1. Introduction

Micro total analysis systems (µTAS), also known as “lab on a chip” (LOC) devices, have recently taken off, thanks to the ever-increasing technological improvements. From the very beginning, the main goal of this class of sensing/analysis devices has been the shrinking of common bench-top laboratory techniques down to the microscale, to increase sensitivity, automation, and portability, aiming at achieving a multifunctional point-of-care (POC) assembled unit [1,2]. Among all the physical principles implemented on LOC platforms, optical analysis has gained a prominent role, giving rise to a new class of techniques, which falls under the name of “optofluidics” [3,4]. Indeed, the integration of optical technologies with microfluidic platforms pledge to provide accurate, compact, and flexible devices [5] able to explore new measurement skills. Optical methods present some key advantages over the other approaches – i.e., they are extremely sensitive, non-destructive, and remarkably robust [6] – e.g., being immune to external electromagnetic interferences. For these reasons, they have been implemented in a widerange of applications, ranging from the environmental monitoring [7] to the bio-diagnosis [6]. The latter is gaining importance, as it allows genetic, protein and single-cell analysis, crucial for the identification of illnesses such as malaria [8], anemia [9], and cancer [10,11]. In this last context, being able to identify the low concentration of circulating tumoral cells – i.e. 1-100 CTCs over 10⁶-10⁹ other cells present in 1 ml of blood [12] – could certainly be the next frontier of early stage cancer detection (up to 4 years before other techniques [13]). The combination of microfluidic and optical technologies has the right qualifications to tackle the challenge to accurately detect them. Spectroscopic techniques, such as Raman, have gained much interest in this topic thanks to the high accuracy and the completely label-free working principle. However, this powerful analysis tool is well-known to typically display extremely weak intensity that forces to come to a trade-off between bulky external optical setups – e.g., optical microscopes [14] and high acquisition times [15], incompatible with the operating clinical use. Using external optical instruments cancels the advantages of using LOC devices and, on top of that, introduces significant disadvantages, such as optical losses and the need for a precise alignment. Moreover, since the solid angle of acquisition is very limited – i.e., from 7.5%-13% of the emitted solid angle (4π steradians) to a maximum of 28% in the case of a water-immersed objective [14,16] –, the acquisition times must be in the order of seconds, critically limiting the device throughput [17]. On the other hand, attempts have been made to reduce the footprint of the instrumentation by integrating both the excitation and the collection stages on-chip, by means of two optical fibers [18] or more [15]. Nevertheless, acquisition times are in the order of tens of seconds, making these systems impractical if applied to real-case clinical scenarios. Therefore, a single cell spectroscopic analysis platform that can fulfill the strict requirements for clinical use, to the best of the authors’ knowledge, does not appear to be available yet. One way of improving these systems and, at the same time, adhering at the general requirements of LOC devices – e.g., compactness, portability, and high throughputs – is to act directly on the signal collection. The best strategy is to aim at integrating the optical circuits into the same substrate of the microfluidic network. In this way, the optical path is minimized, and therefore also the losses. Then, there is a clear need for increasing the solid angle of collection and the integration of light harvesting optical structures can effectively address this lack. Several reflectors have been integrated directly onboard microfluidic platforms to catch part of the optical signal escaping the collection angle of the microscope objective. These elements feature different shapes, including freeform [18–20], truncated pyramid [21], and elliptical reflectors [22]. Despite the good functioning, these elements all show the same limiting aspect. Due to their fabrication processes, they are manufactured out-of-plane – i.e., the main axis of the optical element is parallel to the normal to the surface of the substrate where the microfluidic network is inserted. This prevents a real integration with photonics platforms, in which the wave propagation is typically on the (in-) plane, as it usually is in excitation and acquisition systems through waveguides and/or fiber optics. To exploit these out-of-plane systems requires the use of additional bench-top instrumentations. In addition, it is still needful to couple the microfluidic device to external optical elements (source, detector, and so on) to perform any measurement, introducing in this way additional losses, complicating the device setting, hampering its portability, and requiring a precise alignment. Although theoretically some of these structures could achieve a nominal collection efficiency of 50% (i.e. covering a maximum extension of 180° around the region of interest), due to coupling losses, this value practically reduces to about 24%, as numerically computed by Liu et al [20]. Instead, in-plane optics – i.e., the optical axis is orthogonal to the normal to the substrate surface – would allow to overcome all these limitations. Currently, such integrated elements are limited to a 2D manipulation of the light, being able to reflect, collimate, or focalize a beam just on a plane [22–25]. This may not be sufficient to implement single-cell analysis because a significant portion of the flowing medium can be included in both the excitation and acquisition, thus reducing the Signal-to-Noise Ratio (SNR). Vice versa, dealing with integrated in-plane optical structures that can focus and collect light selectively in 3D from a single spot – rather than a plane – could push beyond the detection limit. This boundary could also be further extended by the ability to overcome the 180° collection angle limitation, exploring new frontiers of on-chip analysis. For this purpose, we have employed the high 3D versatility of femtosecond laser irradiation, followed by chemical etching (FLICE) technique combined with the flexibility of the CO₂ laser polishing and inkjet printing technology to fabricate high quality optical structures, monolithically integrated on a fused silica sample. Besides its three-dimensional freedom, the same fs-manufacturing process also enables the integration of microfluidic channels into the same platform. Moreover, we have improved the surface roughness of our optical elements by coupling FLICE technique to a CO₂ laser polishing step [26–28]. Finally, to make them reflective, a silver nanoparticle coating has been deposited through inkjet printing, as already shown in [29]. In this framework, the fabrication and the characterization of such optical elements and the proof of functioning of two mirrors and an optical fiber integrated on the same platform has been presented.

