1 Introduction

Lab-on-a-chip (LOC) devices allow faster and more selective reactions to be carried out for low reagent concentrations all within a compact platform. One direction in LOC device fabrication seeks to develop a fused silica platform to harness the high optical transparency and low fluorescence of glass, combined with its hydrophilic nature, nonporosity, and compatibility with a diverse range of chemicals and biomaterials. To this end, femtosecond laser assisted micromachining is of rapidly growing interest when used in combination with hydrofluoric acid (HF) etching to flexibly create three dimensional (3D) integrated micro-fluidic systems inside the bulk glass [1–7]. Traditionally wet-lab processes only generate micro-fluidic devices in a planar geometry that limits interconnection and routing of isolated channels across the surface of a chip. Extending the chip layout to make full use of all three spatial dimensions is a significant objective to greatly improve the flexibility of LOC devices and the ability to integrate optical waveguides and other sensors [8–14].

Moderately low repetition rate femtosecond lasers, with a 1 to 250 kHz repetition rate and a wavelength of 800 nm, were first used to modify glasses within a small focal volume, such that they became susceptible to a much higher etch rate in dilute HF acid than the surrounding, unexposed material [1–7]. Hnatovsky and coworkers demonstrated that exposures with a laser polarization perpendicular to the scan direction offered a 70-fold increase in HF etching rate compared with parallel polarization [5, 6, 15]. They associated this high differential etch rate with the formation of volume gratings with nano-planes aligned orthogonal to the polarization direction that opens the flow of HF solution along the laser modification track. Circular polarization offered a similar high contrast in etch rates to those achieved with perpendicular polarization due to the disordered nature of the nanostructures formed which in turn do not restrict the flow of HF acid along the laser modification track. Differential etch rates as high as 280:1 between laser modified and unmodified glass have been reported [5, 6] for exposure with 40 fs pulses with an energy of 200 nJ, followed by etching with a 2.5% HF solution. While this femtosecond laser assisted technique allows the formation of high aspect ratio micro-channels, the embedded channel length is often limited to less than 4 mm before tapering and uncontrollable cross-sectional shape evolve to restricted this application within the LOC community [4, 6, 10]. Osellame et al. introduced a conical wobbling technique to create long straight and uniform cylindrical micro-channels but this technique may not be flexible for creating other cross-sectional shapes or non-straight channels [16]. Recently, buried micro-channels of nearly uniform cross section were formed up to 9.2 mm length, but required a high concentration (10 M) of aqueous KOH and elevated temperature of 80∘C to yield only a moderate etch rate of ∼1.7 μm/min [17].

In this paper, we report on a method of extending the use of femtosecond laser processing of fused silica to produce large buried channels with controllable shaping of the cross-sectional profile by forming multi-scan modification tracks with a high repetition rate (1 MHz) femtosecond laser. A comprehensive study of the differential etch rates as a function of laser polarization, laser energy, sample scanning speed, sample scanning angle, and multi-line scanning defined optimized exposure conditions to form exceptionally smooth-walled rectangular micro-channels in addition to circular and elliptical shape of varying size and aspect ratio, demonstrating the control of the cross-sectional shape of the channel. In addition, we report on a new technique of incorporating vertical access shafts to extend the reach of HF etching and form channels of unlimited length, thereby removing the need for an ultrasonic bath in the post-etching process to facilitate the formation of delicate microstructures without inducing cracks in the substrate. Finally, nanograting orientation was exploited for selective etching to combine buried micro-channels with low insertion loss optical waveguides that may serve as a new platform for 3D optofluidic microsystems.

2 Experimental setup

A fiber-amplified femtosecond laser (IMRA Jewel D-400-VR), frequency doubled from 1045 nm to 522 nm, operating at a repetition rate of 1 MHz with a 220 fs (Lorentzian) pulse duration and pulse energy (E p ) ranging from 50 to 200 nJ, was used to expose fused silica (Corning 7980) substrates. The laser beam was focused with a 40× aspherical lens of 0.55 NA into the transparent glass substrate at depths (d s ) varying from 75 to 210 μm below the surface. Laser modified tracks were formed with parallel, perpendicular, and circular polarized beams scanned transversely at speeds (v s ) ranging from 0.2 to 10 mm/s using a 3-axis air-bearing translation stage (Aerotech ABL1000 motion stages) which had a 2 nm resolution and 200 nm position repeatability. Further details of the beam delivery and second harmonic setup can be found elsewhere [18, 19].

