A Monolithic Silicon Optoelectronic Transducer as a Real-Time Affinity

An optical real-time affinity biosensor, which is based on a monolithic silicon optoelectronic transducer and a microfluidic module, is described. The...
0 downloads 0 Views 394KB Size
Download PDF
Anal. Chem. 2004, 76, 1366-1373

A Monolithic Silicon Optoelectronic Transducer as a Real-Time Affinity Biosensor Konstantinos Misiakos,*,† Sotirios E. Kakabakos,‡ Panagiota S. Petrou,‡ and Hans H. Ruf§

Microelectronics Institute and Institute of Radioisotopes and Radiodiagnostic Products, NCSR “Demokritos”, 15310, Aghia Paraskevi, Attiki, Greece, and Fraunhofer Institute of Biomedical Engineering (IBMT) and University of Saarland, 66386, St. Ingbert, Germany

An optical real-time affinity biosensor, which is based on a monolithic silicon optoelectronic transducer and a microfluidic module, is described. The transducer monolithically integrates silicon avalanche diodes as light sources, silicon nitride optical fibers, and p/n junction detectors and efficiently intercouples these elements through a self-alignment technique. The transducer surface is hydrophilized by oxygen plasma treatment, silanized with (3-aminopropyl)triethoxysilane and bioactivated through adsorption of the biomolecular probes. The use of a microfluidic module allows real-time monitoring of the binding reaction of the gold nanoparticle-labeled analytes with the immobilized probes. Their binding within the evanescent field at the surface of the optical fiber causes attenuated total reflection of the waveguided modes and reduction of the detector photocurrent. The biotin-streptavidin model assay was used for the evaluation of the analytical potentials of the device developed. Detection limits of 3.8 and 13 pM in terms of gold nanoparticle-labeled streptavidin were achieved for continuous- and stopped-flow assay modes, respectively. The detection sensitivity was improved by silver plating of the immolilized gold nanoparticles, and a detection limit of 20 fM was obtained after 20-min of silver plating. In addition, two different analytes, streptavidin and antimouse IgG, were simultaneously assayed on the same chip demonstrating the multianalyte potential of the sensor developed. Bioanalytical microsystems based on miniaturized biosensing elements could find wide applications in DNA analysis,1,2 drug discovery,3 medical diagnostics,4,5 and environmental monitoring6,7 * Corresponding author. E-mail: [email protected]. Fax: (++302106511723). † Microelectronics Institute, NCSR “Demokritos”. ‡ Institute of Radioisotopes & Radiodiagnostic Products, NCSR “Demokritos”. § Fraunhofer Institute of Biomedical Engineering (IBMT) and University of Saarland. (1) Lagally, E. T.; Medintz, I.; Mathies, R. A. Anal. Chem. 2001, 73, 565-570. (2) Lehr, H.-P.; Reimann, M.; Brandenburg, A.; Sulz, G.; Klapproth, H. Anal. Chem. 2003, 75, 2414-2420. (3) Cooper, M. A. Nat. Rev. Drug Discovery 2002, 1, 515-528. (4) Christodoulides, N.; Tran, M.; Floriano, P. N.; Rodriguez, M.; Goodey, A.; Ali, M.; Neikirk, D.; McDevitt, J. T. Anal. Chem. 2002, 74, 3030-3036. (5) Fall, B. I.; Eberlein-Ko ¨nig, B.; Behrendt, H.; Niessner, R.; Ring, J.; Weller M. G. Anal. Chem. 2003, 75, 556-562. (6) Bier, F. F.; Schmid, R. D. Biosens. Bioelectron. 1994, 9, 125-130

