1. Introduction

Common top-side ablation does not allow the direct fabrication of rectangular shaped micro-channels on the irradiated substrates. However, transparent materials offer the possibility for alternative machining processes with the laser pass in/through the material to ablate it from its backside. A variety of such techniques as the laser-induced backside wet etching (LIBWE) [1–3], laser-induced backside dry etching (LIBDE) [4–6] or laser-induced plasma-assisted ablation (LIPAA) [7–9] have been demonstrated by the use of nanosecond lasers. Here, the laser is focused on an additional absorbent layer directly on the sample interface i.e. liquid (LIBWE) or solid (LIBDE) or on a solid layer having a gap between the sample and the solid absorbent layer (LIPAA). Beside these nanosecond laser-based processes, also femto- and picosecond pulsed lasers have been applied e.g. for LIBWE [10,11], LIPAA [12] or additionally, modifications such as dual-laser LIPAA [13]. Exemplarily, Hua et al. [14] demonstrated micro-lens array fabrication by focussing an UV femtosecond laser on the glass backside in a dye solution with subsequent annealing to reduce the introduced surface roughness. Furthermore, bottom up drilling of holes has been shown in soda-lime glass using nanosecond laser [15] and in sapphire with femtosecond laser [16]. While for the nanosecond processing cracking is observed, it could be avoided by using the femtosecond laser. Beside these backside ablation approaches, another flexible 3D geometry fabrication process is the selective laser-induced etching (SLE). Here, firstly an ultrashort pulsed laser is used to write the geometry inside the material causing a modification and subsequently apply a wet etching step to remove the modified material using potassium hydroxide (KOH) or hydrofluoric acid (HF) [17–19]. Also a faster vector scanning approach has been demonstrated, only scanning the profile of the residual geometry and subsequent high temperature etching by a mix of concentrated sulfuric and phosphoric acid to remove the inner material for lens fabrication in sapphire [20].

Since ultrashort pulsed lasers have proven their potential in effectively processing transparent materials such as glasses and polymers accompanied by a high machining quality directly due to non-linear absorption, several ultrashort pulsed laser based processes have been demonstrated within the fabrication of optical components [21,22], medical products [23,24], microfluidics [25,26] or photonic structures [27,28].

Regarding ultrashort pulsed laser ablation to fabricate, e.g. microfluidic channels, transparent substrates can be machined on the front side, similar to non-transparent materials. The ablated area reveals a typical trapezoidal geometrical shape with larger opening diameter compared to the bottom resulting in taperd sidewalls [29–31]. Beside these common ablation mechanism, again, transparent materials allow the laser to be focused inside the bulk material or on its backside enabling alternative ablation processes overcoming this limitation and fabricate e.g. microfluidic channels, with rectangular cross section and thus non-tapered, vertical sidewalls. This is of great interest due to enabling a better integration of photonic structures such as waveguides for e.g. particle detection within the fluid and being well-understood in terms of flow modelling compared to channels with various or complex cross sections based on their fabrication process. Bellouard et al. [18] used a two-step process for the fabrication of microchannels with vertical sidewalls in fused silica. A femtosecond laser is firstly used to exposure and thus modify the material which is secondly etched by a hydrofluoric acid solution forming the microchannels. Liebers et al. [32] used a backside ablation process called flow-supported laser ablation (FSLA) to fabricate different geometries with vertical sidewalls in transparent materials. Here, the laser is focused on the backside of the material and further moved into the material with advancing ablation while a fluid is streamed onto the ablation region to remove debris and prevent heat. Xu et al. [33] used a water-assisted femtosecond laser direct-writing process for the ablation of micropatterns on vertical sidewalls which can be processed by electroless plating to be used as metal electrodes. The usage of water allows deeper grooves revealing higher surface quality and sharper edges compared to the ablation process without a liquid. Grossmann et al. [34] investigated the backside ablation process in alkali aluminosilicate glass with 5 ps pulses wherein a blind hole was drilled on the backside revealing straight sidewalls. However, only percussion drilling was applied for their investigations with a stationary laser spot. All of these processes are capable to ablate steep edges (vertically) but require an additional processing step such as chemical etching, depend on a more complex experimental setup including liquid solutions during the ablation process or have been shown for stationary ablation only. Contrary to these processes, we use a femtosecond laser to directly ablate the channels geometry while a complex experimental setup or a second etching step can be avoided by using ordinary ultrasonic cleaning to remove the detached material out of the channels.

