1. Introduction

The synergy between optics and microfluidics represents a powerful tool to investigate samples, by combining the advantages of lab-on-a-chip devices with the versatility and sensitivity of optical characterization techniques [1]. A crucial issue is the integration of optical components in microfluidic devices, trying to overcome the critical alignment between macrooptical and microfluidic elements. To facilitate this process, optical fibers, glued to the lab-on-a-chip, can be used. However, the diverging beam out of a fiber is not suitable in most applications where some processing is typically required, e.g. focusing of the light in a specific portion of a microchannel. For this reason, in recent years an increasing interest in developing integrated microlenses has been shown [2,3]. Among all, integrated microfluidic lenses offer the great advantage of the tunability, giving rise to compact and robust devices, whose optical properties are not fixed, but can be flexibly tuned [4]. The versatility of these devices can be obtained by either changing the refractive index of the liquid filling the lens or modifying the shape of the device [5]. The majority of the microlenses reported in the literature are out-of-plane lenses, which allow to focus the light in a plane orthogonal to the substrate [6,7]. These lenses present quite complex fabrication processes; moreover in order to be integrated in lab on a chip devices, they need an accurate alignment between different layers, which can be tough to obtain. In-plane microfludic lenses that focus light in a plane parallel to the substrate, better fit for lab-on-chip applications [8]. Different configurations have already been reported in literature, including liquid core liquid cladding lenses [9–11], pressure controlled liquid-air interface [12,13] and gradient refractive index lens [14]. The lenses described in these works present many advantages, like the capability of being easily implemented in lab on a chip devices, the great tunability of the optical properties and the smoothness of the lenses surface. Nevertheless it is important to notice that none of these works actually takes into account the impact of spherical aberrations, which highly limits the focusing capability of the device. Only few works in literature face this problem [7,15,16], but they employ out-of-plane lenses that, as previously discussed, are not suitable for lab on a chip applications.

Here we present an in-plane integrated cylindrical microfluidic lens, with tunable optical properties and with a shape optimized to reduce spherical aberrations. Furthermore, by dynamically changing the liquid in the lens, focused light can be modulated and this capability has been exploited by adding a droplet generator module before the lens. A periodic light modulation of a focused beam is thus obtained, which may operate as an integrated optofluidic chopper, and that can find a wide range of applications, both in on-chip detection and imaging. Capillary electrophoresis is a valuable example of an application that could benefit from on chip light detection. In this technique, the different arrival times of the specimens flowing in a fluidic channel allows their identification, and the capability of distinguishing two species with similar properties is related to the spatial resolution and hence to the width of the laser beam [17]. On the other hand to maximize the signal all the channel cross section should be illuminated, for this reason, a cylindrical lens capable to focus the beam in one direction is ideal for this application. Moreover the sample concentrations are typically low in this application, thus reducing the signal to noise ratio; a common technique to enhance the measurement sensitivity is to modulate the light so as to subtract noise and improve the measurement quality [18]. On chip light modulation could favor this technique also on other aspects, indeed a modulation-frequency-encoded multi-wavelength excitation has been successfully proven to allow parallel optical detection in capillary electrophoresis, and with our device it could highly benefit from the integrated approach [19]. As far as the imaging is concerned, an interesting application is selective plane illumination microscopy, which is a powerful microscopy technique used to reconstruct 3D images of biological samples, based on an optical sectioning approach. The axial resolution in this technique is determined by the waist of the focused beam generated by the cylindrical lens; therefore spherical aberrations should be minimized. In a recent work, we presented the first complete on chip integration of this technique [20], anyway a multicolor analysis would improve the device performances. We believe that the on chip light modulator here presented would perfectly suit for this purpose, allowing the alternation of multiple excitation wavelengths, with no need of external instrumentation.

