Liquid Waveguide-Based Evanescent Wave Sensor That Uses Two

Dec 17, 2010 - The evanescent wave sensor based on a liquid waveguide can also be used for real-time monitoring of chemical reactions, because the cor...
0 downloads 0 Views 2MB Size
Download PDF
Anal. Chem. 2011, 83, 585–590

Liquid Waveguide-Based Evanescent Wave Sensor That Uses Two Light Sources with Different Wavelengths Jong-Min Lim,† John Paul Urbanski,‡ Jae-Hoon Choi,† Todd Thorsen,‡ and Seung-Man Yang†,* National Creative Research Initiative Center for Integrated Optofluidic Systems and Department of Chemical and Biomolecular Engineering, Korea Advanced Institute of Science and Technology, Daejeon, 305-701, Korea, and Hatsopoulos Microfluids Laboratory, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States We demonstrate a prototypic optofluidic evanescent wave sensor made of poly(dimethylsiloxane) (PDMS) elastomer in which two light sources with different wavelengths are coupled into an optofluidic liquid-core/liquid-cladding (L2) waveguide. The exponentially decaying evanescent wave interacts with analyte molecules dissolved in the cladding fluids or products formed by in situ reactions at the core-cladding interface. The analyte molecules exhibit distinctly different light absorbance at the two wavelengths during the light-analyte interaction. Therefore, by using the normalized absorbance calculated from the intensity ratio of the two wavelengths instead of the absolute magnitude of either signal, unwanted effects from omnipresent external noise sources can be reduced. In addition, the differential absorption of the two beams by the analyte solutions can be used to enhance the resolution of sample analysis. The evanescent wave sensor based on a liquid waveguide can also be used for real-time monitoring of chemical reactions, because the core and cladding fluids in the L2 waveguide are slightly miscible at the core-cladding interface due to the diffusional mixing. Optofluidics is a new field that seeks to achieve beneficial synergies by combining microfluidics with optics.1-4 Various optofluidic modules are essential building blocks to construct micro total analysis systems (µTAS) based on optical detection. In particular, µTAS based on optical detection require the efficient coupling of light to the analytes. Numerous efforts have been made to develop sensors by integrating waveguides into microchannels. However, most of those studies adopted the traditional design in which light from a waveguide passes through a microchannel filled * Corresponding author: (e-mail)[email protected]. † Korea Advanced Institute of Science and Technology. ‡ Massachusetts Institute of Technology. (1) Psaltis, D.; Quake, S. R.; Yang, C. H. Nature 2006, 442, 381–386. (2) Monat, C.; Domachuk, P.; Eggleton, B. J. Nat. Photonics 2007, 1, 106– 114. (3) Lee, S. K.; Kim, S. H.; Kang, J. H.; Park, S. G.; Jung, W. J.; Yi, G. R.; Yang, S. M. Microfluid. Nanofluid. 2008, 4, 129–144. (4) Fainman, Y.; Lee, L. P.; Psaltis, D.; Yang, C. Optofluidics: Fundamentals, Devices, and Applications; McGraw-Hill: New York, 2009. 10.1021/ac102615z  2011 American Chemical Society Published on Web 12/17/2010