2. Experimental

A Yb:KGW ultrashort pulsed laser (Pharos, Light Conversion) with a second harmonics module emitting a wavelength of 515 nm is used in our experimental study. The pulse duration of the laser is 230 fs (FWHM) while the pulse energy is controlled by an external attenuator to ensure constant processing parameters. The laser is focused by an objective with a NA of 0.5 (Zeiss, EC Epiplan-Neofluar) on the glass/air interface on the backside of the samples as shown in Fig. 1(a). The objective is mounted on a motorized z-stage (ANT95-50-L-Z, Aerotech) for precise focus alignment. Motorized x/y-stages (ANT130-XY, Aerotech) allow precise positioning and translation of the sample during the process. An on-axis CCD-camera (DMK 27AUJ003, Imagingsource) in combination with a tubus lens is used for focus alignment and beam control. Here, a small part of the laser energy will be reflected by the sample backside, passing back through the objective and the dichroic mirror and can be detected by the camera in combination with the tubus lens (shown by the orange color). Figure 1(b) shows the intensities of the focused beam captured with the on-axis camera for different z-layers (position of the z-axis) named as $\Delta$z from -6 µm to +4 µm, with $\Delta$z = 0 µm being on the bottom side of the sample. Obviously, small changes of the focus layer have a significant influence on the beam intensity distribution. The focal layer is specified to the z-position, showing the highest peak intensity ($\Delta$z = 0 µm) while the concentric rings are related to spherical aberration effects. The spot diameter at the focal position is measured to be 2.3 µm (1/e$^2$).

 

Fig. 1. (a) Schematic illustration of the experimental setup with the laser being focused on the glas/air interface on the backside of the sample by a 20x objective and an on-axis camera being used for beam alignment. (b) Beam intensities captured with the on-axis camera for different z-positions.

Download Full Size | PDF

Both sided polished round fused silica samples (GVB solutions in glass) are used in our experiments having a diameter of 14 mm and a thickness of 2 mm. To investigate the experimental results, a laser scanning microscope (LSM, VK-X210, Keyence) as well as a scanning electron microscope (SEM, MAYA3, TESCAN) is applied. The fused silica samples are coated with 100 nm of gold to ensure electrical conductivity before being analysed in the SEM. The presented cross sections are the average of 10 lines while the depths evaluation is done by a set of 5 measurements (each with 10 lines average), being indicated by the errorbars. The intra-cavity roughness R$_a$ of the ablated cavities is measured on their bottom by averaging measurements along 5 lines. All samples are cleaned in an ultrasonic bath with aceton after laser processing to remove residual debris piled up in the ablated geometries.