1366 Analytical Chemistry, Vol. 76, No. 5, March 1, 2004

as well as in protection against bioterrorism.8 Miniaturization, portability, multianalyte potential, and interfacing with electronic functions are critical elements for biosensing devices to meet the demands in these fields. Optical detection in biosensors is superior to other sensing approaches9 due to the versatility of the existing optical transducers and the large variety of labels that could be used (fluorescent or optical loss tags). In addition, the optical devices are immune to a number of interfering effects common in electrochemical transduction, such as, electromagnetic interference, junction potentials, reference electrodes, electrode corrosion, or unwanted currents caused by ionic transport.10 This is a result of the galvanic isolation of the optical transducer from the excitation and detection electronics. Among the various optical devices, planar integrated waveguides seem to be the most promising in terms of potential for multianalyte testing and assay sensitivity.11 Planar geometry facilitates multiple analyte arrays to be created by a variety of probe delivery methods including ink-jet and contact printing, photolithographic patterning, photochemical attachment, microfluidic network patterning, etc.12-16 Besides, planar integrated waveguides could detect binding events in real time through evanescent field optics that permit detection of the surface-bound molecules either using labels (e.g., fluorescent compounds) or following label-free detection approaches. Moreover, due to the many internal reflections in the waveguides, the resonant mode photons interact with the surface-bound species much more intensively than in vertically illuminated single-pass planar devices. The enhancement factor increases as the waveguide thickness decreases while it tends to saturate for a thickness well below the wavelength due to the Goos-Ha¨nchen shift effects.17 Nevertheless, in most cases, a fiber (7) Mallat, E.; Barcelo, D.; Barzen, C.; Gauglitz, G.; Abuknesha, R. Trends Anal. Chem. 2001, 20, 124-132. (8) Koch, S.; Wolf, H.; Danapel, C.; Feller, K. A. Biosens. Bioelectron. 2002, 14, 779-784. (9) Turner, A. P. F. Science 2000, 290, 1315-1317. (10) Marazuela, M. D.; Moreno-Bondi, M. C. Anal. Bioanal. Chem. 2002, 372, 664-662. (11) Plowmam T. E.; Durstchi, J. D.; Wang, H. K.; Christensen, D. A.; Herron, J. N.; Reichert, W. M. Anal. Chem. 1999, 71, 4344-4352. (12) Silzel, J. W.; Sercek, B.; Dodson, C.; Tsay, T.; Obrenski, R. J. Clin. Chem. 1998, 44, 2036-2043. (13) Martin, B. D.; Gaber, B. P.; Patterson, C. H.; Turner, D. C. Langmuir 1998, 14, 3971-3975. (14) Douvas, A.; Argitis, P.; Misiakos, K.; Dimotikali, D.; Petrou, P. S.; Kakabakos, S. E. Biosens. Bioelectron. 2002, 17, 269-278. (15) Pirrung, M. C.; Huang, C.-Y. Bioconjugate Chem. 1996, 7, 317-321. (16) Bernard, A.; Michel, B.; Delamarche, E. Anal. Chem. 2001, 73, 8-12. 10.1021/ac0353334 CCC: $27.50