with an analyte solution.5-7 The optical path length in the traditional approach was reduced as the channel dimension was reduced. In addition, the traditional approach suffered from increased losses, because the light was unguided in the detection cell.8 As an alternative design, evanescent wave coupling from a solid waveguide has been demonstrated by integrating a solid waveguide along the microchannel.9,10 The evanescent wave, which has an exponentially decaying tail of a guided optical mode from the solid waveguide, could efficiently interact with the analyte solution. In addition, the selectivity could be provided in the evanescent wave sensor by the chemistry on the sensing surface.8 In order to provide the selectivity, however, the indicator molecules should be immobilized on the surface of the solid waveguide core. Once the indicator molecules were immobilized, the evanescent wave sensor based on the solid waveguides could not allow changes in their properties to suit different detection demands.11 In addition, a washing step was required when the indicator molecules were saturated. Recently, various types of liquid waveguides adopting laminar flows in microchannels have been developed.12-17 In these systems, the core fluid has a higher refractive index than the cladding fluids. Therefore, the light, which is coupled to the core fluid from external light sources or optical gain media such as fluorescent dye molecules, is guided by total internal reflection through the stratified stream along the microchannel. More (5) Wu, M. H.; Cai, H. Y.; Xu, X.; Urban, J. P. G.; Cui, Z. F.; Cui, Z. Biomed. Microdevices 2005, 7, 323–329. (6) Zhu, L.; Huang, Y. Y.; Yariv, A. Opt. Express 2005, 13, 9916–9921. (7) Llobera, A.; Wilke, R.; Buttgenbach, S. Lab Chip 2005, 5, 506–511. (8) Mogensen, K. B.; Kutter, J. P. Electrophoresis 2009, 30, S92–S100. (9) Hu, J. J.; Tarasov, V.; Agarwal, A.; Kimerling, L.; Carlie, N.; Petit, L.; Richardson, K. Opt. Express 2007, 15, 2307–2314. (10) Jiang, L. N.; Pau, S. Appl. Phys. Lett. 2007, 90, 111108. (11) Li, X. C.; Wu, J.; Liu, A. Q.; Li, Z. G.; Soew, Y. C.; Huang, H. J.; Xu, K.; Lin, J. T. Appl. Phys. Lett. 2008, 93, 193901. (12) Wolfe, D. B.; Conroy, R. S.; Garstecki, P.; Mayers, B. T.; Fischbach, M. A.; Paul, K. E.; Prentiss, M.; Whitesides, G. M. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 12434–12438. (13) Vezenov, D. V.; Mayers, B. T.; Wolfe, D. B.; Whitesides, G. M. Appl. Phys. Lett. 2005, 86, 041104. (14) Lim, J. M.; Kim, S. H.; Choi, J. H.; Yang, S. M. Lab Chip 2008, 8, 1580– 1585. (15) Lim, J. M.; Kim, S. H.; Yang, S. M. Microfluid. Nanofluid. 2010, DOI 10.1007/s10404-010-0649-5. (16) Schmidt, H.; Hawkins, A. R. Microfluid. Nanofluid. 2008, 4, 3–16. (17) Hawkins, A. R.; Schmidt, H. Microfluid. Nanofluid. 2008, 4, 17–32.

Analytical Chemistry, Vol. 83, No. 2, January 15, 2011

585

recently, Li et al.11 adopted the liquid-core/liquid-cladding (L2) waveguide to demonstrate the evanescent wave sensor. The penetration depth (dp), defined as the distance at which the evanescent wave decays to 1/e of the value at the core-cladding interface, can be estimated for a given wavelength (λ) as follows:18 dp )

λ 2π

1

√

ncore2

(1)

sin θ - ncladding2 2

where ncore and ncladding are the refractive indices of the core and cladding fluids, respectively, and the angle θ of the incident light at the core-cladding interface must be larger than the critical angle (θc) estimated from Snell’s law: θc ) sin-1

(

ncladding ncore

)

(2)

