Integrated Optical Stretcher

This chapter describes the design and fabrication of the fully integrated optical stretcher, accomplished with a recently developed technique based on femtosecond laser writing.

Structure of an integrated optical stretcher

The idea of developing an optical stretcher integrated on a chip made of fused silica would represent a great improvement in the analysis of cell mechanical properties. The lab-on-a-chip approach grants very small dimensions, low costs and high reproducibility and it integrates microfluidic and optical functions onto a single chip. The fabricated chip is based on a fused silica glass substrate, thus providing high transparency for cell imaging, and represents a significant improvement in terms of stability, robustness and optical damage threshold over existing optical cell stretchers. Optical trapping and manipulation of red blood cells (RBCs) in the optofluidic chip are obtained by means of two counter-propagating beams coming from two integrated optical waveguides orthogonal to the microfluidic channel, designed with femtosecond laser pulses. The delivery of the cell suspension to the trapping region is accomplished by an easy connection of the microchannel to an external fluidic circuit, which guarantees a controlled flow and a high-throughput analysis. A fiber laser source is butt coupled to the waveguides in the chip, delivering the light required for the trapping and stretching of cells. Since glass absorption in the wavelength range adopted in the experiments (near infrared) is very low, the high powers needed for optical stretching can be easily coupled without appreciably heating the chip. Moreover, the high spatial quality of the trapping beams is guaranteed by the waveguide spatial mode distribution.

Design

In order to optimize the performance of the integrated optical stretcher (IOS) we first performed a careful design through numerical simulations. The design variables are the distance between the waveguide end-faces and the waveguide mode size, which are in principle dependent on the size of the cells under test. First we calculate the work (εTP, work per power unit) that has to be done against the optical forces to move a particle along a straight line connecting the centre of the trap to any possible target point in the surrounding space. The figure below reports the distribution of εTP of a specific dual beam optical trap obtained considering a distance L between the waveguide end-faces equal to 150 μm.

As expected, εTP behaves like a smooth potential well where the stable trapping position lies in the midpoint (z = 0) along the beam axis. The figure below, instead, reports the value of the escape energy εesc calculated for each trapping configuration obtained by changing the distance L and the transversal waveguide misalignment Δy.

It can be easily observed that the most stable trap, corresponding to the maximum εesc, is obtained for L = 148 μm and Δy = 0. It is worth noting that for L < 100 μm and L > 250 μm the value of εesc becomes considerably lower and the trap cannot be considered as stable. A transversal misalignment of Δy ≈ 1 μm already leads to a sensible variation of the trapping stiffness. The fs-laser writing fabrication procedure guarantees accuracy in the waveguides transversal position of the order of 100 nm thus preventing any criticality due to such misalignment.

Fabrication

The critical parameters in the fabrication of the device are the microchannel diameter, the waveguide mode size and the optical distance between the waveguide end-faces. For the microchannel diameter a value of about 100 μm is chosen, since the capillaries used in the discrete element OS have an internal dimension of the same order; the distance between the waveguide end-faces should range between 200 μm and 400 μm; the target mode size for the waveguides is set to 3.5 μm radius in order to match the fiber single mode at 1.07 μm wavelength. Indeed, in the trapping and stretching experiments a laser wavelength of λ ≈1 μm is chosen due to the following reasons: i) availability of compact fiber lasers with average power sufficient to achieve trapping and stretching of cells; ii) very low absorption of glass and cells; iii) possibility to filter out the laser light used for trapping, keeping the full spectrum of visible light for the cell imaging. For the fabrication of the integrated optical stretcher we used a technique called Femtosecond Laser Irradiation followed by Chemical Etching (FLICE). It is a powerful technique able to directly fabricate buried microchannels and waveguides and to create large access holes on the side facets of the chip in order to achieve easy connection with external capillary tubes. The schematic of the set-up used for femtosecond irradiation of the sample is reported below.

