1. Introduction

Sub-bandgap ultrafast laser pulses (pulse width < 10 ps) can induce non-linear photon absorption processes when focused into wide bandgap materials such as fused silica. Depending on the irradiation parameters, various permanent material modifications can occur to material that is placed in the tightly confined laser focus, while the surrounding material is left unchanged. One of the most intriguing and applicable modifications induced by ultrafast laser inscription in fused silica is an enhanced chemical etching rate of the laser modified material [1–3]. This enhancement allows complex three-dimensional micro-components to be fabricated by first laser inscribing a structure into a substrate and subsequently subjecting the substrate to a chemical etchant to preferentially remove the inscribed material. This technique offers several advantages over other microfabrication techniques which make it highly desirable, including micrometre feature resolution, the ability to write embedded features, and maskless etching. Conversely, there are some limitations which must be considered when designing a component for fabrication, including: polarization dependent etching rates, long laser writing times and geometry dependent etching selectivity. These limitations have hindered the translation of ultrafast-laser-induced selective etching from a predominantly research-based tool towards wider adoption as an industrial microfabrication process.

To overcome these limitations, several groups have investigated optimal laser writing parameters and advanced laser writing techniques. A vast parameter space exists in terms of the laser irradiation and etching parameters that affect the resultant selectivity of the etching process. Parameter space investigations typically involve writing arrays of channels or surfaces with different parameters and measuring the subsequent etching rates [4–9]. These studies reveal interesting characteristics of the laser writing and etching process. For example, etching selectivity is found to be highly polarization dependent under certain irradiation conditions [10] due to the formation of directional nanogratings which facilitate etchant transport anisotropically. Nanogratings can facilitate very high etching selectivity but present a challenge of how to modulate the laser polarization while writing complex surface geometries. Recently however, longer pulses (> 2 ps) were shown to overcome the strong polarization dependence [11], achieving high selectivity even for circularly polarized light. Early work on laser-induced selective etching primarily used hydrofluoric acid as the chemical etchant, with optimal writing parameters delivering a selectivity of up to 100 [12]. In more recent years, heated potassium hydroxide has been shown to offer an etching selectivity of over 1000 [13], while maintaining similar etching rates to HF [14], and has therefore become the most commonly used etchant. Etching conditions [6,15], writing speed [6,7], pulse energy [8,16] and pulse repetition rate [4] are just some of the other parameters that have been investigated through parameter studies.

Further to using optimal irradiation and etching conditions, advanced writing techniques have also been explored for improving the fabrication process. These include cross-polarised double-pulse writing [17], multispot writing [18], pulse-train modulation [19], pulse front shaping [20,21] and spatiotemporal focusing [22,23]. Recently, beam shaping has become extremely appealing in laser machining [24], enabled by advances in adaptive optics instrumentation such as spatial light modulators (SLMs), which are now more cost effective, offer higher pixel densities and better optical power handling. SLMs allow near arbitrary phase profiles to be imparted across the laser beam pixelwise, and can mimic the role of traditional optics and compensation plates to perform aberration correction and beam shaping amongst an abundance of other applications. Of particular interest is the ability to shape the laser focus to suit a particular laser processing task more efficiently. For example, Bessel beams, which are “non-diffracting” and maintain a high-intensity central lobe over relatively long axial distances, can be generated by passing a Gaussian beam through an axicon lens (strictly speaking, forming a Gaussian-Bessel (GB) beam). The elongated focus of a GB beam permits more material to be processed in a single laser-pass while maintaining lateral resolution. Bessel beams formed using physical axicon lenses have been used for several glass microfabrication applications including low-loss waveguide writing, volume gratings [25,26], microchannels [27,28] and glass dicing [29]. Beam shaping using physical optics is limited by the discrete nature of the optic form, introduces alignment challenges, and offers less design freedom. SLMs provide a practical solution to generating Bessel beams without introducing the restrictions of incorporating physical axicon lenses into the beam path and have also been utilised for glass dicing [30], and fabricating microchannels [31] and micro-holes [32].