© 2004 American Chemical Society Published on Web 01/20/2004

thickness of ∼100 nm can sustain over 1000 reflections/mm of propagation length on the exposed waveguide surface. Thus, the use of such thin optical fibers can revolutionize portable analytical instrumentation,18 provided that the light source and the detector would eventually be monolithically integrated with the bioactivated waveguide. Such monolithic integration is a long-sought goal of the analytical community since hybrid integration of the light source requires precision alignment to the thin optical fiber, which is expensive, complicates packaging, and becomes impractical in multianalyte devices. Apparently, microfabrication techniques borrowed from microelectronics could be employed to create fully integrated optical biosensing devices in a cost-effective way. The fact that so far this has not been achieved is attributed to the technical difficulty in co-integrating mainstream solid-state light sources (such as compound semiconductor devices) with the mature silicon technology. In addition, another important issue that should be resolved before realizing monolithically integrated optoelectronic devices is the efficient alignment of the integrated optical fiber with the light emitter and the detector. Apart from the co-integration of active and passive optical components, the biological activation of a fully processed and metallized semiconductor wafer is another issue to be settled. Biological activation of glass, silicon dioxide, or silicon nitride surfaces could be achieved by a variety of methods19that require surface hydrophilization followed by silanization. This procedure results in modification of the surface properties and permits adsorptive or covalent coupling of biomolecules. However, when the activation of optical transducers that are monolithically and fully integrated on silicon is sought, hydrophylization steps that involve treatment with strong acidic or alkaline solutions20 are prohibitive. These treatments either destroy the existing metal interconnects or degrade the detector leakage current and the optical quality of the integrated waveguides. Thus, a method for surface activation that will be compatible with the integrated components is required. Here we demonstrate an optical affinity sensor based on a monolithic optoelectronic transducer, which integrates on a silicon die thin optical fibers (silicon nitride) along with self-aligned light-emitting diodes (LED) and photodetectors (silicon p/n junction). The LEDs are silicon avalanche diodes that emit light when biased beyond their breakdown point, a phenomenon known since the mid-1950s.21,22 The LEDs are optically coupled to the corresponding photodetectors through silicon nitride fibers. Specially designed spacers provide for the smooth bending of the fiber at its end points and toward the light source and the detector ensuring high coupling efficiency. This is shown in Figure 1, which illustrates the solid-state device coupled to a microfluidic module. Moreover, the self-alignment of the light emitter and the detector to the fiber was achieved by ion implanting the LED and the detector with boron through the silicon nitride fiber film. As of now, this is the only sensing device having the light source monolithically integrated with the (17) Lotsch, H. K. V. J. Opt. Soc. 1968, 58, 551-561. (18) Bradshaw, J. T.; Mendes, S. B.; Saavedra S. S. Anal. Chem. 2002, 74, 17511759. (19) Weetall, H. H. Appl. Biochem. Biotechnol. 1993, 41, 157-188. (20) Crass, J. J.; Rowe-Taitt, C. A.; Nivens, D. A.; Ligler, F. S. Biosens. Bioelectron. 1999, 14, 683-688. (21) Chynoweth, A. G.; McKay, K. G. Phys. Rev. 1956, 102, 369-376. (22) Newman, R. Phys. Rev. 1995, 100, 700-703.

Figure 1. Schematic drawing of the monolithic transducer coupled to a microfluidic compartment (not in scale). The fiber bending SiO2 spacers are emphasized. P++ are the self-aligned LED and photodetector (PD) emitter regions heavily implanted with boron while N+ is the phosphorus preimplanted LED base region.

waveguide and the detector. The transducer is functionalized into an affinity sensor by properly modifying the optical fiber surface with recognition molecules that bind the analyte. For this purpose, a methodology for hydrophilization of the surface through oxygen plasma treatment and subsequent silanization with (3-aminopropyl)triethoxysilane (APTES) was developed. Recognition biomolecules can be immobilized onto the aminosilanized surface through adsorption or covalent binding. The sensor is accompanied by a microfluidic module (Figure 1), which allows for reagent application and real-time monitoring of the binding reaction. In the proposed sensor, signal transduction occurs by attenuated total reflection of the waveguided modes, which leads to a decrease in the detector photocurrent upon binding of lightabsorbing moieties within the evanescent field at the surface of the sensing optical fiber. The potential of the device developed to perform as a sensitive real-time affinity sensor was tested in single-analyte assay mode using the biotin-streptavidin model assay. Colloidal gold nanoparticles were employed as labels due to their strong surface plasmon resonance absorption in the greenyellow spectral region. Furthermore, the use of gold nanoparticles provided the ability to increase considerably the detection sensitivity by silver plating of the surface-bound gold nanoparticles, which resulted in an increase of their volume and of the associated light transmission losses. The multianalyte capabilities of the device were demonstrated through the simultaneous real-time monitoring of the binding reaction of two different proteins (streptavidin and anti-mouse IgG) with their counterparts (biotinylated bovine serum albumin and mouse IgG) immobilized onto adjacent fibers of the same chip. EXPERIMENTAL SECTION Materials. Bovine serum albumin (BSA), mouse IgG (mIgG), streptavidin labeled with gold nanoparticles with a mean diameter of 8.4 nm (Au-streptavidin), anti-mouse IgG labeled with gold nanoparticles with a mean diameter of 9.5 nm (Au-anti-mIgG), and APTES were from Sigma (St. Louis, MO). LI Silver (LIS) enhancement kit, 6-((biotinoyl)amino)hexanoic acid sulfosuccinimidyl ester (sulfo-NHS-LC-biotin), and AlexaFluor 546-labeled streptavidin were purchased from Molecular Probes (Eugene, OR). All other chemicals and reagents were from Merck. The water used throughout was doubly distilled. Analytical Chemistry, Vol. 76, No. 5, March 1, 2004