3. Results

3.1 Influence of laser and scan parameters

The influence of the laser parameters fluence and writing speed are investigated by hatches consisting of 10 parallel lines with a distance H of 1 µm between each. A detailed description of the writing scheme is inserted in Fig. 2 (right side) with the green arrow marking the scanning direction. However, Fig. 2 shows SEM images for different writing speeds (10 mm/s - 30 mm/s) and fluences (5.60 J/cm$^2$ - 15.48 J/cm$^2$) for single pass ablation. The corresponding pulse energies are within the range of 0.23 µJ and 0.64 µJ while the applied pulse repetition rate of 50 kHz results in a pulse overlap (overlapping diameters) in writing direction of 91.3% for 10 mm/s, 82.6% for 20 mm/s and 73.9% for 30 mm/s, respectively. Due to the higher pulse overlap, it is found that the surface ablation starts at various fluences for the different writing speeds (white rectangles). At the lowest speed of 10 mm/s, a fluence of 7.80 J/cm$^2$ is required to ablate the entire channel, while for high fluences, the ablated region looks damaged or not completely ablated (right side of channel). The minimum ablation fluence is increased from 8.90 J/cm$^2$ for 20 mm/s to 12.19 J/cm$^2$ for 30 mm/s, respectively. We assume, that the lower fluence necessary for ablation at lower scanning speeds is mainly related to incubation which describes a decreasing ablation threshold with increasing number of pulses. However, based on the applied pulse energies, the peak power of the pulses is calculated to be within 0.94 MW (0.23 µJ) and 2.6 MW (0.64 µJ) and thus exceeding the critical power $P_c~=~3.77\lambda ^2/8\pi n_0n_2$ to cause filamentation [35,36]. Here, $P_c$ is calculated to be 0.78 MW with $\lambda$ being the laser wavelength (515 nm), $n_0$ the linear refractive index (1.45) and $n_2$ the nonlinear refractive index of the material (3.54*10$^{-16}$ cm$^2$/W [37], for 800 nm). Further studies have to be done, to explain the mechanisms behind our backside ablation process in detail and clarify the impact of filamentation. Nevertheless, owing to the small process window between not completely ablated and damaged regions for the 10 mm/s as well as the high fluences necessary for 30 mm/s, a writing speed of 20 mm/s is chosen in the further experiments.

 

Fig. 2. SEM images of ablated channels for different writing speeds and fluences. White rectangles mark the first fully ablated channels. Schematic illustration (right side) shows the writing scheme.

Download Full Size | PDF

To investigate the channel ablation in more detail, the same experimental design (hatches with 10 parallel lines with 1 µm separation) is used while keeping the fluence and writing speed constant at the parameters found in the previous part (20 mm/s, 8.90 J/cm$^2$). Figure 3(a) shows the cross section for different numbers of scans from single pass up to 50. Obviously, the channel depth increases with increasing number of scans. For a single pass, an inclined ablation area is visible, while the first line scanned within the hatch was on the left side (lowest ablation) and the last line of the hatch on the right side (deepest ablation). This observation implies, that each subsequently scanned line leads to a deeper ablation than the previous one. For higher repetitions, a similar behaviour is found while the inclined bottom is deeper in the material, leaving straight sidewalls as seen for e.g. 10 repetitions. While the channels become deeper with increased repetitions, the appearance of the inclined ablation region becomes less pronounced and disappears for repetitions of 20 and 30. For the highest numbers of scans, the bottom of the grooves reveals a bulgy geometry with deeper ablation depth on the outer sides and a decreased depth in the middle.

 

Fig. 3. Influence of the number of scans on the ablated channels for a fluence of 8.90 J/cm$^2$ and a writing speed of 20 mm/s (a) LSM cross section and (b) measured ablation depth and mean depth per layer.

Download Full Size | PDF

Measurements of the ablation depth are shown in Fig. 3(b) while the depth is determined on the deepest position of the groove. As previously described, the depth is increased with the number of scans from 1.90 µm for the single scan up to 10.52 µm for the maximum applied scans of 50. The data fits well with a logarithmic function saturating at a higher number of scans. Calculating the ablated depth per layer shows, that a single pass ablates 1.90 µm at the deepest point while for the 50 scans, only 0.21 µm is ablated for each scan. A sharp drop after the first scans is observed following a power regression. Due to the saturation behaviour as well as the different groove geometries shown in the cross section (Fig. 3(a)), a number of 30 scans ablating 0.31 µm with each scan is a good compromise.