The second harmonic (515 nm) of a cavity-dumped Yb:KYW oscillator providing 350-fs laser pulses at repetition rates up to 1 MHz has been used. The laser beam was focused by a microscope objective inside the sample; the latter was translated by a computer-controlled motion stage. The glass is transparent for the used wavelength; however the high peak intensity achieved by focusing the femtosecond laser pulses induces a nonlinear absorption mechanism consisting of a combination of multiphoton absorption and avalanche ionization. The occurrence of this phenomenon is experimentally indicated by the emission of white light from the electron plasma generated at the laser focus. A first consequence of the irradiation of the nanostructured glass is a slight darkening of the glass color in the modified region. This is consistent with a red shift of the absorption spectra of the glass, corresponding to a refractive index variation through a Kramers-Kronig mechanism. Moreover we have an increase in HF etching rate of fused silica, correlated to the decrease of the Si-O-Si bond angle induced by the hydrostatic pressure or compressive stress created in the irradiated region. When fused silica is irradiated, the modifications induced by the femtosecond laser pulses can be classified into three categories depending on the laser processing conditions: a) for a low fluence, a smooth modification is achieved, resulting mainly in a positive refractive index change with a very weak selectivity in etching; b) for a moderate fluence, sub-wavelength nanocracks are produced, yielding a high etching selectivity of the irradiated volume with respect to the pristine one (up to two orders of magnitude); c) for high fluence, a disruptive modification is obtained with the creation of voids and microexplosions. In particular, regime a) is typically suited for waveguide fabrication, while regime b) is the one employed in the first step of the FLICE technique for microchannel production. Regime c) can be used for direct laser ablation. The fs-irradiation technique with moderate fluence is exploited to create not only the microchannel, but also large access holes on the side facets of the chip in order to achieve easy connection with external capillary tubes. The access-hole diameter of 350 μm is designed to exactly match the outer diameter of the capillary tubes; this tailoring is achieved by irradiating multiple coaxial helixes with different radii and with a pitch of 2 μm. The number of coaxial helixes depends on the desired size of the access hole; for a 350 μm diameter, 3 helixes are written with diameters of 80 μm, 160 μm, and 240 μm, respectively. The two access holes are connected by a straight line that, once etched, will provide a slowly tapered microchannel with a uniform central portion of 80 μm diameter where the optical trapping is achieved. The channel walls have a minimum radius of curvature of 40 μm and show the typical surface pattern obtained with this technology providing an estimated roughness in the 300-500 nm range.

Irradiation is performed at 600 kHz repetition rate with a pulse energy of 290 nJ at the second harmonic wavelength of 515 nm. The laser polarization is set perpendicular to the microchannel axis, which is placed at a depth of 400 μm with respect to the top surface. With the high-repetition-rate laser an irradiation speed of 1 mm/s is feasible; therefore, although complex structures are irradiated, the processing of the full chip is quite fast. The chip is then immersed in an ultrasonic bath with 20% of hydrofluoric acid (HF) in water for 4.5 hours to obtain the 3 mm long buried microchannel.

Waveguide writing in the fused silica sample is performed in the same conditions used for the irradiation step in the microchannel fabrication, i.e. focusing through a 50 × objective the frequency-doubled cavity-dumped Yb:KYW laser; however, this time the laser is operated at a repetition rate of 1 MHz since in this regime the fabricated waveguides exhibit lower propagation losses.

The waveguide writing parameters are optimized in order to have the best guiding properties at the operating wavelength of 1 μm. A pulse energy of 100 nJ and a translation speed of 0.5 mm/s allow obtaining single mode waveguides at the operating wavelength with a mode intensity radius at 1/e2 equal to ~4 μm and an ellipticity factor of 1.1. Measured propagation losses at the operating wavelength are equal to 0.9 dB/cm. Multiple sets of waveguides have been fabricated on the two sides of the microchannel, with a separation between the waveguide end-faces of 100, 130, 180 and 300 μm. Each set is composed of 3 waveguides, laterally spaced by 80 μm, that are fabricated at various depths with respect to the axis of the microchannel, i.e. + 5, 0 and −5 μm. In this way different depth positions of the trap are experimentally tested. Moreover, this approach could be exploited to fabricate several parallel traps able to intercept cells flowing at different heights, thus improving the measurement throughput. So, the overall fabrication process can thus be summarized in the following steps: i) the femtosecond laser is set to a repetition rate of 600 kHz and the structures for the microchannels are irradiated (typically several structures are fabricated on the same glass substrate); ii) The laser repetition rate is switched to 1 MHz without losing the alignment and the sets of waveguides are written in each device; iii) the substrate is cut and different chips of 3 mm × 8 mm size are obtained; iv) etching of the microchannels is performed by immersion in the HF solution. Since the irradiation of both microchannels and waveguides is performed before chemical etching, the writing of the waveguides is interrupted 500 μm before the edge of the chip, in order to avoid any etching of the regions corresponding to the waveguides. After the etching the two end-faces are polished in order to expose the waveguide input ends and perform efficient fiber coupling.

