Lab-on-a-chip with integrated optical transducers - Semantic Scholar
PAPER
www.rsc.org/loc | Lab on a Chip
Lab-on-a-chip with integrated optical transducers S. Balslev, A. M. Jorgensen, B. Bilenberg, K. B. Mogensen, D. Snakenborg, O. Geschke, J. P. Kutter and A. Kristensen* Received 6th September 2005, Accepted 6th December 2005 First published as an Advance Article on the web 22nd December 2005 DOI: 10.1039/b512546d Taking the next step from individual functional components to higher integrated devices, we present a feasibility study of a lab-on-a-chip system with five different components monolithically integrated on one substrate. These five components represent three main domains of microchip technology: optics, fluidics and electronics. In particular, this device includes an on-chip optically pumped liquid dye laser, waveguides and fluidic channels with passive diffusive mixers, all defined in one layer of SU-8 polymer, as well as embedded photodiodes in the silicon substrate. The dye laser emits light at 576 nm, which is directly coupled into five waveguides that bring the light to five different locations along a fluidic channel for absorbance measurements. The transmitted portion of the light is collected at the other side of this cuvette, again by waveguides, and finally detected by the photodiodes. Electrical read-out is accomplished by integrated metal connectors. To our knowledge, this is the first time that integration of all these components has been demonstrated.
1. Introduction Optical techniques play a central role in chemical and biochemical analysis and it is attractive to adopt these in lab-on-a-chip microsystems.1,2 During the past decade, a wide range of such microsystems have been realized, based on external light sources and photo-detectors.3 However, the coupling of optical signals in and out of the device is one of the major challenges in integrated optics. Pig-tail coupling of optical fibers to the microchip is a widely used approach,4–8 but the assembly of such systems is expensive, and coupling losses are a significant problem. Another labor intensive approach is to hybridize micro optical systems with laser diodes, requiring precise handling of very small laser devices.9 In contrast, if the light source and the other optical and fluidic components are defined in a single lithographic step, the components are already aligned. Avoiding gluing technology and achieving direct integration of optical transducers is therefore an attractive approach in the field of lab-on-a-chip.10 This article focuses on the integration of an assembly of discrete components representing three different domains: optics, fluidics and electronics. The integration technology is based on polymer components defined in a thin polymer film deposited on top of a functionalized silicon substrate. We have chosen to work with the photoresist SU-8 from Microchem for the polymer components.11–13 This resist is transparent, has a high chemical resistance and is easy to pattern using UV lithography. Since the focus is integration, details regarding the individual components should be found in the references. Preliminary absorbance measurements serve to demonstrate the concerted operation of the integrated functional elements.
MIC – Department of Micro and Nanotechnology Technical University of Denmark (DTU) Oersteds Plads, Building 345 east, DK-2800, Kongens Lyngby, Denmark. E-mail: [email protected]
This journal is ß The Royal Society of Chemistry 2006
Whereas the first microfluidic lab-on-a-chip devices were based on silicon micro-machining,1 polymers have in general become the preferred material.14 Polymers are available with high chemical resistance and good optical properties. They offer a range of possibilities for functionalization and are cheap to process. Since the surface of a silicon wafer with embedded photodiodes defined by selective doping is still essentially flat, it is easy to apply a spin process and a lithographic step on a polymer thin film and thereby obtain integration of polymer components and silicon components. For clarity the following sections are divided into a description of the whole chip system, the individual components, fabrication and preliminary functionality demonstration.
2. System The lab-on-chip system we have fabricated contains active and passive optics, the possibility of integrating on-chip chemistry, microfluidic handling and optoelectronics. It consists of a microfluidic polymer dye laser delivering light to five polymer waveguides that direct the light to five measuring points along a cuvette15,16 (see Fig. 1). On the other side of the cuvette, five waveguides pick up the transmitted light and direct it to a photodiode array, enabling spatially or temporally resolved measurements. The electrical signal from the photodiodes is used for detection and can be read out from metallic pads on the chip. There is a mixer in front of the microfluidic channel forming the cuvette, making it possible to mix two fluids and measure the absorption of the result. Fluids are supplied to the channels in the chip through holes in the glass lid. The complete micro-system with laser, waveguides, microfluidics and photodiodes has a footprint of 15 mm by 20 mm. Fig. 2 shows the cross sectional view of the system. The structure consists of a 500 mm silicon substrate, a 3 mm silicon dioxide layer, a 10 mm SU-8 layer, a 5 mm polymethyl methacrylate (PMMA) layer and a 500 mm borofloat glass lid. Lab Chip, 2006, 6, 213–217 | 213
Fig. 1 Photograph of the lab-on-chip device with integrated microfluidic dye laser, optical waveguides, microfluidic network and photodiodes. The metallic contact pads for the photodiodes are seen on the far right. The chip footprint is 15 mm by 20 mm. The photograph was taken before a lid was bonded to the structures.
Fig. 2 Outline of the cross section of the lab-on-chip device. Photodiodes are embedded in the silicon substrate (right), while optical waveguides and the microfluidic network are defined in the 10 mm thick SU-8 film. The channels are sealed off by a borofloat glass lid bonded to the SU-8 by means of a thin PMMA film.