To shape the cross section of embedded micro-channels, single- and multi-scanned laser modification tracks were written inside the glass plate under variable exposure conditions as shown schematically in Fig. 1. Rectangular cross sections were studied by forming a 5 × 5 array of modification tracks with layer-to-layer separations ranging from d L =1 to 12 μm and track-to-track separations ranging from d T =1 to 3 μm. Elliptical and circular shaped micro-channels were also formed with various array geometries as shown in Fig. 1. All laser tracks were scanned layer-by-layer from bottom to top to avoid propagation of the laser through modified scattering volumes. After laser patterning, the end facets were ground and optically polished before wet etching the substrate in 2.5 to 5% concentration of HF for 1 to 3 hours at room temperature without the use of an ultrasonic bath. In order to fabricate long (>4 mm) embedded micro-channels and minimize the tapering effect seen [10] over such long micro-channels, small access holes were periodically formed as depicted in Fig. 1 by scanning the laser vertically from the surface at a speed of v s =1 mm/s to the transverse modification tracks with a pulse energy of E p =100 nJ and thus provide multiple access points for the HF acid to reach the buried transverse tracks.

Fig. 1
figure 1

Schematic showing the multi-scanned laser modification zones for elliptical/circular and rectangular cross-section patterns scanning in y-axis (θ=0∘) direction where d s is the depth from the surface, d T is the transverse line-to-line separation along the x-axis, d L is the layer-to-layer separation along the optical z-axis. Vertical laser tracks were also formed to provide multiple access ports for HF etching

Full size image

In order to investigate the possibility of creating uniformly shaped micro-channels in circular or arc-shaped pathways, etch rates were examined for a variety of polarization conditions (parallel, perpendicular, and circular) and scan directions in 1∘ increments over θ=−45∘ to +45∘ relative to the y-axis (θ=0∘ in Fig. 1). The single tracks were written at a speed of 0.5 mm/s with energies in the range of E p =75 to 150 nJ and depths of 75 to 210 μm from the surface. The laser-induced nanograting structure was also examined on polished sample facets with a scanning electron microscope (Leo 1540 SEM) prior to HF etching. Finally, to integrate buried optical waveguides within the micro-fluidic substrate that would be single mode and low loss in the visible spectrum, laser exposure conditions were tuned from those reported previously for infrared guiding conditions from Shane and coworkers [20].

3 Results and discussion

3.1 Nanograting formation

For the range of laser exposures applied here to write optical waveguides and micro-fluidic channels, the orientation of the nanogratings aligned to be perpendicular (or parallel) with respect to the scanning direction when the laser polarization (E-vector) was parallel (or perpendicular) [5–7, 15, 21]. Circular polarization was also examined, yielding randomly oriented nanogratings that appeared as a porous structure in the SEM. For perpendicular polarization, the nanograting period was observed to decrease from ∼180 to ∼120 nm as the pulse energy increased from 75 to 150 nJ, which is contrary to the observation by Shimotsuma et al. for their condition of linear polarization with static exposure and longer 800 nm wavelength [22], but consistent with Kazansky and coworkers for linear polarization with wavelengths of 522 nm, 800 nm and 1045 nm [21]. The nanoplasmonic model developed by Bhardwaj et al. [15] predicts an energy-independent nanoplane spacing of λ 0/2n=180 nm for our λ 0=522 nm wavelength, which is only consistent with our exposure of 75 nJ pulse energy. More advanced models are clearly necessary to explain these disparate observations.