Once a chip is fabricated, it is connected to external fluidic and optical circuits. Using a set-up composed by an optical microscope and accurate translation stages, external capillaries are inserted in the access holes.

Once the capillary is firmly inserted it is glued by a drop of UV-curable resin. The external circuit is essentially made of two butterfly needles glued to the capillaries; the tubes at the other hand of the butterfly needles act as reservoirs. Cell suspension is transported through the trapping region by a controlled microfluidic flow; this is obtained, as in the case of the discrete elements OS, by varying the relative heights of the two reservoirs and can be finely adjusted with a micromanipulator. Optical fibers are aligned to the waveguides input-facets by means of two translation stages. Butt-coupling is presently used in order to have a flexible set-up, able to test all the waveguides in the chip; however, in a final device the fiber will be permanently pigtailed to the waveguide following the standard procedure developed for photonic devices in telecommunications (typical additional losses ~0.5 dB). The chip connected to the capillaries is also glued by UV-curable resin on a thin glass coverslip to increase robustness of the connections but still allowing imaging of the channel content with a high magnification objective. Here is the final result.

Experimental setup

The schematic diagram of the experimental set-up used to demonstrate the effectiveness of the integrated optical stretcher is shown below. A CW ytterbium fiber laser with an emitting power up to 5W at 1070 nm, is used as light source. The beam coming from the laser is split in two branches by means of a 50%-50% fiber coupler (FC1). The optical power in each arm is then controlled by variable optical attenuators (VOAs) and monitored using the 1% port of a 99%-1% fiber coupler (FC2a); this enables to finely balance the optical power at the output of the two fibers. In order to optimize the light coupling into the chip-integrated optical waveguides, a second 99%-1% fiber coupler (FC2b) is added in the fiber line: the power coupled into one waveguide, transmitted through the microchannel and collected by the second waveguide, is thus monitored in the opposite branch. All the fiber components are single mode at the working wavelength as well as the spliced bare end-fibers. The VOAs are specified for operation up to 2 W of optical power, while we verified the FCs up to 4 W. Given the high optical threshold of the fused silica chip, the current set-up can also be used to stretch cells other than RBCs, where higher power may be needed.

The chip is mounted on an inverted microscope equipped for phase contrast microscopy (TE2000U, Nikon). Phase contrast images of optical trapping and stretching are acquired by a CCD camera (DS-Fi1, Nikon). The pixel size for all the employed magnifications was calibrated with a grating; this allows for absolute distance measurements with a resolution of 0.055 μm/pixel in the case of a 40 × objective. The trapping and stretching capabilities of the chip have been tested on RBCs. The cell suspension is prepared by diluting 10 μL of blood in 8 ml of hypotonic solution in which the RBCs acquire a quasi-spherical shape with a radius of ~4 μm and then inserted in the microfluidic circuit with a syringe. For an easy imaging of the flowing cells, the typical value of the cell speed is set in the 10-50 μm/s range.

Experimental results with a round-section microchannel

First experiments have been performed on the circular cross-section microchannel chip, whose fabrication process has been described above. RBCs optical trapping is achieved with an estimated optical power at each waveguide output of about 20 mW. The figure below shows a sequence of a few frames demonstrating how the trapped RBC is stable in its position even if a background flow is present (a flowing cell, indicated by the white arrow, is out of focus since it’s travelling at different height in the microchannel).

