provides a better resistance to thermal dilatation and resulting shear stress [24]. One may notice in figure 2(a), we observe waveguide-like modifications in the continuation of the hole drilling for the first holes from the left (indicated by black arrows for the first two holes), confirming a more sensitive material response of sodalime. The holes displayed in figure 2 present a similar morphology and profile as those obtained with CO 2 laser drilling [14]. However, as the CO 2 laser interaction is purely thermal, the holes have larger diameters (smallest of 70 µm) and present a heat affected zone all along the hole. In contrast, GHz-burst mode drilling enables the attaintment of fine holes of small diameter and no visible heat affected zone. However, we previously observed the appearance of a heat affected zone when increasing the inter-burst repetition rate [19]. Indeed, for sodalime at a too high repetition rate, the hole collapses under its own weight which indicates that the softening temperature has been reached. Furthermore, bubbles in the heat affected zone indicate that the decomposition temperature of the glass is reached. Note, the hole morphologies are completely different for femtosecond laser single-pulse drilling as we have shown in [19].
On figure 3, the evolution of the hole depth is displayed as a function of the number of bursts received by the two materials. This experiment was done for different values of burst fluence ranging from 136 to 339 J cm −2 . The experiments were automatized with the DMCpro software applying the loop option. We programmed five loops, which allows us to conveniently investigate the drilling depth in the range from 1 single burst to 19 000 bursts (inter-burst repetition rate of 1 kHz).
We separated the holes by 100 µm in order to segregate each hole from the thermal effects produced by the previous drilling. Note that the first points, with a low number of bursts, are not displayed on the graphs as there was no  clear drilling observable. Indeed, a minimum of 20 bursts is required at 136 J cm −2 to get a visible onset of drilling in both materials. This value decreases for increasing fluence, for instance to 10 bursts at 171 J cm −2 and down to 1 burst at 369 J cm −2 in sodalime. One may notice that the lowest burst fluence of 136 J cm −2 , shown in figure 3, corresponds to a pulse fluence of 2.7 J cm −2 as there are 50 pulses per burst, which is lower than the single-pulse ablation threshold for both materials. Processing at a fluence value below the single-pulse ablation threshold is characteristic for the GHz-burst regime and proves the outstanding accumulative character of the interaction process [4,5,19]. Indeed, in ablation studies using pulse fluences within the burst exceeding one of the single-pulse ablation thresholds even leads to less efficiency and potentially detrimental effects on the surface quality [8]. We further notice that the hole depth first increases and then saturates at certain values depending on the laser fluence for both materials (figures 3(a) and (b)). The uncertainty of the side-view measures with the microscope (±2.2 µm + 0.02L, with L in mm) is neglected as it is an order of magnitude lower than the spreading due to the front surface roughness, which has an average surface roughness (Sa) of 0.8 µm, and a peak-to-valley surface roughness (Sz) of 20 µm (measured with a Zeiss profilometer, objective 20x/NA 0.70). Figure 4 represents the evolution of the hole depth as a function of the number of bursts sent at a fixed fluence of 151 J cm −2 and an inter-burst repetition rate of 1 kHz in sodalime. The error bars (±20 µm) on this graph originate from the front surface roughness and the accuracy of the focus position. The insert in figure 4 shows a zoom on the evolution of the hole depth for small burst numbers and gives a closer look on the drilling mechanism.
The graph reveals three stages in the hole formation which are indicated by three different linear fits in figure 4 and are schematically illustrated in figure 5. The first stage corresponds to surface ablation (red linear fit on figure 4, stage 1   figure 5). The ablation plume can expand freely in the ambient air above the target. The drilling rate is high. The second stage corresponds to deep ablation (blue linear fit on figure 4, stage 2 on figure 5). The ablation plume is confined by the inner walls leading to a decrease in ablation efficiency. The drilling rate is lower compared to surface ablation (0.70 µm/burst for surface ablation versus 0.15 µm/burst for deep ablation). The drilling rate is almost constant during each of the first two stages. This hypothesis is supported by assuming a beam propagation inside the hole according to previous work reported in literature where the laser beam is subject to reflection under grazing incidence on the inner walls [25] and multiple scattering [26]. However, at each reflection, a part of the energy is lost by refraction, and supplementary losses are also caused by multi-directional scattering. Therefore, the available energy for drilling decreases with increasing depth and the depth finally saturates as has been observed in polymers [26,27]. In the third stage (green linear horizontal fit on figure 4, stage 3 on figure 5) the drilling process is over as the fluence value at the hole bottom is below the ablation threshold. The intersection between the red and blue fits corresponds to the transition between surface and deep ablation (170 µm in sodalime) whereas the intersection between the blue and green fits corresponds to the termination of the drilling (745 µm in sodalime).