3.2 Etch rate dependence on polarization, pulse energy, scanning speed and scanning angle

An average HF etch rate was assessed for single-scan modified tracks formed in fused silica over a large range of laser exposure conditions. Perpendicular (x-axis in Fig. 1), parallel (y-axis), and circular (x–y plane) polarization of the laser were tested for laser scanning along the y-axis to form tracks perpendicular to the facet (θ=0∘). The results are shown in Fig. 2 for a scanning speed of v s =0.5 mm/s and a focal depth of d s =150 μm after etching for 1 hour in 5% HF. The etch rate for the perpendicular polarization was ∼5.5 μm/min over a wide exposure range of 50 to 200 nJ. A slightly lower etch rate of ∼5.0 μm/min was observed for circular polarization for pulse energies above 75 nJ. In sharp contrast, the etch rate for the parallel polarization was 3 to 28 fold lower, varying from ∼0.2 μm/min at 50 nJ pulse energy to ∼1.5 μm/min at 200 nJ pulse energy. The considerable contrast of ∼28:1 in differential etch rate between perpendicular and parallel polarizations at E p =50 nJ follows the expected outcome where the orientation of the nanograting perpendicular to the scanning direction for parallel polarization has stop-layers hindering the progress of the etchant along the modification track while nanograting planes running along the laser track for perpendicular polarization are effectively open to direct the etchant rapidly along the laser modified track. Higher differential etch rates of 70:1 were reported by Hnatovsky et al. [5, 6] using very different exposure conditions of laser wavelength (800 nm vs. 522 nm), pulse duration (40 to 500 fs vs. 220 fs), and scanning speed (0.03 to 0.11 mm/s vs. 0.2 to 10 mm/s). Nevertheless, the overall etch rate for the perpendicular polarization case here is similar to the ∼5 μm/min etch rate reported by Hnatovsky et al. [5, 6]. The relatively constant etch rate of ∼5.5 μm/min for perpendicular polarization throughout a broad processing window of E p =50 to 200 nJ indicates formation of high contrast nanograting planes aligned along the scanning direction. However, the channel cross section increased from 5 μm by 12 μm to 7 μm by 22 μm over this pulse energy range. For all polarizations, chemical etching was not observed for laser pulse energy of 25 nJ.

Fig. 2
figure 2

Etch rate (5% HF for 1 hour) of single laser modification tracks (v s =0.5 mm/s, d s =150 μm) with three different polarizations (perpendicular, parallel and circular) versus pulse energy

Full size image

The polarization and pulse energy dependence on etch rates for other scanning speeds were similar to the trend shown in Fig. 2, but with etch rate decreasing as the scanning speed increased. At a 100 nJ pulse energy, the etch rate remained constant at scanning speeds from 0.2 to 1.0 mm/s for perpendicular and circular polarizations, but decreased ∼50% for parallel and circular polarizations, and decreased ∼10% for perpendicular polarization as scanning speed increased from 1.0 to 10 mm/s. The reduced etch rate can be attributed to an expected lower volume of material modification with increased scanning speed (decreasing dwell time).

The scanning depth was also found to affect the HF etch rate. Using a constant pulse energy E p =100 nJ, an increase in the focusing depth from d s =150 to 210 μm produced different etching trends, decreasing the etch rate by ∼3% for perpendicular polarization, and increasing by ∼50% for parallel polarization and increasing by ∼5% for circular polarization. The increasing spherical aberration produced by the glass surface at increased focusing depth [23] presumably weakens the nanograting formation by spreading out the focused energy resulting in diminished nanogratings. Therefore, the less well defined parallel nanogratings cause a nearly insignificant drop in etch rate for the perpendicular polarization but a large increase for the case of parallel polarization as the stop-layers become less effective.

Figure 3 shows optical images of laser modification tracks written with 100 nJ pulse energy, 210 μm focal depth, and circular polarization obtained over a range of scanning angles =−34∘ to +34∘ to the facet normal and following one hour of etching in 5% HF solution. For all observed angles from θ=−45∘ to +45∘, we inferred a ∼4.7 μm/min etch rate from an average hole depth of ∼280 μm that was independent of the scan angle (θ). This etch rate invariance follows from the anticipated random orientation of grating nanostructures for circularly polarized light and is attractive for directionally independent formation of uniform micro-channel shapes in any substrate direction (x–y plane).