The penetration depth (dp) of evanescent wave can be used as a simple criterion to examine the feasibility of the optical detection methods based on the evanescent wave. Detection methods based on image analysis have limited resolution, despite the use of a high-resolution charge-coupled device (CCD) detector,11 because the penetration depth (dp) of the evanescent wave calculated from eq 1 was on the order of 100 nm. In addition, the evanescent wave sensor based on the L2 waveguide was sensitive to external noise because the sample fluid to be analyzed and reference fluid for normalization were forced to flow at spatially different positions in the microfluidic channel. In the present study, we developed a prototypic optofluidic evanescent wave sensor by adopting the differential optical absorption method.18-20 In particular, two external light sources with distinctly different wavelengths were coupled to the L2 waveguide. The two light sources were absorbed to different degrees by the analyte solution, with strong absorbance of the resonant beam and weak absorbance of the nonresonant beam. Light interacted with the analyte molecules dissolved in the cladding fluids by evanescent wave coupling during propagation through the core of the L2 waveguide. Unwanted effects from omnipresent external noise sources could be reduced by using the normalized absorbance calculated from the intensity ratio of the two wavelengths instead of the absolute magnitude of either signal.18 In the case of the present optofluidic evanescent wave sensor, the normalization could be facilitated with an additional light-emitting diode (LED). Therefore, the present approach may represent a more suitable option for low-end and miniaturized optofluidic sensors to construct disposable µTAS compared to the other methods adopting a bulk and expensive optical system such as a lock-in amplifier. Indeed, it is the first demonstration that an integrated optofluidic sensor adopts the differential optical absorption method. The stratified flow of liquid streams through microchannels can serve as a microfabrication tool that is applicable to a broad range of materials including metals, polymers, inorganic crystals, (18) Pollock, C. R. Fundamentals of Optoelectronics; Irwin: Chicago, 1995. (19) Yakymyshyn, C. P.; Pollock, C. R. J. Lightwave Technol. 1987, 5, 941–946. (20) Schmidt, J. R.; Sanders, S. T. Appl. Opt. 2005, 44, 6058–6066.

586

Analytical Chemistry, Vol. 83, No. 2, January 15, 2011

and ceramics.21 The evanescent wave sensor can be used as a powerful tool for real-time monitoring of the local reactions at the core-cladding interface, because the core and cladding fluids are slightly miscible due to diffusive mixing. Compared to the aforementioned traditional design in which reagents are premixed, continuous operation of the sensor can be achieved in an L2 waveguide evanescent wave sensor because an additional sample preparation process is not required. Also, the proposed approach will reduce errors caused by handling and environmental sources during the sample preparation process.10 When an additional microfluidic mixer is integrated upstream immediately before analysis, the traditional design also may be operated continuously without an additional sample preparation process. However, complete mixing within microfluidic devices is another challenging area of research because laminar flow is favorable within the microchannel.22 In addition, such modification cannot remediate the innate drawbacks of reduced optical path length and increased optical losses. In order to show the feasibility of real-time monitoring of reactions, a miniaturized pH sensor was demonstrated that used the L2 waveguide evanescent wave sensor. Most chemical reactions are significantly influenced by the pH of solution; therefore, the development of miniaturized optical pH sensors has been a demanding issue for many years. Among the various approaches, fiber-optic pH sensors have been of great interest due to their advantages such as safety for in vivo pH measurement, small size, and remote sensing capability.23-26 However, fiber-optic pH sensors require the careful removal of the cladding layer and the deposition of a pH-sensing layer or acid-base indicator molecules on the surface of the optical fiber core because of the short penetration depth (dp) of the evanescent wave. Here, we demonstrate that the L2 waveguide evanescent wave sensor can remediate the major drawbacks by use of diffusional mixing between core and cladding fluids of the waveguide. The saturation of indicator molecules, which provide the selectivity, can also be inhibited with the L2 waveguide evanescent wave sensor as the molecules are continually replenished. In addition, the different types of detection demands can be achieved simply by varying the indicator molecules dissolved in the core fluid, since the selectivity in sensing of the L2 waveguide evanescent wave sensor originates from the different degrees of absorption of the resonant beam and the nonresonant beam during analyte (or reaction product) diffusion into the liquid core. Because the L2 waveguide evanescent wave sensor can provide an in situ detection strategy and a lower interfacial loss compared to the solid waveguide sensor, the L2 waveguide evanescent wave sensor can be utilized for the real-time monitoring of various biological and chemical interactions.11,21 Also, the miniaturized optofluidic evanescent wave sensors have (21) Atencia, J.; Beebe, D. J. Nature 2005, 437, 648–655. (22) Park, S. G.; Lee, S. K.; Moon, J. H.; Yang, S. M. Lab Chip 2009, 9, 3144– 3150. (23) Ge, Z. F.; Brown, C. W.; Sun, L. F.; Yang, S. C. Anal. Chem. 1993, 65, 2335–2338. (24) Egami, C.; Takeda, K.; Isai, M.; Ogita, M. Opt. Commun. 1996, 122, 122– 126. (25) Corres, J. M.; del Villar, I.; Matias, I. R.; Arregui, F. J. Opt. Lett. 2007, 32, 29–31. (26) Gu, B.; Yin, M. J.; Zhang, A. P.; Qian, J. W.; He, S. L. Opt. Express 2009, 17, 22296–22302.