2 Integrated micro-optics

2.1 Numerical simulations and optics design

It is straightforward that the solid angle of acquisition plays a critical role in any optical analysis technique, especially in spectroscopic applications (e.g., the Raman signal is emitted isotropically over 4π steradians). The solid angle of collection (Ω) depends on the area (A) of the collection optic and the distance r from the source of the scattered radiation, according to the differential relationship [30]:

The closer to the source the optical element is, the greater the solid angle, for a fixed area. Therefore, reducing the distance between the element and the light source allows to harvest great solid angles without the need for too large optical surfaces. Although the underlying physical process is different, similar deductions about the solid angle and the optical path length can be derived also in the case of fluorescence, as the collected intensity I_cf can be estimated as [17]:

The currently most used technologies for acquiring such optical signals on microfluidic platforms are still based on bulky external instrumentation – e.g., microscopes and optical setups. In these cases, high NA objectives are used, but their contribute is still not adequate. The relationship between the objective NA and the collected solid angle Ω_ob is [31]:

2.2. Micro-reflectors fabrication process

The manufacturing process is based on three main steps: femtosecond irradiation, followed by chemical etching (FLICE), CO₂ laser polishing, and silver coating inkjet printing.

 

Fig. 1. (a) Up left: geometrical properties of the parabola; the light coming from its focal point is collimated back along its axis and vice versa. Up right: the parabola working in off-axis configuration. Down left: ellipse captures the light coming from a source placed in one of its focal points and focuses at the second focal point. Down right: ellipse in off-axis configuration. (b) 3D geometry used in the numerical simulations to maximize the signal acquisition. On the side of the first parabolic mirror a window is left to let the excitation beam pass through. (c) Numerical ray tracing of the acquisition branch optical circuit. The light source is modelled as a sphere of 15 µm at the center of the microfluidic channel, to resemble a flowing cell. The light is collected by the first parabolic mirror and focused by the second mirror towards an integrated optical fiber. By computing the number of rays collected at the bottom of the fiber, 56% of collection efficiency is achieved. (d) Ray tracing if the excitation branch: the light is shined by an integrated SM optical fiber and focused in one reflection-step by an elliptical mirror towards the microfluidic channel.

