We present a fabrication method for the monolithic integration of microfluidic channels into semiconductor laser material. Lasers are designed to couple directly into the microfluidic channel, allowing submerged particles pass through the output beams of the lasers. The interaction between particles in the channel and the lasers, operated in either forward or reverse bias, allows for particle detection, and the optical forces can be used to trap and move particles. Both interrogation and manipulation are made more amenable for lab-on-a-chip applications through monolithic integration. The devices are very small, they require no external optical components, have perfect intrinsic alignment, and can be created with virtually any planar configuration of lasers in order to perform a variety of tasks. Their operation requires no optical expertise and only low electrical power, thus making them suitable for computer interfacing and automation. Insulating the pn junctions from the fluid is the key challenge, which is overcome by using photo-definable SU8-2000 polymer.
©2006 Optical Society of America
The combined use of laser technology and microscopy has fuelled a revolutionary advance in cellular and molecular biology. Real-time observation and tracking of cellular processes has yielded a wealth of bioscience, but optical methods are not solely restricted to imaging. Light may also be used to trap and interrogate single cells, for instance through Raman  spectroscopy, or via the mechanical interactions of light and matter , and treatment may be conducted by using photoporation to introduce foreign DNA or drugs into individual cells . In addition, the forces exerted by light are sufficient to move, trap and manipulate biological matter ; used together, these techniques allow the creation of all-optical toolkits.
In parallel there have been advances to allow the analysis of samples within microfluidic environments, so-called ‘lab-on-a-chip’ systems, where these optical methods could play an important role. The reliance upon external laser sources and on the bulk optics that to couple their light into the sample chambers presents a problem of size and portability. However, in this paper we explore the monolithic integration of microfluidic channels and semiconductor lasers, a step that could overcome this problem and hence advance the microfluidic agenda. The microfluidic channels are created directly on the GaAs-based semiconductor material in which the lasers are fabricated. The photolithography gives intrinsic alignment to the lasers that are small enough (< 2mm × 20μm each) to allow parallel systems on a single chip. Simple electrical connections allow easy operation and even computer interfacing. The system can be viewed with any microscope system, offering the potential to conduct optical tasks without either specialist optics knowledge or dedicated trapping apparatus.
The divergent beams that naturally emerge from the semiconductor diode lasers offer certain advantages over tightly-focused beams. The low power densities are less likely to cause damaging two-photon effects in cells, and the large encatchment volumes allow them to interact with more particles, increasing the efficiency of detection in a flow of colloids. When configured to face one another two beams create a dual-beam trap that can hold large objects, and whose trapping position can be varied by simply altering the relative beam powers. Such trapping in our device is detailed elsewhere , but in this present paper the main application we present is detection. We use the lasers, in either forward or reverse bias, in a “bio-cavity laser” type operation  to detect particles and to differentiate between them by size. We believe that this fabrication process is broadly applicable, and can be tailored to satisfy specific requirements, both in the configuration of the lasers and in their output wavelengths. We have worked with two different GaAs-based materials that emit at 980nm and 1290nm, whereas GaN emits in the violet part of the spectrum and could be processed to create a device for exciting fluorescence in green fluorescence protein (GFP), a commonly used biological dye. In all manifestations, be they for manipulation or interrogation, our approach offers several distinct advantages that stem naturally from the use of integrated optics.
2. Device overview
The original concept diagram is shown in Fig. 1, which illustrates two basic processes that are envisaged: propulsion and trapping. The radiation pressure of a single, unopposed laser (E) drives particles along the channel, and the dual-beam traps (A-B and C-D) draw the particles to the beam axes and hold them at the equilibrium point between the two independent beams. These integrated lasers can also be used to detect the presence of the particles, either as they pass through the beams or once they have been trapped. The device consists of the lasers and the microfluidic channels, and these are dealt with in turn.