Figure 1. Optical microscopic images of the evanescent wave coupling-based optofluidic sensor at the (a, d) flow focusing region, (b, e) midstream, and (c, f) downstream. (a-c) Ethylene glycol (nEG ) 1.432) and aqueous red food coloring dye solution (nwater ) 1.331) were used as the core and cladding fluids, respectively. (d-f) Phenolphthalein (1 wt %) in ethylene glycol solution and alkaline aqueous solution (pH 11) were used as the core and cladding fluids, respectively. Here, the external light sources were on. Because the core and cladding fluids were mixed by diffusion, the boundaries between core and cladding became blurred as the flow went downstream. The different scattering of the external light at the core-cladding interface caused the different color saturation levels.

great potential in food industry, medicine, and hazardous/toxic waste management, which require real-time and in situ detection.10 EXPERIMENTAL SECTION Channel Design and Fabrication. Figure S1 (Supporting Information) shows a schematic top view and digital camera image of the evanescent wave coupling-based optofluidic sensor. The mold for the microfluidic channel was fabricated by conventional photolithography using negative photoresists (SU-8 50 and SU-8 100, MicroChem). The microfluidic channel was fabricated by a conventional soft-lithographic procedure using poly(dimethylsiloxane) (PDMS 184-A and B, Dow Corning). The negative photoresists were patterned by the method specified by the manufacturer, with slight modification to optimize the thickness and sharpness of the master. The width and height of the microfluidic channel for the L2 waveguide were 400 and 50 µm, respectively. Because multimode optical fibers (step-index fibers, numerical aperture ) 0.22, core diameter dcore ) 105 µm, outer diameter douter ) 125 µm) were used for the coupling and detection of light, an additional layer of negative photoresist (SU-8 100) thicker than 75 µm was patterned only at the couplers for light-in and light-out of the master by use of a mask aligner (BS-205, Shinumst). As shown in the inset of Figure S1b (Supporting Information), the width and height of the empty space acting as a prealigned coupler for optical fiber insertion were 126 and 200 µm, respectively. Therefore, the optical fiber could be manually inserted and precisely aligned to the L2 waveguide as shown in the inset of Figure S1b (Supporting Information) and Figure 1c,f. The master pattern was molded by a soft-lithographic procedure using PDMS for fabrication of microfluidic chips. PDMS 184-A (prepolymer) and 184-B (cross-linker) were mixed at a ratio of 10:1 (w/w) and poured on the master pattern. The PDMS was cured on the master pattern at 70 °C for 2 h. The coupler for optical fiber insertion was opened by use of a razor blade. The PDMS channel was bonded to 2 mm thick flat PDMS by oxygen

Figure 2. Absorption spectra of aqueous red and green food coloring dye solutions and of aqueous pH 10.01 buffer solution mixed with 1 wt % phenolphthalein in ethylene glycol solution (1:1 w/w ratio) measured on a conventional UV-vis spectrophotometer, and the emission spectra of a red LD (λ1 ) 635 nm) and a green LED (λ2 ) 515 nm). The absorbance of ethylene glycol was negligible in the visible light region.