Download Full Size | PDF

2.2.1. FLICE

The FLICE technique allows to freely shape in three dimensions both microfluidic and optical structures at the same time, integrating them into a single fused silica substrate. This is due to the two main steps in which this process is divided: a 3D modification of the material, by means of an ultrashort-pulses irradiation, and its following removal through a wet chemical etching. The physical principle laying behind this process is multiphoton absorption which allows to modify selectively only the material in the focal spot (Fig. 2(a)). This modification consists in the creation of nanostructures, also called nano-gratings, whose alignment depends on the laser beam polarization [32]. The directionality of these structures is a key parameter because it promotes the chemical etching. To facilitate the following fabrication steps, i.e., the optical polishing and the reflective coating, our strategy for irradiating the material is to pile up modified lines, spaced by 4 µm of unmodified fused silica. The optimization of the writing process, involving the modification of the substrate to markedly reduce the surface roughness, is achieved by using the side lobe of the laser beam, rather than its bottom part. The irradiation time for a single optical element is very fast (about 30 minutes). The etching step has been performed in a water-based solution with a concentration of 20% of hydrofluoric acid (HF), helped by an ultrasonic bath (Fig. 2(b)). As a result of the writing strategy, the time taken was about 45 minutes.

 

Fig. 2. (a) First step of FLICE: femtosecond laser irradiation. The multiphoton absorption phenomenon causes the fused silica to be modified just within the laser beam spot volume. Thus, 3D patterning of microstructures is allowed by this direct writing technique. (b) Second step of FLICE: chemical etching. A water-based solution of hydrofluoric acid (20% concentration) selectively removes the modified material with a greater etching rate, thus complex 3D hollow structures can be integrated in fused silica substrates. (c) CO2 laser polishing. An infrared laser beam (λ=10.3 µm) is shined on the optical element surface. The energy absorbed from the beam is converted in heat and a thin layer of almost-melted material creates. During the cooling phase, the peculiar rearrangement of the material surface, driven by the surface tensions of the glass, results in a residual roughness – left over from the etching step – significantly reduced. (d) Inkjet printing. To make the manufactured micro-optics reflective, a silver nanoparticle coating is delivered on the elements surface by means of a nozzle, precisely controlled by the printer. (e) The silver coating is fixed permanently on the elements surface by a sintering step on hotplate at 150° for 5-10 minutes.

Download Full Size | PDF

2.2.2. CO₂ Laser polishing

It is already known that, as a drawback, FLICE technique involves relevant surface roughness, mainly due to the chemical etching phase [31,33]. When aiming at integrating in-plane micro-optics, the residual roughness can be a severe limiting factor, as it can completely spoil the optical properties of the embedded element. To overcome this main issue, we pioneered contactless local thermal annealing by means of a CO₂ laser source (Synrad, Series 48 M, wavelength 10.3 µm) (Fig. 2(c)). This radiation can be absorbed up to the 80% by silicate glasses with a penetration depth of about 20 µm at room temperature. Once absorbed, the laser energy is converted into heat in a thin surface layer that easily reaches the melting point without exceeding the sublimation, resulting in viscosity decrease. Thus, the synergy between the surface melting and the surface tension of the glass material smooths the surface by reducing the roughness [34]. In our polishing setup the laser beam is focused close to the sample by a mounted gold coated parabolic mirror (Thorlabs, MPD169-M01). The remaining setup consists of a high-speed magnetic translation stage (PI V-738), to move the sample in 2D with sub-micrometric accuracy and to scan large areas (100 mm x100 mm) in short times (speeds and accelerations up to 500 mm/s and 1 g), a z-axis manual translation stage to set the correct position of the sample relative to the focal point of the laser and a pyrometer (Pyrospot SU 10GV (t 500-2500°C)) to measure the temperature locally on the fused silica substrate. The latter information is important as it can be used as feedback to automatically control the energy deposited on the sample. According to Weingarten and Schwarz, [26,28] the scanning has been performed in a defocused configuration, at a working distance of 2.5 cm below the actual focal point. The laser pattern is a square and the scanning track is a serpentine. Each polishing session consists of several superimposed patterns. The first is always set to deposit higher amount of energy (fixed power and lower scanning speeds), to heat up the material close to the melting point. This ensures that the greatest peak-valley distances are smoothed out. Immediately after, other three scanning patterns are repeated at higher speeds – each one with a different starting point and a different path orientation. In this way, also smaller surface defects are corrected. The four passages just described are the basic scan set – with a duration of about 1.5 minutes –, which may be repeated from 1 to 30 times, depending on the surface to polish. When treating the micro-optics, special care must be taken to avoid significant and irreversible alterations to the 3D curved profile. In these cases, the delivered laser power has been set at low values, around 17%-18% of the maximum power (10W), and the number of scan sets ranged from 20 to 30. To evaluate the polishing procedure results we have carried out two characterization measurements. First, we have verified through a mechanical profilometer that the design curve of the optics, critical for their performances, has been maintained. Then, to quantify the effect of the thermal treatment on the nanoscale, Atomic force microscopy (AFM) measures have been carried out on 100 µm × 100 µm areas, centered in the middle of the reflector.