1367

Preparation of Biotinylated BSA. Biotinylated BSA (B-BSA) was prepared according to a published method.23 Briefly, BSA was reacted with sulfo-NHS-LC-biotin in 0.3 M sodium carbonate buffer (pH 9.2) at a molar ratio of 1:120 for 2 h at room temperature. Then, the mixture was extensively dialyzed against 0.1 M NaHCO3 solution (pH 8.5), containing 0.15 M NaCl and 15 mM sodium azide, and was stored at 4 °C. Instrumentation and Signal Recording. All the measurements were performed in a semiautomatic wafer prober. The LED bias was provided by the Keithley 220 current source. The photocurrent was measured by the HP 4140B Pico-Amperometer and downloaded on a PC by the LabView software at a maximum rate of two samples/s. Fabrication of the Transducers. Arrays of the monolithic optocouplers were fabricated at the clean room facility of the Microelectronics Institute of NCSR “Demokritos”. The six-mask microelectronic process cycle included a total of three lowpressure chemical vapor deposition (LPCVD) steps, two for SiO2 (field oxide and spacers) and one for Si3N4 (fiber), and two implantations (phosphorus and boron) for the formation of the avalanche diode and the photodetector on the n-type substrate. The base (N+) side of the avalanche junction is formed by phosphorus implantation in lithographically defined windows in the thermally grown field oxide and prior to the SiO2 spacer deposition. The spacers were created next to the field oxide vertical walls by SiO2 deposition and anisotropic plasma etching (Figure 2A). The vertical walls were made by a 2-µm tetraethyl orthosilicate (TEOS) deposition of SiO2, (LPCVD at 710 °C) over the thermal field oxide, lithographic patterning, and anisotropic reactive ion etching in CHF3. Then, a second thick (2.5 µm) TEOS deposition followed, and the spacers were created by anisotropic plasma etching (without patterning, Figure 2A). The TEOS deposition was performed in discrete steps of up to 0.7 µm followed by annealing at 900 °C to avoid excess mechanical stress that can result in film cracking.24 An overall field oxide thickness of 2.5 µm assured relatively smooth bending and enough distance between the long horizontal segment of the fiber and the silicon substrate to minimize substrate losses. After spacer formation, the nitride film was deposited (LPCVD in a mixture of NH3 and SiH2Cl2) and lithographically patterned and etched in CHF3 anisotropic reactive ion plasma. Then boron was implanted through the nitride film to form the self-aligned avalanche junction by phosphorus compensation, as shown in Figure 2B. The implant energy was adjusted so that the range of the ions would match the nitride film thickness. For a 150-nm nitride film, an implant energy of 65 keV was selected. The process was completed by rapid thermal annealing of the boron implant, contact hole, and aluminum patterning. Chips with fibers of variable length and thickness were fabricated. Those used in the assays were 900 µm long, 25 µm wide, and 150 nm thick. Finished devices are shown in Figure 2C, where two adjacent fibers are displayed with their respective LEDs on. Microfluidic Module. The optical microchip was coupled to a microfluidic channel (Figure 3), which permitted reagent application on the functionalized fiber. The channel (length 700 µm, width 240 µm, height 140 µm) was made by silicone (PDMS (23) Kakabakos, S. E.; Christopoulos, T. K.; Diamandis, E. P. Clin. Chem. 1992, 38, 338-342. (24) Bulla, D. A. P.; Morimoto, N. I. Thin Solid Films 1998, 334, 60-64.