There are two common approaches for generating GB beams using an SLM: The first involves displaying a simple annular grating on the SLM, which transforms into a GB beam in the far field [33], i.e. 1-f after a lens positioned 1-f away from the SLM. In fact, any complex amplitude can be obtained in the far-field by generating its Fourier Transform (FT) at the SLM plane. The second method requires a conical phase profile to be imposed on the SLM, analogous to a physical axicon lens, forming a GB beam in the Fresnel plane immediately upon reflection from the SLM. The former method is inefficient as it makes use of only a small portion of the SLM display, while the second approach is efficient but typically incompatible with the 4-f imaging system used to relay image the SLM onto the back aperture of the writing objective to perform aberration correction or write with a diffraction limited Gaussian focus. For many manufacturing applications, it would be beneficial to write with both a traditional Gaussian focus and a GB focus depending on the component geometry.

Recently, zero-order vortex beams, formed by performing the FT of a GB beam [34] (by passing through a lens, for example) were demonstrated for laser-based microfabrication, exhibiting an elongated high-intensity focal region [35,36]. Here, we explore the use of a vortex beam generated by introducing a conical phase front to a Gaussian beam via an SLM, forming an optical vortex at the laser processing focal plane after propagation through a 4-f lens relay system and a flat field objective lens – the ideal optical path for several beam shaping and aberration correction applications used during microfabrication. Crucially, we find that material modification is confined to an elongated high-intensity region preceding the focal plane, with the degree of elongation determined by the magnitude of conical phase front imparted on the beam.

In this paper, we present the results of a thorough investigation into the effects of laser irradiation parameters and SLM generated conical phase front on the etching selectivity and cross-section of microchannels written in fused silica. Writing parameters investigated include the pulse duration (180–5900 fs), pulse energy (0.12–14.4 µJ), pulse repetition rate (25–500 kHz) and sample translation speed (1–128 mm/s). The axial laser depth-of-focus (DOF) was varied by applying phase fronts with conical tilt ranging between 0 to 4.27 mrad. To illustrate the benefit of controllable DOF during laser writing, the rapid fabrication of various geometric fused silica forms by laser writing and subsequent chemical etching is presented.

2. Materials and methods

2.1 Laser inscription and etching

For this work, ultrashort pulses with tuneable pulse width (180–5800 fs) were generated by a diode-pumped ytterbium-doped bulk laser system (Pharos, Light Conversion), with a central wavelength of 1030 nm. The pulse repetition rate was varied between 25 and 500 kHz and an average power of up to 10 W was available. Figure 1 depicts the functional components of the laser inscription system.

 

Fig. 1. A schematic of the laser inscription system. BE: 3 × beam expander, L1: 1000 mm biconvex lens, L2: 400 mm biconvex lens, inscription lens: 0.6 NA long-working-distance objective. Inset: Three phase masks are combined to form the desired beam: a flatness correction mask to compensate for SLM imperfections, a blazed grating to efficiently diffract light into the first order, and an axicon lens for generating the vortex beam.

Download Full Size | PDF

The laser was first passed through a Galilean beam expander with 3× magnification to increase the beam width from 4.5 to 13.2 mm (to 1/e²) in order to sufficiently fill the aperture of the SLM. We used a reflective liquid crystal on silicon (LCOS) SLM (X13138-08, Hamamatsu) with a 1272 × 1024-pixel array measuring 16 × 12.8 mm. The laser was directed onto the SLM at an angle of incidence of 5.5°. The SLM was configured to diffract light into the first order, the details of which are discussed in Section 2.2. We observed a dramatic reduction in the SLM’s diffraction efficiency when the average incident power was above ∼ 2.4 W and therefore limited the average power to below this value. We speculate this drop in efficiency was due thermal effects, a known limitation of SLMs and particularly those relying on an aluminium reflective coating such as that used here, as opposed to a dielectric coating.