The SU-8 layer forms the core of a planar waveguide, as it is sandwiched between materials of lower refractive index (nPMMA = 1.49, nSiO2 = 1.46, nSU-8 = 1.6). The thickness of the SU-8 layer is a compromise between the laser architecture and functionality and the fluidic resistance posed by the microfluidic network. Rectangular multimode waveguides are formed in the SU-8 by defining 30 mm wide strips of SU-8 (see Fig. 3). The microfluidic channels are also defined in the SU-8 layer and are sealed by the PMMA-bonded glass lid. The PMMA and SU-8 layers thus have dual purposes (optical and fluidic), which has been important for our choice of these materials. There is a 70 nm silicon nitride layer between the photodiode silicon surface and the SU-8 (not shown in Fig. 2). The purpose of this layer is to electrically passivate the front surface of the photodiodes as well as insulate the structures electrically. In order to facilitate coupling of light from the SU-8 waveguides into the photodiodes, the silicon dioxide layer has been removed at the photoactive part of the photodiodes (Fig. 2 right). The light in the waveguides will couple to the photodiode via leaky mode coupling.17 The underside of the silicon substrate contains a common Al electrode for the photodiodes.
3. Components The laser light source is based on a high order distributed feedback grating containing the laser dye Rhodamine 6G 214 | Lab Chip, 2006, 6, 213–217
dissolved in ethanol at a concentration of 20 mM15 (Fig. 3). The size of the laser is 1 mm by 1 mm and it is embedded in a 1 mm wide fluidic channel. Both the channel and the grating are defined in the 10 mm SU-8 layer. During operation the laser dye solution is supplied through the channel at a flow rate of 1 to 10 mL h21 and the dye is optically pumped by an external pulsed light source. The laser emits a parallel beam of light at 576 nm into the SU-8 layer slab waveguide. Five rectangular waveguides16 are defined in the SU-8 and a part of the slab wave is coupled into the individual waveguides (Fig. 3). The waveguides are tapered toward the laser.18 The propagation loss for this kind of SU-8 waveguides was measured in16 to 2.5 dB cm21 at 576 nm. This loss is acceptable since the waveguide length on the chip is short, close to 1 cm. The design of the waveguides is the result of a compromise between curvature, length and core width (the height is fixed by the SU-8 layer thickness). The following conflicting interests have been taken into account: The wider the waveguide core, the more light is picked up from the laser, but the wider the core width, the larger the radius of curvature needs to be in order to avoid bending losses. The larger the radius of curvature is, the longer the waveguides need to be, in order to distribute the light along the cuvette, but long waveguides increase the propagation loss and size of the system. We have settled for a core width of 30 mm, a maximum radius of curvature of 10 mm and a waveguide length of approximately 10 mm. The waveguides terminate 25 mm before the cuvette channel and continue on the other side with a similar buffer distance to the cuvette (Fig. 3).19 The waveguides direct the light they have picked up on the far side of the cuvette to the 2 mm long active area of the photodiodes (Fig. 3). Apart from the channel delivering laser dye solution to the laser, the microfluidic system on the chip consists of a mixer that functions by diffusion20 and a cuvette for the absorption measurement (Fig. 4). The two mixer inlet channels are 50 mm wide and 8 mm long and yield a fluidic resistance of 22 6 1015 Pa s m23 for water. The mixing meander where the two fluids meet and mix by diffusion is 100 mm wide and 5.7 mm long. The meander yields a fluidic resistance of 6 6 1015 Pa s m23 and the 500 mm wide cuvette has a fluidic resistance of 1 6 1015 Pa s m23. The mixer requires only small fluid amounts and is designed for flow rates of 1 mL h21. The diodes are 50 mm wide and 2 mm long and are designed for light with a wavelength around 600 nm, close to the dye laser wavelength. The design and fabrication involves the reduction of the photodiode quantum efficiency for both longer and shorter wavelengths than 600 nm in order to optimize performance for the system. The photodiodes are planar n-type diodes embedded in a p-type substrate. Because boron is used to dope the substrate, care must be taken during processing since boron will tend to segregate into silicon dioxide coatings.21 This may lead to surface inversion during operation and thus poor performance. To avoid this, both surfaces are doped to a higher level than the bulk during fabrication. The backside is doped to a high level to allow for Ohmic contact with minimal contact resistance on this side. This journal is ß The Royal Society of Chemistry 2006
Fig. 3 (a) Electron microscope image close-up of a section of a waveguide. The light is guided in the central strip of SU-8. Inset: drawing of waveguide structure (not to scale). (b) Electron microscope image of the microfluidic dye laser (center) and the beginning of the waveguides (lower left). The laser structure is situated in a 1 mm wide channel carrying the liquid gain media. (c) Electron microscope image of the termination of a waveguide at the side of the cuvette, forming a trough towards the cuvette. (d) Electron microscope image of the waveguide to photodiode coupling region. The waveguides picking up the light from the cuvette (upper right) directs the light to the photoactive region making up the last 2 mm under the waveguides which then terminates.
higher resistivity material. The result is that the bulk diffusion length becomes smaller and this reduces the long wavelength response since minority electrons generated deep in the bulk will not reach the junction before recombining.
4. Fabrication
Fig. 4 Design of the microfluidic mixer. There are two inlet holes connected to the inlet arms that meet at the beginning of the mixing meander. The meander opens up to form the cuvette (top right). The curved double line is the waveguide closest to the beginning of the cuvette.