Fig. 3
figure 3

Laser scanned fan structure written at v s =0.5 mm/s with circular polarization at E p =100 nJ/pulse and d s =210 μm, followed by 5% HF etching for 1 hour. The darker lines show the laser modification tracks that were opened by the HF acid and the lighter tracks were unetched

Full size image

A similar analysis for linear polarization yielded etch rates that were strongly dependent on the angle, ϕ, measured between the laser polarization and the channel scan direction (θ), yielding the plots shown in Fig. 4(a) for various focal depths and Fig. 4(b) for various laser pulse energies. The full angle range of ϕ=0∘ to 180∘ was obtained by stitching together data from laser exposures scanned from ϕ=45∘ to 135∘ with polarization perpendicular to the y-axis (Fig. 1) and laser exposures scanned from ϕ=0∘ to 45∘ and ϕ=135∘ to 180∘ with polarization fixed parallel to the y-axis. A peak etch rate of ∼5.5 μm/min was observed for all pulse energies E p =50 to 200 nJ and focal depths d s =75 to 210 μm shown here for perpendicular polarization (E ⊥, ϕ=90∘), but seen to decrease strongly by up to 5 fold for E p =75 nJ and d s =75 μm over only a narrow 10∘ change in polarization angle (ϕ=90∘ to 80∘ or 90∘ to 100∘). In order to take advantage of an etch rate within 96% of the maximum value, the laser polarization must be accurately oriented within a narrow angle range of ϕ=89∘ to 91∘ to the scanning direction that may be challenging when forming looped or curved paths for the micro-channels. The strong etching rate contrast of 28× between perpendicular (E ⊥, ϕ=90∘) and parallel (E ∥, ϕ=0∘ and 180∘) polarizations is also clearly evident in Figs. 4(a) and 4(b) and consistent with the values reported above.

Fig. 4
figure 4

Average etch rate (5% HF after 1 hour) of single-scan modification tracks written at v s =0.5 mm/s versus angle between laser polarization (linear) and scanning direction for exposures of (a) E p =75 nJ and depths of d s =75 μm, 150 μm and 210 μm, and (b) d s =150 μm and pulse energies of E p =75 nJ, 100 nJ, 125 nJ and 150 nJ. The black dashed lines are secant fits according to (1)

Full size image

The etch rate discontinuity at ϕ=45∘ and 135∘ in both Figs. 4(a) and 4(b) arise from the orthogonal orientation of nanogratings formed either parallel (y-axis polarization) or perpendicular (x-axis polarization) to the facet when scanning at either θ=±45∘ direction to the surface. For both parallel and perpendicular polarization at this scan angle, nanogratings are aligned at 45∘ to the modification track that should yield identical etch rates in the bulk. However, there is an anomaly where the initial etching rate is inhibited at the facet for parallel polarization which positions nanograting stop-layers parallel with the facet, while nanogratings planes oriented perpendicular to the facet would enhance the initial etching for perpendicular polarization.

A geometric consideration of the orientation of nano-planes with the track direction (subtending angle, 90∘−ϕ) suggest that the slow etch rates found for parallel polarization, R ∥=0.2 μm/min (E p =75 nJ, d s =75 μm) at ϕ=0∘ or 180∘, will scale up trigonometrically as follows:

$$ R (90^{\circ} - \phi) = R_{\|}\mathrm{csc}(90^{\circ} -\phi) = R_{\|}\mathrm{sec}(\phi).$$
(1)