Figure 4 shows the influence of the applied fluence on the ablation depth and roughness R$_a$ for a number of 30 scans. Please note that the writing speed is reduced to 2 mm/s, by using a repetition rate of 5 kHz by applying a pulse divider of 10 keeping the pulse overlap as well as the beam properties constant. This reduction is tested due to being used later for a more complex structuring, where a writing speed of 20 mm/s is too fast for the x- and y-axis resulting in an inaccurate positioning of the fused silica sample within the laser ablation process. Figure 4(a) shows the cross section for different fluences from 5.60 J/cm$^2$ to 15.48 J/cm$^2$. Apparently, all channels reveal the same geometry, straight sidewalls and a flat bottom, also for the smaller fluences being not sufficient for ablation in the single pass (below 8.90 J/cm$^2$). The ablation depth is increased from 2.64 µm to 13.46 µm with increased fluence following a smooth logarithmic trend shown in Fig. 4(b) and the channel roughness is measured to be between 0.20 µm and 0.33 µm, marked by the coloured area. While the depth can be controlled by the applied laser fluence, no clear trend can be identified for the corresponding roughness.

 

Fig. 4. Influence of the fluence on the ablated channels for a writing speed of 2 mm/s (5 kHz) and a number of 30 scans (a) LSM cross sections and (b) measured ablation depth and roughness (dashed line is a guide for the eye).

Download Full Size | PDF

3.2 Increasing the dimensions of microchannels

To increase the width of the ablated channels, additional lines have been added in steps of 10 up to a width of 50 lines. Figure 5 shows SEM images for single and 30 scan passes, while a writing speed of 20 mm/s (50 kHz) and fluence of 8.90 J/cm$^2$ is used. For the single scan (Fig. 5(a)) it is shown that the depth of the ablated channel can be increase when using a hatch of 20 lines, a behaviour observed before, that subsequent scanned lines increase its ablation depth compared to the previous one. For the ablation of 30 lines, this trend can not be continued for the entire area and a new inclined channel starts to form with smaller ablation between the two ablated channels. A further increase of the channel width shows the second channel is developed while a third one can be seen for 50 lines. It is assumed that for our experimental setup (writing speed and fluence) with increased number of ablated lines and thus increased ablation depth, a critical point is reached at about 4 µm (see cross section of Fig. 5(a)), where the fluence of the subsequent line is absorbed within the material due to the formed channel and the applied fluence is not sufficient. Furthermore, the remaining material between the individual inclined channels looks rough (inset in Fig. 5(a)) while the cross section shows that material removal occurs. However, using a repetition of 30 scans (point of repetitions for homogeneous channels), a significantly changed ablation is observed, where all channels show homogeneously ablated channels with an adequate geometry having straight side-walls and an almost flat bottom. Therefore, the direct ablation of glass from its backside shows a saturable ablation behaviour while for a successively high number of scans, well formed channels can be fabricated revealing almost vertical sidewalls.

 

Fig. 5. SEM images and LSM cross sections of different channel widths between 10 and 50 parallel lines for (a) single pass and (b) 30 scans. Insets show magnifications of the 50 lines channels.

Download Full Size | PDF

To increase the depth of the ablated channels, the focal position of the laser is changed after each scan to extend the saturation depth. A number of 90 scans is applied while the focal position is changed after each layer for $\Delta$z between 0.25 µm and 0.50 µm in steps of 0.05 µm. Figure 6 depicts the measured cross sections showing channels with vertical sidewalls for all values of $\Delta$z. The contour of the ablated channel reveals a rectangular shape only for $\Delta$z between 0.25 µm and 0.35 µm, otherwise an inclined bottom is observed. The depth of the channels increase with increasing $\Delta$z up to 49.77 µm for $\Delta$z = 0.35 µm while the largest depth is observed for 0.40 µm and decreases afterwards. Due to the inclined channel geometry, the maximum depth of the rectangular channels is observed for $\Delta$z = 0.35 µm which corresponds to an ablation depth of 0.55 µm per layer. The corresponding ablation efficiencies are calculated to be 0.032 mm$^3$/min/W for $\Delta$z = 0.30 µm and 0.036 mm$^3$/min/W for $\Delta$z = 0.35 µm which is significantly lower as being observed for frontside ablation of fused silica at a wavelength of 1030 nm where ablation efficiencies of 0.41 mm$^3$/min/W for non-burst and 3.05 mm$^3$/min/W applying GHz-burst mode has been demonstrated [38]. Comparing our results to other backside ablation processes, the etch rate is calculated by dividing the channel depth by the number of pulses. Therefore, following Farson et al. [39], the accumulated fluence $\Phi _{acc}$ is calculated to be 5.2 which means, each area is irradiated with the fluence of 5.2x the single pulse fluence and thus effectively by 5.2 pulses (for each of the 90 layer). The etch rate is then calculated to 95 nm/pulse and 106 nm/pulse for $\Delta$z = 0.30 µm and 0.35 µm, respectively. Compared to the LIBWE process for fused silica, the etch rates are typically in the range of several tens or hundreds of nm/pulse [1–3] while also rates close to to 1 µm/pulse are reported using liquid gallium as absorbent layer [40]. For the LIBDE process on fused silica, etch rates of 180 nm/pulse [4] and several tens of nm/pulse for the LIPAA [8] can be found. Thus, our results are in the range of other backside ablation processes, yet providing advantages in terms of a simplified process flow.