2.2.3. Inkjet printing

To make our structures completely reflective in a broad range of wavelengths, we have chosen a metal-based ink as coating (silver nanoparticles, ANP SilverJet DGP 40LT-15C), thus covering both the visible and the near-infrared spectra. Inkjet printing allows to print fluid inks with a very high precision in the volume drop casting (Fig. 2(d)). The solution has been ejected by a single nozzle by means of a DMATIX printer (Fujifilm, DIMATIX DMP-2831) on a plate at 60°, depositing the coating in several passes (up to 10) to attain a uniform coverage. Since our element surfaces are very small, few microliters of silver nanoparticles are enough to achieve a 100% reflective coating. To make the solvent evaporate, the device has been put on a hotplate at 150° for 5-10 minutes, obtaining a permanent coating on the element surface (Fig. 2(e)).

3. Optical structures characterization

3.1. Micro-optic surface roughness improvement

The FLICE technique enables the integration of complex 3D structures into fused silica substrates and has already proven to be a manufacturing process suitable for the incorporation of micro-optics on board LOC platforms [29]. However, photonic devices must have an optical quality surface (Root mean square (RMS) < λ/10) to preserve their optical performances. Although residual surface roughness, introduced by the chemical etching process [31,35–38], acts as a perturbation factor for micro-optical elements, maintaining the design curve also plays a key role, especially in the case of 3D optics. To meet both requirements, we used a CO2 laser polishing step in the reflective structure integration procedure (Fig. 2(c)). Compared with typical annealing processes, which require precise temperature control for long times (> 100 h) [39] and a geometry-tailored procedure, our proposed strategy ensures a marked improvement in roughness (Fig. 3(a)), in a short time (30-45 min) while preserving the 3D curvature (Fig. 3(b)). Profilometer measurements have confirmed that the polishing process does not introduces bumps or holes in the designed curve. Moreover, they have shown that the differences in quote – i.e., on the z-axis – of few tens of microns, due to curvature, are well tolerated if a defocused laser beam is used. Nonetheless, roughness on the nanoscale could still compromise the element performance. For this reason, a complete characterization of the polishing efficacy has been provided (as described in the Supporting Information) not only by means of RMS roughness, but also in terms of the spatial frequencies content of the surface topography and its level of isotropy before and after polishing (Fig. S2 – SI).

 

Fig. 3. (a) Stereomicroscope image of a mirror before and after the optical polishing procedure. (b) Profilometer analysis of the micro-optics curvature before and after the polishing step. No bumps are introduced by the procedure and the design curve is well preserved. (c) Optical microscope image (5x magnification) of a parabolic mirror before the optical polishing. Inset. AFM measurement of the surface roughness, the RMS is 174.7 nm (figure S1.a, Supporting Information). (d) Optical microscope image (5x) of the same parabolic mirror after the polishing process. Inset. AFM measurement of the surface roughness after laser polishing. Due to the drastic reduction in roughness, the color bar has been scaled down by almost 3 orders of magnitude compared to the previous inset. Thanks to the polishing procedure, the AFM map results significantly more uniform than that of the unpolished case and the RMS has dropped to 3.4 nm (figure S1.b, Supporting Information), achieving ultra-high quality optical structures.