After the SLM, the beam was passed through two biconvex lenses in a 4-f configuration (L1 = 1000 mm, L2 = 400 mm), to demagnify and image the SLM display onto the entrance pupil of the writing objective. A 4-f system is required to relay not only the amplitude profile of the image but also its phase, as imaging with a single lens introduces an undesirable quadratic phase term. The 4-f system also provides a convenient intermediate image plane at which an aperture can be positioned to filter out unwanted diffraction orders. After L2, the collimated beam was passed through a zero-order quarter-waveplate followed by a zero-order half-waveplate mounted within motorized rotation mounts (DDR25/M, Thorlabs) for polarization control. Finally, the beam was passed through a long working distance (10 mm) objective with NA = 0.6 (PAL-50-NIR-HR-LC00, OptoSigma) and an effective focal length of 4 mm, and focused into the sample. Sample translation was provided by a 3-axis air-bearing direct-drive stage (ABL1000, Aerotech). A useful guide to implementing SLMs for laser processing and microscopy is offered by the Dynamic Optics and Photonics Group, University of Oxford [37].

The glass used in this study was 1 mm thick high purity fused silica (7980 0F, Corning). After laser inscription, samples were submerged in an etching bath containing 8 mol/L potassium hydroxide (KOH) heated to 85°C and magnetically stirred throughout the etching period. These conditions have been investigated previously and provide high etching selectivity [14]. The thinning rate of pristine material under these etching conditions was previously reported by us to be 0.38 ± 0.02 µm/h, and this value was used to determine the etching selectivities reported here.

2.2 Vortex beam generation

The SLM imposes a phase front across the laser beam by controlling the effective path-length of light passing through the liquid crystal display, pixelwise. Each pixel typically allows up to 2π of phase delay. For most applications, the phase profile required spans a larger range than 0 – 2π and this is achieved by wrapping the required phase value to the modulo of 2π. To apply the phase profile, an 8-bit (256 intensity levels) bitmap with phase encoded intensity values was simply projected on the SLM display via a digital video interface (DVI) cable connected to a computer.

The bitmap is a summation of various phase maps carrying out specific functions: Firstly, it is necessary to correct for any physical flatness deviation on the SLM display itself by applying a flatness correction mask which is often supplied by the SLM manufacturer, though can be determined experimentally [38]. Secondly, it is often beneficial to apply a multilevel blazed diffraction grating via the SLM to diffract the laser light efficiently into the first diffraction order. This prevents unmodulated light in the zeroth order from degrading the desired beam profile. Here, a 16-level blazed grating was applied with an effective grating period of 6.25 lp/mm, providing a diffraction efficiency of ∼82% into the first order. Finally, to mimic the function of an axicon lens, a positive (i.e., positive maximum phase shift on the optical axis) conical phase mask was generated. The magnitude of phase shift at the centre, φ (radians), is related to the conical phase base-angle, α (radians) and semi-diameter of the active area of the SLM imaged onto the pupil plane of the objective, R, for a given wavelength, λ, by Eq. (1):

Note that the base-angle of a physical axicon lens producing a phase shift, φ, is equal to α/Δn, where Δn is the refractive index difference between the lens and the surrounding medium. The base-angle is a common axicon parameter and is useful for comparing systems utilizing SLM-generated and physical lenses. In our investigation, we generated phase masks with peak phase shifts of 0, 37.5, 75, 112.5 and 150 radians, corresponding to conical phase base-angles of 0, 1.07, 2.13, 3.20 and 4.27 mrad for a semi-diameter, R, of 5.76 mm. The combined phase mask is simply the sum of the flatness correction, diffraction grating and conical front, wrapped at 2π, as depicted in Fig. 1. When the laser reflects off the SLM, the conical phase mask produces a GB beam in the near-field, which is reimaged to the back aperture of the objective lens and transformed into a vortex beam at the laser-writing focal plane. A mathematical description of a vortex beam generated by the FT of a GB beam is provided in Ref. [34].

2.3 Parameter investigation

A parameter-space investigation was carried out to explore the effects of laser irradiation on both the laser-induced etching selectivity and the elongation of the modified region, for a range of conical phase fronts. Three parameter space subsets were investigated: firstly, a preliminary scan of pulse duration, pulse energy and axicon gradient; secondly, a finer scan of pulse duration, pulse energy and writing speed, and thirdly, a comprehensive study of pulse energy, writing speed, conical phase front and pulse repetition rate. As noted, previous studies have demonstrated that high etching selectivity can be achieved with circularly polarised pulses under certain conditions, overcoming the need for polarization alignment during fabrication. This benefits the fabrication technique greatly, and therefore we used circularly polarized light at the substrate throughout the work presented here.