plasma treatment. The flat side was bonded to a slide glass by oxygen plasma treatment. Experimental Procedures and Characterizations. A twophase stratified flow composed of the core phase (ethylene glycol with refractive index nEG ) 1.432) and the cladding phase (water; nwater ) 1.331) was generated in the PDMS (nPDMS ) 1.406) microfluidic channel by hydrodynamic flow focusing. The flow rates of the core and cladding fluids could be precisely controlled by using syringe pumps (KDS 200, KD Scientific). The flow rates of the core and cladding fluids were 1 and 8 mL/h, respectively. Figure 1a-c shows optical microscopic images of the stable two-phase stratified flow when ethylene glycol and aqueous red food coloring dye solution were used as the core and cladding fluids, respectively. Here, the external light sources were on. Because the core and cladding fluids were mixed by diffusion, the boundaries between core and cladding became blurred as the flow went downstream. The different scattering of the external light at the core-cladding interface caused the different color saturation levels. In order to demonstrate the realtime monitoring of chemical reactions at the core-cladding interface via diffusional mixing, 1 wt % phenolphthalein in ethylene glycol solution and alkaline aqueous solution were used as the core and cladding fluids, respectively. The alkaline aqueous solutions were prepared by adding dilute aqueous sodium hydroxide solution or aqueous hydrochloric acid solution to aqueous pH 10.01 buffer solution (Orion 910110, Thermo Scientific). The pH of alkaline aqueous solutions was measured with a pH meter (model 520 A+, Thermo Orion). Figure 1d-f shows optical microscopic images of the stable two-phase stratified flow when 1 wt % phenolphthalein in ethylene glycol solution and alkaline aqueous solution, pH 11, were used as the core and cladding fluids, respectively. The absorption spectra of the aqueous red and green food coloring dye solutions, and the aqueous pH 10.01 buffer solution mixed with 1 wt % phenolphthalein in ethylene glycol solution were measured on a conventional UV-visible spectrophotometer (Cary 100 Conc, Varian) and the results are shown in Figure 2. The absorbance of ethylene glycol was negligible in the visible Analytical Chemistry, Vol. 83, No. 2, January 15, 2011

587

light region. The range of measurement was 800-200 nm with 1 nm resolution. To measure the absorption spectra, 0.1 wt % aqueous food coloring dye solutions and the aqueous pH 10.01 buffer solution mixed with 1 wt % phenolphthalein in ethylene glycol solution (1:1 w/w ratio) were used. The chemical structures of the red (trisodium salt of 3-hydroxy-4-(4-sulfonaphthylazo)-2, 7-naphthalenedisulfonic acid), yellow (trisodium salt of 4, 5-dihydro-5-oxo-1-(4-sulfophenyl)-4-[4-sulfophenylazo]-1H-pyrazole-3-carboxylic acid), and blue (benzenemethanaminium, N-ethyl-N-[4[[4-[ethyl[(3-sulfophenyl)methyl]amino]phenyl] (2-sulfophenyl)methylene]-2,5-cyclohexadien-1-ylidene]-3-sulfo-, inner salt, disodium salt) food coloring dye molecules are reproduced in Figure S2 (Supporting Information). The green food coloring dye is a mixture of yellow and blue food coloring dyes. The aqueous green food coloring dye absorbed red light significantly compared to green light, whereas the red food coloring dye and alkaline phenolphthalein solution absorbed green light but not red light. Therefore, a fiber-pigtailed red laser diode (LD) (LPM-635-SMA, Thorlabs) and a fiber-coupled green light emitting diode (LED) (Black-LED-515, Prizmatix) were selected as the two light sources LD and LED with different wavelengths, λ1 ) 635 nm and λ2 ) 515 nm, respectively. The powers of LD and LED could be controlled by using a laser diode controller (LDC 205C, Thorlabs) and a black LED current controller (BLCC-02, Prizmatix), respectively. The powers of red LD and green LED for the experiments were 0.10 and 0.16 mW, respectively. The beams from the two distinct light sources were coupled into a multimode fiber by use of a 1 × 2 multimode fiber coupler and measured on a fiber-coupled optical spectrometer (OSM-100UV-NIR, Newport) as shown in Figure 2. The spectral integration time was 500 ms. The reagent consumptions for the core and cladding fluids at the given the data acquisition period (500 ms) were 139 and 1111 nL, respectively. The volume and time required for the complete replacement of sample solution were 217 nL and 98 ms, respectively. Therefore, the entire channel volume could be completely replaced between two distinct measurements. RESULTS AND DISCUSSION The total loss of the L2 waveguide is contributed from the propagation losses and the coupling losses. The optical fiberto-waveguide and waveguide-to-optical fiber coupling losses arise mainly from the size mismatch between fiber optics and L2 waveguide. The lateral dimension of L2 waveguide could be matched with the diameter of the optical fiber core simply by changing the flow rates of the constituting fluids. Also, the vertical dimension of L2 waveguide could be adjusted to the fiber optics during the fabrication of the master pattern. However, because the coupling loss is constant throughout the analysis, it does not affect the results. As sample fluids, 0-10 wt % aqueous green or red food coloring dye solutions and pH 7-12 alkaline aqueous solutions were used. Notably, there were significant differences in the measurable ranges of dye solution concentrations; about 10 wt % with evanescent wave sensors and 0.1 wt % with conventional spectrophotometers. The sensitivity of the evanescent wave sensor can be increased by using the waveguide with longer path length and lasers with different wavelengths that have stronger absorptions for the target analytes.9,10 Excessive sample dilution is not required for high concentration samples for evanescent wave sensors as 588