The fabrication of the devices presented in this paper was performed by Femtosecond Laser Irradiation followed by Chemical Etching (FLICE) [21,22], a simple process that permits to obtain microchannel in fused silica glass substrates in two steps. First the device is irradiated with a focused femtosecond laser beam, which modifies the material only in the focal point. Thanks to the localization of this modification the laser can define the desired 3D layout in the substrate volume by simply translating the glass with respect to the focus of the laser beam. Subsequently, the chip is exposed to an aqueous solution of hydrofluoric acid, which (for a fused silica substrate) will preferentially attack the modified areas, leading to the formation of microchannels. This technique presents interesting advantages with respect to standard lithographic approaches, such as the fast prototyping capability, due to the maskless approach, and the possibility to create innovative 3D structures directly buried in the glass volume with no bonding requirement[23–25]. The realization of our devices takes advantages from both these properties.

2. Lens design

The schematic design of the device is reported in Fig. 1: an optical fiber is coupled to the glass substrate and the diverging laser beam is focused in the sample channel by the microfluidic lens, which is itself a microchannel, filled with a proper refractive index oil. The lens is capable to confine the beam in one direction only, giving rise to a sheet of light that fully intercepts the sample channel; this feature might be useful to perform sample analysis with high spatial selectivity and high signal to noise ratio.

 

Fig. 1 Schematic design of the device respectively viewed from top (a) and side (b). The diverging beam output from a fiber is focused in one dimension into the sample channel thanks to the presence of a cylindrical microlens with aspheric profile.

Download Full Size | PDF

The design is optimized considering that the lens optical properties should have well defined features: i) the confocal parameter of the focused beam should be at least 400 μm, to fully and uniformly cover typical channel widths (100–300μm), where the sample will flow; ii) the waist should be as narrow as possible, which, compatibly with the confocal parameter requirement, is ≈ 6 μm; iii) the position of the focused beam should be centered in the sample channel iv) the beam in the direction where it is not focused should be larger than the channel dimension to produce a uniform illumination of the microchannel cross-section.

One of the main benefits of the femtosecond laser fabrication technique is the fast prototyping capability that facilitates the development of different geometries, such as an optimized lens profile that would allow reducing the impact of spherical aberration and improving the focused beam quality. We therefore exploited a simple ray-optics model to design a profile for the cylindrical lens producing an aberration-free focus that in addition satisfies the above mentioned constrains.

The geometry of the desired lens is reported in Fig. 2. The first interface is chosen to have a curvature that does not deflect the diverging beam, being orthogonal to the beam rays. The second curvature is instead obtained by imposing the same focal point to all the rays that are refracted by this surface. By using an incremental approach and starting from the origin O, Fig. 2, the lens profile is found using the following equations:

In this way we can develop an acylindrical lens optimized for our specific device. The dimension of the lens is larger than the beam diameter to avoid any problem related to the blunted edges of the structure after the exposure to HF. Moreover, the irradiation pattern has been defined by including a precompensation for the lens profile, which has been determined empirically and allows obtaining the designed radius of curvature taking into account variations due to different etching rates. In particular we fabricated a lens with the theoretical profile and we measured the profile obtained after the etching step, so as to evaluate the amount and the direction of excess/lack of etching. This measurement allowed us to accordingly modify the irradiation pattern by enlarging or shrinking it where we respectively observed a lack or excess of etching.

 

Fig. 2 Schematic layout of the acylindrical lens, the first interface of the lens does not modify the beam propagation. The second interface instead is designed to focus all the beam rays to the same point.

Download Full Size | PDF

3. Device fabrication and characterization

The device is fabricated in fused silica by a two step process: first, the desired pattern is irradiated by a focused femtosecond laser beam; subsequently the device is exposed to an ultrasonic bath of 20 % HF acid solution that attacks preferentially the irradiated region [26,27], thus allowing the formation of microchannels.

The irradiation is performed by focusing, through a 50× − 0.6 NA microscope objective, the second harmonic of a commercial femtosecond laser (femtoREGEN HIGHQlaser) with an emission wavelength of 1040 nm and a pulse energy up to 23 μJ. Scan velocities and pulse energies are varied in relation to the depth of the single structure with respect to the glass surface, so as to compensate the spherical aberrations introduced by the air-glass interface in the writing process. The geometry of the device is obtained by properly translating the sample with respect to the laser beam with a system of high precision air bearing translation stages (Fiberglide 3D, Aereotech).