The junction depth is fairly large, approximately 1 mm, with a high surface dopant concentration. Due to the top surface being highly doped the hole mobility in the first 50 nm is very low. This means that the vast majority of photo-electrically generated electron–hole pairs in this region will recombine and not contribute to the photocurrent of the diode, forming a ‘dead layer’.22 As a result the response from short wavelength light is suppressed. The doping level of the substrate is around 1015 cm23 of boron. This is chosen for two reasons. First of all, it lowers the diode saturation current and thus noise.23 Secondly, it means that the minority carrier lifetime in the bulk is lower than for This journal is ß The Royal Society of Chemistry 2006
The fabrication of the photodiodes makes use of p-type boron doped (1015 cm23), 100 mm diameter float zone (100) silicon wafers. First, the wafers had alignment marks transferred using 100 nm of silicon dioxide using an LPCVD TEOS process, patterning by UV-lithography and subsequent etching using hydrofluoric acid (HF) and potassium hydroxide (KOH). After the KOH etch the silicon dioxide was stripped in HF and with the wafers still wet from the post-HF rinse they were RCA cleaned. This step is crucial to avoid a drop in the minority carrier lifetime due to iron and copper precipitates on the wafers.24 Both sides of the wafers were then coated with PECVD silicon dioxide. On the front surface there was a boron doped silica glass closest to the surface and it was capped with a layer of undoped silicon dioxide (1 mm in total). The backside was covered with another boron-doped glass with an order of magnitude higher boron concentration than the front surface layer. This layer was also capped with an undoped layer bringing the total thickness of the back side layer to 3 mm. The front side layer was then patterned with the contact doping areas using UV-lithography and HF etching. The wafers were Lab Chip, 2006, 6, 213–217 | 215
then cleaned and the contact doping was performed using POCl3 at 950 uC for 20 min followed by 20 min of in situ annealing. This high temperature step also meant that the boron doping from the surface layers was partly transferred to the silicon surface. Following this the frontside was patterned with the photosensitive areas (emitters) again using UV-lithography and HF etching. The emitters were doped using POCl3 at 850 uC for 20 min with a 20 min in situ anneal. At this point the PECVD glass was etched from the surface and the wafers were RCA cleaned. The front surface was then given a new coating of PECVD silicon dioxide to act as a buffer for the waveguiding layer. This layer was then patterned with the waveguide to photodiode couplers, which were realized by HF etching of the buffer layer. In fact, all areas used for emitters/contacts or leads to the diodes were etched clean. The wafers were then given a coating of 70 nm high quality PECVD silicon nitride. This layer serves as electrical passivation of the front surface as well as electrical insulator. Contact holes were etched through the nitride with HF using a UV lithography pattern. The front surface was patterned with an image reversed photoresist pattern for lift-off of the front side metal. Then the metallization was applied. The front side received a coating of titanium and aluminum with thicknesses of 20 and 100 nm, respectively. The backside was coated with 100 nm aluminum. Following this, the front side pattern was finalized by lift-off. Finally, the wafers were annealed in forming gas (4% H2 in N2) at 450 uC for 20 min. The 10 mm layer of SU-8 was deposited via a spin process and soft baked at 95 uC for 2 min with slow cooling. The SU-8 was UV exposed through the lithographic mask with 550 mJ cm22 and post-exposure baked at 95 uC for 25 min with slow cooling. The SU-8 was developed for 4 min in propylene glycol monomethyl ether acetate (PGMEA) and rinsed in isopropyl alcohol (IPA). In order to get rid of cracks formed in the SU-8 thin film during processing, the cracks were healed by placing the wafer on a 160 uC hotplate for 60 s with rapid removal.25 The silicon wafer containing photodiodes and the SU-8 polymer layer was bonded to a 0.5 mm thick glass wafer via PMMA mediated bonding.26 The glass wafer was prepared with a 5 mm thick layer of PMMA and the two wafers were bonded in a custom bonder. We have used a bonding force of 2000 N applied on the 40 wafers for 10 min and a bonding temperature of 150 uC with slow cooling to 65 uC before the pressure was released. The finished bonded wafer was diced into chips and holes were drilled using a diamond drill. In the dicing process the glass above the metallic contact pads on the chip was removed by dicing half-way through the glass and breaking it off manually.
for the optical pumping. The Nd:YAG laser emitted 5 ns pulses at 532 nm at a frequency of 10 Hz and the pulse energy density reaching the on-chip dye laser was 80 mJ mm22 during characterization. To operate the laser we injected a 20 mM solution of Rhodamine 6G in ethanol at a rate of 10 mL/hr in the laser channel. The spectrum of the emitted laser light can be seen in Fig. 5. The response of the photodiodes is illustrated in Fig. 6. During measurement, the photodiodes were coupled in parallel with a 1 nF capacitor and a 1 kV resistor in order to measure the charge generated with each laser pulse.27 We measured the noise equivalent power of the diodes to 7.09 6 10213 W (10 Hz bandwidth) and the sensitivity to 390 mA W21 (at 580 nm). The clear indication of the lasing threshold at 14 mW (corresponding to a pump energy density on the dye laser of 50 mJ mm22) in the graph in Fig. 6 demonstrates a low influence of scattered 532 nm pump light on the photodiode signal. The inset of the figure shows the diode response to a single laser pulse.
Fig. 5
Emission spectrum from the embedded liquid dye laser.
5. Evaluation of functionality For testing, the chip was mounted in a polycarbonate sample holder with electrical and fluidic connections. The electrical connection was provided by spring loaded probes that made contact to the connector pads on the chip. A window in the polycarbonate lid gave optical access to the on-chip dye laser for the pulsed frequency doubled Nd:YAG laser that was used 216 | Lab Chip, 2006, 6, 213–217
Fig. 6 Photodiode response as function of optical pump power of the laser pumping the on-chip dye laser. The inset shows the open circuit response from a photodiode to a single pump pulse (vertical 10 mV division21, horizontal 10 ms division21).