For each parameter combination, a single transverse channel was written into the substrate, with the focal plane positioned 500 µm nominally beneath the surface, i.e., a stage movement of 500 µm into the glass. To account for beam clipping at the substrate edge, the samples were mechanically diced 2 mm in from the edge to expose the channel facets prior to etching. The channels were then etched for 3 hours before micrographs were taken of the channels’ cross-section and the etched lengths.

3. Results

3.1 Parameter space investigation

The first parameter investigation was carried out to determine the influence of pulse energy, pulse duration and conical phase angle on the etching selectivity and channel cross-section. For this parameter set, the repetition rate was fixed at 50 kHz, the beam was circularly polarised and the sample was translated at 4 mm/s. For these parameters, the number of spatially overlapped pulses irradiating the material was approximately 19 (see Supplement 1 for calculation). In order to reduce the occurrence of self-focusing and consequent filamentation due to nonlinear Kerr lensing, the pulse energy was limited to 350 nJ for the Gaussian focused beam (conical phase angle = 0 mrad), compared to 2200 nJ for the vortex beams. Light emitted from the partial free-electron plasma, formed within the laser focus, was imaged from the side to indicate the modification threshold and focus elongation. Micrographs of the plasma emission with and without a conical phase font are presented in Fig. 2 (a) and (b) respectively. Selected channel cross-sections written with both Gaussian and vortex beams for different pulse energies are presented in Fig. 2 (c). Although the vortex beam is expected to form a ring at the focal plane, it is evident that ionisation is dominant within a central high-intensity region just prior to the focal plane, resulting in the elongated modification channel observed. We observed no significant difference in the channel widths for channels formed with Gaussian and vortex beams, and rather the width was determined predominantly by the pulse energy.

 

Fig. 2. (a) and (b) are micrographs of partial free-electron plasma emission from the focal region inside fused silica under irradiation with a Gaussian and vortex focus (generated by imparting a 3.20 mrad conical phase front) respectively. (c) is a series of micrographs showing etched channel facets written with a range of pulse energies and conical phase masks to highlight the DOF elongation. The repetition rate was 50 kHz, polarisation was circular, translation speed was 4 mm/s and the pulse duration was 1440 fs when writing the channels shown. Inset: A corresponding channel cross-section prior to etching (at same scale).

Download Full Size | PDF

Graphs displaying the selectivity and elongation of the etched channels from the first parameter investigation are presented in Fig. 3. The most significant uncertainty in the selectivity originates from the variation in temperature of the etching bath which dramatically affects the etching rate of the pristine material. Therefore, the uncertainty in the selectivity is approximately equal to the percentage uncertainty in the pristine etching rate of silica at 85 ± 2 °C, i.e. 0.38 ± 0.02 µm/h or ± 5.3%. Note that some degree of variance in the laser writing and etching process is expected beyond this, but repeated measurements for each parameter set was impractical given the size of the parameter-space investigated. Several clear trends were observed, and others more subtle: As expected, a minimum pulse energy threshold was required to induce selective etching; this threshold increased for longer pulses and steeper conical front. The selectivity generally plateaued at higher pulse energies for the range investigated. The selectivity was generally higher for longer pulses and particularly, noticeably lower for 180 and 360 fs pulse durations (note, polarisation was circular). This observation is in agreement with previous research suggesting that for longer pulse durations, the nanogratings that enable high selectivity lose their directionality and become randomly orientated, facilitating highly selective etching even for circularly polarised light [11]. The height of the channels after etching, i.e., the DOF, denoted as the elongation here, was found to increase with pulse energy gradually to a broad peak before falling again slightly in most cases. Most significantly, introducing conical phase profiles via the SLM resulted in channel elongation proportional to the conical phase front base-angle. This observation was true up to a conical angle of 3.20 mrad, beyond which the elongation began to decrease. Prior research has demonstrated that the focal volume intensity distribution of a vortex beam is a complex combination of writing depth, conical base-angle and direction and pulse energy, and that the elongation increases with base-angle only up to a point, at which the focus becomes less well-defined and the elongation reduces [35]. We also note that for steeper conical phase fronts, the diameter of the propagated annular beam increased and became susceptible to aperturing. It is also apparent that simply increasing the pulse energy increases the DOF even when no conical phase is applied. For the pulse energy range investigated here, this elongation can be attributed simply to an increase in the extent of the focal volume above the modification threshold. At higher pulse energies, filamentation is also expected to contribute to the elongation of the modified region and this was explored further in the third set of parameters investigated.

