1. Introduction

Particle sorting plays an important role in medical diagnosis and the environmental sciences. Conventional large-scale sorting systems such as fluorescence-activated cell sorters (FACS) [1] have been developed and commercially available for decades. Alternative microfluidic-based lab-on-a-chip approaches have also attracted much attention due to their small sample and reagent consumption, high efficiency and low cost [2]. More importantly, multiple functionalities such as sample preparation, chemical analysis, bio-analysis or other assays of the sorted populations can be integrated on the same chip.

Various microfluidic based sorting devices have been demonstrated using the electrokinetic mobilization of fluid [2], dielectrophoretic forces [3], and hydrodynamic flow control [4]. However, these techniques can suffer from problems that include electrophoretic damage due to high voltages, the requirement that the buffer has particular attributes, and the need for fluorescent labeling for active sorting to function. A passive particle sorting based on intrinsic properties such as size and refractive index could be useful in many applications due to the simplicity of the system and sample preparation.

Optical forces have been used for particle manipulation due to their non-invasive and non-contaminating nature. Our group has investigated optical manipulation using various platforms including diffractive optics, plasmonics, and Si photonics [5–11]. Several optical force based particle sorting devices have been reported [12–14]. Particles with different sizes, shapes and refractive indexes experience different optical forces and can be separated into output channels based on their intrinsic properties without fluorescent labeling [15–17]. However, a displacement of several microns is often needed to deliver particles into different microfluidic outlets. This requires a large incident power for optical force to push or drag the particles for such a long distance, which is especially difficult for sub-micron particles [13].

Here, we demonstrate a passive sorting system using the near-field optical forces exerted by a waveguide-based optical splitter. Because waveguides are employed, the light input to the device can interact with the particles over a long distance, as the mode maintains its profile and does not spread as it would for a laser beam in free space. The trapped particles can therefore be delivered over long distances with relatively low power consumption. More importantly, particles are sorted into different waveguides, rather than into different microfluidic channels. A displacement on the order of λ/2 is therefore needed, one or two order of magnitude smaller than previous microfluidic based sorting devices, which sort particles into different outlet channels [14, 18]. Therefore, a lower power is required to switch the particles, which opens up new opportunities for sorting sub-micron particles. The planar nature of the system makes it possible to implement massive parallel processing and integrate other pre- or post-sorting functionalities into the system.

In the following sections, we first present the experimental results, including fabrication of the device and its demonstration for particle sorting. Then we explain the mechanism by which the 3-dB splitter achieves sorting, supported by the results of simulations and the position tracking. The paper concludes with a discussion of how this work could be applied and extended.

2. Design and fabrication

Figure 1 shows the schematic diagram of our sorting system. The sorting is performed using a 3-dB optical splitter consisting of a channel waveguide (CWG) and a slot waveguide (SWG). We have reported a compact polarization splitter using a similar structure [19]. Transverse magnetic (TM) mode incident light from the CWG is split equally into two outlets of the splitter. The splitter is embedded in a PDMS microfluidic channel oriented perpendicular to the waveguides. Fluorescent polystyrene particles are delivered to the device through the microfluidic channel and trapped by the gradient force associated with the evanescent field of the waveguides when they come into their vicinities. The trapped particles are pushed along the waveguide by the scattering force and delivered to the coupling region. There, the smaller particles are switched to the SWG with the help of an obstacle (a stuck particle), while the larger particles remain on the CWG. As discussed in Section 4, this is due to the different potential wells experienced by the large and small particles. The stuck particle in red plays a critical role for transferring the small particles. More details will be discussed in section 4.

 

Fig. 1 Microparticle sorting device: schematic diagram.

Download Full Size | PDF

The device is fabricated on a silicon on insulator (SOI) wafer with a 220 nm top silicon layer and a 3 µm thick buried oxide. Hydrogen silsesquioxane (HSQ) e-beam resist is coated on the wafer and pre-baked for 5 minutes at 180 °C. The patterns are exposed using electron beam lithography with an acceleration voltage of 100 kV (Elionix ELS-7000). After developing the chip using 25% tetramethylammonium hydroxide (TMAH) for 1 minute, the patterns are transferred to the top Si layer by dry etching using hydrogen bromide (HBr). A PDMS chip containing microfluidic channels (50 μm high, 100 μm wide) is bonded to the SOI chip. Figure 2 shows a scanning electron microscope top view image of the fabricated device comprising of a CWG and a SWG. The inset shows a magnified image of the coupling region. The CWG has a width of 420 nm. The SWG is 450 nm wide with a 50 nm nanoslot in its center. The coupling coefficient depends on the size of the gap and the length of the coupling region. We choose a gap of 200 nm and a coupling length of 4.6 µm to achieve a 3 dB splitting of the TM mode. Fiber couplers consisting of silicon inverse tapers embedded in 2 x 2 µm SU8 waveguides are employed at both ends of the waveguides to improve the coupling with the lensed fiber that delivers the laser input and to reduce the Fabry-Perot effect by lowering the reflection from the device output. The chip is placed in an experimental set-up for sorting demonstrations. The output of a tunable laser (HP8168, near λ=1550 nm) is amplified by a high power C-band erbium-doped fiber amplifier (EDFA) and launched into the waveguide to excite TM modes. The output from the device is collected by a lens onto a photodetector. A homebuilt upright fluorescence microscope equipped with a CCD camera is used to monitor the particle sorting process.