Here, dramatically faster (28×) etching along the nano-planes enhances the slow R ∥ etch rate as the apparent stop planes rotate with sec(ϕ) to align with the channel direction. The secant lines in Fig. 4 closely match the experimental etch rates for exposure conditions of E p =75 nJ with d s =75 μm and 150 μm. The secant fit is capped at the peak etch rate of R ⊥=5.5 μm for perpendicular polarization in the angle range ϕ=87.8∘ to 92.2∘. As the etching depth (Fig. 4(a)) and pulse energy (Fig. 4(b)) increase, the observed etch rates deviated from the secant fit where the uniformity and quality of nanograting stop-layers were apparently degraded by higher optical aberration expected when focusing deeper into the glass [23, 24] as well as by higher pulse energy. Hence the strongest contrast in parallel and perpendicular polarization etch rates were found in channels formed at shallow depth of 75 μm and low pulse energy of 50 nJ to provide the most uniform and highest contrast nanograting structures for etching straight channels. However, the angle-invariant etch rate found for circular polarization (Fig. 3) is potentially more attractive for fabricating circular or arc-shaped micro-channels whereas linear polarization requires the rotation of polarization to follow the changing scan direction.

In summary for single track laser writing, a wide laser processing window was found for enabling fast HF acid etching of 5.5 μm/min for perpendicular polarization and 5.0 μm/min for circular polarization over a wide range of focusing depth, from 75 to 210 μm, and scanning speeds ranging from 0.2 to 1.0 mm/s with pulse energy in respective ranges of E p =50 to 200 nJ and 75 to 200 nJ. The combination of the results from Figs. 2 and 4 indicate a new fabrication window of low laser pulse energy for efficient nanograting generation to provide high polarization contrast etching of micro-channels. Further, the high etching contrast is attractive for integrating optical waveguides with such channels, however, rectangular cross-sectional shaped channels with smooth sidewalls and without cracks would be highly desirable.

3.3 Flexible cross-sectional shaping

In this section, we present a method of creating arrays of multiple laser exposure lines that are optimized in their vertical, d L , and transverse, d T , track-to-track spacing to control the cross-sectional profile of micro-fluidic channels with the objective of forming crack-free and open channels with smooth morphology.

Figure 5(a) provides a reference showing a backlighting microscope image of an end view of a single laser track written with a pulse energy of 50 nJ observed before (top) and after HF etching (bottom). The etched single track micro-channel appears dark under backlighting after 1 hour in 5% dilute HF and increased in cross-sectional area compared with the original laser modification size from 2 μm by 6 μm to 5 μm by 11 μm. The elliptically shaped cross section approximately follows the laser beam profile of ∼1-μm focal spot diameter and ∼6-μm depth-of-focus expected with a 40× aspherical lens.

Fig. 5
figure 5

Cross-section microscope images of laser exposed and etched (5% HF after 1 hour) (a) single track and (b)–(f) rectangular shaped micro-channels laser written with perpendicular polarization at depth d s =150 μm, scanning speed v s =0.5 mm/s and pulse energy (a)–(e) E p =50 nJ and (f) E p =125 nJ. The line-to-line spacing in rectangular channels were (b), (f) d T =2 μm and d L =3 μm, (c) d T =3 μm and d L =3 μm, (d) d T =3 μm and d L =6 μm, and (e) d T =3 μm and d L =12 μm. The laser was incident from the bottom

Full size image

After sufficient etching time, arrays of single-line modification tracks would open up to each other to form into a large single channel over long lengths as shown in the images Figs. 5(b) to 5(f). These representative samples of multi-track arrays examine a range of horizontal spacing of d T =2 to 3 μm, and vertical spacing of d L =3 to 12 μm. With a large spacing of d T =3 μm and d L =12 μm as shown in Fig. 5(e), the top and bottom walls are strongly corrugated as channels in horizontal rows have opened, while etching was not completed between these layers, creating a stack of thin and parallel micro-channels of interest for parallel channel cytometry. A single open channel with improved side wall smoothness is found as the vertical offset was decreased to 4.5 μm and the horizontal spacing was decreased to 2 μm as shown in the trend from Figs. 5(e) to 5(c). A trade off balancing the formation of relatively smooth wall surfaces against longer laser scan time was found over a large range of line-to-line separations of d T =1 to 2 μm and d L =1.5 to 6 μm.