 

Fig. 3. Line graphs displaying etching selectivity (left) and DOF (right) for several SLM generated conical phase masks for varying pulse energy and pulse duration. In each case, the laser polarisation was circular and the translation speed was 4 mm/s. The uncertainty in the selectivity is approximately 5.2%.

Download Full Size | PDF

The elongation of the preferentially etched material was also less for longer pulses with comparable pulse energy, particularly for pulse durations of 2880 and 5900 fs. We suspect that this is due to the reduction in peak intensity within the focus, leading to a smaller laser-affected volume. An optimum pulse duration exists, then, in which both selectivity is high for circularly polarized light (≥ 720 fs) and the modification volume is efficiently elongated (≤ 2880 fs).

The substrate translation speed is another important factor when considering efficient microfabrication. To investigate the influence of translation speed on the etching selectivity and focus elongation, it too was investigated as a function of pulse energy and pulse duration. The results of this investigation are presented as line plots in Fig. 4. Channels were again written at a nominal depth of 500 µm, with translation speeds ranging from 1 to 8 mm/s (corresponding to 75 to 10 overlapped pulses) and pulse energies from 500 to 2250 nJ. Since the first parameter investigation determined that a minimum pulse duration between 720 and 1440 fs is required for optimum etching performance, pulse durations from 1080 to 5900 fs were investigated here. A fixed conical phase angle of 3.20 mrad was used, as this resulted in longest elongation after the first investigation.

 

Fig. 4. Line graphs displaying etching selectivity (left) and DOF (right) for several pulse durations with varying pulse energy and writing speed. In this case, the laser polarisation was circular and the axicon base-angle was fixed at 3.20 mrad. The uncertainty in the selectivity is approximately 5.2%.

Download Full Size | PDF

We observed that the substrate translation speed was only significant for pulse durations of 1080 and 1440 fs, for which slower writing speeds resulted in significantly lower etching rates. We also observed that the modification elongation decreased slightly as the pulse duration increased. The results suggest that to achieve longest modification elongation along with highly selective etching, pulses with durations from 1080 to 1440 fs with translation speeds > 3 mm/s are optimum. For scenarios in which the translation speed is limited, longer pulses may give better results.

With the optimum parameters narrowed down, a third and final parameter set was investigated, exploring a wide range of pulse repetition rates, pulse energies and writing speeds for several conical phase angles starting at 0 mrad. Again, single-pass channels were inscribed in fused silica at a nominal depth of 500 µm and chemically etched for 3 hours. The achieved selectivities and DOFs are presented as heat maps in Fig. 5. For this investigation, the pulse duration was kept constant at 1440 fs, because this produced highest selectivity and elongation for the broadest set of irradiation parameters in the initial studies. The dark blue regions represent parameters in which no material modification was observed, and the grey regions represent inaccessible parameters due to the average power surpassing the maximum efficiently handled by the SLM as discussed earlier. Each pulse energy increment along the x-axis of the heat map is subdivided into eight increments which represent eight translations speeds use: 1, 2, 4, 8, 16, 32, 64 and 128 mm/s.

 

Fig. 5. Heat maps displaying the results of the final parameter space investigation. The selectivity (top) and DOF (bottom) were measured for channels written with several pulse energies, writing speeds, conical phase base-angles and pulse repetition rates. The minor tick marks along the pulse energy axis represent doubling sample translation speeds from 1 to 128 mm/s, while the minor tick marks along the repetition rate axis denote base-angles from 0 to 4.27 mrad. Circularly polarised light was used throughout. The uncertainty in the selectivity is approximately 5.2%.