This journal is ß The Royal Society of Chemistry 2006
In a first demonstration of the concerted operation of the integrated components (laser, wave-guides, microfluidic cuvette and photodiodes), we performed an absorbance measurement on two different concentrations of xylenol orange dye.28 Before each particular xylenol orange dye solution was injected into the cuvette, a reference measurement with pure water was performed. The xylenol orange solution was diluted in an ammonia buffer (pH = 10) with Ca2+ ions. The photodiode signal, I, for different xylenol orange dye solutions, was referenced to the photodiode signal, I0, for pure water in three tests. As expected, a test with pure water and pure ammonia buffer gave essentially the same signal readout on the photodiodes (so I/I0 # 1). A xylenol orange dye concentration of 0.06 mM gave I/I0 = 0.7 and a concentration of 0.12 mM gave I/I0 = 0.3. Although this examination of the chip is not sufficient to determine the limit of detection and the linear range of the device, it proves the successful integration of the individual components. In summary, we have demonstrated a fabrication platform for integration of five functional components (liquid dye laser, waveguides, fluidic channels, measurement cuvettes and photodiodes). Higher integration is much more than just a demonstration of fabrication skills: it improves our understanding of how lab-on-a-chip devices will evolve in the coming years. In the future, the presented device will be tested for more application-oriented problems.
Acknowledgements This work was supported by the Danish Technical Research Council (STVF) grant no.: 26-02-0064, grant no. 26-00-0220 and the H. C. Ørsted Foundation.
References 1 A. Manz, N. Graber and H. M. Widmer, Sens. Actuators B, 1990, 1, 244–248. 2 S. Verpoorte and N. F. De Rooij, Proc. IEEE, 2003, 91, 930.
This journal is ß The Royal Society of Chemistry 2006
3 K. Uchiyama, H. Nakajima and T. Hobo, Anal. Bioanal. Chem., 2004, 379, 375. 4 J. Hu¨bner, K. B. Mogensen, A. M. Jorgensen, P. Friis, P. Telleman and J. P. Kutter, Rev. Sci. Instrum., 2001, 72, 229. 5 H. P. Fischer, K. Peters, R. Ziegler, D. Pech, A. Kilk, Th. Eckhardt, G. G. Mekonnen and G. Jacumeit, Opt. Fiber Technol., 2000, 6, 1, 68. 6 I. Hardy, P. Grosso and D. Bosc, Photon. Technol. Lett., IEEE, 2005, 17, 5, 1028. 7 J. N. McMullin, Proc. SPIE, 2000, 4087, 1050. 8 L. Eldada, K. M. T. Stengel, L. W. Shacklette, R. A. Norwood, C. Xu, C. Wu and J. Yardley, Proc. SPIE, 1997, 3006, 344. 9 N. M. Jokerst, M. A. Brooke, S. Cho, M. Thomas, J. Lillie, D. Kim, S. Ralph, K. Dennis, B. Comeau and C. Henderson, Proc. SPIE, 2005, 5730, 226–233. 10 S. Verpoorte, Lab Chip, 2003, 3, 42N. 11 SU-8 10 from MicroChem Corp., Newton, MA. www.microchem. com. 12 K. Y. Lee, N. LaBianca, S. Zolgharnain, S. A. Rishton, J. D. Gelorme, J. M. Shaw and T. H. P. Chang, J. Vac. Sci Technol. B, 1995, 13, 3012. 13 H. Lorenz, M. Despont, N. Fahrni, N. LaBianca, P. Renaudy and P. Vettiger, J. Micromech. Microeng, 1997, 7, 121. 14 H. Klank, J. Kutter and O. Geschke, Lab Chip, 2002, 2, 242. 15 S. Balslev and A. Kristensen, Opt. Express, 2005, 13, 344. 16 K. B. Mogensen, J. El-Ali, A. Wolff and J. P. Kutter, Appl. Opt., 2003, 89, 19, 4072. 17 M. Nathan, O. Levi, I. Goldfarb and A. Ruzin, J. Appl. Phys., 2003, 94, 12, 7932. 18 Yi-Fan Li and John W. Y. Lit, J. Opt. Soc. Am. A, 1985, 2, 3, 462. 19 P. Friis, K. Hoppe, O. Leistiko, K. B. Mogensen, J. Hu¨bner and J. P. Kutter, Appl. Opt., 2001, 40, 6246. 20 T. T. Veenstra, T. S. J. Lammerink, M. C. Elwenspoek and A. van den Berg, J. Micromech. Microeng., 1999, 9, 199. 21 S. M. Sze, in Semiconductor Devices. Physics and Technology, John Wiley & Sons, New York, 1985. 22 M. A. Green, in High Efficiency Silicon Solar Cells, Trans Tech Publications Ltd, Aedermannsdorf, 1987. 23 M. Razeghi and A. Rogalski, J. Appl. Phys., 1996, 79, 7433. 24 C. B. Nielsen, C. Christensen, C. Pedersen and E. V. Thomsen, J. Electrochem. Soc., 2004, 151G338. 25 Crack healing process from personal communication with MicroChem Corp., Newton, MA. www.microchem.com. 26 B. Bilenberg, T. Nielsen, B. Clausen and A. Kristensen, J. Micromech. Microeng., 2004, 14, 814. 27 B. R. Clemesha, J. Phys. E, 1972, 5, 859. 28 C. Gay, J. Collins and J. M. Gebicki, Anal. Biochem., 1999, 273, 2, 143.