1368 Analytical Chemistry, Vol. 76, No. 5, March 1, 2004

Figure 2. (A) Schematic illustration of the SiO2 spacer formation. A composite film consisting of thermal and chemical SiO2 is patterned by lithography and CHF3 anisotropic reactive ion etching.. A thick TEOS deposition over the patterned SiO2 layer follows (top schematic), and subsequent anisotropic reactive ion etching creates the spacers (bottom schematic). (B) Boron implantation through the Si3N4 fiber layer to form the self-aligned p/n junction by phosphorus compensation. (C) Microphotograph showing a pair of fibers with the LEDs on (pointed by the arrows as vertical white segments near the fiber left ends). The fibers connect to the detectors on the right (not shown). The white horizontal strips on the left are the metal interconnects. The fibers had a width of 25 µm and a thickness of 150 nm.

micromachining) on a glass support and formed a reversible seal when pressed against the die. The total volume of the channel was 24 nL. Surface Pretreatment. The dies were silanized on a wafer scale so that proteins can be immobilized through adsorption onto their surface. Prior to silanization, the surface of the wafer was

Figure 3. A single optocoupler with the fluidic device on top. The emitter and detector planar aluminum contact pads are shown on the left and the right, respectively, and next to them are the respective electrical probe contacts to the current source and the current meter. The fiber is shown in the middle of the microfluidic channel, the length of which defines the exposed fiber length (700 µm).

hydrophilized by 30-s treatment with oxygen plasma (10 mTorr). Then, the wafer was immersed in a 2% (v/v) aqueous APTES solution for 20 min followed by gentle washing with water and drying for 20 min at 120 °C. After drying, the wafer was sonicated for 1 min in a water bath, washed with water extensively, and dried under N2 flow. Biotin-Streptavidin Binding Assay. The silanized wafer was incubated with a 25 µg/mL B-BSA solution in 0.05 M sodium carbonate buffer, pH 9.2, for 1 h at room temperature. After washing with water, the wafer was immersed in a 1% (w/v) BSA solution in 0.05 M NaHCO3, pH 8.5 (blocking solution), for 30 min at room temperature. Finally, the wafer was washed with distilled water and dried under N2 flow. Blank chips were also prepared by coating only with BSA in order to determine the nonspecific binding of the labeled analyte onto the fiber surface. To perform the assay, the microfluidic module was applied onto the chip surface. Distilled water, assay buffer (0.05 M sodium phosphate buffer, pH 6.5, 4% (w/v) BSA, 0.05% (w/v) bovine γ-globulins, 0.15 M NaCl), and finally the Au-streptavidin solutions (0.05-830 pM) in assay buffer were sequentially delivered to the chip surface and the photocurrent was recorded in real time. The stopped-flow assay was performed by interruption of the Au-streptavidin solution flow as soon as the fiber was completely covered. In the continuous-flow assay mode, the Austreptavidin solution was continuously pumped through the channel at a constant rate of 0.45 µL/s using a peristaltic pump (Pharmacia LKB Biotechnology). After 30 min of reaction, under stopped- or continuous-flow conditions, the fiber was washed with 2 mL of washing solution (0.05 M sodium phosphate buffer, pH 6.5, 0.15 M NaCl, 0.05% (v/v) Tween 20) and 2 mL of distilled water. Dose-response curves for both continuous- and stoppedflow assay conditions were obtained by plotting the ratio of the initial detector photocurrent (S0) measured during initial washing of the fiber with the assay buffer to the photocurrent (S30) measured after 30 min of reaction. Dual-Analyte Assay. The simultaneous dual-analyte assay was performed on the same chip by selectively coating adjacent fibers, one with B-BSA and the other with mIgG, using the fluidic device placed with its long axis along each fiber. The respective coating solutions, 25 µg/mL B-BSA and 25 µg/mL mouse IgG were pumped for 30 min at a rate of 0.45 µL/s. The fibers were blocked by passing blocking solution for 30 min at the same flow rate and washed extensively with distilled water. Then, the fluidic device was rotated 90° to cover part (240 µm in length) of the two fibers