 

Fig. 2 SEM image of the 3-dB splitter; the inset shows an enlarged image of the coupling region.

Download Full Size | PDF

Figure 3 shows the transmission spectra of the device from both outputs of the splitter measured using the fiber end-fire technique. Our design gives a 3 ± 0.2 dB splitting of the TM mode across the entire C-band from 1530 to 1565 nm. A wavelength of 1550 nm is used in our sorting demonstration. Figure 3 shows that, at this wavelength, the power coupled into the SWG is 7% higher than the power remaining in the CWG.

 

Fig. 3 Output power (μW) versus wavelength (nm) from SWG and CWG when input to device is TE mode (orange and blue curves) and when it is TM mode (red and black curves).

Download Full Size | PDF

3. Demonstration of microparticle sorting

We experimentally demonstrate the passive sorting of particles with diameters of 320 nm and 2 µm using the optical splitter. Particles are mixed together in de-ionized (DI) water and injected into the microfluidic channel using a syringe. The behavior of particles encountering the splitter is captured by the fluorescence microscope and included as the Supplemental Movie. Figure 4 shows selected frames taken from this movie. Both the large and small particles are observed to be pulled onto CWG by the optical gradient force and propelled to the coupling region by the scattering force. A long section of the CWG before the splitter is situated inside the microfluidic channel, which increases the probability that it captures particles. Over the course of the movie, all of the smaller particles that are trapped by the CWG are switched to the SWG by the splitter, while all of the larger particles remain on the CWG. The smaller particles are switched to the SWG after bouncing off an obstacle particle (320 nm diameter) stuck on the CWG in the coupling region (indicated in the second frame of Fig. 4). The stuck particle provides a perturbation that allows small particles to be ejected from the CWG potential well and be trapped in the SWG potential well. The sorting mechanism is further discussed in Section 4. All particles are correctly sorted during the observation period of more than 10 minutes. A sorting rate of ~30 particles/min is observed, which can be improved by increasing the density and speed of trapped particles on the waveguide. Parallel processing by integrating multiple channels could further boost the sorting rate.

 

Fig. 4 CCD image of the sorting with a guided power of 20 mW. The waveguides are indicated by the gray lines in the first frame. The trapped particles are marked by white arrows with the indicators S and L for the 320 nm and 2 µm particles, respectively. The obstacle particle is marked in the second frame (Media 1).

Download Full Size | PDF

The guided power in the channel waveguide incident to the device is estimated to be 20 mW. This is one to two orders-of-magnitude smaller than the optical powers used in some previous optical sorting approaches [14,15,17,18]. The laser power is roughly ten times larger than the guided power due to coupling losses. It has been shown, however, that such losses can be reduced considerably [20]. It should therefore be possible for the approach we introduce to employ both laser and guided powers that are far smaller than previous methods.

4. Mechanism of the sorting phenomenon

To investigate the mechanism by which particles can be switched, we simulate the gradient force along the y-direction (Fy) as a function of particle position, also along the y-direction. We do not know the precise value of the gap between the particles and the waveguide surface, but make the same approximation as that of previous work [9,21], and take it to be 30 nm, slightly larger than the Debye length. The particles are situated at the start of the bending region of the CWG. We assume a total guided power of 20 mW in the incident waveguide. Trapping potentials are calculated by integrating force with distance along the y-axis. We take the potential at 2 μm from the center of the structure to be zero. Calculation results are plotted in Figs. 5(a) and (b) for 320 nm and 2 µm particles, respectively. Figure 5(a) shows that a double-well distribution with two minima, each around the waveguide centers, occurs for the smaller particles (320 nm diameter). On the other hand, as shown in Fig. 5(b), there is only one minimum located at roughly the midpoint of the two waveguides for the larger particles (2 µm diameter). We believe that the dramatic differences between the potential vs distance curves play an important role of the sorting. For example, if the stuck bead causes one of the 320 nm particles to be displaced by some means to a y-position greater than ~250 nm (see Fig. 5(a)), it will become trapped on the SWG. On the other hand, it is entirely possible that the stuck bead has no effect on the delivery route of 2 µm particles because only one potential minimum exists for them. In the bending region, the distance between CWG and SWG increases. This modifies the potential wells. We simulate the gradient force and the potential depth along the y-direction when particles are located 7 µm after the start of the bending region as in Figs. 5(c) and (d). The further splitting of two potential wells for 320 nm particles prevents the particles jumping back to the CWG. From Fig. 5(d), it can be seen that the potential well experienced by 2 µm particles splits as the gap between the waveguides enlarges. The CWG exerts a larger force on the particles than the SWG, and provide a deeper potential well (Fig. 5(d)). The 2 µm particles fall into the deeper well and follow the CWG when it departs from the SWG.