Download Full Size | PDF

As a remarkable result, a significant reduction in roughness was achieved from an RMS value of 174.7 nm (unpolished case) to 3.4 nm (polished case), far below the standard optical quality (λ/10). This result is comparable to other works using the same technique for unetched surfaces [28] and lower than the values obtained for etched samples but polished by oven annealing [39]. The values of the correlation lengths (Lx and Ly) before and after polishing clearly show a reduction in the level of anisotropy due to the increase in the bandwidth of their spectrum. Thus, it proves that FLICE technique can be a powerful tool for creating embedded micro-optics when coupled with CO2 laser polishing.

3.2. 3D parabolic micro-reflector

To prove the functioning of our micro-optics we have setup several experiments. First, we have fabricated a platform containing just one 3D parabolic mirror to characterize its focusing behavior. The platform and the experiment are sketched in the inset in Fig. 4(a). The layout is very compact, including just an in-plane 3D parabolic mirror in off-axis configuration and a scattering tank embedded in the fused silica platform. The laser polishing step has been applied to the micro-optic and to both the substrate wall facing the entering laser beam and the visualization tank wall facing the mirror. In this way, we have ensured that residual roughness of the etching process does not affect the visualization tests. The paraboloid main property is to reflect a collimated beam, incoming along its axis, and to focus it at its focal point, which in our device has been placed at the center of the tank. The device has been put on a tailored setup to align it with the collimated laser beam (λ=473 nm), coming from an external source. In Fig. 4 the clear focusing effect due to the parabolic reflector is reported and its behavior is compared to the un-focused beam image, obtained by shining the laser beam directly inside the visualization tank, without hitting the mirror. The two images have been taken by fixing the exposure time and at the same laser input power. In this way, the comparison between the intensity maps in the two cases is consistent (Fig. 4(b)). It is appreciable how the light intensity is generally spread along the whole unfocused beam, reaching at maximum a pixel intensity of 60. On the contrary, if reflected by the mirror, the same beam is well concentrated around the focal spot, doubling the pixel intensity values. A comparison of the two cross sections highlights that the collimated incoming laser beam, about 380 µm wide, is focused in only 50 µm into the focal spot by the micro-optics, making it fully compatible with a single-cell analysis application – e.g., single CTCs can range from 15-40 µm in diameter [40]. By observing Fig. 4(c) it is remarkable that the focusing behavior also occurs in the orthogonal plane, proving a complete 3D light manipulation. In this plane the focal spot size is around 40 µm, showing a good symmetry of the focused beam. Since the design presented in the numerical simulations requires the integration of two parabolic mirrors working synergically, we have fabricated a second platform including two parabolic reflectors and one optical fiber. As shown in the inset of Fig. 5(a), the diverging cone of the optical fiber is placed in the theoretical focal spot of the first reflector, so that the light, after being collimated by it, hits the second mirror, and is focalized inside the visualization chamber. Again, the light (λ=520 nm) is correctly focused inside the scattering tank and shows a spot size around 40 µm. The distance from the fiber tip to the second focal spot is 6.7 mm in air, which corresponds to an optical path of 4.59 mm in fused silica (refractive index = 1.458). Making a comparison, if we consider light traveling in glass in a straight line without hitting any mirror, the divergent beam from the SM fiber (NA = 0.14) would have had a diameter of 894 µm, ∼ 22 times larger than the case including the reflector. It is noteworthy that, because the coating is made by silver, the reflector can handle a broad spectrum of wavelengths, being a versatile optical element.