Analytical Chemistry, Vol. 83, No. 2, January 15, 2011

Figure 3. Measured intensity from the red LD and normalized absorbance (A) calculated from the ratio of the measured intensity (IR,x) of the red LD to that of the green LED as a function of the concentration of green food coloring dye in the presence of external noises such as temporal fluctuations in input and/or output coupling efficiency (e.g., vibrations). Here, the Peclet number was 7.35 × 104. (Inset) Optical microscopic image of the evanescent wave couplingbased optofluidic sensor with aqueous green food coloring dye solution.

demonstrated in this paper. In addition, we can monitor in real time chemical reactions via diffusional mixing between the core and cladding fluids in the evanescent wave sensor. As a result, the evanescent wave sensors provide a promising option for realtime and in situ detection devices and can find a wide range of practical applications. In addition, the proposed approach will reduce errors caused by handling and environmental sources during the sample preparation process.10 Because the green dye absorbed red light, as shown in Figure 2, the red LD was used as a light source for the resonant beam. In an ideal environment, the stratified flow in the PDMS microfluidic channel was stable because the fluid flow rate did not fluctuate. As the concentration of food coloring dye increased, more light was absorbed. Therefore, under ideal conditions the intensity of light, which was measured by the spectrometer connected to the multimode optical fiber at the light-out coupler, decreased as the concentration of food coloring dye increased. However, when unwanted external noise was present, such as temporal fluctuations in input and/or output coupling efficiency (e.g., vibrations), the stability of the stratified flow could not be sustained. As a result, the measured intensity of the resonant beam (red LD) did not show a definite trend as a function of the dye concentration (see Figure 3). However, by using the additional nonresonant beam (green LED), which was absorbed to only a small degree by the sample solution, the normalized intensity ratios (IR,x) could be obtained as follows:

IR,x )

Ix,red Ix,green

(3)

Here, Ix,red is the measured intensity of the resonant beam and Ix,green denotes the intensity of the nonresonant beam. The subscript x was used to denote the weight percent of aqueous food coloring dye solution used for calculating the intensity ratio (IR,x). Then, the normalized absorbance (A) could be calculated from the normalized intensity ratios as follows:

Figure 4. Normalized absorbance (A) calculated from the ratio (IR,x) of measured intensities of the red LD and green LED as a function of the concentrations of red and green food coloring dyes, respectively. Here, the Peclet number was 7.35 × 104.