simultaneously. Solutions containing Au-anti-mIgG or a mixture of Au-streptavidin and Au-anti-mIgG in 0.05 M Tris-HCl buffer, pH 7.2, 4% (w/v) BSA, 0.05% (w/v) bovine γ-globulins, and 0.15 M NaCl were sequentially pumped to the chip surface and the photocurrents from both detectors were recorded simultaneously. Silver Plating of the Gold Nanoparticles. After completion of the binding reaction, silver enhancement solution, prepared according to the instructions of the LIS enhancement kit manufacturer, was applied onto the chip surface. The silver enhancement solution was incubated for 5 min in the dark, and then the chips were extensively washed with distilled water and dried under N2 flow. The photocurrents of the treated fibers were measured, and the silver enhancement cycle was repeated several times. Detector photocurrents before (SP0) and after silver plating (SP) were normalized to corresponding photocurrents of a blank fiber, and the ratios of the normalized signals (NSP0/NSP) were plotted against the respective Au-streptavidin concentrations. RESULTS AND DISCUSSION Design and Performance of the Optical Transducer. The transducer developed integrates thin SiN3 fibers along with selfaligned LEDs and photodetectors on a silicon chip. The light emitted by the LED junction upward (Figure 1) enters the fiber, bends over the spacer, and is guided to the detector where the fiber abrupt end break delivers the photons to its p/n junction, a point with maximum photogenerated carrier collection efficiency. The dimensions of the LED (Figure 2C) and detector junctions were 100 × 100 µm2 and could be easily shrunk down, limited only by the fiber width (25 µm) that, in turn, could be reduced if intense packing of the optocouplers was sought. Efficient alignment and optical coupling of the light emitter and the detector with the integrated optical fiber were essential in order to realize the proposed monolithic optoelectronic transducer. Self-alignment here was critical since both the emitting junction width and the fiber thickness were in the 100-nm range. Given that the base (N+) side of the avalanche junction was already formed by phosphorus implantation prior to the SiO2 spacer deposition, self-alignment of the up-going segment of the fiber with the LED was achieved by ion implanting the emitter with boron through the silicon nitride fiber film to form the p/n junction by phosphorus compensation. The vertical segment of the fiber and the spacer masked the boron implant, and thus, the avalanche junction was placed exactly under this segment (Figure 2B). In addition, to efficiently couple the emitter and the detector Analytical Chemistry, Vol. 76, No. 5, March 1, 2004

1369

Figure 4. Quantum efficiency and spectral response. (A) Detector photocurrent as a function of the LED bias current for a 900-µm-long fiber (9) and I-V reverse bias characteristic of the LED (solid line). The breakdown voltage is 6.5 V and the operating bias in the vicinity of 10 V. (B) Emission spectrum of the LED (curve 1) and absorption spectrum of a 920 pM Au-streptavidin solution (curve 2).

through the silicon nitride fiber, rounded spacers (Figure 2A) were created to the end points of the optical fiber. These spacers allowed smooth bending of the fiber from the field oxide toward the LED and the detector. Moreover, by employing rapid thermal processing after boron implantation to prevent junction shifts, excellent stability, optical coupling performance, and device yield were achieved. The device yield in a carefully processed wafer batch in our microelectronics laboratory conditions was nearly 95%. The LED I-V characteristic and the linear photocurrent response versus the LED bias current are shown in Figure 4A. For a 10-mA LED bias current, a detector photocurrent of nearly 800 pA was obtained on a 900-µm fiber. This amounts to an external LED-photodetector optical coupling efficiency of 40%, since the LED quantum efficiency was measured to be 2 × 10-7. Considering the fiber thickness and the LED spectral width, which was approximately 500-800 nm (Figure 4B, curve 1), such coupling efficiency is excellent and should be compared with the 1% coupling efficiency obtained on thicker fibers when integrating compound semiconductor LEDs on silicon by microbonding techniques.25 Additionally, it can be further improved by increasing the deposited silicon dioxide thickness and the associated spacer radius. Despite the low quantum efficiency of the LED, the photocurrents achieved (sub-nA) could be measured very accurately by modern portable instrumentation. Due to the small detector capacitance, of ∼1 pF, the main noise source for signal (25) Nagata, T.; Namba, T.; Kuroda, Y.; Miyake, K.; Miyamoto, T.; Yokoyama, S.; Miyazaki, S.; Koyanagi, M.; Hirose, M. Jpn. J. Appl. Phys. 1995, 34, 12821285.