 

Fig. 4. (a) Characterization of a single parabolic mirror. A collimated beam (λ=473 nm) is shined towards the parabolic mirror, which reflects the incoming light inside a visualization chamber. The incoming beam is correctly focused at the focal point of the designed parabola. Next to the focused beam is reported the unfocused beam obtained by letting it pass through the scattering tank without hitting the mirror. The two images are captured by a stereomicroscope with the same exposure times, so that the intensity can be compared. Inset: Sketch of the experiment (b) Intensity map comparison between the unfocused and the focused beam. The focal spot concentrates the light, and the intensity is doubled with respect to the unfocused beam. The spot size in the xy-plane is 50 µm, computed as the full width at half maximum (FWHM) of the line profile cutting the focal spot. (c) Side view of the platform (zy- plane). The focusing behavior is present also in the zy-plane, where the spot size is around 40 µm.

Download Full Size | PDF

 

Fig. 5. (a) Sketch of double parabolic mirror experiment. The platform embeds an optical fiber and two 3D offset parabolic mirrors. The light coming from the fiber is collimated by the first reflector and then focused inside the scattering tank by the second reflector. (b) Stereomicroscope image of the experiment. The optical circuit works as designed and correctly focuses the beam (λ=520 nm) coming from the integrated optical fiber. Inset: Zoom on the focalized beam and relative intensity map. Again, the focusing behavior can be appreciated as the light intensity is higher around the focal spot. (c) Line profile of the pixel intensity map shown in the inset. The spot size is around 40 µm.

Download Full Size | PDF

3.3. 3D Elliptical micro-reflector

The design proposed in the numerical simulation also includes an elliptical mirror for the excitation branch of a totally integrated optical circuit. The reason underlying this is to compact and simplify the geometry as much as possible, by reducing the number of optical elements. To this aim we have exploited the unique geometrical properties of the ellipsoid. As the inset in Fig. 6(a) shows, diverging light from the optical fiber is shined from the first focal point of the ellipsoid. By rotating the fiber to point against a section of the ellipsoid, it is possible to focalize the incoming beam in the second focal spot using a single reflector. Even in this case, it is possible to see how the mirror can manipulate the light beam and concentrate it in a focal area (Fig. 6(b)). As for the previous micro-optic element, the focalization occurs in both planes, achieving a complete 3D focusing. The focal spot is smaller in the xy-plane (∼ 30 µm) but more elongated. This depends on the sensitivity of the elliptical geometry to finite sources.

 

Fig. 6. (a) Sketch of 3D elliptical mirror experiment. The laser light (λ=520 nm) is shined against an elliptical mirror, that focuses it, in a single reflection-step, inside the visualization chamber. (b) Stereomicroscope image of the experiment on the xy-plane. The mirror correctly focuses the incoming beam inside the scattering tank. On the side, the line profiles at different cuts across the beam are reported. The maximum intensity is reached in the focal spot area, which shows a size of 30 µm. (c) Side view of the platform. The beam is focalized also in the zx-plane, thanks to the 3D structure of the reflector. Again, on the side the cut profiles are shown. The spot size in this plane is around 49 µm.

Download Full Size | PDF

Indeed, the less concentrated the source is at a point, the lower the focusing performance. Since the light source can be approximated to a sphere with the diameter of a cell, this geometry would be less suitable for the acquisition branch, compared to parabolic mirrors. Concerning the excitation, by using a SM fiber with a 4 µm core, the light is sufficiently confined to allow good performances. Moreover, the spot size is well suited for the selective excitation of a single flowing cell.

4. Experimental section

Scattering solution: The visualization tanks on board the fused silica platforms are filled up with a scattering solution, based on a mixture of deionized water (3 mL) and Al₂O₃ nanoparticles (ALFA AESAR, Aluminum oxide 20% in H₂O, colloidal dispersion, 0.05-micron particles, 30 µL).