One limitation with the multi-scan cross-section shaping of micro-channels is the higher stress accumulated at the corners of the arrays, as seen from birefringence observations under an optical microscope. Under sufficient laser exposure, such stresses induce micro-cracks as shown in Fig. 5(f) that results in the finger-like open structure seen after HF etching. For perpendicular polarization, crack formation was not evident in the present array geometries written at depth d s =150 μm, energy of E p <125 nJ, and scanning speeds ranging from 0.2 to 10 mm/s. Micro-cracks only become apparent as the depth increased to d s =210 μm at speeds of v s <1 mm/s. On the other hand, for parallel and circular polarizations, cracks were observed at depths of d s =150 μm and 210 μm with E p >100 nJ and scanning speeds of v s <1 mm/s. Parallel and circular polarizations are more susceptible to initiate cracks as the nanogratings forming perpendicular and randomly along the scanning direction, respectively, build up a higher stress along the channel compared with the case where the nanogratings align along the scanning direction. Laser exposure in this range is known to generate a positive refractive index change that is associated with densification of the glass substrate [25] and manifests in the stress fields seen as the bright regions in the surrounding glass as observed in Fig. 5. The cracks were mostly concentrated along sharp low-radius corners of rectangular structures where stress fields flare out diagonally as observed in Fig. 5(b) to (f). Spontaneous cracking after exposure and HF etching is undesirable and limits the total exposure acceptable, favoring exposure at pulse energies of E p <125 nJ for perpendicular polarization, and E p <75 nJ for parallel and circular polarizations at moderate scanning speeds of 0.5 to 1 mm/s. In order to fabricate high contrast and crack-free micro-channels of flexible cross-sectional size and in a rapid etching time, laser polarization perpendicular to the scanning direction is preferred to produce nanogratings aligned parallel with the channel with exposure conditions in the following ranges: scanning speed of 0.5 mm/s<v s <1.0 mm/s, pulse energy of 50 nJ<E p <125 nJ, and focusing depth of 75 μm <d s <150 μm.

Figure 6 demonstrates the advantage of the multi-scan method in flexibly shaping the cross section of micro-channels. The microscope images show various rectangular and nearly circular shaped micro-channels laser written with the laser parameters of E p =75 nJ, v s =0.5 mm/s and d s =150 μm, followed by chemical etching of 1 hour in 5% HF. The rectangular shaped micro-channels in Figs. 6(a) and 6(b) were laser written with a perpendicular polarization, and the circular/elliptical shaped micro-channels in Figs. 6(c) and 6(d) were written with a circular polarization.

Fig. 6
figure 6

Cross-section microscope images of laser exposed and etched (5% HF after 1 hour) (a), (b) rectangular and (c), (d) circular/elliptical shaped micro-channels laser written with E p =75 nJ, v s =0.5 mm/s and d s =150 μm. Rectangular channels and circular/elliptical channels were written with perpendicular laser polarization and circular laser polarization, respectively. The line-to-line spacing was (a) d T =1 μm and d L =3 μm, (b) d T =2 μm and d L =3 μm, and (c), (d) d T =2 μm and d L =1.5 μm. The laser was incident from the top

Full size image

The above optical microscope assessment provided a broad laser exposure window (0.5 mm/s <v s <1.0 mm/s, 50 nJ <E p <125 nJ, d s <150 μm, 1 μm <d T <2 μm and 1.5 μm <d L <6 μm) for creating crack-free and smooth surfaces in the rectangular micro-channels. Further optimization based on SEM analysis and atomic force microscopy (AFM) yielded the best surface smoothness as shown for a micro-channel fabricated with perpendicular polarization in Fig. 7(a) for the channel bottom and Fig. 7(b) for the channel sidewall. In contrast, micro-channels fabricated with parallel or circular polarizations presented much rougher surfaces as the nanograting planes were not aligned with any of the channel walls as in the case of perpendicular polarization. The smoothest walls were found for perpendicular polarization and laser exposure of pulse energy E p =50 nJ and scanning speed v s =0.5 mm/s, yielding a bottom surface roughness of ∼200 nm (rms based on 50 μm × 50 μm area) in Fig. 7(c), and a sidewall surface roughness of ∼10 nm (rms based on 3 μm × 3 μm area) in Fig. 7(d). The limited viewing area (3 μm × 3 μm) available for the sidewall may suggest a larger overall wall roughness, however, the AFM image encompasses two laser tracks that were very precisely positioned (2 nm resolution) by the motion stages. The 20× smoother sidewalls are attributed to the parallel alignment of nanogratings which otherwise form orthogonally to the bottom/top channel wall and lead to corrugation of ∼200 nm period that matches the 180 nm period that we reported above for the bulk nanograting. This highly smooth sidewall with near optical quality will permit the integration of optical waveguides crossing the micro-channel with relatively low scattering loss from each sidewall facet.