1370 Analytical Chemistry, Vol. 76, No. 5, March 1, 2004

shaping times, τsh, of the order of 1 ms will be the shot noise due to the detector junction leakage current, Io. A leakage current of 1 pA and a τsh of 1 ms correspond to an equivalent noise charge (reduced at the input) of (e2qIoτsh/4)1/2 or ∼12 electrons rms, which when divided by τsh gives a current noise of 1.9 fA rms. Therefore, the current uncertainty due to the detector leakage current was well below 0.1 pA. This was due to the small leakage current and capacitance of the detector, and thus, the accuracy in the current reading was basically limited by the stability of the LED. The stability was very good and corresponded to an uncertainty of less than 0.5% over a long period of time, enabling the use of the present optoelectronic device as an accurate signal transducer despite the inherently low silicon LED quantum efficiency. The presented optoelectronic characteristics of the transducer allow for the realization in a monolithic form of an entire class of optical sensors presently manufactured by hybrid integration techniques and, in fact, enable new multisensory applications by lab-on-a-chip devices. Here, stability, manufacturability, cost, reproducibility, and array potential are the key issues. They are all addressed by the employment of the mature microelectronics technology including self-alignment for efficient optical coupling and miniature detector elements for optimum noise performance. Surface Activation. Surface pretreatment before the protein probe immobilization is critical.26 Additionally, it must be electron device compatible. Several methodologies have been used for silicon surface activation. In general, they involve cleaning and hydrophilization of the surface to expose hydroxy groups followed by treatment with a silanizing agent. Cleaning and hydrophilization is usually achieved by treatment with concentrated acidic or alkaline solutions.20 Most of these methods, however, cannot be applied in our devices since they would either destroy the aluminum pads or etch the fiber surface causing losses of the waveguided modes. As an alternative, oxygen plasma treatment of silicon nitride surfaces was evaluated since it is compatible with microelectronic fabrication procedures. The surface protein adsorption capacity and the optoelectronic transducer performance were used as the evaluation criteria. The protein absorption capacity of APTES-treated silicon nitride surfaces hydrophilized using oxygen plasma treatment was compared with that determined for surfaces hydrophilized using piranha treatment (a 1:1 mixture of H2SO4 and H2O2 for 30 min). This was carried out by spotting B-BSA on the silanized surfaces and detecting the immobilized protein using streptavidin labeled with AlexaFluor 546. Fluorescence intensity measurements were performed by employing the ImagePro Plus 4.0 software (Media Cybernetics) on images received with an epifluorescence microscope (Zeiss Axioscope) equipped with a digital camera (Sony Cybershot). We found that the fluorescence intensity values determined for the amino-silanized surfaces obtained following a 30-s oxygen plasma treatment at 10 mTorr were 20-30% higher than those obtained after piranha treatment hydrophilization. On the other hand, the effect of oxygen plasma treatment on the performance of the optoelectronic transducer was evaluated by performing photocurrent measurements of the transducers prior to and after repeated activation cycles by oxygen plasma treatment and silanization. It was found that at least five cycles could be performed without (26) Houseman, B. T.; Huh, J. H.; Kron, S. J.; Mrksich, M. Nat. Biotechnol. 2002, 20, 270-274.

Figure 5. Biotin-streptavidin binding reaction. (A) Real-time response plotted as the ratio of the initial detector photocurrent (S0) to the instantaneous photocurrent (St) for continuous- (curve 1) and stopped- (curve 2) flow conditions. Curve 3 is the signal from a blank fiber coated with BSA. The Au-streptavidin particle concentration was 104 pM in all cases. (B) Dose-response plots showing the ratio of the initial (S0) to the final photocurrent value (S30) obtained under stopped- (open symbols) and continuous-flow conditions (solid symbols) after a 30-min reaction. The inset is a blowup of the lowconcentration region. Each point represents the mean value of three different fibers; the error bars represent (SD.