Surface topography measurements: The evaluation of the surface topography is performed by scanning the sample through an Atomic Force Microscope (model NT-MDT Solver Pro P47). The system is working in semi-contact mode and equipped with a high-resolution silicon probe (model NSG03), having a typical radius of curvature around 6 nm and force constant around 1.74 N/m. The measurements have been acquired in semi-contact mode at a working frequency of about 90 kHz (resonant frequency of the silicon tip). In this configuration the probe is scanned over the surface while oscillating at its resonant frequency. The topography is retrieved by monitoring the amplitude changes through a lock-in detection. The advantages of this technique rely on its high sensitivity, rejection to noise and the additional capability to monitor phase changes of the recorded signal, which are in turn related to possible changes of the material surface property. All the images have been acquired over a scanning grid of 512 × 512 points and a size ranging from 100 µm to 100 µm, after applying only a first-order correction to the acquired scan line, to remove the effect of the tilted plane. The phase contrast maps, recorded during the topography acquisition, do not highlight any relevant local change induced by the manufacturing process in the dielectric property of the surface. The correlation lengths L_x and L_y have been retrieved through the auto-correlation function of the AFM topography and fitted with Gaussian and exponential statistical models, as reported in Figure “ACF functions”.

Single parabolic reflector characterization setup: A collimated laser beam (wavelength: 473 nm, Lambda opbl-9010f CW 200mW) is shined against the reflector and focused inside the visualization chamber. To ensure a good alignment between the beam and our micro-optic, a special setup has been mounted on an optical breadboard, including three tilting mirrors and a 3-axis translation stage provided with a rotational stage (Thorlabs, XRNR1/M) to correct any tilt of the platform containing the micro-optic. Since the laser beam cross-sectional area (∼ 3.14 mm²) is greater than the area covered by the mirror (1.4 mm²), the beam diameter has been restricted to ∼ 400 µm by a pinhole (Thorlabs, ID25/M) before entering the device.

Double parabolic and single elliptical reflector characterization setup: The excitation beam is provided by a platform-integrated optical fiber connected to a fiber-laser system. The laser setup is composed by a controller (Thorlabs, ITC4001) and a fiber-connected laser diode (Thorlabs, LP520-SF15, wavelength: 520 nm) coupled with a SM fire with an 4µm-core and 0.14 NA (Thorlabs, P1-460B-FC-2) entering the devices. Due to the light source integration, no alignment setup has been necessary.

Vision setup: In all the experiments the vision has been given by a stereomicroscope (Optika microscopes Italy SZM-T) and the images have been captured by a CCD camera coupled with it.

Data analysis: The pixel intensity maps are obtained by processing the stereomicroscope images with a custom-made image analysis algorithm on MATLAB (MATLAB, 2018. version R2018b, Natick, Massachusetts: The MathWorks Inc).

5. Conclusions

In this work, we propose the breakthrough design and fabrication of in-plane micro-optical elements to manipulate in 3D light beams directly on-board microfluidic platforms. The fabrication process couples three direct writing and maskless techniques – i.e., FLICE, CO2 laser polishing, and inkjet printing –, achieving ultra-high optical quality reflective structures (RMS= 3.4 nm, below the standard quality factor λ/10). Two basic elements – i.e., 3D parabolic and elliptical mirrors – either individually or coupled to dielectric waveguides (optical fibers) to form a simple integrated optical circuit, have been characterized in depth. All the devices show remarkable focusing performances, with focal spots ranging from 30 µm to 50 µm in 3D, that would fit perfectly into the perspective of integrated single-cell optical analysis. Moreover, through numerical simulations, we have estimated that by integrating our micro-structures it is possible to improve the solid angle of acquisition up to the 56% of the total emitted angle, doubling the currently performances reported in literature. In addition, the ability to integrate such optical structures extremely close to any isotropic photonic source plunges the insertion losses. In this way, the signal to noise ratio could increase significantly, aiming at breaking down the time needed to acquire spectrophotometric measurements of single cells in real-case scenarios. These features could help overcoming the hurdles hindering in-flow spectroscopy techniques, such as Raman spectroscopy, on board LOC devices.