Fig. 7
figure 7

SEM images of the (a) bottom and (b) sidewall of an etched micro-channel laser written with perpendicular polarization, E p =50 nJ, v s =0.5 mm/s, d s =75 μm, d T =1 μm, and d L =1.5 μm and the corresponding AFM images (c) and (d), respectively

Full size image

The SEM and AFM assessment for smoothest channel walls favored the lowest laser pulse energy (E p =50 nJ) at the threshold of HF etching where one expects the lowest stress accumulation together with the fewest defects in generating parallel nanogratings. Further, this low exposure level may facilitate a smooth extension of nanogratings into new exposure zones as the laser scans forward as reported by Taylor et al. [26]. A slightly higher exposure of 75 nJ <E p <100 nJ was used below to ensure the rectangular channels could be completely etched at this above threshold exposure with only slightly less smooth sidewalls and without cracking over long channel lengths.

In brief, this rapid prototyping technique in cross-section shaping allows flexible and arbitrary patterning by merely changing the laser scanning pathway to form the desired cross-sectional shape while also reducing tapering along the micro-channel length.

3.4 Micro-channel length

The present multi-scan arrays assist in holding the shape of the resulting open channel over long lengths unlike the distortions and tapering effects reported previously for single track lines [1–6, 10]. However, the HF etching was limited as in prior reports [4, 6, 10] to producing channels of mm’s in length. Figure 8(a) shows a schematic of a buried rectangular micro-channel where vertical holes have been introduced to open to the channel periodically to the glass surface to facilitate HF penetration over unlimited channel length. Here, θ T indicates a tapering angle anticipated for a buried channel. Figures 8(b) to 8(e) are the top view microscope images of a section of rectangular micro-channels of 8 μm × 26 μm cross-sectional area formed over 30 mm length and have ∼7 μm diameter vertical holes positioned periodically with 100 μm (Figs. 8(b) and 8(c)) and 200 μm (Figs. 8(d) and 8(e)) separation. The channel was written at a depth of d s =75 μm using a pulse energy of E p =100 nJ and etched for 3 hours in 5% dilute HF.

Fig. 8
figure 8

A schematic of a micro-channel (a) with vertical access holes, and the top view microscope images of completely etched (5% HF after 3 hours) multi-scanned tracks laser written with perpendicular polarization at E p =100 nJ and d s =75 μm with vertical holes laser written at E p =100 nJ and separations of (b), (c), (f) 100 μm and (d), (e), (g) 200 μm. Imaging was aligned with the buried channel in (b) and (d) and the top glass surface in (c) and (e)

Full size image

For the 200 μm separation of access holes, a slight tapering angle of θ T =∼2∘ is noted in Fig. 8(d), no taper or width variation is discernible by the optical microscope within the <1 μm resolution for the 100 μm hole separation case shown in Fig. 8(b). Such access holes are clearly advantageous for unlimited extension of the micro-channel length while only covering an insignificant 0.6% of the wall surface area. Such holes can easily be sealed on the glass surface, but further investigation is required to ensure no disturbance of the flow dynamics would limit potential micro-fluidic applications. In many applications, opened holes would effectively become self-sealing due to strong surface tension of liquid reaching the hole opening, but the additional capillary forces from numerous access holes may strongly modify the flow dynamics in applications such as electrophoresis.