 

Fig. 5 In-plane gradient forces (Fy) and potential depth plotted as a function of particle position for (a) 320 nm and (b) 2 µm particles at start of bending region. Similarly, Fy and potential depth are plotted for (c) 320 nm and (d) 2 µm particles at distance of 7 µm from start of bending region. Optical forces are calculated by Maxwell stress tensor method using fields found from 3D FDTD simulations. To indicate the positions of the waveguides, the cross sections of the CWG and SWG are shown in (a)-(d). The guided power in the incident waveguide is taken as 20 mW.

Download Full Size | PDF

To confirm this mechanism, we analyze the trajectories of particles that traverse the coupling region and encounter the stuck bead. The analysis is performed on the Supplemental Movie, by finding the center-of-mass of the particles in each movie frame. The results are presented in Fig. 6 for two small and two large particles. The trajectories of these particles are shown as colored lines overlaid on one frame of movie. The stuck particle is indicated by a white arrow. The traces of small particles show that they move from the CWG to the SWG at around the position of the stuck bead. After that, these particles are move along the SWG, and do not return to the CWG. This behavior is consistent with the double potential well discussed previously. For larger particles, the traces shift to positions between two waveguides near the stuck particle, which is also consistent with the potential distribution in Fig. 5(b). These particles follow the CWG as the two waveguides split at a position of ~7 µm to the right of the stuck particle, which is again consistent the deeper potential well on the CWG simulated in Fig. 5(d). The results of the particle tracking analysis are therefore consistent with the particle sorting phenomenon suggested by the analysis of simulations.

 

Fig. 6 Traces of two small and two large particles passing through the coupling region. The positions are tracked using center of mass analysis of each frame of the Supplemental Movie. The stuck particle is indicated by an arrow.

Download Full Size | PDF

As a first generation device, there are still many possibilities to improve the performances. The size of the gap between the CWG and SWG determines the upper limit on the size of the particles experiencing a double well potential. Therefore, engineering the gap size can be used to control the size threshold of the sorting process. Particles with a smaller size than this threshold could be separated from larger particles. A multilevel sorting system providing a higher size resolution could be achieved by cascading several sorting units with different size thresholds. Switching of the smaller particles requires the perturbing obstacle within the coupling region, which comprises a stuck particle in our demonstration. However, more robust alternatives could be applied; for example, Si nanoparticles fabricated by e-beam lithography and etching, or even features formed in resist by photo- or electron-beam lithography. The planar nature of this waveguide based sorting device could enable it to be integrated with other waveguide-based optofluidic devices to provide pre- or post-sorting functionalities such as buffering and sensing [22].

5. Conclusion

We demonstrate a novel sorting device based on near-field optical forces using a waveguide-based 3-dB optical splitter. In the fabricated device, 3-dB splitting of TM mode incident light is achieved via careful engineering of the coupling length. After injecting DI water containing 320 nm and 2 μm polystyrene particles to the microfluidic channel, particles are pulled onto the CWG by optical gradient forces, and propelled to the splitter by optical scattering forces, where they encounter a perturbation comprising a stuck bead. The 2 μm particles continue to be propelled along the CWG after leaving the splitter, while the 320 nm particles are transferred to the SWG. Optical force calculations based on the Maxwell stress tensor are performed to understand the sorting mechanism. In the coupling region, the smaller particles experience a double-well trapping potential. A structural perturbation can switch them from the CWG to the SWG, while the larger particles, experiencing a single potential well, are not affected by the perturbation. In the bending region, the larger particles follow the CWG, as its potential well is deeper. Position tracking analysis of particles passing through the device also provides evidence for this sorting mechanism. The planar nature of this waveguide based sorting device makes it easy to implement massive parallel processing and integrate other pre- or post-sorting functionalities into the system. Given its passive nature and ability to sort sub-micron particles, we believe that the all-optical sorting mechanism we introduce could lead to a variety of lab-on-a-chip applications in the life sciences, medical diagnosis and the environmental sciences.