Lab Chip, 2006, 6, 213–217 | 217
www.rsc.org/loc | Lab on a Chip
Lab-on-a-chip with integrated optical transducers S. Balslev, A. M. Jorgensen, B. Bilenberg, K. B. Mogensen, D. Snakenborg, O. Geschke, J. P. Kutter and A. Kristensen* Received 6th September 2005, Accepted 6th December 2005 First published as an Advance Article on the web 22nd December 2005 DOI: 10.1039/b512546d Taking the next step from individual functional components to higher integrated devices, we present a feasibility study of a lab-on-a-chip system with five different components monolithically integrated on one substrate. These five components represent three main domains of microchip technology: optics, fluidics and electronics. In particular, this device includes an on-chip optically pumped liquid dye laser, waveguides and fluidic channels with passive diffusive mixers, all defined in one layer of SU-8 polymer, as well as embedded photodiodes in the silicon substrate. The dye laser emits light at 576 nm, which is directly coupled into five waveguides that bring the light to five different locations along a fluidic channel for absorbance measurements. The transmitted portion of the light is collected at the other side of this cuvette, again by waveguides, and finally detected by the photodiodes. Electrical read-out is accomplished by integrated metal connectors. To our knowledge, this is the first time that integration of all these components has been demonstrated.
1. Introduction Optical techniques play a central role in chemical and biochemical analysis and it is attractive to adopt these in lab-on-a-chip microsystems.1,2 During the past decade, a wide range of such microsystems have been realized, based on external light sources and photo-detectors.3 However, the coupling of optical signals in and out of the device is one of the major challenges in integrated optics. Pig-tail coupling of optical fibers to the microchip is a widely used approach,4–8 but the assembly of such systems is expensive, and coupling losses are a significant problem. Another labor intensive approach is to hybridize micro optical systems with laser diodes, requiring precise handling of very small laser devices.9 In contrast, if the light source and the other optical and fluidic components are defined in a single lithographic step, the components are already aligned. Avoiding gluing technology and achieving direct integration of optical transducers is therefore an attractive approach in the field of lab-on-a-chip.10 This article focuses on the integration of an assembly of discrete components representing three different domains: optics, fluidics and electronics. The integration technology is based on polymer components defined in a thin polymer film deposited on top of a functionalized silicon substrate. We have chosen to work with the photoresist SU-8 from Microchem for the polymer components.11–13 This resist is transparent, has a high chemical resistance and is easy to pattern using UV lithography. Since the focus is integration, details regarding the individual components should be found in the references. Preliminary absorbance measurements serve to demonstrate the concerted operation of the integrated functional elements.
MIC – Department of Micro and Nanotechnology Technical University of Denmark (DTU) Oersteds Plads, Building 345 east, DK-2800, Kongens Lyngby, Denmark. E-mail: [email protected]
This journal is ß The Royal Society of Chemistry 2006
Whereas the first microfluidic lab-on-a-chip devices were based on silicon micro-machining,1 polymers have in general become the preferred material.14 Polymers are available with high chemical resistance and good optical properties. They offer a range of possibilities for functionalization and are cheap to process. Since the surface of a silicon wafer with embedded photodiodes defined by selective doping is still essentially flat, it is easy to apply a spin process and a lithographic step on a polymer thin film and thereby obtain integration of polymer components and silicon components. For clarity the following sections are divided into a description of the whole chip system, the individual components, fabrication and preliminary functionality demonstration.
2. System The lab-on-chip system we have fabricated contains active and passive optics, the possibility of integrating on-chip chemistry, microfluidic handling and optoelectronics. It consists of a microfluidic polymer dye laser delivering light to five polymer waveguides that direct the light to five measuring points along a cuvette15,16 (see Fig. 1). On the other side of the cuvette, five waveguides pick up the transmitted light and direct it to a photodiode array, enabling spatially or temporally resolved measurements. The electrical signal from the photodiodes is used for detection and can be read out from metallic pads on the chip. There is a mixer in front of the microfluidic channel forming the cuvette, making it possible to mix two fluids and measure the absorption of the result. Fluids are supplied to the channels in the chip through holes in the glass lid. The complete micro-system with laser, waveguides, microfluidics and photodiodes has a footprint of 15 mm by 20 mm. Fig. 2 shows the cross sectional view of the system. The structure consists of a 500 mm silicon substrate, a 3 mm silicon dioxide layer, a 10 mm SU-8 layer, a 5 mm polymethyl methacrylate (PMMA) layer and a 500 mm borofloat glass lid. Lab Chip, 2006, 6, 213–217 | 213
Fig. 1 Photograph of the lab-on-chip device with integrated microfluidic dye laser, optical waveguides, microfluidic network and photodiodes. The metallic contact pads for the photodiodes are seen on the far right. The chip footprint is 15 mm by 20 mm. The photograph was taken before a lid was bonded to the structures.
Fig. 2 Outline of the cross section of the lab-on-chip device. Photodiodes are embedded in the silicon substrate (right), while optical waveguides and the microfluidic network are defined in the 10 mm thick SU-8 film. The channels are sealed off by a borofloat glass lid bonded to the SU-8 by means of a thin PMMA film.