Hole diameter and aspect ratio
In the previous section, we observed that the maximum hole depth depends on the burst fluence and on the number of bursts. These parameters also have an impact on the diameter at the entrance of the hole. Figure 6 depicts the evolution of the hole diameter at the entrance as a function of the burst number for sodalime with burst fluences ranging from 136 J cm −2 to 369 J cm −2 and at an inter-burst repetition rate of 1 kHz. When the drilling begins, the diameter increases rapidly with increasing burst number. At a second time, it reaches a saturation value that increases with the burst fluence. The results for the maximum depth, diameter and aspect ratio of the holes in sodalime and fused silica for an inter-burst repetition rate of 1 kHz are summarized in table 1. The entrance diameter value is measured at 20-30 µm below the front surface under sideview with a measurement microscope (Mitutoyo, model MF-B1010D, objective 20×). Regarding fused silica, we were not able to measure the diameters at the entrance of the hole due to the uneven surface of the sample and the resulting shadow effect ( figure 2). Therefore, the values displayed in this table for the hole diameters in fused silica correspond to a measurement realized below the shadow part and not to the real entrance diameter. The uncertainty on the diameter measures corresponds to the accuracy of the measurement microscope.
We observe that both the diameter and the depth increase with the fluence. The aspect ratios obtained in these experiments were up to 37 for sodalime. However, the quality of the inner hole surface, which will be discussed in the next section, became rather poor with increasing fluence in sodalime. Figure 3(b) attests that fused silica seems to follow the same tendency as sodalime regarding the evolution of the maximum depth with an increasing fluence. Moreover, for this material, the maximum aspect ratio obtained in this experiment is 41 and the quality of the inner surface does not change with the fluence nor with the number of bursts applied on the sample. Note that a compromise has to be found between burst fluence and inter-burst repetition rate. In order to optimize the aspect ratios, we chose to work at the highest fluence available with our setup, 369 J cm −2 . The inter-burst repetition rate was kept at 1 kHz to avoid any heat affected zone [14], and we kept the bursts containing 50 pulses at an intra-burst repetition rate of 1 GHz. Figure 7 displays the corresponding holes for drilling times of 16, 17 and 18 s (from the left to the right).
On these images, we show holes up to 1520 µm deep in sodalime and 1602 µm in fused silica with entrance diameters about 50 µm in sodalime and 22 µm in fused silica  corresponding to aspect ratios of 30 and as high as 73, respectively. The images were taken with the measurement microscope and a 3× objective in order to get the full size of the holes.
In fused silica, the holes are thinner compared to sodalime which we attribute to the higher ablation threshold. However, in both materials, the depth of the obtained holes is similar. As discussed in section 3.1, we believe that thanks to reflection and scattering, the laser beam is guided along the hole and allows for residual energy deposition up to the bottom of the hole. Despite the higher ablation threshold of fused silica, comparable hole depths are reached in both glasses. We might explain this observation by a closer look on the hole quality.

Hole quality
For more detailed observations, we used the optical microscope with a 20× objective allowing for more precise morphology analyses of the inner surface. Figure 8(a) depicts the holes obtained in sodalime for drilling times of 800, 900 and 1000 ms at a burst fluence of 270 J cm −2 . Figures 8(b) and (c) show images taken for drilling parameters set to a burst fluence of 270 J cm −2 and for drilling times from 1 s to 15 s (from the left to the right) in sodalime and fused silica, respectively. The images framed in red are a zoom of the entrance part of the two holes for a drilling time of 7 s. We observe a very different quality of the inner walls for the two materials (figures 8(b) and (c)). In sodalime, the inner walls are structured ( figure 8(a)), whereas in fused silica the inner surface remains smooth all along the hole. In sodalime, we distinguish three zones, at the tip the surface is bumpy (A), further up the hole in a transition zone, the surface appears rippled (B). Finally, the inner walls feature a groove-liked structure (C). These surface patterns are comparable to those of surface structuring of metals [28] and silicon [29]. Moreover, the inner texture of the hole depends on the drilling time as only the drilling time, and thus the burst number, increases from the first hole in figure 8(a) (left: 800 ms) to the third hole (right: 1000 ms).
The difference of the inner surface quality induces different scattering losses of the beam during the drilling process in these two materials. The glossy surface of fused silica (figure 8(c)) allows for low-loss reflections and therefore a more efficient beam transmission towards the tip of the hole increasing its depth. This results in a compensation for the larger energy amount needed for drilling due to the higher ablation threshold of this material. This is the reason why we measure similar lengths in sodalime and in fused silica (figure 8).
One may notice that the aspect ratios obtained for sodalime in this section are lower. Indeed, in figure 3(a) the maximum hole depth of 1660 µm was drilled at a burst fluence of 339 J cm −2 and led to a hole diameter of 45 µm, resulting in an aspect ratio of 37 (see table 1), which is the highest we obtained in our experiments for sodalime. Based on this observation, we believe that there is an optimum fluence depending on the material characteristics which leads to the highest aspect ratio.

Conclusion
In conclusion, we demonstrated novel results on top-down percussion drilling in sodalime and fused silica by femtosecond laser GHz-burst micromachining. Our comparative study, based on the influence of different laser parameters on the depth as well as on the diameter allowed us to optimize the drilling conditions and to reach aspect ratios up to 37 in sodalime and 73 in fused silica. The quality of the holes was investigated and revealed a different behavior for the two materials. The inner surface of the drilled holes in sodalime is structured, whereas the hole quality in fused silica stays excellent with a smooth inner surface even at extreme aspect ratios. Studying the hole geometries leads to the hypothesis of a beam propagation by multiple internal reflections within the hole during the drilling, where the reachable hole depth at a fixed burst fluence is only limited by refraction losses. These impressive results of percussion drilling in femtosecond laser GHz-burst mode allowing for extreme hole geometries may pave the way for future applications in photonics devices or micro-electronics.