3.5 Optofluidic integration

Polarization contrast is demonstrated here to be highly beneficial in optofluidic integration. A single laser scanning step was applied to form a buried rectangular micro-channel with perpendicular polarization and define nearly optically smooth sidewall surfaces through which an optical waveguide was made to intercept. The waveguide was laser written at a depth d s =75 μm from the surface, scanning speed v s =0.2 mm/s, and E p =50 nJ pulse energy. Parallel polarization was used to form perpendicular nanogratings that aligned parallel with the channel wall and thus serve as strong chemical stop-layers, preventing the etching of the waveguide precisely at the channel wall as shown in the optical microscope image of Fig. 9(a). The buried micro-channel of 10-μm × 20-μm rectangular cross section and 20 mm length was fabricated with vertical access holes spaced by 100 μm, d s =65 μm and E p =100 nJ. After laser exposure, the sample was immersed in 5% dilute HF for 4 hours.

Fig. 9
figure 9

A microscope image (a) of a buried optical waveguide written with E p =50 nJ to cross an embedded 10-μm × 20-μm rectangular micro-channel formed in bulk fused silica glass after 4 hours of etching in 5% HF. The mode profile (b) of 633 nm light in the waveguide

Full size image

The optical waveguide was single mode at 633-nm wavelength yielding a mode field diameter (MFD) of ∼7.0 μm as seen in Fig. 9(b) and a propagation loss as low as 1.1 dB/cm. An insertion loss as small as 2 dB was measured through a water filled channel. Considering an expected Fresnel loss of 0.02 dB and a mode mismatch loss of 1.84 dB, one infers a scattering loss of approximately 0.07 dB from each of the side wall surfaces attesting to their exceptional wall smoothness. The polarization-dependent etching offers a convenient means of laser patterning to integrate micro-fluidic and optical components in a single exposure step without chemical etching damage into the optical components. This approach is advantageous compared with post-writing of waveguides [11] into pre-existing micro-fluidic channels that introduce extra process steps and optical aberrations in the laser focusing. The present method also eliminates the need for a stop-layer between the micro-channel and the waveguide [12].

3.6 Future implications and direction

The present results provide an efficient and rapid prototyping technology for fabricating 3D optofluidic devices. Polarization controlled orientation of nanogratings was instrumental for enabling single-step laser patterning of all buried micro-fluidic channels and optical circuits prior to the HF etching step, harnessing nanograting stop-layers to prevent degradation of optical devices at the micro-fluidic interfaces. Very low laser exposure conditions were identified to generate very smooth channel walls that dramatically reduce scattering loss for crossing optical waveguides. Such optically smooth channel sidewalls are attractive for further applications requiring internal reflection to probe analytes or to integrate micro-mirrors as demonstrated by Sugioka and coworkers [13] in Foturan glass. Polarization and scanning-angle control to create such flat and smooth surfaces would also improve the precision of integrated micro-mechanical and optical devices for optical sensing applications as reported by Bellouard et al. [14]. Finally, the formation of periodic vertical access holes for HF etching significantly opens the domain for writing buried micro-fluidic channels over unlimited length and in any desired 3D routing direction in which multi-level network of micro-channels, optical circuits, and micro-mechanics can be further densified and integrated.

4 Conclusion

To summarize, a broad range of femtosecond laser exposure conditions were explored experimentally, and found to enable pulse-energy independent and high polarization contrast HF acid etching with perpendicular polarization whereas circular polarization offered scanning-angle-invariant etching. These polarization approaches provide the means to fabricate straight channels and curved channels, respectively. Low energy laser patterning of buried opened micro-channels of flexible cross-sectional shape with extended length facilitated the fabrication of crack-free rectangular micro-channels with smooth side walls, allowing low loss integration of polarization controlled optical waveguides. This extensive study on channel shape optimization together with the integration techniques promise an efficient prototyping technology for fabrication of complex 3D optofluidic microsystems.