The SU-8 layer forms the core of a planar waveguide, as it is sandwiched between materials of lower refractive index (nPMMA = 1.49, nSiO2 = 1.46, nSU-8 = 1.6). The thickness of the SU-8 layer is a compromise between the laser architecture and functionality and the fluidic resistance posed by the microfluidic network. Rectangular multimode waveguides are formed in the SU-8 by defining 30 mm wide strips of SU-8 (see Fig. 3). The microfluidic channels are also defined in the SU-8 layer and are sealed by the PMMA-bonded glass lid. The PMMA and SU-8 layers thus have dual purposes (optical and fluidic), which has been important for our choice of these materials. There is a 70 nm silicon nitride layer between the photodiode silicon surface and the SU-8 (not shown in Fig. 2). The purpose of this layer is to electrically passivate the front surface of the photodiodes as well as insulate the structures electrically. In order to facilitate coupling of light from the SU-8 waveguides into the photodiodes, the silicon dioxide layer has been removed at the photoactive part of the photodiodes (Fig. 2 right). The light in the waveguides will couple to the photodiode via leaky mode coupling.17 The underside of the silicon substrate contains a common Al electrode for the photodiodes.
3. Components The laser light source is based on a high order distributed feedback grating containing the laser dye Rhodamine 6G 214 | Lab Chip, 2006, 6, 213–217
dissolved in ethanol at a concentration of 20 mM15 (Fig. 3). The size of the laser is 1 mm by 1 mm and it is embedded in a 1 mm wide fluidic channel. Both the channel and the grating are defined in the 10 mm SU-8 layer. During operation the laser dye solution is supplied through the channel at a flow rate of 1 to 10 mL h21 and the dye is optically pumped by an external pulsed light source. The laser emits a parallel beam of light at 576 nm into the SU-8 layer slab waveguide. Five rectangular waveguides16 are defined in the SU-8 and a part of the slab wave is coupled into the individual waveguides (Fig. 3). The waveguides are tapered toward the laser.18 The propagation loss for this kind of SU-8 waveguides was measured in16 to 2.5 dB cm21 at 576 nm. This loss is acceptable since the waveguide length on the chip is short, close to 1 cm. The design of the waveguides is the result of a compromise between curvature, length and core width (the height is fixed by the SU-8 layer thickness). The following conflicting interests have been taken into account: The wider the waveguide core, the more light is picked up from the laser, but the wider the core width, the larger the radius of curvature needs to be in order to avoid bending losses. The larger the radius of curvature is, the longer the waveguides need to be, in order to distribute the light along the cuvette, but long waveguides increase the propagation loss and size of the system. We have settled for a core width of 30 mm, a maximum radius of curvature of 10 mm and a waveguide length of approximately 10 mm. The waveguides terminate 25 mm before the cuvette channel and continue on the other side with a similar buffer distance to the cuvette (Fig. 3).19 The waveguides direct the light they have picked up on the far side of the cuvette to the 2 mm long active area of the photodiodes (Fig. 3). Apart from the channel delivering laser dye solution to the laser, the microfluidic system on the chip consists of a mixer that functions by diffusion20 and a cuvette for the absorption measurement (Fig. 4). The two mixer inlet channels are 50 mm wide and 8 mm long and yield a fluidic resistance of 22 6 1015 Pa s m23 for water. The mixing meander where the two fluids meet and mix by diffusion is 100 mm wide and 5.7 mm long. The meander yields a fluidic resistance of 6 6 1015 Pa s m23 and the 500 mm wide cuvette has a fluidic resistance of 1 6 1015 Pa s m23. The mixer requires only small fluid amounts and is designed for flow rates of 1 mL h21. The diodes are 50 mm wide and 2 mm long and are designed for light with a wavelength around 600 nm, close to the dye laser wavelength. The design and fabrication involves the reduction of the photodiode quantum efficiency for both longer and shorter wavelengths than 600 nm in order to optimize performance for the system. The photodiodes are planar n-type diodes embedded in a p-type substrate. Because boron is used to dope the substrate, care must be taken during processing since boron will tend to segregate into silicon dioxide coatings.21 This may lead to surface inversion during operation and thus poor performance. To avoid this, both surfaces are doped to a higher level than the bulk during fabrication. The backside is doped to a high level to allow for Ohmic contact with minimal contact resistance on this side. This journal is ß The Royal Society of Chemistry 2006
Fig. 3 (a) Electron microscope image close-up of a section of a waveguide. The light is guided in the central strip of SU-8. Inset: drawing of waveguide structure (not to scale). (b) Electron microscope image of the microfluidic dye laser (center) and the beginning of the waveguides (lower left). The laser structure is situated in a 1 mm wide channel carrying the liquid gain media. (c) Electron microscope image of the termination of a waveguide at the side of the cuvette, forming a trough towards the cuvette. (d) Electron microscope image of the waveguide to photodiode coupling region. The waveguides picking up the light from the cuvette (upper right) directs the light to the photoactive region making up the last 2 mm under the waveguides which then terminates.
higher resistivity material. The result is that the bulk diffusion length becomes smaller and this reduces the long wavelength response since minority electrons generated deep in the bulk will not reach the junction before recombining.
4. Fabrication
Fig. 4 Design of the microfluidic mixer. There are two inlet holes connected to the inlet arms that meet at the beginning of the mixing meander. The meander opens up to form the cuvette (top right). The curved double line is the waveguide closest to the beginning of the cuvette.