A microscope image of the lens is reported in Fig. 3, where panels (a) and (b), show the top views of the device profile respectively before and after the exposure to the HF solution. Figure 3(c) instead shows the side view of the same lens after the etching step. Figure 3(a) reports the irradiated pattern where the lens acylindrical profile designed to reduce spherical aberrations is clearly visible. In particular it is worth noticing the fractal design: multiple scaled lens profiles are irradiated with the aim to facilitate the acid penetration in the lens volume. To obtain the complete lens cavity, which is 550 μm high as can be observed from the side view image, the concentric lens profiles have been irradiated at different depth from the glass surface, with a reciprocal separation of 6 μm. Moreover, it is possible to observe the presence of two lateral access holes, designed to receive the peek tubing (with an external diameter of 360 μm) used to fill the lens with the desired refractive index oil. The pulse energy used for the irradiation of the lens is 270 nJ, while for the access holes is 420 nJ and 320 nJ respectively for the lower and upper one with respect to the glass surface. The translation speed is equal to 1 mm/s. The fabrication geometry exploited in the irradiation process allows realizing channels parallel to the writing beam propagation direction from the bottom to the top of the substrate [28]. This allows obtaining a spatial transverse resolution of about 1 μm, as determined by the focal spot size, in the irradiated lines. On the other hand this configuration, because of the ellipticity of the focused spot size, which being elongated along the beam direction allows a smoother modified area due to the overlap of the irradiated lines, permits a reduction of the microchannel roughness after the etching process down to few nanometers [28,29], thus causing negligible light scattering related to the surface quality.

 

Fig. 3 Microscope images of the top view of the lens before (a) and after (b) the exposure to the acid solution. (c) Microscope images of the side view of the lens microchannel after the etching step. The scale bar, for all panels, is 100 μm.

Download Full Size | PDF

The cylindrical lens is first filled with water to avoid bubble formation, then a transparent oil with refractive index n_oil = 1.56 is inserted, which is higher than the fused silica refractive index to produce a converging lens. To characterize the lens optical properties we fabricated a cuvette in front of it that is 800 μm long and ∼1 mm high. The cuvette is centered at the expected focusing position, and filled with Rhodamine, to easily observe through a fluorescence signal the focused beam. The length exceeds on purpose the confocal parameter length, allowing to better study the beam propagation. The laser used to characterize the device is a frequency-doubled Nd:YAG laser (Uniphase) with emission wavelength at 532 nm, coupled to a single mode optical fiber. The measurements are performed on an inverted microscope (LEICA DMI 3000M), with a 10×, 0.25 N.A. objective. A proper filter is introduced in the acquisition setup to remove the contribution of the excitation wavelength (532 nm). The acquired images are subsequently analysed with a specifically developed Matlab routine to retrieve the waist of the focused beam. The result is reported in Fig. 4. The images are acquired through an objective with a depth of field smaller than the beam dimension, and are then subsequently analyzed with a specifically developed Matlab routine, which implement a deconvolution process to retrieve the correct waist of the beam along the propagation direction, by taking into account the beam divergence. The result, reported in Fig. 4(b), is perfectly matched by a Gaussian fit (red line) and shows a minimum waist equal to 6.2 μm, which is in very good agreement with the designed one. Indeed, considering that the lens focal length is equal to 0.7 mm, and the fiber mode diameter is 4 μm, we retrieve a theoretical focused waist equal to 6 μm. It is worth mentioning that we fabricated also a cylindrical lens with the same focal length, whose beam waist resulted to be larger of about 20%. The comparison among these two values highlights the advantages related to the lens profile, which allows a perfect match between theoretical and experimental results, avoiding the spherical aberration contribution.

 

Fig. 4 Beam waist characterization. Panel (a) shows the acquired microscope image of the Rhodamine fluorescence, while panel (b) shows the correspondent waist analysis. The minimum waist retrieved for the beam focused by the acylindrical lens is about 6.2 μm.