( )

A ) -log10

IR,x IR,0

(4)

where IR,0 denotes the intensity ratio when deionized water without food coloring dye was used as a sample fluid. In contrast to the lack of a clear trend in the intensity of the resonant beam alone, the normalized absorbance (A) increased linearly with the concentration of green dye as noted from Figure 3. As mentioned above, the red dye absorbed green light but not red light (see Figure 2). Therefore, when a sample solution containing red dye was used, the measured intensity from the green LED decreased and the measured intensity from the red LD did not change significantly as a function of dye concentration. The ratio (IR,x) of the red LD to green LED intensities increased with increasing concentration of red dye. Therefore, the normalized absorbance (A) decreased linearly. This coupled optical behavior enhanced the resolution of analysis, allowing analytes to be analyzed more efficiently as demonstrated in Figure 4. As mentioned above, the L2 waveguide evanescent wave sensors have a useful structure for real-time monitoring of chemical reactions at the core-cladding interface. However, there are several constrictions for monitoring of chemical reaction by use of the L2 waveguide evanescent wave sensors. First, the core fluid should have a higher refractive index than the cladding fluid and PDMS substrate for confining and guiding of light. Second, the reaction should neither contain any sediment nor produce gases in order to maintain a stable stratified flow in the microfluidic channel. Finally, the chemical reaction should induce appreciable change in absorbance at a specific wavelength of light. The light sources to be coupled to the L2 waveguides are determined by the final constriction. In order to demonstrate the real-time monitoring of reactions at the core-cladding interface by diffusional mixing, 1 wt % phenolphthalein in ethylene glycol solution (nEG ) 1.432) and alkaline aqueous solution with different pH (nwater ) 1.331) were used as the core and cladding fluids, respectively. When the core and cladding fluids were mixed by diffusion, the chemical reaction of indicator molecules and alkaline aqueous solution

occurred (see Figure 1e,f) and the absorbance at specific wavelength increased significantly (see Figure 2). When the diffusion length (l) is longer than the penetration depth (dp) of evanescent wave, the region of L2 waveguide can be used for evanescent wave sensing. The critical angle of the incident light at the core-cladding interface can be calculated from eq 2 as 68.35°. The penetration depth (dp) of the evanescent wave calculated from eq 1 becomes shorter than 1 µm when the incident angle of light exceeds 68.63°. Here, the wavelength (λ) of light is assumed to be 515 nm because the evanescent wave emitted from green LED (λ2 ) 515 nm) is strongly absorbed by the alkaline phenolphthalein solution (see Figure 2). In addition, the highest-order mode tends to be coupled to the lower-order mode due to mode coupling. As a result, most of the guided modes have shorter penetration depth (dp) than 1 µm. Note that our L2 waveguide should support numerous transverse optical modes (i.e., V number of the waveguide is >160), due to its large refractive index difference and large length scales of the waveguide cross-section, and that each guiding modes carries only a small fraction of the total energy.18 Two-dimensional diffusion length (l) can be estimated for a given diffusion coefficient (D) as follows:27 l2 ) 4Dt

(5)

When the mass fraction of ethylene glycol is 0.5, the mutual diffusion coefficient for water and ethylene glycol is 6.80 × 10-10 m2/s.28 Then, from eq 5, the time (t) required for diffusional penetration length (l) of 1 µm is 0.368 ms. The average velocity of the fluids calculated from the mass flow rate is 12.5 cm/s. As a result, diffusional penetration over 1 µm can be achieved at about 50 µm downstream of the microfluidic channel. As noted previously, most of the guided modes have shorter penetration depth (dp) than 1 µm. Therefore, 1.995 cm out of a 2-cm-long microfluidic channel can be fully used for evanescent wave sensing and the actual volume of sample solution used for sensing was 5.25 nL, assuming that the typical penetration depth (dp) of evanescent wave was 1 µm. The present L2 waveguide safely be assumed to have a step index compared to the conventional graded-index waveguides. The most common refractive index profile for a graded-index fiber is very nearly parabolic. As a result, light rays follow sinusoidal paths down the fiber.18 However, only the edge of the core layer has an index gradient in the present L2 waveguide due to the large width of the core layer and the high flow rate of constituent fluids. Therefore, most light rays follow a rectilinear path down the fiber. The alkaline phenolphthalein solution absorbed green light but not red light (see Figure 2). The noise on the absorbance of alkaline phenolphthalein solution around 510-570 nm may originate from the saturation of absorbance during the measurement. As a result, measured intensity from the green LED decreased and measured intensity from the red LD did not change significantly as the pH of alkaline aqueous solution increased, when the alkaline aqueous solution and the phenolphthalein in (27) Bruus, H. Theoretical Microfluidics; Oxford University Press: New York, 2008. (28) Ternstrom, G.; Sjostrand, A.; Aly, G.; Jernqvist, A. J. Chem. Eng. Data 1996, 41, 876–879.