3. Laser structure and fabrication
The cross-section of a typical laser is shown in the SEM image in Fig. 2(a), and Figs. 2(b) and 2(c) show the arrangement of lasers around the channel. Light emission is due to electrical injection from the top and bottom contacts into ten layers of InAs quantum dots that emit at around 1290nm. The active layer is bounded by an AlGaAs cladding, 60% Al and 900nm thick on the top and a 60% Al and 1.5μm thick on the bottom. The material was grown at Nanosemiconductor GmbH, Dortmund, Germany, and all etching was done in our in-house CAIBE (Chemically-Assisted Ion Beam Etching) machine with Ar:Cl2. Horizontal waveguiding is provided by etching to a depth of 750nm on either side of a photolithographically defined 3–4μm-wide ridge; this supports only the zero-order transverse mode. A layer of SU8-2000.5 polymer provides electrical insulation between adjacent lasers. The length of the SU-8 strips is defined lithographically to be the same as that of the ridges, to prevent interference with the facet etching. The sample is heated in a 180°C oven which causes the SU8 to reflow from the ridges and then hardens the SU8 between the ridges. Any remaining residue is removed from the ridges by a simplified version of chemical-mechanical polishing. The top contacts (20nm Ni, 200nm Au) are patterned by a lift-off mask, and deposited in an electron beam evaporator. The back contact (14nm Au, 14nm Ge, 14nm Au, 11nm Ni, 200nm Au) is applied over the underside of the material. Finally, both facets of each laser are defined by CAIBE etching facets to ~2μm depth, using mask of SU8 or SR1818 photoresist [Fig. 2(b) and 2(c)]. The typical cavity length is ~1800μm, giving threshold currents of ~15mA and stable powers of ~10mW (CW) per facet at 100mA.
4. Microfluidic channels
As shown in Figs. 2 and 3 the laser facets and the deep microfluidic channel are etched separately. This is because the facets are required to be only ~2μm deep, smooth and vertical, which is best achieved by a gentle etch (beam: 1200V, 12mA; chemistry 120°C, 1.95sccm Cl2), whereas the channel must be ~15μm deep in order to accommodate large particles, which is best achieved by a faster etch whose quality is poorer (beam: 1450V, 17mA; chemistry: 150°C, 2.5sccm Cl2). Etching the deep channel requires a thicker mask, but the intense ion bombardment also denatures the polymer mask making it hard to remove afterwards. Therefore, a lift-off mask is used, in which three layers of LOR-7B are applied to cover the facets and the channel is defined in an overlying layer of SR1818, the charred remains of which are removed by dissolving the LOR-7B in its MF319 developer.
The key challenge in this work is insulating the pn junction from the fluid. This is accomplished by lining the channel with SU8-2000 polymer, as shown in Fig. 3. There are two steps, one for the base and the other for the walls. For the base, SU8-2000.5 is spin-coated and such that it is neither thick enough to fill the channel nor thin enough to leave bald patches on the base. The walls are photodefined in SU8-2050, and this thick layer of polymer also covers the front section of the lasers and contacts. The overhang (~10° negative slope) that is seen in Fig. 3(b) is a consequence of the limited resolution of the technique, but actually causes the useful effect of lifting the opposing beams upwards thus raising the point at which they overlap to form the trap and reducing the likelihood of trapped particles approaching, and adhering to, the base of the channel.
A metal layer (20nm Ni, 200nm Au) is deposited onto the base of the channel; this serves the dual purposes of apparently reducing the adhesion of particles to the base, and of improving the image contrast by reflecting. The small size (5mm × 5mm × 0.5mm) of a typical device is seen in Fig. 4, along with the following additions. To contain the fluid, a glass lid is added (1mm × 4mm × 0.2mm) that is sealed in place by dispensing Norland Optical Adhesive (NOA) 71 around the edges of the lid. The device is mounted on a printed circuit board (PCB) onto which wire-bonds are made; electrical connections can be soldered from the PCB to a power supply and meters. The circuit board also provides support for the capillaries via which fluid is introduced into the channel. The joints are sealed with the NOA 71, and the outer ends are attached, via silicone tubing, to whatever pump, syringe or reservoir is desired. A brass block provides heat sinking from the back of the GaAs.