Download Full Size | PDF

The results highlight that the pulse energy threshold for modification increases with larger conical base-angle, as expected. An interesting behaviour was observed when varying the translation speed parameter: For lower pulse energies, faster writing speeds resulted in a lower etching rate, suggesting that the optimum fluence was not reached at those speeds. However, we also observed etching rates were sub-optimal when slowest writing speeds were used in many cases, suggesting that pulse overlap also has an optimum value. This behaviour was not observed for the elongation however, where an inverse relationship between writing speed and elongation was observed. Further, we observed that the optimum translation speed for high etching selectivity increased with higher repetition rates for similar pulse energies, suggesting the total fluence determines the morphological change rather than the individual pulse energy, in agreement with previous studies [16]. This observation is significant for industrial manufacturing applications as it implies that fabrication time can be reduced simply by increasing the writing speed and repetition rate proportionally. Challenges remain however in overcoming the average power limits handled efficiently by modern LCOS-SLM’s. Indeed, we observed that only the lowest repetition rate, corresponding to lowest average power, resulted in fully elongated channels.

Again, channel elongation was proportional to the conical base-angle where sufficient pulse energy was available. For this parameter set, channels were also written at high pulse energies for a base-angle of 0 mrad. In this regime, filamentation due to Kerr self-focusing is likely, and indeed we recorded DOF elongation up to 147 µm for 14.4 µJ pulses. However, this is still substantially less than the DOF observed when writing with a vortex beam.

An alternative representation of the data provided by the heatmap in Fig. 5 can be found in Supplement 1. Here, the data is offered in terms of the number of overlapped pulses and the net fluence for each pulse energy investigated. It is common to consider pulse repetition rates below 1 MHz to be thermally non-cumulative; therefore, one may expect the repetition rate and translation speed to be inherently correlated, with the material modification influenced by the pulse overlap alone. We observed this to be the case in general, however, we note a trend for the modification threshold to be more readily reached with higher repetition rates for a given pulse energy and pulse overlap. We also observe a stark reduction in selectivity when writing with many overlapped pulses (approximately >50). We suggest that this may in fact be due to thermal accumulation leading to the spontaneous annealing of the material defects that drive selective etching. The observation of thermal accumulation at repetition rates < 1MHz [39], and even as low as 1 kHz [40], supports this. As expected, the threshold for modification in terms of net fluence increases with applied conical phase. Generally, once the threshold is reached, the selectivity achieved is consistently around 700, regardless of the degree of elongation.

In terms of the focus elongation, we observed the DOF to increase with conical phase as expected. We further observed the degree of elongation to increase with the number of overlapped pulses for a given pulse energy. At lower pulse energies, the elongation achieved by applying steeper conical phase decreased, and was eventually surpassed by that achieved with shallower conical phase applied. This is likely due to the pulse energy distributed at the extremities of the focal region dropping below the modification threshold.

3.2 Demonstration of glass dicing

To demonstrate the benefit of controllable DOF during inscription, we compared the time taken to inscribe simple geometric shapes, namely an ellipse, star, circle and triangle, into fused silica, both with and without beam elongation. The components were chosen as they demonstrate highly selective polarisation insensitive etching, rapid laser inscription via beam elongation, and are not manufacturable using traditional methods. Firstly, we determined the maximum raster scan layer spacing that resulted in sufficient overlap of the modified regions to facilitate highly selective chemical etching across layers. Surfaces were written through 2 mm thick fused silica substrates with a length of 2 mm and a variable raster spacing. Both Gaussian and vortex (generated with a 4.27 mrad conical phase front) foci were used for comparison. For both beam types, the pulse duration was 1440 fs and the repetition rate was 50 kHz. The pulse energy was 1620 nJ for the Gaussian focus, and 8100 nJ for vortex focus. For the Gaussian focus, the raster spacing was initially set to 2 µm and gradually increased to 20 µm in steps of 2 µm and for the vortex focus, the raster spacing was varied between 100 and 450 µm is steps of 50 µm.