Download Full Size | PDF

4. Light modulation

The optical properties of the proposed lens strictly depend on the refractive index of the liquid flowing in it. For this reason by simply changing the liquid we can easily tune the focus position. Furthermore we can modulate the focused light by substituting oil with water, whose refractive index (n_w = 1.33) is lower than the one of the surrounding glass (n_g=1.46). This allows to focus and defocus the light, and consequently it permits to switch on and off the fluorescence signal in the sample channel, which is a useful capability when performing sample analysis. In literature a few works are present, showing microfluidic lenses with a similar integrated focusing and defocusing behaviour [10,30]. However, these are implemented with complex geometries that need the simultaneous control of multiple inlet channels or electric field applications, and none of these take into consideration spherical aberration effects. We have first tested the light modulation capability of our optofluidic lens by fabricating a cuvette on the same substrate, again with the FLICE technique, in front of the lens. The experiment, as in the previous case, is performed at a wavelenght of 532 nm, and an optical fiber is butt coupled to the sample. The cuvette is filled with Rhodamine, in order to characterize the focusing capability of the lens by monitoring the fluorescence signal in the cuvette. The fluidic network of the lens has been filled half with oil and half with water; thanks to the immiscibility of these liquids, it is possible to maintain a clear separation between the two. External tubes are used to control the fluids in the device; by exploiting a pressure driven pump (MFCS-FLEX Fluigent) we have shifted with high precision the position of the interface between the two liquids before and after the lens. Due to the distance between the optical fiber and the cuvette (≈ 3.5 mm), and to the divergence of the laser beam, whenever the beam is not focused, no fluoresce signal is detected, Fig. 5.

 

Fig. 5 Working principles of the device when either oil or water is filling the lens; water is colored in red to be easily distinguished from oil. In panel (a) the microlens is filled with oil and the beam is focused in the Rhodamine cuvette, switching on the correspondent fluorescence signal (b). In panel (c) red water is filling the lens; in this case the beam is not focused and no fluorescence signal is detected in the test channel (d). Scale bar is 100 μm. (See Visualization 1)

Download Full Size | PDF

This approach allows to dynamically switch on and off the fluorescence signal on demand (see Visualization 1). Anyway, this method is not suitable for a frequent switching, which would indeed require continuosly inverting the pressure to change the liquid in the microchannel, being in this operation limited by the inertia of the fluids.

In order to have a periodic modulation of the fluorescence, and an easier handling of the pressures, we have added a droplet-generator module in the device [31,32]. The purpose is to obtain a stream of oil droplets in water, which will be able to temporally focus the beam in the sample channel, while flowing in the microfluidic lens. The droplet generator consists in a 2D flow focusing device fabricated by femtosecond laser micromachining [23]. As a first step we have characterized the droplet generation capability, by inserting an aqueous solution in the two sheath channels, and oil in the central one. We have observed that by simply keeping constant the insertion pressures of the two liquids with respect to the outlet channel, a periodic formation of droplets takes place. Thus this device could work without the use of high-precision and programmable pressure driven pumps. From our characterization experiments, we found that the droplet dimension shows a dependence on the ratio between the insertion pressure of the two liquids, for example in Figs. 6(a) and 6(b) it is possible to note the different dimensions of the droplet in the main channel, according to two different pressure ratios. In panel (a) the insertion pressures for oil and water are respectively 99 mbar and 76 mbar, while in panel (b) the pressures are both equal to 76 mbar. On the other hand, the rate at which the droplets are generated is influenced mainly by the value of the pressure difference between the channel input and output: the higher is the pressure difference, the higher is the droplet generation rate.

 

Fig. 6 Validation of the droplet generator, consisting in a 2D hydrodynamic focusing device. (a) Droplet formation with insertion pressure for oil and water respectively of 99 mBar and 76 mBar. (b) Droplet obtained with both pressures equal to 76 mBar. The scale bar is equal to 100 μm.