5. Performance and detection
The critical device characteristic is the laser power that enters the microfluidic channel. This power cannot be measured directly, and must be inferred from the power emitted from the outer facets. However, these two powers are not equal due to the SU8 coating on the inner facets. If the reflectivities of the bare and coated facets are R1 and R2, respectively, and the power of the mode within the GaAs cavity just before it reaches the bare facet is Po then it is the transmitted fraction, (1-R1) Po, that can be measured from a device. The reflected fraction, R1Po, grows exponentially as it travels towards the SU8-coated facet. By equating the facet losses and the round-trip gain, the total growth in one length of the cavity is (R1R2)-1/2 so the power that is transmitted into the SU8 coating is Po R1 (R1R2)-1/2(1-R2), and the ratio of the powers that enter into the air and SU8 is given by
Refractive indices of 1.0, 1.55 and 3.4 for air, SU8-2000 and GaAs, respectively, at near-normal incidence, give values of R1 = 0.30 and R2 = 0.18, and a predicted ratio of 1.79. Of the light that enters the SU8 coating, a certain fraction is reflected at the SU8-fluid interface and therefore does not enter the trapping channel. However, the non-vertical slope of the SU8 [Fig. 3(b)] means that this reflected fraction does not efficiently feed back into the laser but is ‘lost’ from the system. At near-normal incidence this fraction is ~4.7% for air and ~0.6% for water of refractive index 1.33, and the corrected ratios are 1.71 and 1.78, respectively. This was tested experimentally by cleaving a device along the channel and measuring the output powers from both ends of several lasers. Figure 5(a) shows the data for a typical laser before and after the SU8 is added to one facet. Due to decreased overall reflectivity, the total power emitted from both facets of a device laser is actually 88±1% that of a normal laser, but the helpful effect is that the power coupled through the SU8 into the channel is higher than the output of a normal laser. Figure 5(b) shows the ratios of the output powers, measured at two different currents, for nine lasers before and after the addition of the SU8 to one facet, with air beyond the SU8. The slope for the normal lasers is 0.99 +/- 0.02, which is close to the expected value of 1, and the slope for the device lasers is 1.63 +/- 0.09, which agrees with the predicted value of 1.71. It is not possible to make accurate measurements with water beyond the SU8, but the agreement in the case of air lends confidence to the predicted ratio of 1.78, which is used to estimate the powers that enter the channel.
Our device allows the exploitation of intrinsic properties of the semiconductor material to access interesting detection methods. These are distinct from those that use additional apparatus to measure some property, such as fluorescence signals. The first method uses the fact that, although the sloped SU8 walls refract the beams upwards, some of the light from each laser will enter into the facing laser, affecting its performance. We find that the presence of a particle in the beam path blocks some of this mutual feedback, thereby reducing the power emitted from the outer facets of each laser by approximately 0.1-0.2mW, depending upon the size of the particle. Figure 6(a) shows that the power drop depends roughly upon the laser currents, which is because higher currents provide more power at the edges of the beams that can bypass the particle. The second method uses one laser in forward bias; the second laser is normally forward-biased to maintain the trap, but is periodically switched to reverse bias, when it acts as a photodetector. The measured photocurrent is of the order of 100μA, and is reduced by the presence of a particle in the beam path. A larger particle blocks more of the light, giving a larger reduction in the photocurrent, as seen in Fig. 6(b). This second method requires no external optical apparatus, and only the simple electrical connections that are already present, making it a truly integrated, size-specific detection technique.
6. Conclusions and further work
We have presented a novel fabrication technique that integrates microfluidic channels onto semiconductor laser material so that the lasers beams couple directly into the flow channel. The key step that has been demonstrated is the ability to insulate the pn junction from the fluid, as well as combining high-quality facet etching with deep etching for microfluidic channels. The polymer insulation has the useful consequence of coupling more power into the channel, providing larger forces for manipulation. The ability to detect particles and to discriminate between different sizes using only the intrinsic properties of the lasers could be fine-tuned for accurate detection  and opens up the possibility of entirely integrated optical manipulation devices. The device requires no optical alignment, it is compact and portable, and can easily be interfaced to a computer and automated. It could make optical manipulation accessible to a wider scientific audience, with particular applications in lab-on-a-chip systems for biological testing.
We acknowledge the financial support of the U.K. Engineering and Physical Sciences Research Council and the European Union fp6 NEST programme “ATOM_3D”.
References and links
1. P. R. T. Jess, V. Garcés-Chávez, D. Smith, M. Mazilu, L. Paterson, A. Riches, C. S. Herrington, W. Sibbett, and K. Dholakia, “Dual beam fibre trap for Raman micro-spectroscopy of single cells,” Opt. Exp. , 14, 5779–5791 (2006). [CrossRef]
2. P. L. Gourley, “Biocavity laser for high-speed cell and tumour biology,” J. Phys. D: Appl. Phys. 36, R228–239(2003) [CrossRef]
3. L. Paterson, B. Agate, M. Comrie, R. Ferguson, T. Lake, J. Morris, A. Carruthers, C. T. Brown, W. Sibbett, P. Bryant, F. Gunn-Moore, A. Riches, and Kishan Dholakiaet, “Photoporation and cell transfection using a violet diode laser,” Opt. Exp. , 13, 595–600 (2005). [CrossRef]
5. S. Cran-McGreehin, T. Krauss, and K. Dholakia, “Integrated monolithic optical manipulation,” Lab-on-a-Chip, DOI: 10.1039/b605237a (2006).