Analytical Chemistry, Vol. 83, No. 2, January 15, 2011

589

Figure 5. Normalized absorbance (A) calculated from the ratio (IR,x) of measured intensities of red LD and green LED as a function of the pH of aqueous cladding fluid when 1 wt % phenolphthalein in ethylene glycol solution was used as the core fluid. Here, the Peclet number was 7.35 × 104.

Figure 6. Normalized intensity ratios (IR,x) of measured intensities of red LD and green LED as a function of Peclet number when 1 wt % phenolphthalein in ethylene glycol solution and three distinct aqueous solutions (i.e., pH 7, 10, and 11) were used as the core and cladding fluids, respectively.

ethylene glycol solution were used as cladding and core fluids, respectively. The ratio (IR,x) of red LD to green LED intensities increased with increasing pH of alkaline aqueous solution. Here, the subscript x is used to denote the pH of alkaline aqueous solution. Therefore, the normalized absorbance (A) decreased linearly as shown in Figure 5. The relative importance of diffusion and convection can be indicated by the Peclet number (Pe ) LV/D). Here, L, V, and D denote the characteristic length, average velocity, and diffusion coefficient, respectively. The diffusional mixing is reduced as the Peclet number increases. Absorption of the resonant beam decreases, whereas the absorption of nonresonant beam is not influenced significantly as the Peclet number increases. Therefore, the normalized intensity ratios (IR,x) of measured intensities of red LD and green LED decreased as a function of the Peclet number when 1 wt % phenolphthalein in ethylene glycol solution and alkaline aqueous solution (pH 10 and 11) were used as the core and cladding fluids, respectively (see Figure 6). On the other hand, the normalized intensity ratios (IR,x) did not change significantly when neutral aqueous solution (pH 7) was used for cladding fluid (see Figure 6). Here, the Peclet number could be controlled simply by changing the flow rate of the constituent fluids. The flow rate ratio of the constituent fluids did not change throughout the experiments in order to maintain the V number of the L2 waveguide.

strongly absorbed by the analyte solution and the other had a wavelength that was barely absorbed by the solution. The differential absorption of the two beams by the analyte solutions can be used to enhance the resolution of sample analysis. Therefore, the optofluidic evanescent wave sensor enables the normalization of unwanted effects such as external noise and contamination in the optics. Moreover, the various detection demands can be satisfied and the saturation of indicator molecules can be inhibited because the constituting fluids are replenished in the L2 waveguide evanescent wave sensor. We also demonstrated that the liquid-waveguide evanescent wave sensor can be used for real-time monitoring of chemical reaction at the core-cladding interface. The L2 waveguide evanescent wave sensor can be utilized for real-time and low-cost detection for various applications including chemical interaction analysis, biological measurement, food and medicine safety testing, and environmental monitoring.

CONCLUSION A novel evanescent wave sensor based on a liquid waveguide has been developed by introducing two light sources with different wavelengths into soft lithographically featured PDMS microfluidic channels. The exponentially decaying evanescent wave interacts with analytes dissolved in the cladding fluids or products formed by chemical reactions at the core-cladding interface. The light sources were chosen such that one had a wavelength that is

590

Analytical Chemistry, Vol. 83, No. 2, January 15, 2011

ACKNOWLEDGMENT This work was supported by a grant from the Creative Research Initiative Program of the MEST/KOSEF for “Complementary Hybridization of Optical and Fluidic Devices for Integrated Optofluidic Systems”. We also appreciate partial support from the Brain Korea 21 Program. SUPPORTING INFORMATION AVAILABLE Two figures, showing the evanescent wave coupling-based optofluidic sensor and chemical structures of red, yellow, and blue food coloring dyes. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review October 4, 2010. Accepted November 29, 2010. AC102615Z