affecting the initial photocurrent value. On the contrary, as was expected, even one treatment of the transducers with piranha solution resulted in complete aluminum removal from the contact pads. Thus, oxygen plasma treatment was adopted in our final protocol for the integrated fiber surface activation. Single-Analyte Assay. The biotin-streptavidin model assay was employed in order to demonstrate the bioanalytical capabilities of the sensor presented. Gold nanoparticles were selected as optical loss labels due to their absorption spectrum that overlapped in a great extent with the emission spectrum of the LED (Figure 4B). Fluorescent compounds or chromophores can be also used though they are much less efficient compared to gold nanoparticles (data not shown) due to their relatively narrow absorption spectrum and smaller effective photon capture cross section. The combination of the optical device with a microfluidic cover that protected its electrical contact pads from the reaction fluid permitted real-time monitoring of the detector photocurrent. The signals recorded in real time using a 104 pM Au-streptavidin concentration under either continuous- (curve 1) or stopped-flow (curve 2) conditions are presented in Figure 5A. In addition, the signal obtained using a blank fiber assayed under continuousflow conditions is presented since identical results were obtained with blank fibers assayed under stopped-flow conditions. As shown, under both assay conditions, the real-time signal developed with an initial fast phase and reached a value of 1.05 (or 5% photocurrent drop) in ∼10 s. A slower phase followed, though the continuous-flow assay signal developed much faster than in

the stopped-flow case. The signal was specific to the Austreptavidin binding on the biotinylated fiber since no signal was observed in blank fibers. The observed binding kinetics can be explained on the basis of diffusion-controlled analyte transport, which governs the binding reactions on the fiber surface. The high association rate constant of the biotin-streptavidin27 reaction makes bulk analyte diffusion the rate-limiting factor, especially under stopped-flow conditions. In continuous flow, convection decreases the thickness of the diffusion layer at the binding surface resulting in faster diffusion and reaction rates. The plots of the signal against the concentration of Austreptavidin obtained after 30 min of reaction for the continuousand the stopped-flow assay modes are presented in Figure 5B. Since the blank values were negligible, we considered as detection limit the concentration corresponding to a signal of 1.05, which is 10 times the uncertainty (0.5%) in the photocurrent readings. Thus, the calculated detection limits were 3.8 and 13 pM for the continuous- and the stopped-flow assay modes, respectively. The real-time monitoring of signal evolution permitted the performance of kinetic measurements by calculating the initial reaction rate. This was done by application of polynomial leastsquares fit of the data obtained for the first minutes of the reaction. We found that the initial reaction rate calculated for the first 5 min of reaction provided a dose-response curve that was comparable with that obtained after 30 min of reaction (Figure S-1, Supporting Information). In addition, regression analysis of the signals obtained after 30 min of reaction with the initial reaction rate values showed that they were linearly correlated (R2 ) 0.998). Thus, the initial rate could be used as a faster alternative measure of the analyte concentration. To verify that the signal drop obtained under real-time measurements originated in the surface plasmon resonance of the gold nanoparticles, we employed a ray optics model, which relates the metal nanoparticle absorption in a bulk solution to the propagation losses of a waveguided electromagnetic mode. The signal as the normalized photocurrent drop is given by the relation

S0/S )

∫I(λ) dλ

∫{I(λ)/exp[ln10A(λ)]} dλ

(1)

where I(λ) is the spectral intensity of the waveguided photons and A(λ) is the absorption of the gold particle-coated fiber. In eq 1, we assume that the spectral response of the photodetector is flat, a realistic assumption since the photons are delivered right at the photodetector junction and the wavelengths considered are strongly absorbed by silicon. To relate A to the cuvette volume absorption, Α(λ), shown in Figure 4B, a treatment similar to the one in ref 27 is employed:

LfCs 2nc(nf2 - N2) A(λ) ) A′ CvLcteff (n 2 - n 2)N f

(2)

c

where Lf and Lc are the active fiber length (700 µm) and cuvette optical path length, respectively, Cs is the surface concentration of gold nanoparticles on the fiber, and Cv is the particle volume (27) Diamandis, E. P.; Christopoulos, T. K. Clin. Chem. 1991, 37, 625-636.

Analytical Chemistry, Vol. 76, No. 5, March 1, 2004

1371

concentration in the cuvette. Also, nc and nf are the refractive indices of the superstrate (water) and the fiber and N is the mode effective index. Finally, teff is the waveguide effective thickness, which is the fiber thickness (150 nm) plus the substrate and superstrate increments due to the Goos-Ha¨nchen shift effects.28 From eqs 1 and 2 and after numerical integration of the data in Figure 4B (LED emission and gold absorption spectrum), the signal in the linear region of Figure 5B, for small Cs values (