The junction depth is fairly large, approximately 1 mm, with a high surface dopant concentration. Due to the top surface being highly doped the hole mobility in the first 50 nm is very low. This means that the vast majority of photo-electrically generated electron–hole pairs in this region will recombine and not contribute to the photocurrent of the diode, forming a ‘dead layer’.22 As a result the response from short wavelength light is suppressed. The doping level of the substrate is around 1015 cm23 of boron. This is chosen for two reasons. First of all, it lowers the diode saturation current and thus noise.23 Secondly, it means that the minority carrier lifetime in the bulk is lower than for This journal is ß The Royal Society of Chemistry 2006
The fabrication of the photodiodes makes use of p-type boron doped (1015 cm23), 100 mm diameter float zone (100) silicon wafers. First, the wafers had alignment marks transferred using 100 nm of silicon dioxide using an LPCVD TEOS process, patterning by UV-lithography and subsequent etching using hydrofluoric acid (HF) and potassium hydroxide (KOH). After the KOH etch the silicon dioxide was stripped in HF and with the wafers still wet from the post-HF rinse they were RCA cleaned. This step is crucial to avoid a drop in the minority carrier lifetime due to iron and copper precipitates on the wafers.24 Both sides of the wafers were then coated with PECVD silicon dioxide. On the front surface there was a boron doped silica glass closest to the surface and it was capped with a layer of undoped silicon dioxide (1 mm in total). The backside was covered with another boron-doped glass with an order of magnitude higher boron concentration than the front surface layer. This layer was also capped with an undoped layer bringing the total thickness of the back side layer to 3 mm. The front side layer was then patterned with the contact doping areas using UV-lithography and HF etching. The wafers were Lab Chip, 2006, 6, 213–217 | 215
then cleaned and the contact doping was performed using POCl3 at 950 uC for 20 min followed by 20 min of in situ annealing. This high temperature step also meant that the boron doping from the surface layers was partly transferred to the silicon surface. Following this the frontside was patterned with the photosensitive areas (emitters) again using UV-lithography and HF etching. The emitters were doped using POCl3 at 850 uC for 20 min with a 20 min in situ anneal. At this point the PECVD glass was etched from the surface and the wafers were RCA cleaned. The front surface was then given a new coating of PECVD silicon dioxide to act as a buffer for the waveguiding layer. This layer was then patterned with the waveguide to photodiode couplers, which were realized by HF etching of the buffer layer. In fact, all areas used for emitters/contacts or leads to the diodes were etched clean. The wafers were then given a coating of 70 nm high quality PECVD silicon nitride. This layer serves as electrical passivation of the front surface as well as electrical insulator. Contact holes were etched through the nitride with HF using a UV lithography pattern. The front surface was patterned with an image reversed photoresist pattern for lift-off of the front side metal. Then the metallization was applied. The front side received a coating of titanium and aluminum with thicknesses of 20 and 100 nm, respectively. The backside was coated with 100 nm aluminum. Following this, the front side pattern was finalized by lift-off. Finally, the wafers were annealed in forming gas (4% H2 in N2) at 450 uC for 20 min. The 10 mm layer of SU-8 was deposited via a spin process and soft baked at 95 uC for 2 min with slow cooling. The SU-8 was UV exposed through the lithographic mask with 550 mJ cm22 and post-exposure baked at 95 uC for 25 min with slow cooling. The SU-8 was developed for 4 min in propylene glycol monomethyl ether acetate (PGMEA) and rinsed in isopropyl alcohol (IPA). In order to get rid of cracks formed in the SU-8 thin film during processing, the cracks were healed by placing the wafer on a 160 uC hotplate for 60 s with rapid removal.25 The silicon wafer containing photodiodes and the SU-8 polymer layer was bonded to a 0.5 mm thick glass wafer via PMMA mediated bonding.26 The glass wafer was prepared with a 5 mm thick layer of PMMA and the two wafers were bonded in a custom bonder. We have used a bonding force of 2000 N applied on the 40 wafers for 10 min and a bonding temperature of 150 uC with slow cooling to 65 uC before the pressure was released. The finished bonded wafer was diced into chips and holes were drilled using a diamond drill. In the dicing process the glass above the metallic contact pads on the chip was removed by dicing half-way through the glass and breaking it off manually.
for the optical pumping. The Nd:YAG laser emitted 5 ns pulses at 532 nm at a frequency of 10 Hz and the pulse energy density reaching the on-chip dye laser was 80 mJ mm22 during characterization. To operate the laser we injected a 20 mM solution of Rhodamine 6G in ethanol at a rate of 10 mL/hr in the laser channel. The spectrum of the emitted laser light can be seen in Fig. 5. The response of the photodiodes is illustrated in Fig. 6. During measurement, the photodiodes were coupled in parallel with a 1 nF capacitor and a 1 kV resistor in order to measure the charge generated with each laser pulse.27 We measured the noise equivalent power of the diodes to 7.09 6 10213 W (10 Hz bandwidth) and the sensitivity to 390 mA W21 (at 580 nm). The clear indication of the lasing threshold at 14 mW (corresponding to a pump energy density on the dye laser of 50 mJ mm22) in the graph in Fig. 6 demonstrates a low influence of scattered 532 nm pump light on the photodiode signal. The inset of the figure shows the diode response to a single laser pulse.
Fig. 5
Emission spectrum from the embedded liquid dye laser.
5. Evaluation of functionality For testing, the chip was mounted in a polycarbonate sample holder with electrical and fluidic connections. The electrical connection was provided by spring loaded probes that made contact to the connector pads on the chip. A window in the polycarbonate lid gave optical access to the on-chip dye laser for the pulsed frequency doubled Nd:YAG laser that was used 216 | Lab Chip, 2006, 6, 213–217
Fig. 6 Photodiode response as function of optical pump power of the laser pumping the on-chip dye laser. The inset shows the open circuit response from a photodiode to a single pump pulse (vertical 10 mV division21, horizontal 10 ms division21).