Moreover, we observe a controlled movement of the trapped RBC along the beam axis obtained by unbalancing the optical forces applied on the two sides of the dual beam trap. The force unbalance is easily achieved by varying the output power of one of the two waveguides, which can be finely tuned by adjusting the corresponding VOA.

When a single cell is stably trapped in the microchannel it can be stretched along the trap axis by simultaneously increasing the optical forces applied to the cell by the two counterpropagating beams. Experimentally this is achieved by raising the power emitted from the laser source. Therefore, the trap is still stable and a progressive stretching of RBC is observed. Below, a sequence of frames demonstrating the optical stretching of a single RBC. The cell can be elongated up to 25% of its initial size when increasing the waveguide output power to 300 mW. However, in order to achieve such a clearly visible elongation the cell is stretched into its plastic deformation regime. By stretching the cell with lower optical power smaller deformations are observed in the elastic regime.

Fabrication of a square-section microchannel and experimental results

The lens effect induced by the curvature of the microchannel prevents from a reliable retrieval of the cell contour. To solve the lens-effect problem, a new chip with a square cross-section microchannel has been fabricated. To this aim, we designed and implemented an irradiation path to obtain a square cross-section channel (SC).

This is obtained by irradiating two coaxial helixes, with a pitch of 2 μm, with rectangular cross-section one inside the other. The irradiated helixes have a cross-section height and width of 45 μm and 30 μm, respectively, for the inner one and 70 μm and 60 μm, respectively, for the external one. The microchannel is then terminated by two access holes with circular cross-section that are obtained by irradiating three coaxial helixes with diameters of 60 μm, 130 μm and 200 μm, respectively, a pitch of 2 μm and a length of 800 μm per side. While in the portions of the microchannel closer to the access holes the etching smoothes out the corners of the rectangular cross-section, in the central portion of the channel, where the HF solution arrives later and acts for a shorter time, the channel closely follows the irradiation path with a sharp rectangular shape.

Irradiation is performed with a pulse energy of 300 nJ and a translation speed of 1 mm/s, leading to an overall irradiation time of about 60 minutes for the complete structure. The chemical etching is executed immersing the chip in an ultrasonic bath with 20% of HF in water. A 4.5 hours etching produces the microchannel, which is 400-μm buried under the top surface, has a 2-mm length, a central rectangular cross-section of 85x75 μm, and two 800-μm-long access holes. To test this new device we first characterized the coupling losses and the trap quality of the waveguides. We evaluated the coupling losses facing couple the light in each couple of waveguides, in order to evaluate the coupling losses inside the chip when the microchannel is empty, then we do the same measurements filling the channel with the RBC suspension. The estimate losses from the left fiber to the trapped cell are in the order of 3 dB in each waveguide. The trapping quality of each set of waveguides has been characterized by trapping particles flowing at different velocities and then lowering the optical power until the particle escapes from the trap. From the distance covered in a specific amount of time we measured the escape velocity of the cell at different trapping power for each set of waveguide.

By exploiting the measurement of the escape velocity we can calculate also the transverse trapping force by taking into account the viscosity of the solution. Figure below reports the trapping forces calculated from the measured escape velocities.

The experimental results are in good matching with the simulations. Indeed the figure below shows the simulated trapping force as a function of the power emitted by two facing beams.

After the characterization of the trapping efficiency, we tested the stretching capabilities of the integrated OS by exploiting red blood cells. First we trapped a red blood cell at a power of 35 mW inside the microchannel.

Then we increased the optical power to 300 mW and the particle stretched.

Data on the elongation of the trapped red blood cells as a function of the laser power have been recorded on an Excel graph.

Although it has been possible to perform stretching experiments with the square-channel microchannel chip, it still has a big limitation due to the roughness of the channel walls due to the etching process. This roughness can prevent from acquiring a good imaging of the sample, thus making the retrieving of the cell contour more difficult. Currently we’re investigating possible solutions to the roughness problem ranging from the use of a different concentration of hydrofluoric acid in the etching solution to the choice of a different substrate or even the use of laser ablation for the microchannel fabrication.