The substrates were then etched in 8M KOH at 85°C for 2 hours to allow the etchant to penetrate through the inscribed material. Afterwards, brightfield micrographs were taken of the etched surfaces to clearly identify the surfaces with overlapped modification regions through the substrate, as shown in Fig. 6 (a) and (b). We found that a spacing greater than 16 µm resulted in disconnected channels that prevented etching through the sample when writing with a regular diffraction limited Gaussian focus. Conversely, by using a vortex focus, we observed continuous modification of the glass and subsequent highly selective etching at a layer spacing of up to 200 µm. We note that the theoretical confocal parameter of the Gaussian beam waist when focused with an NA of 0.6 in fused silica is 4.4 µm (for a beam quality, M, of 1.2), less than the 16 µm scan spacing that resulted in continuous modification. This observation may be attributed to spherical aberration-induced focus elongation within the substrate [41], self-focusing, or simply because the modification threshold energy extended beyond the confocal parameter in this case.

 

Fig. 6. (a) and (b) show chemically etched tracks written with a vortex beam to determine the maximum z-spacing for continuous modification through silica glass. The surfaces were formed by raster scanning 2 mm long tracks in 2 mm thick glass. Note the micrographs were taken with the surfaces tilted. (c) shows a star, triangle and circle within an ellipse, each inscribed into a 1 mm thick fused silica substrate. The separated components, written with a vortex beam in 72 s, and subsequently etched for two hours, are shown in (d). (e) and (f) are surface profiles of the 1 mm thick etched surfaces inscribed with a Gaussian and vortex focus respectively.

Download Full Size | PDF

With the maximum beam elongation established, geometric shapes were laser written in fused silica substrates, measuring 10 × 10 × 1 mm, and etched in KOH for 2 hours. The full thickness of the substrate was modified layer by layer, with each layer separated by 16 and 200 µm when using a Gaussian and vortex beam respectively. The inscribed substrates and post-etch shapes are presented in Fig. 6 (c) and (d) respectively. With a writing speed of 4 mm/s, the total inscription time was approximately 15 minutes when using the Gaussian focus, and just 72 seconds when using the vortex focus.

The 12-fold reduction in inscription time highlights the potential for ultrafast-laser-induced selective etching to be used for industry level manufacturing tasks. The two-hour etching time is compatible with rapid prototyping and, as a serial process, does not present a barrier for large scale manufacturing.

The surface quality of laser fabricated components may also be of interest for certain applications, such as when manufacturing micro-optics and microfluidic devices. The form of etched surfaces written with a Gaussian focus and a vortex focus (with the irradiation parameters stated above) through 1 mm thick fused silica was measured by focal-variation microscopy (S neox, Sensofar), and the corresponding surface plots are presented in Fig. 6 (e) and (f). The roughness measured was significant in both cases: the arithmetic mean roughness (ISO 25178) was 336 nm and 460 nm for Gaussian- and vortex-written surfaces respectively. It is evident from the depth dependence of the surface texturing when writing with a Gaussian beam (Fig. 6 (e)) that laser damage was at least partly responsible for the roughness; unsurprisingly, given the relatively high pulse energy (1620 nJ) used. Perhaps more significant is the clear appearance of laser-scan layers when writing with the vortex beam (Fig. 6 (f)). The layers are formed with the same period as the layer spacing, 200 µm in this case, and are indicative of a non-uniform energy distribution over the DOF. In this instance, the peak-valley variation on the surface between adjacent layers was approximately 6 µm.

It is worth noting that lower roughness has been achieved by writing with a diffraction limited focus and lower pulse energy, particularly in the nanograting regime [5,42]. For surfaces requiring a smooth finish, writing with less energy and a smaller layer separation, at the expense of longer writing times, may be beneficial.

4. Discussion

Over the past two decades, ultrafast-laser-induced selective etching has been refined to the point where it may now be seriously considered for industry-level manufacturing; aided by the development of robust and less-expensive femtosecond laser systems. Advanced techniques, such as beam shaping, continue to improve the efficiency of laser energy deposition during writing, leading to significant reductions in laser processing times. Adaptive optics offers unparalleled control of the beam intensity profile; facilitating near arbitrary spatial control of the focal volume in real-time, generally without the need to alter the position or alignment of physical optics in the beam delivery path. A significant exception to this occurs when generating a Bessel beam using a diffractive axicon via an SLM, since the Bessel beam forms in the near-field of the axicon rather than in the laser processing plane.