This journal is ß The Royal Society of Chemistry 2006
In a first demonstration of the concerted operation of the integrated components (laser, wave-guides, microfluidic cuvette and photodiodes), we performed an absorbance measurement on two different concentrations of xylenol orange dye.28 Before each particular xylenol orange dye solution was injected into the cuvette, a reference measurement with pure water was performed. The xylenol orange solution was diluted in an ammonia buffer (pH = 10) with Ca2+ ions. The photodiode signal, I, for different xylenol orange dye solutions, was referenced to the photodiode signal, I0, for pure water in three tests. As expected, a test with pure water and pure ammonia buffer gave essentially the same signal readout on the photodiodes (so I/I0 # 1). A xylenol orange dye concentration of 0.06 mM gave I/I0 = 0.7 and a concentration of 0.12 mM gave I/I0 = 0.3. Although this examination of the chip is not sufficient to determine the limit of detection and the linear range of the device, it proves the successful integration of the individual components. In summary, we have demonstrated a fabrication platform for integration of five functional components (liquid dye laser, waveguides, fluidic channels, measurement cuvettes and photodiodes). Higher integration is much more than just a demonstration of fabrication skills: it improves our understanding of how lab-on-a-chip devices will evolve in the coming years. In the future, the presented device will be tested for more application-oriented problems.
Acknowledgements This work was supported by the Danish Technical Research Council (STVF) grant no.: 26-02-0064, grant no. 26-00-0220 and the H. C. Ørsted Foundation.
References 1 A. Manz, N. Graber and H. M. Widmer, Sens. Actuators B, 1990, 1, 244–248. 2 S. Verpoorte and N. F. De Rooij, Proc. IEEE, 2003, 91, 930.
This journal is ß The Royal Society of Chemistry 2006
3 K. Uchiyama, H. Nakajima and T. Hobo, Anal. Bioanal. Chem., 2004, 379, 375. 4 J. Hu¨bner, K. B. Mogensen, A. M. Jorgensen, P. Friis, P. Telleman and J. P. Kutter, Rev. Sci. Instrum., 2001, 72, 229. 5 H. P. Fischer, K. Peters, R. Ziegler, D. Pech, A. Kilk, Th. Eckhardt, G. G. Mekonnen and G. Jacumeit, Opt. Fiber Technol., 2000, 6, 1, 68. 6 I. Hardy, P. Grosso and D. Bosc, Photon. Technol. Lett., IEEE, 2005, 17, 5, 1028. 7 J. N. McMullin, Proc. SPIE, 2000, 4087, 1050. 8 L. Eldada, K. M. T. Stengel, L. W. Shacklette, R. A. Norwood, C. Xu, C. Wu and J. Yardley, Proc. SPIE, 1997, 3006, 344. 9 N. M. Jokerst, M. A. Brooke, S. Cho, M. Thomas, J. Lillie, D. Kim, S. Ralph, K. Dennis, B. Comeau and C. Henderson, Proc. SPIE, 2005, 5730, 226–233. 10 S. Verpoorte, Lab Chip, 2003, 3, 42N. 11 SU-8 10 from MicroChem Corp., Newton, MA. www.microchem. com. 12 K. Y. Lee, N. LaBianca, S. Zolgharnain, S. A. Rishton, J. D. Gelorme, J. M. Shaw and T. H. P. Chang, J. Vac. Sci Technol. B, 1995, 13, 3012. 13 H. Lorenz, M. Despont, N. Fahrni, N. LaBianca, P. Renaudy and P. Vettiger, J. Micromech. Microeng, 1997, 7, 121. 14 H. Klank, J. Kutter and O. Geschke, Lab Chip, 2002, 2, 242. 15 S. Balslev and A. Kristensen, Opt. Express, 2005, 13, 344. 16 K. B. Mogensen, J. El-Ali, A. Wolff and J. P. Kutter, Appl. Opt., 2003, 89, 19, 4072. 17 M. Nathan, O. Levi, I. Goldfarb and A. Ruzin, J. Appl. Phys., 2003, 94, 12, 7932. 18 Yi-Fan Li and John W. Y. Lit, J. Opt. Soc. Am. A, 1985, 2, 3, 462. 19 P. Friis, K. Hoppe, O. Leistiko, K. B. Mogensen, J. Hu¨bner and J. P. Kutter, Appl. Opt., 2001, 40, 6246. 20 T. T. Veenstra, T. S. J. Lammerink, M. C. Elwenspoek and A. van den Berg, J. Micromech. Microeng., 1999, 9, 199. 21 S. M. Sze, in Semiconductor Devices. Physics and Technology, John Wiley & Sons, New York, 1985. 22 M. A. Green, in High Efficiency Silicon Solar Cells, Trans Tech Publications Ltd, Aedermannsdorf, 1987. 23 M. Razeghi and A. Rogalski, J. Appl. Phys., 1996, 79, 7433. 24 C. B. Nielsen, C. Christensen, C. Pedersen and E. V. Thomsen, J. Electrochem. Soc., 2004, 151G338. 25 Crack healing process from personal communication with MicroChem Corp., Newton, MA. www.microchem.com. 26 B. Bilenberg, T. Nielsen, B. Clausen and A. Kristensen, J. Micromech. Microeng., 2004, 14, 814. 27 B. R. Clemesha, J. Phys. E, 1972, 5, 859. 28 C. Gay, J. Collins and J. M. Gebicki, Anal. Biochem., 1999, 273, 2, 143.
Lab Chip, 2006, 6, 213–217 | 217