Download Full Size | PDF

Visualization 2 shows the corresponding device characterization at different insertion pressures, demonstrating the flexibility of this device in generating droplets with different dimensions and at different repetition rates. It is important to note that there is a minimum droplet size that allows a good light modulation. Considering the beam divergence and the height of the sample channel, which is 35 μm, we need to have a minimum droplet length equal to 30 μm. Moreover, we observed that after an initial instability of the droplets rate, which lasts a couple of minutes and is due to the fluidic inertia of oil, the system stabilizes and we observed no significant deviation of the droplet frequency. Indeed, by analyzing the system modulation rate over a time period of 20 minutes, we observed a standard deviation of about 0.5 Hz

The layout of the final device, which combines the droplet generation with the acylindrical lens, in order to obtain an integrated optofluidic chopper, is reported in Fig. 7, and presents a necessary 3D geometry. It is constituted by a droplet generator followed by an acylindrical lens that faces a sample channel. The lens is fabricated along the z direction to ensure smooth wall roughness. The validation of the chip is performed on an inverted microscope; the sample channel is filled with Rhodamine to monitor the lens behaviour and the insertion pressures are controlled by pressure driven pumps.

 

Fig. 7 Schematic layout of the device for periodic light modulation: a droplet generator module (1), is used to create droplets of oil, which flow in the acylindrical lens (2), focusing the laser in the test channel (3). A fiber (4) is butt-coupled to the device.

Download Full Size | PDF

In order to study the device behaviour at higher modulation rate, we collected the intensity of the fluorescence signal from one of the microscope outputs. The light was then focused on a fast photodiode (Visible DC-125 MHz, low noise photoreceiver, New Focus) and the electric signal was analysed by means of a digital oscilloscope (Tektronix DPO2024B).

First we have characterized the device working at low droplet generation frequency, thus allowing to clearly visualize the correspondence between the oil droplet flowing inside the lens and the modulation of the fluorescence signal. As reported in Figs. 8(a) and 8(b) (frames from Visualization 3), we observed an excellent agreement between the two phenomena, confirming that by means of the train of droplets the periodical modulation of the fluorescence signal can be achieved. Figure 8(c) shows the corresponding periodic modulation observed by analysing the signal with the oscilloscope.

 

Fig. 8 Panels a, b and c show the device characterization performed at low flow rate to better observe the correspondence between the droplet generation and the light modulation. Droplets that periodically flow in the lens have been created (a), and the corresponding periodic modulation is observed in the sample channel (b), (see Visualization 3). Scale bar is 150 μm. Panel (c) shows the periodic modulation of the signal acquired with the digital oscilloscope. Panel (d) shows the result of the experiments performed at increasing droplet rates, reaching a value of about 60 Hz at the maximum pressure supplied by our pump system. Error bars are obtained by repeating each measurements three times.

Download Full Size | PDF

The validation has been then performed by increasing the insertion pressures, but keeping the ratio of the water and oil pressures equal to one, so as to increase the droplet generation rate without changing the droplet size. The results of the measurements are reported in Fig. 8(d). The graph shows that the repetition rate of the light modulator can be increased by changing the insertion pressure, reaching a maximum value of ≈ 60 Hz for 1 bar pressure difference. This limit is due to the range of pressures achievable with our pump system, with respect to the fluidic resistance of the device. The resistance is related not only to the device geometry, but also to the choice of external tubes, opening to the possibility to further enhance this value by using larger tubes and access holes. In fact droplet generation at repetition rates of the order of few tens of kilohertz has been demonstrated in the literature [33–35].

5. Conclusions

In this work we presented an in-plane integrated microfluidic lens, an optical component that is ideally suited for lab-on-chip applications. The proposed lens, fabricated by femtosecond laser micromachining, has an optimized geometry that reduces the spherical aberrations, and also shows interesting tunable optical properties. Indeed the focusing capability of the lens can be modified by simply changing the fluid flowing in the device. On the basis of the refractive index of the liquid, focusing or defocusing of the laser beam can be achieved by modulating the fluorescence signal in a test channel filled with Rhodamine. In order to easily achieve a periodic light modulation, a droplet generator module has been added before the lens; the results show that this device can work as an integrated light modulator, switching on the fluorescence signal periodically with a repetition rate that depends on the insertion pressure. The result is obtained implementing a simple 3D geometry that would not have been possible with standard lithographic techniques.