In this work, we have used SLM-generated conical phase fronts, analogous to small-angle diffractive axicon lenses, to generate a high aspect-ratio laser focus extending to the focal plane and demonstrated that the degree of elongation can be controlled simply by increasing the base-angle of the conical phase front. We applied this technique to ultrafast-laser-induced selective etching of fused silica and found irradiation parameters for which the etching-rate enhanced region formed by a single-pass of the laser focus can be elongated by 12.5 times that inscribed with a Gaussian focus. Crucially, the etching selectivity remains high (approximately 800 - 1000) and polarization insensitive for the optimum parameters found. These findings are beneficial in both research and industrial settings: the significant reduction in laser writing time resulting from the elongated modification volume further enhances the rapid-prototyping appeal of laser-assisted etching and enables the fabrication of larger components and quantities which would otherwise be impractical. Further, circumventing the need for strict feature-dependent polarisation alignment while writing greatly reduces the complexity of the laser inscription and hardware control systems.

Interesting trends were observed from the parameter investigation: For a given pulse energy, the elongation of the DOF increased with the number of overlapped pulses, particularly when a conical phase was applied, as long as the pulse energy threshold requirement was met. A consistently high selectivity was achieved for the full range of conical phase amplitudes investigated when sufficient pulse energy was used. However, the selectivity also displayed a characteristic drop-off for each pulse energy investigated when the number of overlapped pules was high. This suggests that there is a point at which additional laser pulses undo the modification that facilitates increased etching rates. Since circular polarisation was used throughout this study, we expect that the enhanced etching rate was not driven by the formation of directional nanogratings, but rather the formation of molecular defects, nanopores and nanocracks within the laser affected zone. Therefore, we propose that when sufficient pulse overlap is reached, thermal accumulation locally anneals the laser affected zone leading to the drop in etching rate.

One may expect the spatial resolution when writing with a vortex beam to decrease compared to writing with a Gaussian beam. However, we found that this is not generally the case, and that high aspect-ratio channels can be formed irrespective of the conical phase applied. Writing with a vortex beam does affect the surface quality, however, and therefore, this approach is best suited to fabricating large structural surfaces in dramatically reduced laser inscription times, and less suited to forming surfaces in which high surface quality is needed, for example forming optical flats.

The vortex focus offers a key advantage compared to a more typical GB focus, in that the optical path is compatible with several adaptive-optics-enabled beam-shaping applications without any necessary mechanical changes during laser writing, for example, aberration correction [43], adaptive slit shaping [44], multispot writing [45] and accelerating beam writing [46,47]. This implementation facilitates the fabrication of complex components consisting of a variety of surface types and features such as intricate elements requiring high resolution, deep surfaces, and large planes. The reduction in inscription time by 12-fold highlights the potential for industrial level glass cutting, drilling and shaping.

Challenges remain in terms of the power handling capabilities of LCOS-SLM’s, however, recent progress in thermal management promises significant improvement in SLM performance at high powers [48].

5. Conclusions

In this paper, we have demonstrated a method of varying laser focus DOF using an SLM for applications in direct laser writing and selective chemical etching. By introducing a conical phase delay to the laser beam via an SLM, a vortex beam was formed at the focal plane after propagation through a 4-f lens relay and objective lens, forming an elongated high-intensity focal region permitting high-aspect ratio glass modification of up to 350 µm along the optical axis. Meanwhile, the chemical etching selectivity of the laser irradiated material remained high, above 1000 in some cases, as confirmed by comprehensive parameter-space investigations. Further, laser inscription was performed with circularly polarized light, demonstrating that the need for maintaining strict polarization alignment to achieve highly selective etching can be overcome. To demonstrate the benefit of elongating the DOF for glass microfabrication applications, a selection of fused silica geometric shapes were fabricated by ultrafast laser induced selective chemical etching with writing time reduced by 12-fold when compared to writing with a more typical Gaussian focus.