 

Fig. 6. Influence of the focus adjustment $\Delta$z after each scan pass (a) LSM cross section, (b) measured ablation depth and roughness (dashed lines are a guide for the eye) and (c) SEM images (40$^\circ$ tilted view) for $\Delta$z = 0.35 µm and 0.40 µm (white squares mark the enlarged area).

Download Full Size | PDF

The roughness of the channels increase form 0.21 µm to 0.86 µm with increasing $\Delta$z while a gentle regime is found within the range of the non-inclined bottoms ($\Delta$z between 0.25 µm and 0.35 µm). Figure 6(c) shows channels ablated with $\Delta$z = 0.35 µm (largest depth with rectangular cross section) and 0.40 µm (largest depth). Apparently, channels with vertical sidewalls and sharp edges are fabricated in glass. The enlarged areas (white squares) show that the roughness on the bottom of the channel is significantly increased for $\Delta$z = 0.4 µm while both reveal smaller damage on the edges.

3.3 Demonstrating the versatility of the ablation process

Demonstrating the versatility of the process, not being limited on straight channels, a sample with 5 concentric rings was designed and ablated using 15 parallel lines, a scan number of 30 and a fluence of 12.19 J/cm$^2$. To ensure a high contour accuracy, the writing speed is set to 2 mm/s by applying a pulse divider of 10 (5 kHz). The rings are designed to have inner diameters between 100 µm (smallest) and 800 µm (largest). Figure 7 shows the SEM image of the ablated structures. Obviously, all rings are accurately ablated while the smallest circle (most difficult due to the highest angular speed) is shown enlarged. It has successfully been demonstrated that the here presented backside ablation process is not limited to straight lines and can be flexibly adapted to versatile geometries. Therefore, the presented process is applicable to fabricate rectangular shaped microchannels (cf. Figure 6(c)) directly on the backside of fused silica by focussing the laser on the glass/air interface without using additional absorbent layer or time-consuming etching.

 

Fig. 7. SEM image (40$^\circ$ tilted view) of ablated rings with different diameters using a writing speed of 2 mm/s (pulse divider 10), a number of 30 scans and a fluence of 12.19 J/cm$^2$.

Download Full Size | PDF

4. Conclusion

In this study, a backside ablation process to fabricate microchannels on fused silica with vertical sidewalls is demonstrated. Compared to other femtosecond laser based microchannel fabrication techniques, this process does not rely on etching solutions which are hard to handle and very time consuming or on additional materials to generate an artificial interface on the glass surface to trigger ablation. In our study, rectangular microchannels having a maximum depth of 49.77 µm are fabricated. A saturation behaviour is found, making the use of multiple scans inevitable when generating channels with flat bottom. However, the depth can be controlled by altering the applied fluence. A further expansion of the channel depth can be achieved by applying focus adjustment after each scan while the applied fluence can be kept small.